le transfert multi échelle des produits agrochimiques dans les sols

Université Joseph Fourier - Grenoble I

Spécialité : Sciences de la Planète

Synthèse des Travaux de Recherche en vue de l'obtention de

L’Habilitation à Diriger des Recherches

Le transfert multi échelle des produits

agrochimiques dans les sols tropicaux d’origine volcanique

Céline Duwig

Chargée de Recherches à l’IRD

LTHE (UMR 5564), Equipe Transpore CERMO, 460 Rue de la piscine

BP 53, 38041 Grenoble Cedex 9

Soutenue le 20 octobre2009 devant le jury composé de :

Mermoud André, Prof. EPFL Rapporteur Vanclooster Marnik, Prof. UCL Rapporteur Voltz Marc, DR INRA Rapporteur Charlet Laurent, Prof. Univ. Grenoble 1 Examinateur Valentin Christian, DR IRD Examinateur Vauclin Michel, DR CNRS Examinateur

2

Le transfert multi échelle des produits agrochimiques dans les sols d’origine volcanique

Céline Duwig, LTHE (UMR 5564), CNRS, INPG, IRD, Univ. J. Fourier, Grenoble I

Résumé : Les sols dérivés de cendres et d’éjectas volcaniques couvrent environ 1% de la surface de la terre mais ils sont généralement situés dans des régions à très forte densité de population et plus particulièrement dans des pays en voie de développement. Comme partout ailleurs, ces sols sont affectés depuis plusieurs dizaines d’années, par une intensification des activités agricoles, qui conduit généralement à une dégradation du milieu physique. Mes travaux de recherche m’ont amené à travailler sur deux types de sols d’origine volcanique : les sols ferrallitiques des îles Loyauté en Nouvelle-Calédonie et les Andosols du Mexique. Ces sols ont des propriétés physico-chimiques uniques, et sont caractérisés par la présence de charges électriques variables sur la surface des grains. L’étude du transfert des intrants appliqués à leur surface nécessite la mise en place de techniques spécifiques, essentiellement pour pouvoir étudier les interactions matrice du sol/soluté. J’ai donc développé une approche interdisciplinaire et multi échelle pour étudier l’influence des pratiques agricoles sur la dynamique de l’eau, l’érosion et le transfert des agrochimiques vers les horizons profonds. Mes travaux de recherche associent des observations quantitatives sur le terrain où sont appliquées les pratiques agricoles locales, ainsi que des expériences en conditions contrôlées en laboratoire qui permettent d’améliorer nos connaissances sur les propriétés minéralogiques, physiques et chimiques des sols d’origine volcanique et l’interaction de leurs constituants avec les produits agrochimiques. L’objectif est double : améliorer la connaissance (en couplant résultats expérimentaux et description mathématique) des processus de transfert des solutés réactifs à différentes échelles et évaluer les modèles correspondants. La première partie de ce mémoire concerne l’étude in situ des transferts d’eau et de nutriments et des pertes en sédiments en respectant les pratiques agricoles locales. Au Mexique, malgré des pentes prononcées, le ruissellement et l’érosion mesurés en surface sont faibles (Viramontes et al., 2008), due à la forte conductivité hydraulique à saturation du sol (1.5 m j-1) et à sa capacité de rétention en eau importante. Pour les deux sites d’étude (la Loma au Mexique et l’île de Maré en Nouvelle-Calédonie), le transfert d’eau et de nutriments se fait essentiellement de manière verticale, le drainage représentant plus de 50% de la pluie incidente et les pertes en nitrate atteignant 100% des apports. Alors que les deux sols étudiés possèdent une capacité d’échange anionique et retardent le transfert du nitrate vers les horizons de profondeur, les fortes intensités de pluies ainsi que la nature perméable des sols font que le nitrate est rapidement lixivié (Duwig et al., 1998). Les pertes en nitrate sont fortement diminuées lors d’apports en azote fractionnés au moment des besoins de la plante (Duwig et al., 2000). Les résultats de terrain obtenus dans les îles Loyauté ont également permis l’évaluation des capacités de prédiction du modèle mécaniste WAVE. Le modèle s’est avéré suffisamment robuste pour simuler correctement les variables d’état du sol (humidité, pression de l’eau, concentration en nitrate), alors qu’il a été utilisé dans des conditions pour lesquelles il n’avait pas été conçu (intensité de pluie importantes), surtout pour les années très humides (processus dominés par le transfert convectif). Pour une meilleure simulation des variables, il faudrait utiliser le module SURCROS de croissance de la plante (Duwig et al., 2003b). Afin de paramétrer les différents processus inclus dans le transfert des solutés dans ces sols d’origine volcanique, des études « statiques » ont été menés pour : (i) mieux connaître la réactivité de ces sols à charges variables et leur capacité de rétention, (ii) étudier la porosité et la connexion de pores afin d’estimer les chemins de transfert accéléré dans les sols. Les principaux produits d’altération dans l’Andosol sont les allophanes et la ferrhydrite (composés non cristallisés) (Prado et al., 2007), alors que ceux du sol ferrallitique sont cristallisés et dominés par les oxydes d’aluminium (gibbsite et boehmite) et de fer (goethite) (Becquer et al., 2001). La sorption des agrochimiques dans ces deux sols a été étudiée par expériences en batch, en faisant varier le pH, la concentration du soluté, en présence ou non d’ions compétiteurs, ainsi qu’avec ou sans matière

3

organique (Prado et al., 2009, en préparation1). Le réseau poral de l’Andosol a été étudié par analyse d’image de lames minces qui a permis d’obtenir les caractéristiques morphologiques des pores, ainsi que des valeurs de tortuosité et de connectivité (Prado et al., 2009). Des travaux sont en cours pour obtenir le réseau poral en 3D par tomographie X, afin d’étudier l’effet des conditions initiales et aux limites, pour à terme modéliser le transfert de solutés inertes ou réactifs à travers le réseau. Les résultats découlant de l’étude in situ m’ont amené à étudier en conditions contrôlées en colonnes de sols les différents processus entrant en jeu dans le transfert des produits agrochimiques. Différents dispositifs expérimentaux en colonnes de sol ont été mis en place pour obtenir les paramètres de transfert : (i) en condition de flux en régime permanent, par l’analyse des courbes de sorties de colonnes de sol remanié ou intact (Prado et al., 2006) et (ii) en régime transitoire, par l’utilisation des tubes de Perroux en étudiant la teneur en eau et la concentration du nitrate dans différentes sections de tubes de Perroux (Duwig et al., 2003a) ), et par l’analyse du signal des TDR dans des lysimètres (Vogeler et al., 2000). Les résultats obtenus montrent que le nitrate est retenu dans les deux sols (facteur retard de 1.2 à 1.7) et augmente avec la profondeur. Dans le sol ferrallitique, la charge positive est due aux oxydes de fer et d’aluminium présents dans tout le profil, mais cette charge est contrebalancée par la matière organique dans l’horizon de surface (Duwig et al., 2003a). Dans l’Andosol, l’augmentation du facteur retard est corrélée à la capacité d’échange anionique et à la teneur en allophanes qui augmentent en profondeur (Prado et al., 2009 soumis2). La sorption d’un pesticide (2,4-D) dans l’Andosol est non linéaire et est corrélée principalement au taux de carbone organique et au pH : de plus, la présence d’allophanes augmente la sorption en comparaison avec des sols non allophaniques (Müller et al., 2007). La comparaison des résultats en conditions statiques (batch) et dynamiques (colonnes) ainsi que les résultats obtenus sur sols remaniés et intacts ont montré que l’organisation porale joue un rôle essentiel : le transfert du nitrate en colonnes intactes d’Andosol s’est trouvé accéléré par rapport aux colonnes remaniées, dû à la présence de flux préférentiels qui a été montrée par l’utilisation d’H2

18O comme traceur de l’eau (Prado et al., 2006). Des infiltrations d’eau avec du colorant fluorescent dans des monolithes intacts ont aussi mis en évidence les chemins de flux préférentiels qui ont été reconstruits en 3D grâce au couplage de l’analyse de la concentration du colorant par spectrométrie et par traitement d’image (Duwig et al., 2008). L’approche pluridisciplinaire et multi échelle (du pore à l’échelle macroscopique) développée dans mes travaux a permis d’améliorer nos connaissances sur (i) les propriétés uniques des sols d’origine volcanique et l’interaction de ces propriétés sur les transferts de produits agrochimiques et (ii) le devenir de engrais et des herbicides appliqués sur ces sols.

1 Prado et al. Nitrate sorption depending on pH, organic matter and the presence of competitive anions in a Mexican Andosol. En préparation. 2 Prado B., Duwig C., Hidalgo C., Gaudet J.P., Etchevers Barra J., Vauclin M., 2009. Preferential flow and nitrate fate

in a Mexican Andosol. Agricultural Water Management, soumis.

4

Table des Matières

Abréviations et sigles utilisés 6 Définition des deux types de sols étudiés 7 DOSSIER ADMINISTRATIF A. CURRICULUM VITAE ET RESUME DES ACTIVITES LIEES AU METIER DE CHERCHEUR 8

1. Etat Civil 9 2. Etudes et diplômes 9 3. Expérience professionnelle 9 4. Formations complémentaires 10 5. Prix et Distinctions 10 6. Mobilité 10 7. Activités d’animation de la recherche 11 8. Activités d’évaluation de la recherche 11 9. Participation à des contrats de recherche 11 10. Collaborations nationales et internationales 12

B. ENCADREMENT ET ACTIVITES D ’ENSEIGNEMENT 13

1. Encadrement de master et projet d’ingénieur (PFE) 14 2. Encadrements de doctorats 14 3. Participations a des jurys de thèse 14 4. Etudiants 1er cycle et ITA 15 5. Activités d’Enseignement et de Vulgarisation 15 6. Séminaires 15

C. PUBLICATIONS 17 1. Revues internationales à comité de lecture 18 2. Ouvrages 19 3. Conférences 20 4. Rapports de recherche 22

DOSSIER SCIENTIFIQUE D. SYNTHESE DES TRAVAUX DE RECHERCHE 23

1. Introduction et cadre général des recherches 24 1.1. Le transfert de solutés 24 1.2. Les sols d’origine volcanique 25 1.3. La problématique « Recherche pour le Développement » 25 1.4. Une approche multidisciplinaire et multi échelle, proche des conditions in situ 27

2. Le transfert de l’eau, des sédiments et des engrais dans les sols d’origine volcanique: mesures in situ et modélisation 30

2.1. Introduction 30 2.2. Le milieu physique 30 2.3. Les dispositifs expérimentaux 34 2.4. Résultats expérimentaux 36 2.5. Evaluation du modèle WAVE 43 2 .6. Conclusions 43

3. Propriétés physiques, minéralogiques et chimiques des sols d’étude 45 3.1. Introduction 45 3.2. Eléments de théorie 45 3.3. Propriétés physiques 47 3.4. Propriétés minéralogiques 49 3.5. Propriétés chimiques 50 3.6. Conséquences quant à la classification 56 3.7. Effets des activités agricoles 56 3.8. Conclusions 56

5

4. Le transfert et la sorption des produits agrochimiques dans les sols d’origine volcanique : études dynamiques en conditions contrôlées 58

4.1. Introduction 58 4.2. Eléments de théorie : transport de solutés dans la zone non saturée du sol 58 4.3. Traceurs de l’eau 59 4.4. Expériences en régime permanent 59 4.5. Expériences en régime transitoire 64 4.6. Conclusions 67

5. Conclusion générale 67

E. PERSPECTIVES 69 1. Préambule 69 2. Cadre général 69 3. Questions scientifiques 71 4. Méthodologies et objets étudiés 71

4.1. Méthodes non invasives 71 4.2. Application à l’évoluyion spatio-temporelle de l’architecture du sol 71 4.3. Evaluation des risques de contamination du sol et des eaux souterraines par les produits pharmaceutiques 73 4.4. Transport de nutriments dans les bassins versants agricoles d’altitude 73 REFERENCES 76 ANNEXE A : principales caractéristiques physiques, minéralogiques et chimiques du Ferralsol de Maré et de l’Andosol de la Loma 83 ANNEXE B : Photos des profils de sol 84 ANNEXE C : Photos des différents dispositifs d’étude en colonne de sol 85 ANNEXE D : Approche modélisatrice mécaniste déterministe 1D 86 DEUXIEM E PARTIE : Sélections d’articles en référence avec le travail décrit dans ce rapport 90

6

Abréviations et sigles utilisés

CDE Convection Dispersion Equation

CEC Capacité d’Echange Cationique

CEA Capacité d’Echange Anionique

CO Carbone Organique

COD Carbone Organique Dissous

COP Carbone Organique Particulaire

CXTFIT Code for Estimating Transport Parameters from Laboratory or Field Tracer Experiments

EA Exchangeable Acidity

Feox, Al ox, Siox Fe, Al et Si dans les amorphes extraits après agitation avec une solution d’oxalate d’ammonium

Fed, Ald Fe, Al sous forme d’oxydes extraits par une solution de citrate-dithionite

Fep, Alp Fe, Al complexés avec la matière organique extraits avec une solution de pyrophosphate de sodium

Koc Coefficient de partage carbone organique-eau

Kd Coefficient de Distribution

LGIT Laboratoire de Géophysique interne et Tectonophysique

MET Microscopie Electronique de Transmission

MIM Mobile Immobile Model

MO Matière organique

PIE Point Isoélectrique

UNAM Universidad Nacional Autonoma de Mexico

WAVE Water and Agrochemicals in the Vadose Zone Environment

f Fraction d’eau mobile

α Coefficient de transfert de masse entre les phases mobiles et immobiles

q Vitesse de Darcy (cm s-1)

RF Facteur de retard

λ Dispersivité (cm)

θ Teneur en eau volumétrique (cm3 cm-3)

ν vitesse de pore (cm s-1)

ρ densité apparente (g cm-3)

7

Définitions des deux types de sol étudiés

(Chesworth, 2008 ; extrait de « Encyclopedia of Soil Science ») ANDOSOLS

Les Andosols sont généralement présent dans des régions volcaniques actives, ou ayant été actives. Ils présentent une extension de 1 à 2 % et sont présent dans des régions fortement peuplées. Ils ont des propriétés uniques qui les placent à part par rapport aux autres sols. Le terme « Andosol » vient du Japonais, « an » signifiant « noir » et « do » signifiant sol. Le matériau parental le plus commun des Andosols est le tephra. Le tephra est un terme générique pour les éjectas aériens volcaniques, sans distinction de morphologie, taille et composition. Par exemple, la cendre volcanique est un tephra de taille < 2 mm. La nature des tephras varie substantiellement selon la nature de l’éroption volcanique qui produit le tephra.

La dégradation rapide des tephra est ce qui distingue le plus la genèse des Andosols. Ce processus est parfois appelé andosolization. Elle résulte en une solution du sol sur saturée en Al, Si et fréquemment Fe, ou en molécules organiques, ce qui a pour conséquence la précipitation des constituants colloïdaux. La nature des matériaux parentaux des Andosols et le climat qui influence la vitesse de dégradation et par conséquence la libération de l’Al, Fe and Si, sont les facteurs dominants pour la formation des Andosols.

Les composants minéralogiques génériques dans la fraction argileuse (taille granulométrique) des Andosols sont l’ allophane, l’imogolite, la ferrihydrite, and l’halloysite.

Ces constituants colloïdaux fournissent à ces sols des propriétés particulières comme une densité apparente faible, des caractéristiques de charges variables, la thixotropie et une forte rétention du phosphate.

Le concept original dans le mot « Andosol » reflète sa couleur noire, qui résulte principalement de l’accumulation de la matière organique. De forts taux de matière organique caractérisent les Andosols bien développés, et peuvent être rencontrés à n’importe quelle profondeur, et la distribution est souvent erratique. Cette accumulation est due à deux processus : la formation de complexes allophane-humus ou de complexes métal-humus.

FERRALSOLS

Ce sont des sols rouges ou jaunes fréquents sous les tropiques humides. Leur nom dérive de L. Ferum et alumen, indiquant une minéralogie riche en fer et aluminium.

Les Ferralsols sont des sols très dégradés, fortement à excessivement lessivés, avec une minéralogie fine (taille granulométrique des argiles) dominée par la kaolinite, et les oxydes de Fe et Al. Ce sont des sols profonds à très profonds. La profondeur commune varie entre 3 et 10m. Ce sont des sols homogènes du point de vue de la couleur, de la texture, et de la minéralogie tout au long du profil, et il est donc difficile de distinguer les transitions entre les horizons B et les horizons inférieurs.

Ils dérivent de matériaux parentaux divers. Beaucoup de ces sols dérivent de matériaux non consolidés et retravaillés, qui ont pu présenter des caractéristiques ferralic dès leur origine. Ils peuvent aussi dériver de cendres volcaniques.

La formation des Ferralsols se produit sous des conditions climatiques qui favorisent la degradation et le lessivage intenses : ces conditions sont communes aux régions tropicales humides, ou avec mousson, subtropicales ou sous climat tempéré chaud.

Le processus de formation principal s’appelle la ferralitization, ou sous des régimes de dégradation moins intenses, la fermonosilization, qui conduit à un assemblage minéralogique avec un mélange de kaolinite et d’oxydes de fer, d’aluminium et de titanium. Ce sont des sols qui présentent moins de 4% de minéraux primaires dégradables dans la fraction sableuse, et ont donc une réserve minérale très faible.

Ils occupent environ 750 million ha, et sont surtout présent dans les tropiques humides, en Amérique du Sud (Brésil), Afrique (Zaïre, Centre Afrique et Afrique du Sud, Guinée et Madagascar.

8

A.A.A.A. CURRICULUM VITAECURRICULUM VITAECURRICULUM VITAECURRICULUM VITAE

9

1. Etat Civil Céline DUWIG Chargée de Recherche IRD Née à Strasbourg, le 09/12/1970 Deux enfants Adresse professionnelle : LTHE, BP 53, 38041 Grenoble Cedex 09 Tel : 04 76 63 56 50, Email : [email protected] 2. Etudes et Diplômes Doctorat de Mécanique des Milieux Géophysiques et Environnement, Université Joseph Fourier, Grenoble 1: “ Etude des transferts d'eau et de nitrate dans les sols ferrallitiques de Maré (Nouvelle-Calédonie) : risques de pollution des lentilles d'eau douce ». 8 Septembre 1998. Directeur de thèse : Michel Vauclin. DEA de Mécanique des Milieux Géophysiques et Environnement, Université Joseph Fourier, Grenoble 1 : « Etude de l’interception des pluies par une forêt de type méditerranéen, intégration dans le cycle de l’eau du bassin versant ». Septembre 1994. Diplôme d’ingénieur de ENSHMG (Ecole Nationale Supérieure d'Hydraulique et de Mécanique de Grenoble), option « Ressources en Eau et Aménagement ». Juin 1994. Langues : Français, Anglais et Espagnol (courant), Allemand (scolaire). 3. Expérience professionnelle Depuis mai 2001: Laboratoire d’étude des Transferts en Hydrologie et Environnement (LTHE, UMR 5564 (CNRS, INPG, UJF, IRD). Chargé de recherche IRD (passage en CR1 en 2006). Equipe TRANSPORE. 2001 : Ecole Polytechnique Fédérale de Lausanne (EPFL, Lausanne, Suisse), laboratoire de microbiologie des sols. Post-doctorante. 1998-2000: HortResearch Institute (Palmerston North, Nouvelle-Zélande), Environment and Risk Management Group. Post-doctorante. 1998: Ambassade de France en Nouvelle-Zélande. Assistance pour la mise en place d'un projet, financé par la CEE, d'adduction d'eau en zone rurale aux Samoa Occidentales. 1994-1998: Institut de Recherche pour le Développement (IRD, Nouméa, Nouvelle-Calédonie), Laboratoire d’Agro-pédologie. Doctorante. 1994 : CEMAGREF, Aix-en-Provence. Etude de l’interception des pluies par une forêt de type méditerranéen, intégration dans le cycle de l’eau du bassin versant. Stage de D.E.A.

10

4. Formations complémentaires Formation en Science du sol à Nancy par J. Dufey sur l’interaction entre solutés et surfaces solides. Octobre 2002. 100 heures de cours d’espagnol 2001-2002, Université Stendhal, Grenoble 5. Prix et Distinctions Thèse-Pac award, Nouvelle Calédonie, 1998. Une conférence invitée. 6. Mobilité 2007-2008 : Affectation à la UNAM (Universidad Nacíonal Autonoma de Mexico) au CIEco (Centro de Investigaciones en Ecosistemas), Morelia, Mexique. 2004-2007: Affectation au Colegio de Postgraduados au Laboratoire de Fertilité des Sols, Montecillo, Mexique. 2004 : Séjour de 5 mois comme chercheur invitée à AgResearch, Hamilton, Nouvelle-Zélande, financé par une bourse du programme « Biological Resource Management for Sustainable Agriculture Systems » de l’OCDE. 2001-2003 : Missions de courtes et de longues durées au Mexique auprès de nos partenaires, Colegio de Postgraduados, Montecillo, et IMTA, Cuernavaca. 1998-2001 : Chercheur post-doctorante à HortResearch, Nouvelle-Zélande. 1994-1998 : Doctorante à l’ORSTOM (ex IRD) à Nouméa, Nouvelle-Calédonie. 7. Activités d’animation de la recherche 2004-2008 : Formation de techniciens et d’étudiants du Colegio de Postgraduados, Montecillo, Mexique, à l’étude expérimentale et à la modélisation du transfert de solutés en colonnes de sol. Implémentation d’un sujet d’enseignement spécial au Colegio de Posgraduados sur le transfert de solutés dans les sols. 2007 : Formation de techniciens et étudiants à la mesure des caractéristiques hydrodynamiques par infiltrométrie à disque, dans le cadre du programme européen DESIRE, Chili. 2005-2006 : Création et animation d’un club JRD à Texcoco/Mexico, sur le thème des ressources en eau dans la zone centrale du Mexique. Organisation d’un atelier pour les Ecoles primaires et secondaires durant la Semaine de la Science en 2005. Participation au Forum Mondial de l’Eau à Mexico en avril 2006. 2002-2006 : Responsable du suivi hydrologique et des transferts de nutriments sur le bassin de la Loma et de Valle de Bravo au Mexique, ainsi que des travaux en laboratoire. Conception et organisation des recherches et coordination des collaborations avec les laboratoires mexicains partenaires. Formation des techniciens et étudiants des instituts mexicains partenaires à la mesure in situ et en laboratoire des variables du bilan hydrique et de nutriments. 2002-2003 : Chargée des abonnements et accès électroniques des revues scientifiques pour le LTHE

11

2000 : Formation à la gestion de l’irrigation de techniciens agricoles des Iles Loyauté. Conseil Régional de Nouvelle-Calédonie. 8. Activités d’évaluation de la recherche 2006 : Membre du comité d’évaluation du « Pacific Development and Conservation Trust » Nouvelle-Zélande. Depuis 2001 : Membre du Comité Editorial du journal « Agricultural Water Management » : révision d’environ 9 articles par an. Correctrice occasionnelle pour Geoderma, Water Resources Research, Journal of Contaminant Hydrology. Participation à des jurys de L3, Master et Doctorat. 9. Participation à des contrats de recherche Programmes et contrats que j’ai coordonnés : 2005-2008 : Programme CONACYT en collaboration avec le Colegio de Postgraduados (C.P.): « Etude du processus de transfert des pesticides dans les sols volcaniques : détermination des paramètres effectifs de transfert » (responsable scientifique : Céline Duwig) a été financé par le SEP-CONACYT (ministère de l’enseignement et de la recherche mexicain, 1% des projets retenus) à hauteur de 100 000 euros pour 3 ans. 2002-2006 : Programme AMHEX, financé par l’IRD. Etude des pertes en érosion et en nutriments dans un petit bassin versant volcanique (la Loma) au centre du Mexique. Programmes et contrats auxquels je participe ou j’ai participé : 2007-2012 : Le programme européen DESIRE (Desertification Mitigation and Remediation of Land, coordinateur, Coen Ritsema, Alterra, Pays-Bas) a été financé par l’Union Européenne (FP6 : EU Global Change and Ecosystems Programme) à hauteur de 7 millions d’euros, dont 225 000 euros pour l’IRD (responsable : Christian Prat, IRD) pour 5 ans. Ce programme regroupe plusieurs instituts européens et du pourtour méditerranéen. Plusieurs sites d’études ont été sélectionnés, en Méditerranée, ainsi qu’au Mexique (bassin versant de Morélia), Chine, Australie, Botswana et aux Etats-Unis 2007-2010 : Pogramme STREAMS (Sediment Transport and Erosion Across Mountains, responsable scientifique : Michel Esteves), financé par l’ANR – Programme Blanc 06 pour 3 ans par le LTHE en collaboration avec le Laboratoire des Sciences du Climat et de l'Environnement (UMR 1572), EDF – DTG (Département Surveillance – Service Environnement Aquatique) et HYDROWIDE (Grenoble). 2000 : Participation à la rédaction du projet Européen CROPPRO : « Development of integrated farming approaches for sustainable crop production in environmentally-constrained systems in the Pacific Regions ». Collaboration entre la Hollande, la Nouvelle-Zélande, les Samoas occidentales, Fiji et Tonga. Financé par le FP5 (coordinateur : Coen Ritsema).

2000 : « Risk assessment of discharge of organic wastes to land ». Projet financé par le Conseil Général de Gisborne (Nouvelle-Zélande).

12

2000 : « Study of nitrate leaching under intensive land use in Ohau ». Projet financé par le Conseil Général du Manawatu (Nouvelle-Zélande). 1993-1998 : Etude des risques de dégradation de la fertilité des sols et de pollution des lentilles d’eau douce. Contrats de Développement Etat-Province des Iles Loyauté (Opération n°6 DEF « Sol et nappe phréatique des Iles Loyauté). Soutien financier du Fond de la Recherche et de la Technologie.

10. Collaborations nationales et internationales

Laboratoire de Géophysique Interne et Tectonophysique (LGIT) (Grenoble, France) :

détermination des capacités d’échange des sols à charge variable (collaborateurs : Lorenzo Spadini et Laurent Charlet).

Laboratoire d’Agropédologie de l’IRD (Nouméa, Nouvelle-Calédonie) : étude des sols

ferrallitiques (Thierry Becquer, maintenant dans l’UMR « Biodiversité et fonctionnement du sol » à SupAgro à Montpellier).

Universidad Nacional Autónoma de México (UNAM), Centro de Investigaciones en

Ecosistemas (CIEco) (Morelia, Mexique): suivi des pertes en sédiments et de la qualité chimique et biologique des eaux de surface dans le bassin du Cointzio. Cycle de l’azote et du carbone (collaborateurs : Felipe Garcia, Catherine Mathuriau).

Universidad Nacional Autónoma de México (UNAM), Instituto de Geologia (Mexico,

Mexique) : analyses de pesticides et métaux lourds dans les eaux de surface et dans les sols. Mesures par Microscopie Electronique de Transmission avec le laboratoire d’analyses isotopiques pour l’analyse de l’oxygène 18 (traceur de l’eau) (collaborateurs : Christina Siebe, Blanca Prado, Pedro Morales).

Colegio de Postgraduados, Laboratorio de Fertilidad de Suelo (Montecillo, Mexique) : étude des

composants amorphes des Andosols, interactions sols/solutés, analyses chimiques des eaux et des sols, étude minéralogique des sols (collaborateurs : Jorge Etchevers, Claudia Hidalgo).

Instituto Mexicano de Tecnología del Agua (IMTA) (Cuernavaca, Mexique) : suivi des

paramètres du bilan hydrique sur le bassin de la Loma et Valle de Bravo au Mexique (collaborateurs : David Viramontes, Benjamin de Leon).

Département de Chimie des Matériaux, Faculté de Chimie et de Biologie, Université de

Santiago de Chile (Santiago, Chili): charges superficielles du sol, migration électrophorétique (collaborateur: Mauricio Escudey)

Plant and Food Research (ex HortResearch) (Palmerston North, Nouvelle-Zélande): infiltrométrie à disque, analyses des courbes de sorties des expériences en colonnes de sol, propriétés physiques des sols volcaniques (collaborateurs: Brent Clothier, Iris Vogeler). University of Auckland, Communication and Information Technology Research (Auckland, Nouvelle-Zélande): analyses d’images de lames minces, de mesures par tomographie à rayon X, de photos d’expériences en colonnes de sol avec traceurs fluorescents (collaborateur: Patrice Delmas) Agresearch (Hamilton, Nouvelle Zélande): transfert et sorption de pesticides dans les sols volcaniques (collaboratrice: Karin Müller).

13

BBBB. ENCADREMEN. ENCADREMEN. ENCADREMEN. ENCADREMENT D’ETUDIANTS ET T D’ETUDIANTS ET T D’ETUDIANTS ET T D’ETUDIANTS ET ACTIVITES D’ENSEIGNEMENTACTIVITES D’ENSEIGNEMENTACTIVITES D’ENSEIGNEMENTACTIVITES D’ENSEIGNEMENT

14

1. Encadrement de master et projet d’ingénieur (PFE) 1. Francois Thouvenel, 2009. Détermination de la courbe de rétention en eau d’un Andosol

mexicain et calibration du modèle WAVE. Stage de Master 1, Université Joseph Fourier. 2. Jorge Antonio Lopez, 2005-2008. Estudio de la sorción de plaguicidas en suelos volcanicos con

técnicas estáticas y dinámicas. Thèse de Master au Colegio de Postgraduados. Bourse du CONACYT, Mexique. Superviseurs: Céline Duwig et Jorge Etchevers. Soutenance Janvier 2009.

3. Hugo De Palmas 2006. Sorption et mouvement de l’atrazine dans un Vertisol mexicain. Stage césure de deuxième année de l’INAPG du 01/07 au 20/12/2006. Superviseur : Céline Duwig et Jorgé Etchevers.

4. Mathilde Colletta, 2005. Comportement hydrologique d’un petit bassin versant agricole. La Loma, Etat de Mexico, Mexique. Mémoire de stage de deuxième année de l’INAPG. Stage césure effectué du 15/06 au 15/12/2005. Superviseur : Céline Duwig.

5. Hélène Morin, 2005. Statistical analysis of volcanic soil column dyed images. Application to Dye Concentration calibration and estimation. Mémoire de stage de deuxième année de l’Ecole des Mines de Paris. Stage effectué du 15/07 au 15/09/2005. Superviseurs : Céline Duwig et Patrice Delmas.

6. Alexandre Dubois, 2005. Bilan hydrique et de nutriments durant 3 années du sous-bassin versant la Loma, Mexique. Mémoire de fin d’études de l’ISTOM. Stage effectué du 15/06 au 15/12/04. Soutenance en juillet 2005. Superviseur : Céline Duwig.

7. Anne Mandron, 2004. Evaluation des risques de pollution nitrique des eaux de la Loma (Mexique). Mémoire de fin d’études de l’ISTOM, Stage effectué du 15/06 au 15/12/03, soutenance le 13/07/2004. Superviseur : Céline Duwig.

8. Fernando Rojas Rojas, 2004. "Cuantificación y caracterización del escurrimiento superficial y erosión hídrica utilizando variables pluviométricas en una microcuenca de la cuenca Valle de Bravo", Thèse de Master de la Universidad Autónoma del Estado de México. Stage effectué du 15/06/03 au 15/12/04. Soutenance le 10/12/04. Superviseurs : Céline Duwig, David Viramontes, Michel Esteves.

9. Carine Raeppel, 2003. Caractérisation hydrodynamique des sols volcanique d’un petit bassin versant du Mexique. Mémoire de stage de 3ième année IUP Environnement, Géo – Ingénierie et Développement, Bordeaux. Superviseur : Céline Duwig. Participation au jury de soutenance.

2. Encadrements de doctorats Elias Raymundo Raymundo. 2004-2008. Parámetros de transporte de atrazina en un Andosol y Vertisol de México. Doctorat du Colegio de Postgraduados, Montecillo, Mexique. Etudiant du Guatemala avec une bourse de la Fondation Ford. Directeur de thèse: Céline Duwig (75%), Blanca Prado, Claudia Hidalgo. Soutenance en août 2008. Blanca Prado Pano. 2002-2006. Etude du mouvement de l’eau et du transfert réactif du nitrate dans les sols volcaniques du bassin versant élémentaire de la Loma, Mexique. Doctorat de l’Université Joseph Fourier, Grenoble. Etudiante mexicaine avec une bourse du CONACYT, Mexique. Directeurs de thèse : Céline Duwig (100%) et Michel Esteves. Soutenance Juillet 2006. 3. Participations à des jurys de thèse Participation en tant que rapporteur :

Rogerio Cichota. 2007. Modelling sulphate dynamics in soils – The effect of ion-pair adsorption. PhD in Soil Science, Massey University.

15

Participation en tant qu’examinateur:

Elias Raymundo Raymundo. 2008. Colegio de Postgraduados. Direction: Céline Duwig, Blanca Prado, Claudia Hidalgo Moreno. Blanca Prado Pano. 2006. Université Joseph Fourier. Direction : Céline Duwig, Michel Esteves.

2. Etudiants 1er cycle et ITA Etudiants 1er cycle: Florent Momal, 2007. Dye tracing and Image Analysis for Quantifying Water Infiltration in soil

column. Thèse de licence, Saint-Cyr. Superviseurs: Patrice Delmas and Céline Duwig. Liliana Terracas, 2007. Analyse de la demande en engrais azoté d’un sol volcanique (bassin versant

la Loma). Thèse de licence. Universidad Autonoma de Chapingo, Mexique. Superviseurs : Céline Duwig et Blanca Prado.

Manuel Arce Romero, 2004. Transfert de nitrate dans des colonnes de sols volcaniques. Stage de licence de Chapingo, Mexique. Superviseurs : Blanca Prado et Céline Duwig.

Personnels techniques : Aurelio Baez, 2007-2008. Ingénieur du Colegio de Postgraduados. Formation à l’hydrodynamique

des sols. Israël Castro Luna, 2005. Technicien du Colegio de Postgraduados. Formation à l’utilisation de

colonnes expérimentales de laboratoire appliquée à l’étude des transferts de masse dans les sols. Marco Antonio Varias, Luis Martin Espinoza, 2004. Techniciens de l’IMTA. Formation à la mesure

des paramètres du bilan hydrique sur le terrain. 5. Activités d’Enseignement et de Vulgarisation 2007 : Cours sur le transfert réactifs dans les sols. Théorie et Modelisation. Colegio de Postgraduados. 10h. 2001-2003 : Cours de micrométéorologie en 2ième année de l’Ecole Nationale Supérieure d’Hydraulique et de Mécanique de Grenoble, option Eau et Environnement. 6h/an. 1999 : Travaux pratiques en science du sol au niveau licence, à Massey University, Palmerston North, Nouvelle-Zélande. 15h.

6. Séminaires 2008. Estudio del transporte de plaguicidas en vertisoles y andosoles de México. Séminaire donné

au Colegio de Postgraduados, 11 Mars, Montecillo, Mexique. 2005. Estudio del proceso de migración de plaguicidas en los suelos volcánicos: Determinación de

los parámetros reales de transporte con las técnicas no destructivas. Séminaire donné à Instituto de Geografía, Universidad Nacional Autonoma de México (UNAM), 16 novembre, Mexico, Mexique.

2005. Transporte de agua y solutos en un Andisol de México: de las columnas de suelo a la cuenca hidrológica. Séminaire donné au Colegio de Posgraduados, 31 mai, Montecillo, Mexique.

2005. Impact of agricultural activities on water quantity and quality in a mexican volcanic basin. Séminaire donné au National Institute for Agro-Environmental Sciences, 16 mai, Tsukuba, Japon.

16

2004. Impact of agricultural activities on surface water quantity and quality in a mexican volcanic basin. Plant Protection Group. Séminaire donné à l'Institut AgResearch, 21 janvier, Hamilton, Nouvelle-Zélande.

2003. Efectos de las prácticas agrícolas sobre la calidad del agua en la cuenca de La Loma, Amanalco de Becerra. Séminaire donné à l'Instituto Mexicano de Technologia del Agua (IMTA), dans le cadre de la présentation des résultats préliminaires du projet AMHEX, 15 avril, Cuernavaca, Mexique.

2002. Nitrate leaching through ferrallitic soils in the Pacific. Séminaire donné à l'Institut für Hydraulik und Landeskulturelle Wasserwirtschaft dans le cadre du programme de recherche franco-autrichien (Programme AMADEUS), 13 décembre, Vienne, Autriche.

17

C. C. C. C. PUBLICATIONSPUBLICATIONSPUBLICATIONSPUBLICATIONS

18

1. Revues internationales à comité de lecture Prado B., Duwig C., Hidalgo C., Gaudet J.P., Etchevers Barra J., Vauclin M., 2009. Preferential

flow and nitrate fate in a Mexican Andosol. Agricultural Water Management, soumis. Lopez J., Duwig C., Prado B., Müller K., Etchevers Barra J., 2009. Destino de la atrazina en columnas

intactas de un Andosol Mexicano. Revista Internacional de Contaminación Ambiental, soumis. 1. Raymundo E., Nikolskii I., Duwig C., Prado B., Hidalgo C., Reyes F.G., Figueroa B., 2009.

Transporte de atrazina en un Andosol y un Vertisol disturbados de Mexico. Interciencia, 34(5), 330-337.

2. Prado B., Duwig C., Márquez J., Delmas P., Morales P., James J., Etchevers J., 2009. Image Processing-based study of soil porosity and its effect on water movement through Andosol intact columns. Agricultural Water Management, 96 (10), 1377-1386.

3. Duwig C., Delmas P., Müller K., Prado B., Morin H., Ren, K., 2008. Quantifying fluorescent tracer distribution in allophanic soils to image solute transport. European Journal of Soil Science 59: 94-102.

4. Viramontes D., Esteves M., Descroix L., Duwig C., Rojas-Rojas F., Gutiérrez A., de León-Mojarro B., 2008. Cuantificación del escurrimiento y erosión hídrica en andosoles de una microcuenca experimental en Valle de Bravo. Ingeniería hidráulica en México, 23 (3): 89-103.

5. Müller, K., Duwig C., 2007. The transport and sorption of 2,4-D in allophanic soils. Soil Science 72(5):333-348.

6. Prado B., Duwig C., Hidalgo C., Gómez D., Prat C., Etchevers J. D., Esteves M., 2007. Characterization, classification and functioning of two profiles under different land uses in a volcanic sequence in Central Mexico. Geoderma 139: 300-313.

7. Duwig C. Müller K., Vogeler, I., 2006. 2,4-D movement in allophanic soils from two contrasted climatic regions. Communications in Soil Science and Plant Analysis 37 (15-20): 2841-2855.

8. Prado B., Duwig C. , Escudey M., Esteves M., 2006. Nitrate sorption in a mexican allophanic andisol using intact and packed columns. Communications in Soil Science and Plant Analysis 37 (15-20): 2911-2925.

9. Höhener P., Duwig C., Pasteris G., Kaufmann K., Dakhel N., Harms H., 2003. Biodegradation of petroleum hydrocarbon vapors: laboratory studies on rates and kinetics in unsaturated alluvial sand. Journal of Contaminant Hydrology 66: 93-115.

10. Duwig C., Becquer T., Charlet L., Clothier B.E., 2003. Nitrate retention in a variable charge soil from the Loyalty Islands, New Caledonia. European Journal of Soil Science 54: 505-515.

11. Duwig C., Normand, B., Vauclin, M., Green, S.R., Becquer, T., Vachaud, G., 2003. Evaluation of the WAVE-model on two contrasted soil and climate conditions. Vadoze Zone Journal 2:76-89.

12. Robinson, B., Duwig, C., Bolan, N., Kannathasan, M., Saravanan, A., 2003. Uptake of arsenic by New Zealand water cress (Lepidium sativum). The Science of the Total Environment 301:67-73.

13. Becquer T., Pétard J., Duwig C., Bourdon E., Moreau R., Herbillon A.J., 2001. Mineralogical, chemical and surface properties of Geric Ferralsols of New Caledonia. Geoderma 103: 291-306.

14. Vogeler I., Green S., Nadler A., Duwig C., 2001. Measuring and Modelling Transient Solute Transport through the Rootzone using Time Domain Reflectometry. Australian Journal of Soil Research 39: 1359-1369.

15. Duwig C., , Becquer T., Vogeler I., Vauclin M., Clothier, B.E., 2000. Water dynamics and nutrient leaching through a cropped Ferralsol in the Loyalty Islands (New Caledonia). Journal of Environmental Quality 29: 1010-1019.

16. Vogeler I., Duwig C., Clothier B.E., Green S.R., 2000. Time Domain Reflectometry and Solute Transport : Measurements and Modelling. Soil Science Society of America Journal 64: 12-18.

17. Duwig C., Becquer T., Clothier, B.E., Vauclin M., 1999. A simple method to estimate anion retention in unsaturated soil. Comptes Rendus de l’Académie des Sciences, rubrique : Sciences de la terre et des planètes 328, 759-764.

19

18. Duwig C., Becquer T., Clothier B.E., Vauclin M., 1998. Nitrate leaching through oxisols of the Loyalty Islands (New Caledonia) under intensified agricultural practices. Geoderma 84: 29-43.

2. Ouvrages Duwig C., 2006. Sol. In 101 mots : L’environnement de la Nouvelle-Calédonie. Ouvrage collectif

sous la direction de Bernard Capecchi. Editions Île de Lumière, Nouméa. pp. 205-206. Viramontes D., Rojas Rojas F., Duwig C., Esteves M., 2006. Caracterización hidro-erosiva de

suelos de tipo andosol e importancia de este recurso en El Estado de Michoacán. En, "Las ciencias del agua en Morelia, aplicaciones frente a los retos del siglo XXI".

Descroix L., Esteves M., Viramontes D., Duwig C., Lapetite J.M., 2005. Eau et espace à Valle de Bravo : la bataille pour l'eau. In Descroix Luc (ed.), Estrada J. (ed.), Gonzalez Barrios J.L. (ed.), Viramontes D. (ed.). La Sierra Madre occidentale, un château d'eau menacé. Latitudes 23, Paris : IRD, p. 283-293.

Clothier B. E., Green S.R., Vogeler I., Robinson B., Mills T.M., Duwig C., Roygard J.F.K., Walter M., van den Dijssel C.W., Scotter D.R., 2004. Contaminants in the soil rootzone: transport, uptake and remediation. Waste Management. Eds A. L. Juhasz, G. Magesan and R. Naidu. Enfield, USA & Plymouth, UK, Science Publishers, Inc.: 3-34.

3. Conférences

• Conférences internationales Strozzi, A.G., Marquez J., Trujillo F., Duwig C., Prado B., Gamage P., Delmas P., 2009. 3D porous

media liquid-solid interaction simulation using SPH modeling and tomographic images. IAPR Conference on Machine Vision Applications, 20-22 May, Yokohama, Japon.

Müller K, Duwig C., Prado B., Raymundo E., Hidalgo C., Etchevers J., Siebe C., 2008: Impact of effluent irrigation on the filtering of atrazine. EUROSOIL, 24-29 August 2008, Vienna, Austria.

Duwig C., Prado B., Raymundo E., Lopez J., Hidalgo C., Müller K., Etchevers J., 2008. Herbicide fate in an allophanic soil: comparison of several experimental and data analysis techniques. EGU General Assembly 2008, 14-18 April, Vienna, Vol. 10, EGU2008-A-04906 (oral).

Delmas P., Duwig C., Müller K., Prado B., Baez A., 2008. Infiltration of a Fluorescent Dye in Allophanic Soils using Image Analysis: Calibration Procedure. EGU General Assembly 2008, 14-18 April, Vienna, Vol. 10, EGU2008-A-06000 (poster).

Raymundo E., Nikolskii I., Duwig C., Prado B., Hidalgo C., Morales P., Mendoza R., 2008. Transporte de atrazina en un vertisol de México. V Congreso Iberoamericano de Física y Química Ambiental. Mar del Plata, 14-18 April, Argentina, 2008 (oral).

Susperregui A.S., Gratiot N., Esteves M., Duwig C., Prat C. 2008. The highly turbid tropical reservoir of Cointzio (Mexico). Part II : Sedimentation at various timescale. 11th IASWS symposium, February. Australia.

Raymundo R. E., Hernández V. J., Nikolskii I., Duwig C., Prado B., Hidalgo C., Morales P., Mendoza H. R., 2007. Lixiviación de plaguicidas en suelos irrigados del Bajío Mexiquense. XIV Congreso Nacional de Irrigación, ANEI, 3-5 October, Morelia, Mexique (oral).

Lopez J., Nikolskii I., Prado B., Duwig C., Müller K., Etchevers J., 2007. Evaluación del transporte de atrazina en andosol : estudios en columnas inalteradas. XVII Congreso Latino-americano de la ciencia de suelo (CLACS), 17-21 September, Léon Guanajuato, Mexique (poster).

Müller K., Duwig C., Prado B., Raymundo E., Hidalgo C., Etchevers J., Siebe C., 2007: Impact of effluent irrigation on the filtering of atrazine. International Summer School Pesticides - Environment. Metaponto, Italy, 7-14 September 2007. GRIFA Quaderno No. 26, pp. 829-856.

Raymundo E., Nikolski I., Duwig C., Prado B., Hidalgo C., 2007. Transporte de atrazina en un andosol y en un vertisol de México. XVII Congreso Latino-americano de la ciencia de suelo (CLACS), 17-21 September, Léon Guanajuato, Mexique (oral).

20

Müller, K., Duwig C., Prado B., Raymundo E., Hidalgo C., Etchevers J., Siebe C., 2007. Impact of wastewater irrigation on atrazine transport through soil. 10th International Symposium on Soil and Plant Analysis, June 11-15, Budapest, Hungary (poster).

Müller K., Duwig C., Delmas P., Morin H., Ren K., Prado P., 2006. Quantifying fluorescent tracer distribution to image solute transport in allophanic soils. Workshop on “Preferential flow and transport processes in soil”. November 4-9, Monte Verità, Ascona, Switzerland (oral).

Prado B., Duwig C., Hidalgo C., Gómez D., Prat C., Etchevers J. D., Esteves M., Terrazas L., 2006 Cambio de las características de un suelo volcánico al pasar de un uso forestal a uno agrícola. IVth Internacional Symposium on deteriorated volcanic soils, “Satellite-Symposium” affiliated to the 18th World Congress of Soil Sciences (WCSS), 1-8 July, Morelia, Mexico (oral).

Prado B., Duwig C., Escudey M., Hidalgo C., Padilla J., 2006. Adsorción de nitrato en un suelo volcánico mexicano: competencia con el cloruro y efecto de la materia orgánica. IV Congreso Iberoamericano de física y química ambiental, 22-26 May, Caceres, Spain (poster).

Prado B., Duwig C., Delmas P., Müller K., Li J., Esteves M., 2006. Combining Image Processing and Displacement Experiments to study Solute Transport in an Andosol. EGUXXIX General Assembly, 2-7 April, Vienna, Austria (oral).

Li J., Delmas P., Duwig C., Prado B., Flores J.M., Liu J., 2005. Mexican volcanic soils porosity analysis using image processing techniques. II Taller de Procesamiento de Imágenes y Óptica (PIO2005), Guanajuato, Mexico, 21-23 November, 4 pages (oral).

Duwig C., Müller K., 2005. 2,4-D movement in allophanic soils from two constrasted climatic regions. 9th International Symposium on Soil and Plant Analysis, 30 January – 4 February, Cancun, Mexico (poster).

Prado B., Duwig C., Esteves M., 2005. Nitrate sorption in an allophanic andisol using intact and repacked columns. 9th International Symposium on Soil and Plant Analysis, 30 January – 4 February, Cancun, Mexico (poster).

Prado B., Duwig C., Hidalgo C., Gutierrez C., Esteves M., 2004. Caracterización Físico Química y Mineralogica de un Andosol mexicano. Análisis de su Influencia en el Movimiento de Aniones Activos en el suelo. XVI Congreso Latino-Americano y XII Congreso Colombiano de la Ciencia del Suelo, 27 September – 1° October, Cartagena de Indias, Colombia (oral communication).

Prado B., Duwig C., Gaudet J.P., Esteves M., 2004. Reactive anions transport through intact and repacked columns of a mexican Andosol. EGUXXVIII General Assembly, 26-30 April, Nice, France (oral communication).

Esteves M., Duwig C., Lapetite J.M., Vandervaere J.P., Viramontes D., Vauclin M., 2003. AMHEX: a soil erosion study in a volcanic mountainous region of Mexico. Érosion et ravinement en montagne, processus, mesures, modélisation, régionalisation, CEMAGREF, 15-18 October, Dignes, France (poster).

Duwig C., Normand, B., Vauclin, M., Vachaud G., Green S.R., Becquer T., 2002. Field evaluation of the WAVE-model for nitrate leaching on two contrasted soil and climate situations. Third International Conference on Water Resources and Environment Research (ICWRER) on « Water Quantity & Quality Aspects in Modelling and Management of Ecosystems », 22-25 July, Dresden, Germany (oral communication).

Duwig C., Vauclin M., 2002. An overview of measurement techniques and modeling approaches of agrochemicals transport in the vadoze zone. International Workshop on innovative soil-plant systems for sustainable agriculutral practices, organised by OECD/Agriculture Cooperative Research Programme, 3-7 june, Izmir, Turkey (invited oral communication).

Höhener P., Duwig C., Pasteris G., Dakhel N., Kaufmann K., Werner D., 2002. Biodegradation of petroleum hydrocarbon vapors in unsaturated alluvial sand. EGSXXVII General Assembly, 22-26 April, Nice, France.

Duwig C., Becquer T., Vauclin M., Clothier B.E., 2001. Nitrate retardation in a Ferralsol from New Caledonia : consequence on nitrate leaching beyond the rootzone using the WAVE model. Soil Structure, Water and solute transport. An international symposium organized by IRD in memory of Michel Rieu, 8-10 October, Paris (poster).

21

Clothier B.E., Green S., Vogeler I., Robinson B., Mills T., Duwig C., Roygard J., Walter M., van den Dijssel C., Scotter D., 1999. Contaminants in the soil of the rootzone: transport, uptake and remediation. 2nd International Conference on Contaminants in the Soil Environment in the Australasia-Pacific Region, 12-17 décembre, New Delhi, India.

Duwig C., Vogeler I., Green S.R., Clothier B.E., Becquer T., 1998. Measurement of reactive solute chemical transport through soil. XVI World Congress of Soil Science, 20-26 August, Montpellier, France (poster).

Vogeler I., Duwig C., Green S.R., Clothier B.E., Becquer T., 1998. Modelling of reactive solute chemical transport through soil. XVI World Congress of Soil Science, 20-26 August, France (oral communication).

Duwig C., Bourdon E., Becquer T., De Blic Ph., Clothier B.E., Vauclin M., 1996. Soil structure and water movement on Maré (Loyalty Islands, New Caledonia). Western Pacific Geophysics Meeting, 23-27 July, Brisbane, Australia (invited oral communication).

Duwig C., Becquer T., Clothier B.E., Vauclin M., 1996. Nitrate leaching in the Loyalty Islands (New Caledonia) under intensified agricultural practices. First International Conference on Contaminants in the Soil Environment in the Australasia-Pacific Region, 18-23 February, Adelaide, Australia (oral communication).

Corniaux C., Becquer T., Danflous J.P., Duwig C., Vernier Ph. et Dulieu D., 1997. Preservation of the environment on coral islands and intensification of agriculture : applied study on Maré, Loyalty Islands, New Caledonia. Applications of Systems Approaches at the Farm and Regional Levels, Vol. 1 (eds P.S Teng, M.J. Kropff, H.F.M. ten Berge, J.B. Dent, F.P. Lansian et H.H. van Laar). Proceedings of the Second International Symposium on Systems Approaches for Agriculture Development (SAAD), 6-8 December 1995, IRRI, Manila, Philippines. Kluwer Academic Publishers, Dordrecht and IRRI.

• Conférences nationales

Susperregui A.S., Gratiot N., Esteves M., Duwig C., Prat C., 2007. El functionnamiento hydro-sedimentario de la presa de Cointzio. Simposio Acciones y resultados para el desarrollo sostenible de la cuenca del lago de Cuitzeo, Michoacán, Oct. 2007, Morelia, Mexico.

Raeppel C., Duwig C., Viramontes D., Esteves M., Lapetite J.M., 2003. Caractérísticas físico-químicas de los suelos de la microcuenca La Loma (Cuenca Valle de Bravo). ). Memorias del XII Congreso Nacional de Irrigación, 13-15 août Zacatecas, Zac., México. Mesa 4. Gestión de recursos naturales en cuencas: 283-290 (oral communication).

Duwig C., Esteves M., Viramontes D., Lapetite J.M., 2003. Transporte de fertílizantes en la microcuenca La Loma (Cuenca Valle de Bravo). Memorias del XII Congreso Nacional de Irrigación, 13-15 août Zacatecas, Zac., México. Mesa 5. Contaminación, tratamiento y uso de agua residual en la agricultura: 93-98 (oral communication).

Clothier B.E., Green S., Roygard J., Vogeler I., Duwig C., 1999. Modelling rootzone processes : a tool for environmental risk assesment. Environmental Aspects of Pesticides Use, 15-16 November, Ruakura Research Centre, Hamilton, New Zealand (oral communication).

Duwig C., Green S., 1999. Evaluation of the WAVE model for nitrate leaching from a corn crop. "Best soil management practices for production", 11th Annual FLRC Workshop, Fertiliser and Lime Research Centre, Massey University, 10-11 February, Palmerston North, New Zealand (poster).

Duwig C., Vogeler I., Clothier B.E., Green S., 1997. Nitrate leaching to groundwater under mustard growing on soil from a coral atoll. "Nutritional requirements of horticultural crops", 10th Annual FLRC Workshop, Fertiliser and Lime Research Centre, Massey University, 4-5 February, Palmerston North, New Zealand (oral communication).

Duwig C., Becquer T., Clothier B.E., Vauclin M., 1996. Nitrate leaching in the Loyalty Islands (New Caledonia) under intensified agricultural practices. "Recent developments in understanding chemical movement in soils : Significance in relation to water quality and efficiency of fertiliser use", 9th Annual FLRC Workshop, Fertiliser & Lime Research Centre, Massey University, 4-5 February, Palmerston North, New Zealand (poster).

22

Bourdon E., Duwig C., Becquer T. Letournel F. et Blavet D., 1997. Travail du sol, état structural et enracinement des végétaux cultivés sur sols ferrallitiques allitiques des Iles Loyauté (Nouvelle-Calédonie). In : R. Pirot, S. Perret et H. Manichon (eds). Le travail du sol dans les systèmes mécanisés tropicaux. 11-12 September 1996, CIRAD, Montpellier, France (poster).

4. Rapports de recherche Duwig C. and Müller K., 2004. Pesticide fate in volcanic soils from two contrasted climatic regions.

Final report for the fellowship at Agresearch, Hamilton, New Zealand, financed by "OECD Co-operative Research Programme: Biological Resource Management for Sustainable Agriculture Systems".

Duwig C., Becquer T., Clothier B.E., Green S., 2000. Recommandations pour une meilleure gestion de l’irrigation sur les Iles Loyauté. CD-rom avec programme de bilan en eau. (Recommendations for a better irrigation management on the Loyalty Islands. CD-rom with a water balance programme). Ed. IRD et HortResearch Ltd.

Deurer M., Duwig C., Green S., Clothier B.E., 2000. Risk assessment of discharge of organic waste to land. Client Report. HortResearch Internal Report No. 2001/56.

van den Dijssel C., Deurer M., Duwig C., Vogeler I., Roygard J., Green T., Mills T., Granel T., Clothier B.E., 2000. Study of nitrate leaching under intensive land use in Ohau. Client Report. HortResearch Internal Report No. 2000/381.

Green S., Clothier B.E., Deurer M., Duwig C., 2000. Modelling the fate of simazine and nitrate from a vieyard in Marlborough. Client Report. HortResearch Internal Report No. 2000/393.

Duwig C., Becquer T., Bourdon E., Nigote W., Dubus I. et Vincent V., 1997. Suivi hydro-chimique sous différents systèmes de culture à Maré : résultats de l'année 1996. Nouméa : ORSTOM. Conv. : Sci. Vie : Agropédol., 38, 37p., multigr.

Duwig C., Becquer T., Bourdon E., Nigote W. et Taputuarai L., 1996. Suivi hydro-chimique sous différents systèmes de culture à Maré. Documents Scientifiques et Techniques, III1, 58p. ORSTOM, Nouméa.

23

D. D. D. D. SYNTHESE DES SYNTHESE DES SYNTHESE DES SYNTHESE DES TRAVAUX DE TRAVAUX DE TRAVAUX DE TRAVAUX DE

RECHERCHE 1994RECHERCHE 1994RECHERCHE 1994RECHERCHE 1994----2009200920092009

24

1. Introduction : cadre général des recherches

1.1. Le transfert de solutés

L’étude des phénomènes de transfert de molécules dans les sols est d’un double intérêt : d’un point de vue agronomique, elle permet de connaître la disponibilité des éléments fertilisants pour le couvert végétal ou l’efficacité d’un pesticide pour améliorer la croissance des plantes ; d’un point de vue plus environnemental, elle permet d’estimer les risques de contamination des sols et des eaux souterraines par les éléments inorganiques (engrais, métaux lourds) ou organiques (pesticides et autres molécules de synthèse).

Les sols représentent le compartiment clef régulant la dynamique globale des nutriments et

polluants. Le sol, en tant que zone fertile et biologiquement active, nous fournit les moyens de base pour la production alimentaire, et nous offre également la possibilité d’éliminer certains déchets bénins. De plus, le sol est la principale interface dans laquelle circule l’eau de pluie avant d’atteindre les eaux souterraines ou les eaux de surface. La qualité des eaux de ces réservoirs est donc fortement influencée par les processus de transport et les mécanismes d’échange qui prédominent lors du passage de l’eau dans le sol (Clothier et al., 1998).

Le mouvement et le devenir des solutés dans la zone non saturée du sol sont affectés par un

grand nombre de processus physiques, chimiques et microbiologiques et leur compréhension et description couvrent une large gamme de domaines scientifiques (mathématique, physique, chimie et biologie). De plus, le transfert de solutés est affecté par la non uniformité de la structure du sol et de sa composition. Cette non uniformité concerne à la fois l’espace et le temps. Les phénomènes qui en résultent sont l’instabilité du flux d’eau, l’existence possible d’écoulements préférentiels, et le non équilibre du transport de solutés. De plus, le devenir d’un élément dans le sol peut devenir complexe si celui-ci est sous plusieurs formes et dans différentes phases et est sujet à des biotransformations (Leij and van Genuchten, 2000). Dans ces dix dernières années, l’étude et la modélisation du transfert de solutés a fortement évolué (aux différentes échelles), grâce, notamment, aux progrès réalisés sur les instruments de mesure sur le terrain et en laboratoire et leur automatisation, sur les analyses chimiques (développement de méthodes spectroscopiques et microscopiques), ainsi que sur la modélisation avec le développement de modèles complexes (combinant les réactions d’adsorption, les équilibres en solution, les biotransformations, les non équilibres physiques et chimiques). Cependant, la plupart des données concernent les caractérisations physico-chimiques de la molécule et les propriétés statiques de l’environnement édaphique (adsorption, porosité), conduisant bien souvent à rechercher des corrélations entre les constituants du sol, considérés comme constants, et le comportement géochimique d’un élément ou d’une molécule (par exemple, utilisation du paramètre global comme le Kd) (Martins, 2008). Encore trop peu d’études sont réalisées en conditions de flux et/ou en prenant en compte l’environnement géochimique de la solution du sol et de la matrice.

Concernant la description mathématique, les équations aux dérivées partielles, comme

l’équation de convection-dispersion, nous ont fourni et nous fournissent toujours une description fondamentale du flux hydrique et du transport dans un sol considéré comme isotrope et uniforme. Leur application à des sols structurés a été entravée par la difficulté à décrire correctement les phénomènes de non équilibre (Clothier et al., 1998). Actuellement, les modèles mécanistes, fondés à l’échelle du pore souffrent de l’absence de mesures directes pour corroborer leurs hypothèses. Les approches macroscopiques (Ma and Selim, 1998 ; Johnson et al., 2003) comme les modèles à multi porosité sont souvent mono dimensionnels, ne considèrent pas toujours la structure du sol et nécessitent l’utilisation de paramètres empiriques. Ils sont aussi affectés par des incertitudes dues à leur paramétrisation par modélisation inverse (Roulier et Jarvis, 2003 ; Wehrer et Totsche, 2005).

25

Les progrès dans la description mathématique et la paramétrisation des processus de flux et de transport ont été limités par notre inhabilité à déterminer indépendamment la géométrie du milieu poreux et sa réactivité de surface. Pourtant, la prise en compte de l’architecture du sol est indispensable pour une prédiction correcte des temps de séjour des contaminants ou nutriments dans la zone non saturée. L’effet du transport dans les macropores a été largement documenté dans des expériences au champ et dans des lysimètres intacts (voir Flury, 1996 pour une revue bibliographique). Les pertes de pesticides par les flux préférentiels ou les flux dans les macropores correspondent généralement à moins de 1% de la dose appliquée, mais peuvent atteindre 5% (FOCUS, 2001). Or, la concentration maximale admissible d’un seul pesticide selon la directive de la Communauté Européenne pour l’eau potable est de 0.1 mg L-1. Pour une dose de 0.2 kg ha-1 et une recharge annuelle de 200 mm, cela peut impliquer une perte maximale par lixiviation de seulement 0.1% de la quantité apportée (Jarvis, 2007).

1.2. Les sols d’origine volcanique

L’étude et la modélisation des transferts de solutés dans des sols d’origine volcanique sont complexifiées par leurs propriétés particulières. Les sols formés à partir d’éjectas et de cendres volcaniques ont plusieurs caractéristiques physiques, minéralogiques et chimiques distinctes qui sont rarement présentes dans des sols dérivés d’autres roches mères. Les propriétés les plus connues des sols aux propriétés andiques que l’on trouve particulièrement dans la Ceinture de Feu du Pacifique (Pacific Ring of Fire) et les zones de subduction comme celles de la mer Méditerranée, en Indonésie, dans les Caraïbes, en Afrique, en Islande et dans les Açores), sont les suivantes : densité apparente faible, rétention en eau importante, grande friabilité, agrégats de sol très stables, charge variable, forte rétention du phosphate (Shoji et al., 1993). Ces propriétés distinctes sont largement attribuables à la formation de matériaux amorphes ou non cristallins (comme l’allophane, l’imogolite, la ferrihydrite, les complexes Al/Fe-humus), ainsi qu’à l’accumulation et la stabilité de la matière organique. La formation de matériaux non cristallins est directement reliée aux propriétés des éjectas volcaniques, soit à la dégradation rapide de verres volcaniques (Dahlgren et al., 2004). Sous les tropiques humides, l’altération d’éjectas et de cendres volcaniques peut conduire à la formation de Ferralsols, qui sont caractérisés par des fortes teneurs en oxydes de fer et d’aluminium. Les Geric Ferralsols représentent une évolution extrême de ce type de sol : ils sont pratiquement dépourvus de silice et donc de minéraux argileux, du fait de la pauvreté en silicium dans la roche mère. Les Geric Ferralsols sont présents sur les atolls coralliens surélevés des Iles Loyauté (Nouvelle Calédonie) et d’autres îles du Pacifique Sud comme dans l’archipel des Salomon, à Tahiti, à Hawaii et Fiji et dans les Caraïbes (Latham, 1980 ; Ségalen, 1995). Par ailleurs, bien que reposant sur un substrat calcaire, ces sols sont généralement non carbonatés, ce qui s’explique par l’origine allochtone des produits d’altération. Ils sont caractérisés par leur richesse en phosphore total, dont les teneurs atteignent souvent 1 à 2% de P2O5. Ce phosphore, qui provient probablement de la transformation d’un guano riche en phosphate d’aluminium, est très peu assimilable par les plantes (Tercinier et al., 1971 ; Dubus et Becquer, 2001). Les sols andiques dérivés d’apports volcaniques récents sont généralement très fertiles, alors que les sols ferrallitiques, très évolués, le sont beaucoup moins. Ils sont généralement déficients dans la plupart des nutriments pour les plantes, ou ceux-ci ne sont pas disponibles.

1.3. La problématique « Recherche pour le Développement » Les sols dérivés de produits volcaniques couvrent environ 1% de la surface de la Terre mais

ils sont généralement situés dans des régions à très forte densité de population (Leamy, 1984) et plus particulièrement dans des pays en voie de développement. Comme partout ailleurs, ces sols sont affectés depuis plusieurs dizaines d’années, par une intensification des activités agricoles, qui conduit généralement à une dégradation du milieu physique, dont les causes peuvent être : l’acidification des sols, la réduction du taux de matière organique, la salinisation, la compaction et

26

l’érosion, ainsi que la pollution des eaux de surface et souterraines (Sumner et McLaughin, 1996). Mes recherches m’ont amené à travailler sur deux types de sols d’origine volcanique : les Ferralsols des Iles Loyauté en Nouvelle-Calédonie et les Andosols du Mexique. Ces sols proviennent de matériaux volcaniques proches, mais ne sont pas au même stade de dégradation et n’ont pas le même environnement géologique et climatique.

Les Iles Loyauté incluent Maré, Lifou et Ouvéa, et sont des atolls coralliens surélevés. Ils

sont supposés être d’origine volcanique, mais la nature du sous-sol est encore inconnue, bien que du basalte ressorte localement au milieu de Maré (Bogdanov et al., 2007). Les Ferralsols des Iles Loyauté, notamment sur l’île de Maré, ne couvre que 20% de la surface totale, mais représentent près de 60% des sols cultivés car ils sont les seuls sols assez profonds présentant une extension significative pour être mis en culture. Ils sont très perméables, peu profonds et reposent sur du calcaire corallien très fracturé. Ces formations calcaires servent de réservoir à une importante nappe d’eau douce, flottant sur l’eau de mer, qui est la seule ressource en eau de l’île, à part l’eau de pluie. Sur ce soubassement calcaire, se sont déposés, selon Tercinier (1971), des matériaux d’origine volcanique (cendres et ponces) qui se sont accumulés en couches plus ou moins épaisses (1 m au maximum) entre les formations rocheuses d’origine corallienne, et qui ont donné naissance aux sols actuels. Etant donnés la nature perméable et peu profonde des sols, le sous-sol calcaire et fracturé, et l’intensité des pluies qui peut être importante (e.g. 7 épisodes de pluie d’intensité ≥ 20 mm h-1 ont été enregistrés en 35 jours en 2006) sous ce climat tropical, les lentilles d’eau douce sont particulièrement vulnérables à toute application de polluants à la surface du sol. Le besoin de maintenir un certain niveau d’autosuffisance alimentaire et d’augmenter les revenus découlant de l’agriculture ont conduit à l’intensification des pratiques. Cette évolution s’est ensuivie de la diminution des pratiques traditionnelles, comme la jachère, qui permettait de restaurer en partie la fertilité des sols. Les engrais et pesticides sont encore peu utilisés, mais bien souvent, sans vraiment de contrôles ou de formations préalables des agriculteurs qui les utilisent. Cet écosystème est fragile, et il est nécessaire d’étudier le devenir de possibles contaminants dans les sols, afin de pouvoir proposer aux agriculteurs des pratiques agricoles respectueuses de leur environnement.

Les Andosols du Mexique se situent pour la plupart sur la ceinture volcanique transversale

(Eje volcanico transveral), qui s’étend sur 900 km d’ouest à est dans la région centre-sud du pays, et qui est aussi la région la plus peuplée. Dans le passé, la ville de Mexico a été une région fertile appelée Anáhuac en náhuatl, qui réfère à la région des lacs du bassin de Mexico et à ses peuples riverains. C’est précisément l’eau et sa gestion intelligente qui ont permis à Tenochtitlan d’être un des centres politiques, religieux et économiques le plus important de Méso-amérique. Au long des 500 dernières années, les lacs ont été drainés et asséchés et les forêts ont été détruites. Actuellement, la ville de Mexico est l’une des villes les plus peuplées du monde : elle concentre 19% de la population du Mexique, environ 20 millions d’habitants, ainsi qu’une grande partie de l’activité commerciale, industrielle et politique du pays. La ville de Mexico doit faire face à un accroissement continu de la demande en eau. Elle est maintenant dépendante de ressources externes au bassin de la ville pour son approvisionnement en eau et ses besoins se montent à plus de 70 m3 s-

1. Deux des principaux problèmes que la ville de Mexico doit résoudre sont la surexploitation de l’aquifère régional et la préservation de la qualité de l’eau. Le système hydrologique Cutzamala a été créé en 1982 pour diminuer la surexploitation des aquifères dans le bassin de Mexico. Pour le moment, il apporte 19 m3 s-1 à la ville de Mexico. C’est un ouvrage d’adduction qui fut le plus ambitieux de son temps, tant par sa taille, que par ses caractéristiques techniques, conjuguant volume, dénivelé et longueur de l’adduction. Il comprend un canal de 160 km de long, et une remontée d’un dénivelé de 1000 m, grâce à 6 usines de pompage. Le lac de barrage de Valle de Bravo constitue la prise d’eau la plus importante du système, car elle apporte 6 m3 s-1. Le maintien de la qualité de l’eau du barrage de Valle de Bravo est donc indispensable pour l’approvisionnement en eau potable de la ville de Mexico. Le bassin versant correspondant (62000 ha), présente une occupation des sols majoritairement agricole, mais la ville de Valle de Bravo connaît également un important essor touristique, avec notamment l’utilisation du lac pour des

27

activités de loisir. Cette importance hydrologique et économique fait que le bassin de Valle de Bravo est stratégique et prioritaire dans les plans nationaux de développement.

Une gestion durable de ces deux milieux nécessite des informations détaillées sur leurs

propriétés. Les caractéristiques minéralogiques et physico-chimiques des Ferralsols sont peu connues (Becquer et al., 2001) et en général, que ce soit sur les Ferralsols ou les Andosols, peu d’études ont été effectuées sur les caractéristiques de ces sols en liaison avec le transfert de nutriments et de contaminants. La modification et l’intensification des pratiques culturales peuvent également affecter les facteurs environnementaux, plus spécifiquement le pH et le contenu en matière organique, ainsi que les propriétés de sorption. Comprendre les propriétés fondamentales de ces sols et le transfert des solutés appliqués à leur surface est essentiel pour développer des moyens adéquats pour leur utilisation et leur protection.

1.4. Une approche multidisciplinaire et multi échelle, proche des conditions in situ L’étude du devenir de possibles contaminants dans les sols nécessite une approche multi

échelle et multi disciplinaire, compte tenu de la forte hétérogénéité du milieu et du couplage entre les processus de transport, de rétention et de transformation microbienne. D’après Sparks (2001), pour comprendre et modéliser les différents processus, il est important de les étudier à différentes échelles, de la molécule au paysage.

Ainsi, par une succession d’allers-retours entre les différentes échelles d’étude, je me suis

attachée à répondre aux objectifs suivants : - Quelles sont les caractéristiques des sols d’origine volcanique influençant les transferts d’eau et de solutés ? - Quels sont les mécanismes prépondérants dans le transfert de l’eau et des solutés, et ce aux différentes échelles étudiées ? - Comment les pratiques agricoles locales influencent-elles le transfert des nutriments et des pesticides et quels sont les risques de dégradation et contamination du milieu ?

Pour ce faire, nous avons dû adapter les approches expérimentales et modélisatrices à la

spécificité des sols d’origine volcanique et de leur climat en menant des études : - in situ du transfert en surface du sol et dans la zone non saturée, avec le suivi des évènements pluvieux intenses et rapides des climats subtropicaux ; - en laboratoire, en statique, en sol humide (le séchage des sols volcanique transforme de façon irréversible leurs caractéristiques chimiques, physiques et minéralogiques) avec des conditions géochimiques stables (les caractéristiques de charge des sols à charge variable varient avec le pH et la force ionique de la solution) ; - en laboratoire, en colonnes de sols, avec traceurs d’eau adaptés aux sols à charge variable. En comparant les colonnes de sol remanié et intact, afin de comprendre l’impact de la destruction de la structure naturelle du sol.

Ce mémoire est divisé en trois parties. La première partie concerne l’étude in situ des transferts d’eau et de nutriments en surface et

dans la zone non saturée de ces sols. Les observations sur le terrain ont été effectuées à l’échelle de la parcelle et du petit bassin versant pour répondre aux problématiques et interrogations qui nous avaient été posées par les agriculteurs et les décideurs. Les études ont été réalisées sur les parcelles des agriculteurs, en respectant les pratiques agricoles locales.

Les résultats découlant des travaux sur le terrain m’ont amené à étudier plus en détail les

caractéristiques de ces sols, sujet de la deuxième partie, et plus particulièrement les interactions et réactions bio-géo-chimiques, en conditions statiques. Elles ont été menées pour, in fine, paramétrer les processus de transfert, en suivant des protocoles adaptés à la spécificité de ces sols.

28

Les études relatives au phénomène de transport menées en conditions contrôlées en colonnes

de sols de laboratoire sont présentées dans la troisième partie. Elles permettent de découpler les différents processus intervenant dans le devenir des agrochimiques, et d’étudier les conséquences des conditions initiales et aux limites sur le transfert.

Cette approche multi échelle est résumée dans la Figure 1.

Un deuxième volume regroupe une sélection d’articles relatifs à ces trois thèmes est fournie, en relation avec les résultats et mécanismes principaux présentés.

29

Figure 1. Etude du devenir de produits agrochimiques dans les sols : approches multi échelle et couplages de processus

Extraction par les plantes

Apport aux cultures Pollution diffuse

Photo dégradation

Ruissellement

Lixiviation Rétention Biotransfor

mation

Contamination eau surface

eau souterraine

1. Échelle métrique : Suivis des bilans de masse et des transferts in situ

2. Échelle décimétrique: Mesures des paramètres de transfert en conditions contrôlées en colonnes de sol

3. Échelles millimétriques et micrométriques: caractérisation de l’architecture poral, du mouvement d’un colorant

4. Échelles nanométriques: déterminations minéralogiques

30

2. Le transfert de l’eau, des sédiments et des engrais à l’échelle de la parcelle et du petit bassin versant: mesures in situ et modélisation

2.1. Introduction Les recherches que j’ai menées jusqu’à présent répondent à une problématique

environnementale : le risque de pollution des ressources en eau et de dégradation des sols par les produits agrochimiques utilisés lors de pratiques agricoles de plus en plus intensives. Pour les deux sites d’étude, la ressource en eau est précieuse et pas toujours facilement accessible ou disponible : forte variation saisonnière d’apports pluvieux, lentilles d’eau douce à grande profondeur, conflits sur l’utilisation de l’eau (par exemple au Mexique, entre l’usage local de l’eau et l’eau réservée pour la ville de Mexico). De plus, cette ressource est fragile vis-à-vis de tout soluté polluant appliqué à la surface du sol : les intensités de pluies sont importantes sous ces climats tropicaux ou sub-tropicaux, les sols sont très perméables, les roches sous-jacentes sont fracturées et perméables, et dans le cas de Maré en Nouvelle-Calédonie, les sols sont très peu profonds.

Les conditions de lixiviation des produits agrochimiques sont également liées à la dynamique

et la variabilité temporelle de certains facteurs environnementaux (pluie, humidité du sol, pratiques culturales comme le labour et les intrants,…). Les observations doivent donc être menées sur une certaine durée et en continu, afin d’obtenir des données les plus représentatives possibles de la variabilité spatiale et temporelle de ces facteurs. Les mesures en continu avant, durant et après les évènements pluvieux permettent également d’identifier les conséquences des conditions initiales et aux limites sur le drainage, la lixiviation et le ruissellement, les seuils de déclenchement du ruissellement, ainsi que les temps de séjour des solutés dans la zone racinaire et la zone non saturée du sol. Le suivi sur plusieurs saisons des pluies tant à Maré (1995 à 1997) que à la Loma (2002 à 2005) nous a permis d’étudier des saisons vraiment contrastées au niveau de l’intensité et la quantité de pluie, et nous avons donc pu obtenir des réponses différentes en termes de flux et de transport des nutriments.

2.2. Le milieu physique Les paragraphes ci-dessous sont en grande partie tirés des thèses de Duwig (1998) et Prado

(2006). Sites expérimentaux En Nouvelle-Calédonie, l’étude a été conduite sur l’île de Maré, l’île la plus au sud des Iles

Loyauté (Figure 2a). Les Îles Loyauté sont une série d’îles calcaires karstifiées qui se sont soulevées et déformées au cours de leur montée sur le bourrelet élastique que forme la plaque Australie, avant sa subduction dans la fosse des Vanuatu (Pacifique Sud Ouest). Des cartes de fracturation indiquent une direction majeure N110 ± 35°. L’orientation des fractures a été modélisée analytiquement comme le résultat de la déformation élastique de la lithosphère australienne avant la subduction (Bogdanov et al., 2007). Sur le plan géomorphologique, Maré dessine un large plateau central, représentant le fond d'un ancien lagon aujourd'hui émergé, entouré par une couronne de falaises (Figure 2b), correspondant à l’ancienne barrière récifale. Au niveau hydrologique, l'île est dépourvue de cours d'eau du fait de la porosité de son substrat calcaire. Les formations calcaires servent de réservoir à une importante nappe d’eau douce (Figure 2c). Sur ce soubassement calcaire, se sont déposés, selon Tercinier (1971), des matériaux d’origine volcanique (cendres et ponces), qui se sont accumulés en couches plus ou moins épaisses entre les formations rocheuses d’origine corallienne, et qui ont donné naissance aux sols actuels.

31

Le Bassin de Valle de Bravo au Mexique a une superficie de 63473 ha. Le bassin est situé à l'Ouest de l'État de Mexico entre les parallèles 19°23’00” et 19°05’30” de latitude Nord et 100°11’40” et 99°52’00” de longitude Ouest. La région d’étude fait partie de la province physiographique de l’axe néo-volcanique central Mexicain (Figure 3a). Elle est caractérisée par la prédominance de roches volcaniques de l’ère cénozoïque, de cendres volcaniques, de bancs sablonneux et de tezontle (pierre volcanique poreuse). En ce qui concerne l'aspect morphologique, prédominent les pentes escarpées, les cônes volcaniques à différents états de dégradation, les plateaux de cendres et de matériaux érodés, des bancs de lave et des petites vallées. L'altitude maximale est de 3760 m au sommet El Calvario à l'extrémité orientale du bassin et la partie la plus basse est le lac du barrage de Valle de Bravo, à 1792 m. Dans le bassin de Valle de Bravo, le bassin versant élémentaire de la Loma (Figures 3a et 3b) a été choisi pour étudier les conséquences des pratiques agricoles sur la disponibilité et la qualité de l'eau. Il est représentatif de la variabilité des conditions environnementales de climat et de végétation, ainsi que de l'utilisation du sol dans le bassin de Valle de Bravo. Il a une surface de 50.1 ha et est situé entre les altitudes 2500 à 3100 m, dans la commune d'Amanalco de Becerra, État de Mexico, entre les parallèles 19° 16' 48.6" et 19° 16' 11.3" de latitude nord et 99° 58' 13.7" et 99° 59' 13.7" de longitude ouest. Hydrologiquement, le bassin de la Loma fait partie du sous bassin versant Rio Amanalco dans le bassin versant de Valle de Bravo.

Eléments de climatologie et d’écologie végétale Le climat de Maré, et des Iles Loyauté en général, est plus homogène que celui de la Grande

Terre du fait des faibles variations altimétriques de l'île. La température annuelle moyenne varie entre 21 et 24 °C avec des maxima de l'ordre de 31 °C (entre janvier et mars) et des minima inférieurs à 10 °C (entre juin et septembre). Il existe toutefois de légères variations entre les divers points d'observations météorologiques marquées notamment par une amplitude thermique plus faible le long des côtes par rapport à l'intérieur de l'île. La moyenne des précipitations annuelles est de 1641 mm à La Roche (pour la période 1956-1985) (Blanchard, 1990). La distribution des pluies montre une certaine variabilité avec un maximum de précipitation au centre de l'île et une décroissance régulière des hauteurs d'eau lorsqu'on se rapproche des côtes. La comparaison des précipitations et des mesures d'évapotranspiration montre que la pluviométrie est généralement supérieure à cette dernière : des périodes de déficit hydrique n'apparaissent, en moyenne mensuelle, qu'entre octobre et décembre (Latham et Mercky, 1983). Toutefois, l'irrégularité des pluies est importante, avec de fortes variations spatiales et interannuelles.

Actuellement, la végétation naturelle la plus représentative de l'île est sans doute la forêt dense humide sempervirente sur calcaire (Latham et Mercky, 1983). Elle couvre actuellement environ la moitié de la surface de l'île. Cette forêt a subi d'importantes modifications d'origine anthropique (Jaffré et Veillon, 1987) liées à la pratique des cultures traditionnelles, qui consiste à utiliser le terrain pendant deux ou trois ans, à la suite de son défrichement (par brûlis ou “à la main”), et ensuite à le laisser en jachère pendant une période plus ou moins longue pouvant varier de quelques années (généralement 8 à 10 ans) à plusieurs décennies. A ces modifications liées à une utilisation agricole du milieu, s'ajoutent bien souvent celles liées au brûlage plus ou moins systématique de certaines parcelles (Becquer et al., 1993). Dans les zones fortement évoluées, comme la station de Tawaïnèdre (Figure 2a) ces modifications du milieu par l’activité humaine ont conduit à une régression importante de la couverture forestière au profit d'une savane plus ou moins herbacée à goyaviers (Psidium guajava), Lantana (Lantana camera), Imperata (Imperata cylindrica), faux gaïac (Dodonea viscosa) et faux poivrier (Schinus terebenthifolius). Dans les zones les moins favorables à l’agriculture, en raison de la pierrosité des sols, la forêt dense originelle s'est beaucoup mieux maintenue.

La Loma a un climat sub-tropical d'altitude. La température mensuelle minimale est de 3.1°C

en février, et la maximale de 25.5°C en avril. La température annuelle moyenne est de 10.7°C. La précipitation annuelle est de 1214 mm, avec la majeure partie (76%) de la pluie qui tombe entre juin et septembre. L'évapotranspiration annuelle potentielle est de 1243 mm. Les sols sont sous un

32

régime hydrique udique3 estimé par le modèle Newhall (van Wambeke et al., 1986), à partir des données climatiques sur 30 années (1962 à 1992) des stations météorologiques autour d'Amanalco.

La végétation naturelle de la région d'étude est composée de Pinus spp., Quercus spp., Abies religiosa, Agave spp., Prunus serotina spp. et Crataegus mexicana. Le stratus herbacé est abondant et comporte principalement des espèces des familles Asteraceae et Gramineae. Les principaux usages du sol de la Loma sont la forêt (46%), l’agriculture (37%) (maïs et avoine principalement) et le reste est laissé en jachère. Pendant les dernières 50 années, la part des sols cultivés a augmenté au détriment de la forêt, ce qui a provoqué une diminution de la surface de la forêt avec un changement des propriétés du sol.

Les risques de pollution des aquifères Une étude entreprise par l’Agence pour l’Eau et l’Environnement du Pacifique (A2EP, 1993)

montre que les concentrations en nitrate dans les eaux des lentilles de Maré étaient faibles et généralement inférieures à 7 mg NO3

- L-1. Certaines stations de pompage à proximité d’habitations ou de zones agricoles étaient néanmoins des eaux dont les concentrations en nitrate sont souvent plus élevées (10-14 mg NO3

- L-1). Lors du recensement agricole de 2002 (ISEE, 2003), sur les trois Iles Loyauté, près de 95% des exploitations déclarent ne pas utiliser d’intrants (engrais, amendements, pesticides). Ce phénomène peut s’expliquer par le fait que la majorité des exploitations ait de faible dimension et que les habitants soient très soucieux de préserver les lentilles d’eau douce. Les résultats et recherches bibliographiques montrent que la lentille d’eau douce de Maré semble être assez bien préservée pour l’instant. L’utilisation d’engrais chimiques est encore très limitée : la consommation est de l’ordre de 10 t d’engrais azotés par an, pour les trois îles Loyauté. La teneur en nitrate dans les eaux souterraines n’est pourtant pas nulle et peut s’expliquer par les apports d’engrais localisés, par la lixiviation de l’azote des cendres après brûlis, ou encore par la présence d’élevage (porcs et vaches) sur des surfaces restreintes.

Il existe très peu de bibliographie sur la qualité de l’eau du réservoir Valle de Bravo, et rien

sur la qualité des eaux souterraines. Dans le document « Plan de Ordenamiento Ecologico Regional de la Subcuenca Valle de Bravo-Amanalco », élaboré par le secrétariat du ministère de l’agriculture de l’Etat de Mexico (SEMARNAT, 2003), il est signalé que des facteurs comme le changement d'usage des sols, les décharges d’eaux résiduaires provenant de l'activité aquicole qui apportent 20550 t/an d'excrétas de poissons ainsi que les apports d’eaux usées au lac de barrage avec un débit de 125 L s-1 équivalent à 700 t de matière organique par an en DBO4, sont les responsables de la détérioration des eaux du barrage. Les auteurs ont relié une telle situation à plusieurs facteurs, l’agriculture étant l’un d’entre eux. Ce document indique également une réduction du 10% de la surface en forêt de 1993 à 2000. Olvera (1990) conclut que le sous-bassin de Rio Amanalco apporte à lui seul 26,5 t P/an au lac de barrage de Valle de Bravo, ce qui constitue 56.7 % du phosphore qui arrive annuellement. Olvera-Viscan et al. (1998) ont estimé que la rivière Amanalco apporte la plus grande partie du nitrate au lac. Le nitrate et l’azote organique sont les formes les plus abondantes dans l’eau, et les concentrations en ammonium sont très basses.

3 Régime d’humidité des sols dans lequel la répartition de l’eau dans le profil est telle qu’elle empêche que

l’horizon sub-superficiel sèche pendant une durée supérieure à 90 jours consécutifs dans l’année et, dans cette période, qu’il ne demeure pas sec plus de 45 jours consécutifs durant l’été

4 Demande biochimique d’oxygène

33

(a) (b)

806040200

-20-40-60-80

-100-120-140

calcaire corallien

cinérites

basaltes

contact eau douce/eau saléed’après sondages électriquescontact eau saumâtre /cinérites dans l’hypothèse de leur existence en un niveau continu

surface approximative de la lentille

niveau dela mer

(c) Figure 2 : (a) carte de Maré et localisation du site (Tawaïnèdre), (b) photo montrant le plateau

calcaire, (c) niveau de contact eau douce/eau salée (tiré de Blanchard, 1990)

(a) (b)

Figure 3 : (a) Localisation de l’axe néo-volcanique au Mexique, du bassin de Valle de Bravo et de la Loma (carte tirée de Prado et al., 2007) et (b) visualisation d’une partie du bassin de la Loma et de la parcelle P23.

P23

cm

34

2.3. Les dispositifs expérimentaux

À l’échelle de la parcelle et du bassin versant, j’ai mené des expérimentations in situ en conditions réelles. Le suivi des flux d’eau et de solutés se fait grâce à une instrumentation en partie automatisée à l'aide de centrales d'acquisition programmables. Ceci nous permet d’utiliser, conjointement et à un pas de temps très fin, des modèles mécanistes tels que WAVE pour l’étude des transferts hydriques et de nutriments ou polluants. Les interactions sols / solutés sont ainsi étudiées spécifiquement dans un contexte agro-pédologique et climatique donné.

L'étude effectuée sur l'île de Maré, est localisée sur le site de la station agronomique du Centre d'appui au Développement Agricole de Tawaïnèdre (Figure 2b). Cette station située à environ 2 km de la tribu de Tawaïnèdre (coordonnées = S 21°30' - E 168°4', altitude 41 m), est constituée de différentes parcelles, dont la surface totale fait environ 32 ha. Il s’agit de sols ferrallitiques allitiques, qui sont les plus intéressants du point de vue de l’agriculture.

Trois parcelles ont été étudiées durant les trois cycles culturaux de 1995, 1996 et 1997 (de

janvier à mai) : (i) une parcelle sol nu fertilisée (B), (ii) une parcelle en rotation maïs-patate douce/maïs-patate douce (C), et (iii) une parcelle sous graminées pérennes (G). Ces trois parcelles reçoivent une fumure minérale "haute" (F2) qui correspond à un apport de 800 kg ha-1 de 13-13-21 (64 kg ha-1 sous forme de N-NH4

+ et 40 kg ha-1 sous forme N-NO3-). Elles sont soumises à un

travail du sol "lourd" (labour et semis mécanisés), tous les ans pour les parcelles B et C, et au semis (en 1993) des graminées pour G.

Les flux d’eau et de solutés dans le sol ont été étudiés grâce à l’installation de sondes TDR (2 séries par parcelle à 10, 20, 30 et 40 cm de profondeur), de tensiomètres (2 séries par parcelle à 10, 20, 30 et 40, 50 cm de profondeur), et de bougies poreuses (16 bougies à 10 cm et 16 à 40 cm par parcelle).

La Figure 4 présente les différents instruments. Les données sont relevées manuellement pratiquement tous les jours durant la saison des pluies. Un pluviomètre à augets basculeurs relié à une centrale d'acquisition a permis d'enregistrer le volume de pluie tombé instantanément. Une station climatologique enregistre le rayonnement solaire global, le rayonnement net, la température, l’humidité relative, la vitesse et la direction du vent.

-400

tensiomètres Guides d’ondes TDR bougies poreuses

céramiques poreuses

pompe à videcapteur de pression

connecteurs

0

10

20

30

40

50cm

aiguille hypodermique

bouchon auto-cicatrisant tube capillaire

Figure 4 : station de mesure du transfert hydrique et de solutés dans le sol à Maré. Photo de la

parcelle sol nu.

A la Loma, au Mexique, trois parcelles, délimitées et isolées pour collecter les flux en surface, ont été étudiées durant les saisons des pluies de 2002, 2003, 2004 et 2005 (juin à octobre) : (i) une parcelle sous forêt, dans la partie la plus pentue (36 %) du bassin (PB01), (ii) une parcelle sous maïs avec une pente de 28 % (P23), suivie à partir de 2003 et (iii) une parcelle sous maïs, cultivée en avoine en 2002 avec une pente de 9 % (P67) (voir Figure 5). Les deux parcelles de maïs reçoivent un apport d’engrais de 90 kg N ha-1 par an et 290 kg P ha-1 sous forme phosphate d’ammonium.

35

Les flux d’eau dans le sol sont calculés grâce à l’installation de 3 sondes reflectometriques CS615 (Campbell) de 30 cm de long entre 0-30, 30-60 et 60-90 cm de profondeur sur chaque parcelle. Les données sont enregistrées toutes les 30 min par une centrale Campbell. Trois pluviomètres à augets basculeurs sur le bassin enregistrent la pluie instantanée et une station micro météorologique installée en 2003 enregistre chaque heure la direction et la vitesse du vent, l’humidité relative, la température de l’air, la radiation solaire globale et calcule l’évapotranspiration potentielle selon l’équation de Penman-Monteith. Le ruissellement sur les parcelles agricoles est mesuré par un canal Ventury de type H associé à un limnigraphe installé au point aval. Les données du limnigraphe sont enregistrées toutes les 10 secondes par un datalogger de type HOBO. Le ruissellement de la parcelle sous forêt est collecté dans un fut de 200 L et mesuré chaque jour. La quantité de sol érodé est calculée en mesurant la concentration en sédiments d’échantillons d’eau prélevés en surface pendant l’épisode ruisselant. A l’exutoire du bassin versant sont installés un piézomètre jusqu’à 5 m de profondeur, et un seuil Parshall dans le canal (voir Figure 5). Un limnigraphe thalimède (OTT) mesure la hauteur d’eau dans le canal. Des prélèvements manuels d’eau ruisselée à la sortie des canaux Ventury durant les événements ruisselants ont permis de calculer la concentration en sédiments et la quantité de nutriments (azote total, phosphore total, carbone organique, nitrate et ammonium analysés dans les eaux brutes et eaux filtrées) perdus par ruissellement et par érosion.

(a)

(b) (c) Figure 5 : (a) carte de localisation des sites de suivi du bilan en eau et en solutés sur le bassin de la Loma, (b) canal Ventury sur P23, (c) un seuil Parshall dans le canal à l’exutoire du bassin.

PB01

36

Des prélèvements réguliers de sol (en général après chaque pluie donnant lieu à ruissellement)

et analysés pour la concentration en nutriments associés au calcul du drainage ont été utilisés pour déterminer les pertes en nutriments par lixiviation.

2.4. Résultats expérimentaux

Dynamique de l’eau La conductivité hydraulique sur Maré (Duwig et al., 1998) a été déterminée en combinant des

mesures par infiltrométrie à disque sur le terrain et la méthode du plan de flux nul (Vachaud et al., 1981). A la Loma, celle-ci a été déterminée uniquement par infiltrométrie à disque (Duwig et al., en préparation5). A Maré, la conductivité à saturation en surface, Ks, est de 9.2, 4.2 et 8.3 × 10-5 m s-1 pour les parcelles B,G, C respectivement. A la Loma, Ks en surface est de 1.7, 1.8, et 1 × 10-5 m s-1 pour les parcelles PB01, P23 et P67 respectivement. La conductivité à saturation à la Loma est donc plus faible qu’à Maré, par contre, la réserve en eau utile du sol est beaucoup plus importante: elle est de 36 mm sur Maré entre 0 et 50 cm, avec peu de variation entre les parcelles alors qu’elle est de 226, 179 et 186 mm entre 0 et 90 cm, pour PB01, P23 et P67 respectivement à la Loma, soit une réserve en eau de 7 mm pour 10 cm de sol sur Maré contre 20 mm ou plus pour 10 cm de sol à la Loma.

La pluviométrie sur les deux sites a été très variable d’une saison des pluies à l’autre. Le Tableau 1 donne la pluie cumulée rapportée à 100 jours au cours des saisons des pluies (de janvier à mai à Maré et de juin à octobre à la Loma). Tableau 1 : pluie et évapotranspiration potentielle rapportées à 100 jours lors des saisons des pluies (janvier à mai sur Maré, et juin à octobre sur la Loma) des différentes années étudiées. Drainage total au cours de la saison des pluies exprimé en pourcentage de la pluie incidente. Site Maré, Nouvelle Calédonie La Loma, Mexique Année 1995 1996 1997 2002 2003 2004 2005 Pluie/100 jours, mm 488 1099 352 743 751 610 401 ETP/100 jours, mm 377 313 338 326 265 217 311 Drainage, % pluie B 57 72 24 PB01 31 44 36 20 Drainage, % pluie C 53 69 17 P23 / 62 58 29 Drainage, % pluie G 54 70 22 P67 42 62 64 27

L’intensité de la pluie : A Maré, en moyenne, on a enregistré 14 fois par an des pluies supérieures à 20 mm j-1. En

mars 1996, le cyclone Béti (28/03) a entraîné 238 mm en 4 jours, et au plus fort du cyclone, 139 mm en 24 h. Durant la saison pluvieuse de 1996, on a enregistré 7 fois des intensités de pluie supérieures à 20 mm h-1. A la Loma, les événements pluvieux de moins de 10 mm en total représentent 73% des pluies annuelles. La durée moyenne d’un événement est de 108 min (déviation standard de 92). Les intensités de pluie maximales en 10 min varient entre 12 et 98 mm h-1, avec une moyenne de 27 mm h-1. Pour pouvoir comparer les deux sites, et étant donné que les pluies instantanées ne sont pas disponibles sur Maré, les cumuls journaliers ont été utilisés. A Maré, le nombre de jours de pluie où le cumul est supérieur à 20 mm j-1 est de 12% des jours avec des pluies enregistrées, alors qu’il n’est que 9 % pour la Loma. Avec un seuil de 30 mm j-1, les fractions passent à 8% pour Maré et 3% pour La Loma. La pluviométrie moyenne durant la saison

5 Duwig C., Vandervaere J.P. et al. Consequences of soil properties and land use on hydrodynamic parameters of volcanic soils of the Mexican Highlands. En preparation pour Geoderma;

37

des pluies est plus importante à la Loma, mais les intensités maximales certainement plus fortes à Maré.

Le drainage a été calculé par la loi de Darcy en 1995 et 1996 sur Maré (Duwig et al., 1998), et par bilan hydrique en 1997, ainsi que sur la Loma (Duwig et al., en préparation6). A Maré, on cumule les risques de pertes de contaminants au bas de la zone racinaire : pluies intenses fortes, sol peu profond et réserve utile faible, et conductivité hydraulique à saturation très forte. Par conséquent, le drainage cumulé représente une part un peu plus importante du bilan hydrique que sur la Loma, pour les années pluvieuses (entre 53 et 70% de la pluie incidente sur Maré (Figure 6a) et entre 31 et 64% sur la Loma (Figure 6b), toutes parcelles confondues). Les années 1997 et 2005 sur Maré et la Loma respectivement, représentent des années particulièrement sèches, et la part du drainage dans le bilan hydrique est largement diminuée (Figures 6a et 6b).

A Maré, une part importante du drainage résulte de quelques événements pluvieux intenses,

en janvier 1995, mars 1996 et février 1997. Ces événements représentent 20 à 23 % du cumul de pluie des 5 premiers mois de l’année, et ils ont conduit à 40-46% du drainage total durant ces 5 mois. Ces événements intenses ne sont bien sûr pas prévisibles et il est impossible de modifier les pratiques culturales pour limiter le drainage. La faible réserve utile du sol et ces fortes intensités de pluie expliquent qu’il y ait peu de différences entre les parcelles sol nu et avec végétation. Dans ces conditions, la consommation de la plante est négligeable, et le sol ne joue pas de rôle tampon vue la faible réserve utile.

A la Loma, la variation de stock reste identique quelle que soit la pluviométrie de la saison,

confirmant la forte capacité de rétention des Andosols. Le drainage est rapide puisque les tensiomètres à 90 cm de profondeur répondent au maximum en une journée après des pluies supérieures à 20 mm. Le drainage est plus faible sous forêt que sous maïs dû à l’interception des pluies (14.4%, Viramontes et al. 2008), à l’évapotranspiration et la réserve en eau utile plus importantes. Les deux parcelles de maïs enregistrent des drainages équivalents, la variation interannuelle étant bien plus importante que la variation entre les parcelles.

Les coefficients de ruissellement à la Loma sont faibles malgré les pentes prononcées, entre

1.8 et 4.5% selon les années et les parcelles (pas de ruissellement à Maré en absence de topographie). Les coefficients de ruissellement sont nuls sous forêt. Viramontes et al. (2008) expliquent que dans la zone d’étude, le ruissellement se développe quand le profil de sol est saturé, à partir de zones préférentielles contributives qui ont des teneurs en eau élevées.

A la Loma, Colletta (2005) a tenté d’extrapoler les résultats des parcelles à la totalité du

bassin versant, en utilisant la carte d’occupation des sols, et en considérant que la moyenne du drainage obtenu sous P23 et P67 était représentatif du drainage sous parcelle cultivée du bassin, et le drainage sous PB01 représentatif du drainage sous la surface boisée. En calculant le bilan hydrique sur la totalité du bassin, en 2003, on peut comparer le ruissellement calculé avec celui mesuré par le limnigraphe installé à l’exutoire du bassin. On a une différence de hauteur ruisselée de 11.6 mm mesurée contre 19.5 mm calculée, ce qui reste dans le même ordre de grandeur.

6 Duwig C. et al. Agricultural impact on a small volcanic watershed from Mexico: Part 1. Water balance and sediment transport. En preparation.

38

(a) A Maré : du 05/02 au 05/05 pour les 3 années

(b) A la Loma : du 17/06 to 10/09 pour les 4 années

Figure 6 : (a) Composantes du bilan hydrique à Maré pour les trois années et les trois parcelles étudiées (b) et à la Loma pour les trois parcelles et les quatre années

Pertes en sol et en nutriments par érosion Les Andosols matures sont considérés comme étant résistants à l’érosion hydrique grâce à

leur forte conductivité hydraulique et la forte stabilité des agrégats (Nanzyo et al., 1993a). Mais quand la végétation est enlevée, les sols nus deviennent susceptibles à l’érosion hydrique et éolienne à cause de leur faible densité apparente (Kimble et al., 2000). Au Mexique, les agriculteurs divisaient leurs parcelles par des haies d’agaves pour diminuer la perte en sol, mais ces pratiques sont de plus en plus abandonnées à cause du labour mécanisé et de la colonisation des agaves par les taupes. Rojas (2004) et Viramontes et al. (2008) trouvent des pertes de sol par érosion de 6.5, 4.4 et 1.6 t/ ha pour P23, P67, et à l’exutoire du bassin en 2003, et nulles sous forêt. Ces pertes sont dues majoritairement à trois évènements pluvieux qui ont donné lieu à 40% du ruissellement et entre 50 et 90% de l’érosion. Bien que les pertes en érosion soient faibles, ce travail montre l’importance de garder les Andosols avec un couvert végétal. Ceci fait que l’Andosol garde sa structure (qui se modifie de façon irréversible avec la dessiccation) et que les agrégats restent stables. Dixon et Schultze (2002) mentionnent que les sols allophaniques offrent une résilience à l’érosion. Ceci est dû à la formation d’agrégats stables entre les allophanes et la matière organique, qui sont résistants aux effets mécaniques des gouttes de pluie et au ruissellement. De plus, les conductivités hydrauliques élevées généralement observées pour les Andosols diminuent le taux de ruissellement.

0

100

200

300

400

500

600

700

800

900

B C G

Dstock

ETR

Drainage

1995mm

0

100

200

300

400

500

600

700

800

900

B C G

1996

0

100

200

300

400

500

600

700

800

900

B C G

1996

0

100

200

300

400

500

600

700

800

PB01 P23 P67

∆stock

Runoff

ETR

Drainage

Water balance 2002

0

100

200

300

400

500

600

700

800

PB01 P23 P67

Water balance 2003

0

100

200

300

400

500

600

700

800

PB01 P23 P67

Water balance 2004

0

100

200

300

400

500

600

700

800

PB01 P23 P67

Water balance 2005

39

Les concentrations en carbone organique dissous (COD) et particulaire (COP) ont été mesurées dans les eaux de ruissellement (Figure 7) et dans les eaux de surface à l’exutoire du bassin versant de la Loma (Prado et al., 2007). Les concentrations en COP sont stables, avec une concentration moyenne de 7.9% (déviation standard 0.8). La perte totale en COP calculée sur la saison des pluies 2003 est de 600 kg C ha-1. Cela représente entre 15 et 20% du stock en carbone organique du premier cm de sol. Les concentrations en COD (dans l’eau de ruissellement filtrée à 0.7 microns) sont plus variables, mais la perte totale ne dépasse pas 10 kg C ha-1 pour la saison des pluies 2003. Pourtant, étant donné les pratiques agricoles à la Loma où la plupart des résidus ne sont pas incorporés au sol mais servent à l’alimentation du bétail, à la fin de la saison de culture, le pool de matière organique n’est pas renouvelé et l’horizon organique perd peu à peu sa fertilité, ce qui explique le taux de CO du sol de la Loma plus faible que d’autres sols volcaniques aux propriétés similaires. Le taux de CO des premiers 15 cm du sol à la Loma est de 54 g kg-1, alors qu’il et de 100 g kg-1 dans les Andosols de l’Ile de Java (Van-Ranst et al., 2002), de 140 g kg-1 dans les Andosols du Sud de l’Equateur (Buytaert et al., 2002), et de 95 g kg-1 dans les sols volcaniques du Chili (Escudey et al., 2004).

Figure 7 : Concentrations moyennes en carbone organique dissous et particulaire dans les eaux de ruissellement de la parcelle P23 en 2003 (tiré de Prado et al., 2007)

De la même manière, des fortes concentrations en azote organique (calculé par différence

entre l’azote total et l’azote inorganique) et en phosphore total ont été observées dans les eaux de ruissellement, et par différence entre les concentrations des eaux brutes et filtrées, on peut en déduire que ces concentrations proviennent des sédiments transportés. L’érosion est relativement faible au niveau du bassin, mais la présence d’une couche de sol de surface très riche en matière organique et nutriments fortement fixés au sol comme le phosphore induit des concentrations élevées dans les eaux de ruissellement (phosphore total dans l’eau brut atteint des valeurs de 200 mg L-1). On retrouve ces mêmes concentrations élevées en azote organique dans les eaux de surface à l’exutoire du bassin de la Loma et d’Amanalco (sous bassin de Valle de Bravo, dans lequel se situe le bassin de la Loma) (Figure 8). Par contre, les concentrations en azote minéral sont toujours faibles (de l’ordre de 4 à 5 mg L-1), que ce soit à l’exutoire du bassin de la Loma (Figure 8) ou dans la totalité du bassin de Valle de Bravo (analysées en 2005). Donc, malgré un ruissellement et une érosion relativement faibles, quelques fortes crues en début de saison suffisent à contaminer les eaux de surface en azote organique et phosphore en début de saison. Ces résultats sont préliminaires et la valorisation des données est en cours.

40

Figure 8 : pluie, lames d’eau ruisselées, niveau de la nappe souterraine, concentrations en nitrate, ammonium et azote total dans les différentes sources d’eau (source, eau souterraine, eau de ruissellement et à l’exutoire du bassin versant d’Amanalco) durant la saison des pluies 2003 à la Loma

0

10

20

3040

50

60

70

28/4 8/5 18/5 28/5 7/6 17/6 27/6 7/7 17/7 27/7 6/8 16/8 26/8 5/9

Inte

nsi

ty,

mm

d-1RAINFALL

0.0

1.0

2.0

3.0

4.0

5.0

Inte

nsi

ty,

mm

d-1 RUNOFF

012345678

Con

cent

ratio

n, m

g N

L-1

spring

groundwater

storm flow

Rio Amanalco

NITRATE

0123456789

10

Con

cent

ratio

n, m

g N

L-1

spring

groundwater

storm flow

Rio Amanalco

AMMONIUM

0

5

10

15

20

25

30

27/4 7/5 17/5 27/5 6/6 16/6 26/6 6/7 16/7 26/7 5/8 15/8 25/8 4/9

Co

nce

ntr

atio

n,

mg

N L

-1

spring

groundwater

stormflow

Rio Amanalco-NF

TOTAL NITROGEN

0

50

100

150

200

250

300

Le

vel,

cm

GROUNDWATER

41

Dynamique des nutriments par lixiviation A Maré, étant donné le peu de différences entre les quantités drainées sous les différentes

parcelles, c’est essentiellement la concentration en nutriments au bas du profil de sol (40 cm) qui fait la différence. L’apport d’engrais a été réalisé en une seule fois en 1995 au moment du semis : les concentrations en N-NO3

- et en K+ diminuent fortement après la première pluie. On note également une concentration plus forte des deux éléments sous maïs que sous prairie, car le maïs est trop petit pour consommer de manière significative. En 1996 et 1997, l’apport d’engrais a été fractionné en deux fois.

Le résultat fondamental de cette étude est que malgré une pluviométrie plus importante en 1996 qu’en 1995, la quantité totale de nutriments lixiviée au bas de la zone racinaire est plus faible en 1996 sous maïs (elle est plus importante sous sol nu) : 50 kg N ha-1 en 1996 contre 150 kg N ha-1 en 1995 sous maïs, ce qui met en évidence l’effet positif du fractionnement d’engrais et de l’apport au moment où la plante est capable de consommer (voir Figure 9). Les pertes en nitrate peuvent surpasser les 100% des apports, montrant que le nitrate stocké par le sol ou résultant de la nitrification peut aussi être perdu par lixiviation. La nitrification calculée par bilan sur la parcelle sol nu sans fertilisation, peut monter à 60 kg N-NO3

- ha-1 en 80 jours en 1996 (année la plus humide). On observe également une lixivation retardée du K+ par rapport à celle du NO3

-. La quantité lixiviée de K+ reste cependant assez importante par rapport à celle dans d’autres sols tropicaux, en raison de l’absence d’argiles 2 :1. Le K+ est donc essentiellement retardé par les forces électrostatiques sur les substances humiques (Duwig et al., 2000).

Figure 9 : Lixiviation cumulée du nitrate et du potassium à 40 cm de profondeur pour 1995, 1996 et 1997 pour les trois parcelles étudiées. En 1995, l’engrais (N) a été appliqué le 11 janvier. En 1996 et 1997, il a été appliqué en deux fois, N1 (mi fevrier) et N2 (mi mars).

A la Loma, les calculs préliminaires de quantité lixivée de nitrate montrent que les pertes par

lixiviation sont très importantes (200 kg N ha-1 en fin de saison des pluies 2003 sur P23 à 90 cm de profondeur, voir Figure 10b). En utilisant un apport de 90 kg N ha-1 (sous forme de diammonium phosphate et d’urée, information récupérée auprès des agriculteurs), et une nitrification de 150 kg N ha-1 (voir chapitre 3) sur les 150 jours du cycle du maïs, le bilan en nitrate boucle avec une consommation par la plante de 150 kg N ha-1, ce qui est certainement trop important vues la densité et la taille des plants. La parcelle P23 est située juste en aval de la forêt et bénéficie peut être du flux de subsurface en provenance de celle-ci. Les pertes en lixiviation de la parcelle P67 sont deux fois plus faibles (Figure 10b), avec pourtant les mêmes apports d’engrais et le même drainage. Le stock d’azote total est identique sur les deux parcelles, et stable au cours du cycle (entre 10 et 14 t/ha)

42

avec une petite baisse avec l’arrivée de la saison des pluies. La quantité d’engrais (ammonium) est trop importante, surtout qu’elle est apportée au moment du buttage (mars) et du semis (15 avril). De plus, on a certainement un pic de minéralisation et de nitrification (Figure 10a) au début de la saison des pluies en juin (le sol devient subitement plus humide et les températures sont encore élevées), alors que les pluies ont été très intenses et rapprochées en juin 2003. A la fin du premier mois de la saison des pluies (fin juin 2003), 60% de la quantité totale lixiviée de nitrate est déjà parti au bas de la zone racinaire (90 cm), soit 100% des apports.

D’après Dahlgren et al. (2004), les Andosols accumulent généralement de grandes quantités de nutriments dans l’horizon A riche en matière organique. Par exemple, un horizon A d’épaisseur de 50 cm a environ 15 Mg N ha-1 et 5.5 Mg P ha-1 (Nanzyo et al., 1993b). Alors que le pool d’azote excède fortement les besoins de la plante, l’azote disponible dépend du taux de minéralisation et de nitrification. Etant donné le pH et l’accumulation de la MO dans l’horizon de surface, la capacité d’adsorption du nitrate dans l’horizon A est faible et il est donc quasiment lixivié librement vers les horizons profonds. Par contre, les horizons de profondeur des Andosols, riches en Al allophaniques peuvent retenir plus fortement le nitrate (environ 15 µg N g-1) et donc retarder la lixiviation du nitrate (voir chapitre 3). Etant donné la forte profondeur racinaire dans ces types de sol (pas de restrictions physiques ou chimiques telles que l’acidité la toxicité aluminique), le retard du nitrate permet un plus grand temps de contact pour la consommation par la plante (Dahlgren et al., 2004). Dans notre cas pourtant, la profondeur racinaire du maïs est encore faible en juin, lorsque les fortes pluies entraînent le nitrate vers les horizons profonds et le retard du nitrate dans les horizons de profondeur ne permet donc pas une meilleure adsorption par la plante. Malheureusement, aucune donnée de concentrations dans les nappes profondes n’est disponible pour corroborer ces fortes quantités de nitrate lixivié.

Figure 10 : (a) Concentrations en nitrate dans les différentes couches de sol et dans les sédiments récoltés à l’exutoire de P23 et (b) quantité de nitrate lixiviée à 90 cm de profondeur sur les parcelles P23 et P67 de la Loma, en 2003.

05

101520253035404550

04/04 24/04 14/05 03/06 23/06 13/07 02/08 22/08 11/09

N-N

O3- le

achi

ng, k

g ha-1

P23

P67b

0

10

20

30

40

50

60

70

80

90

04/04 24/04 14/05 03/06 23/06 13/07 02/08 22/08 11/09

N-N

O3- c

onc

entr

atio

n, m

g k

g-1

0-30 cm

30-60 cm

60-90 cm

sediments

a

0

10

20

30

40

50

60

04/04 24/04 14/05 03/06 23/06 13/07 02/08 22/08 11/09

Dai

ly r

ainf

all,

mm

43

2.5. Evaluation du modèle WAVE

La collecte d’une base complète de données climatologiques, hydrologiques et pédologiques

sur trois années a permis d’évaluer le modèle WAVE (Water and Agrochemicals in the Vadose Environment, Vanclooster et al., 1995), initialement développé pour des milieux pédo-climatiques tempérés, avec les données expérimentales de Maré (Duwig et al., 2003b). Le même exercice est en cours avec les données de la Loma (Thouvenel, 2009). Certains paramètres du modèle ont été calibrés en utilisant les données expérimentales d’humidité, de pression de l’eau et de concentrations en nitrate de l’année 1996. La capacité de prédiction du modèle a été évaluée en comparant les données expérimentales et simulées des années 1995 et 1997. Les prédictions sont généralement bonnes et meilleures lors des années plus humides (Figure 11). Pour ce sol perméable, les flux semblent être surtout fonction du climat. Les prédictions de la concentration en nitrate sont moins bonnes pour les années plus sèches car les paramètres décrivant le cycle de l’azote sont difficiles à estimer directement sur le terrain. La prise en compte des paramètres qui simulent spécifiquement la croissance de la plante (module SUCROS) aurait certainement amélioré les capacités de prédiction du modèle, en reliant le climat, la fertilisation, le sol et la phénoménologie des plantes. Pour cela, il aurait fallu dès le début organiser les mesures sur le terrain ou en laboratoire en fonction des paramètres demandés par le modèle.

Figure 11 : Composantes du bilan hydrique à Maré : mesures et simulations sur les parcelles B (◊, ligne fine) et C (●, ligne épaisse). DAS signifie Days After Sowing. Tiré de Duwig et al. (2003).

2.6. Conclusions

Les deux milieux et sols étudiés ont des caractéristiques qui présentent des risques élevés pour la perte de nutriments en surface et en profondeur dans le sol :

- Des saisons climatiques contrastées, avec une saison des pluies caractérisée par de fortes intensités, et des quantités de pluies importantes.

- Des sols très perméables ; de plus, les fortes pentes à la Loma et une faible couverture végétale en début de saison des pluies (sur les deux sites, les sols sont laissés nus en saison sèche, il n’y a pas de possibilité d’irrigation) engendrent des risques d’érosion et de pertes

44

en nutriments sous forme organique. La faible profondeur du sol et le soubassement calcaire à Maré rendent le milieu très fragile face à l’utilisation déraisonnée d’intrants chimiques ;

- Des apports d’engrais au semis, avec des risques de drainage important quand la plante est trop petite pour consommer le nitrate. A Maré, il a été montré que le fractionnement de l’engrais diminue fortement la quantité de nitrate lixiviée au bas de la zone racinaire. A la Loma, l’apport de nitrate est visiblement trop élevé, étant donné le taux d’azote organique présent naturellement dans le sol, et la forte minéralisation qui peut avoir lieu en début de saison des pluies.

Sur les deux sites, des quantités importantes de nitrate sont perdues au bas de la zone

racinaire, lors des saisons des pluies dépassant la moyenne annuelle, et alors que les engrais sont apportés en une fois au moment du semis. Dans le bassin de Valle de Bravo, les eaux de surface ne semblent pas particulièrement chargées en azote inorganique, mais peuvent présenter de fortes concentrations en azote organique, surtout en saison des pluies, et la plupart des eaux des lacs de barrage sont eutrophisés (Olvera-Viascán et al., 1998). Alors qu’il n’est évidemment pas possible de prévoir les évènements pluvieux intenses, il serait possible de modifier les pratiques agricoles selon les caractéristiques du sol et du climat. La bonne calibration d’un modèle de simulation adapté, permet d’avoir un outil de prédiction utile pour une meilleure gestion des pratiques agricoles. Pour cela, le code WAVE s’est avéré suffisamment robuste pour simuler correctement les variables d’état dans des conditions pour lesquelles il n’était pas conçu au départ.

45

3. Propriétés statiques : caractéristiques physiques, minéralogiques et

chimiques des sols étudiés

3.1. Introduction

L’étude du mouvement de l’eau et du transfert de solutés dans le sol requiert la caractérisation

du sol comme substrat. Le sol est un milieu triphasique : solide, liquide et gazeux. Dans ce chapitre, la phase solide, qui est constituée de particules minérales et organiques est considérée. L’analyse de la phase solide va également nous permettre de caractériser le réseau poral du sol, qui peut être occupé par les phases liquides et gazeuses.

L’analyse du sol a été effectuée dans le but de répondre à deux questions : - quelles sont les propriétés du sol qui affectent le mouvement de l’eau et des solutés ? - quel est l’impact des pratiques agricoles sur le sol, et donc indirectement sur le transfert de

solutés (nutriments et/ou contaminants) ? Sur chacun des sites étudiés, des profils de sols ont été excavés, pour décrire les horizons, la

couleur, la structure. Des échantillons remaniés et intacts ont été prélevés pour analyser en laboratoire la texture, la courbe de rétention en eau, les caractéristiques chimiques et minéralogiques, et des lames minces nous ont permis d’analyser le réseau poral. Compte tenu des propriétés particulières des Andosols, certaines méthodes d’analyse sont inadaptées ou imprécises et donc les résultats ont dû être critiqués. Pour caractériser et classifier le sol (à fin d’inter comparaisons futures), certains paramètres ont dû également être déterminés afin de pouvoir utiliser la classification de la FAO (1998). On présente une synthèse des résultats acquis à Maré et à la Loma et dont la plupart sont tirés des articles de Becquer et al. (2001) et Prado et al. (2007) respectivement. Dans le cas contraire, la publication est précisée.

3.2. Eléments de théorie

Les allophanes L’allophane et l’imogolite sont souvent associés aux sols dérivés de cendres volcaniques,

car la libération de l’Al et Si des matériaux volcaniques comme les verres volcaniques donnent lieu à la précipitation d’aluminosilicates non cristallins. Le facteur clé pour la formation de l’allophane et de l’imogolite est d’avoir suffisamment d’Al et de Si en solution. Les conditions environnementales favorisant leur formation sont une pluviométrie importante et un faible pH (Harsch, 2000).

Les allophanes sont constitués de particules irrégulières sphériques et creuses, de diamètre extérieur de 3.5 à 5 nm, avec une paroi d’épaisseur de 0.7 à 1 nm avec une composition chimique variable. La mesure de la surface spécifique varie de 581 m2 g-1 avec l’azote à 700-1100 m2 g-1 par adsorption de l’éthylène glycol monoethyl éther (Egashira et Aomine, 1974). L’allophane n’a pas de composition chimique définie, et affiche une variation de concentrations en Si et Al allant d’un rapport atomique Al :Si de 1 :1 à 2 :1. Les allophanes sont donc répartis en allophanes riches en Al et riches en Si. Les premiers, aussi appelés proto-imogolites, sont les plus abondants (Parfitt et Kimble, 1989) et sont souvent comparés à l’imogolite, car ayant le même arrangement atomique et la même composition chimique. L’imogolite a une morphologie tubulaire qui peut atteindre plusieurs microns de longueur. La surface externe du tube d’imogolite est composée d’une structure courbée du type de la gibbsite, avec un groupe orthosilicate coordonné par un oxygène à trois atomes d’aluminium à l’intérieur (Figure 12). L’allophane riche en Si a du silicate polymérisé avec des groupes orthosilicate. L’Al apparaît principalement dans des sites octaédriques et dans quelques sites tétraédriques (Dahlgren et al., 1993). Les constituants de type allophanique sont définis comme des aluminosilicates non cristallins, qui sont dissous par le citrate dithionite et 2% de Na2CO3

(Wada and Greenland, 1970). Ils n’ont jamais été isolés et leur existence est seulement déduite à

46

partir de la différence des spectres infrarouges de la fraction fine du sol avant et après le traitement chimique mentionné ci-dessus (Wada and Greenland, 1970). La structure de l’allophane reste un sujet de débats considérables. Des mesures en spectroscopie de fluorescence X ont montré que l’allophane peut contenir de l’Al coordonné à 4 ou 6 oxygènes. Il a été démontré que des pores existent à l’intérieur de la sphérule d’allophane, avec des diamètres entre 0.3 et 2.0 nm, qui semblent provenir de l’omission de deux atomes de Si et deux atomes de Al entre chaque cellule construite de 6 unités, comme le montre la Figure 12. Les aluminosilicates non cristallins présentent plusieurs challenges intéressants pour les sciences du sol. Des études précises in situ de ces matériaux sont toujours d’actualité pour déterminer comment leurs propriétés structurelles, chimiques et physiques influencent les processus tels que le transport de contaminants, la productivité et la stabilité du sol (Harsch, 2000).

Figure 12 : Schéma de la structure de l’allophane et d’un micropore à sa surface (Dahlgren et al., 1993) Altération des cendres volcaniques

Comme je l’ai déjà écrit en introduction, mes recherches portent sur deux sols d’origine volcanique, mais sous conditions climatiques différentes, et à différents stades de dégradation.

Les cendres volcaniques contiennent un certain nombre de composés qui sont très

dégradables dans l’environnement édaphique. Parmi ces minéraux, les verres volcaniques sont les composés les plus facilement dégradables à cause de leur nature amorphe. La dégradation peut se décrire comme une combinaison de cinétiques paraboliques et linéaires (Hodder et al., 1990), qui reflètent l’hydratation des verres et la formation des minéraux argileux, respectivement. La vitesse de dégradation et la genèse des minéraux argileux dans les sols dérivés de cendres volcaniques sont principalement liées à l’interaction des facteurs environnementaux avec la composition physico-chimique et minéralogique de la cendre. Les facteurs environnementaux les plus importants sont ceux qui affectent la concentration de H4SiO4 aqueux et la disponibilité de Al et Fe. Le pH et les acides organiques qui entravent la co-précipitation de Al avec Si jouent aussi un rôle critique. Par exemple, le taux de libération du fer des minéraux primaires, la présence de constituants qui entravent le processus de cristallisation (composés organiques et silice), l’humidité et la température du sol régulent la synthèse des oxy-hydroxydes de fer. Les précipitations et l’épaisseur des dépôts de cendres qui influent sur le régime hydrique du sol et donc du lessivage et la végétation qui contrôle le cycle du carbone sont aussi des facteurs importants. Lors de la dégradation des cendres, la quantité de minéraux primaires facilement dégradables diminue et la minéralogie de la fraction argileuse change. Les concentrations en H4SiO4 soluble sont fortement affectées alors que le degré de dégradation augmente. Les matériaux métastables qui dépendent de l’environnement riche en Si se dissolvent ou se transforment en phases minérales stables comme la kaolinite et la gibbsite.

47

L’incorporation de Al et Fe dans les complexes humiques atteint aussi un équilibre entre l’incorporation de nouveau complexes et la dégradation des vieux complexes métal - humus. La dégradation de l’humus va promouvoir la transformation de la ferrihydrite en goethite grâce à la disparition de l’humus inhibant la cristallisation. La transformation des matériaux non cristallins et des phases amorphes se poursuit jusqu’à ce que les sols dérivés de cendres volcaniques passent des Andosols à d’autres ordres comme les Spodosols, Alfisols, Ultisols et Oxisols (Dahlgren et al., 1993).

Les paragraphes suivants décrivent les résultats des analyses effectuées sur les deux sols. Le Tableau en annexe A donne les principales caractéristiques physiques, minéralogiques et chimiques du Ferralsol de Maré et de l’Andosol de la Loma et la figure en Annexe B présente des photos de deux profils de sol.

3.3. Propriétés physiques du Ferralsol de Maré et de l’Andosol de la Loma Caractérisation morphologique Pour les deux profils cultivés en maïs, la structure est largement modifiée par le labour. A

Maré, on observe à l’oeil la juxtaposition d’une structure grumeleuse et d’une structure sub-anguleuse fine. En profondeur, la structure devient de plus en plus massive, avec des fissures et de la MO, puis vers 25 à 30 cm, une structure micro-agrégée d’aspect massif, avec peu de MO, très cohérente et compacte. Le volume des vides est faible et les pores sont fins et tubulaires (observation visuelle). A la Loma, la structure est non cohésive avec des éléments sub-anguleux, puis elle est plus massive et homogène avec la profondeur.

Texture La texture a été déterminée par la méthode de la pipette « Robinson » (pour les particules de

taille inférieure à 50 µm) après avoir séché le sol à l’air libre, et éliminé la MO avec 30% H2O2. Dans le cas du sol de la Loma, cette méthode a été comparée avec la granulométrie laser (avec ou sans sonification, avec ou sans matière organique).

Le sol de Maré a une texture prédominée par des particules de tailles des argiles, mais qui

sont en fait les oxydes de fer et d’aluminium. Le sol de la Loma a une texture limoneuse (25% sable, 70% limon, et 10% d’argile environ) avec la méthode de la pipette, alors que le granulomètre laser donne un contenu en argile plus élevé (4% sable, 70% limon et 26 argile) et plus proche des observations tactiles de terrain.

Les composants non cristallins jouent un rôle important comme agent ligands. De plus,

chaque colloïde inorganique montre un point de charge nulle différent, ce qui rend la dispersion complète de particules minérales virtuellement impossible (Nanzyo et al., 1993a). C’est la principale raison pour laquelle la taille des grains a été abandonnée pour la classification des Andosols (Soil Survey Staff, 1999). Généralement, les méthodes de laboratoire montrent un contenu en argiles texturales sous-estimé par rapport aux déterminations de terrain. Buurman et al. (1997) après avoir étudié la texture par granulométrie laser de sols volcaniques d’une même toposéquence, après différents traitements pour enlever la MO des allophanes et pour disperser les agrégats obtenaient des pourcentages d’argile appréciables seulement après avoir enlevé une partie des allophanes. Il en concluait qu’il était inutile d’essayer de déterminer un pourcentage d’argile pour les Andosols, ou de calculer la CEC pour la fraction argileuse.

Densité apparente Les sols d’origine volcanique, et en particulier ceux étudiés, sont généralement caractérisés

par une densité apparente faible. Le sol de Maré a une densité apparente qui varie entre 0.7 ± 0.6 g cm-3 en surface à 0.95 ± 0.11 g cm-3 à 40 cm de profondeur, toutes parcelles cultivées

48

confondues. Contrairement aux sols de Maré où la masse volumique sèche diminue avec le labour, à la Loma, celle-ci augmente quand le sol est labouré, car on observe un palier net entre 20 et 25 cm de profondeur : 0.75 ± 0.11 entre 0 et 20 cm et 0.53 ± 0.08 g cm-3 entre 20 et 90 cm de profondeur. Il n’y a pas de variation significative selon les années ni entre les différentes parcelles cultivées.

A Maré, il semble que le travail mécanisé compacte l’horizon de profondeur, la teneur en MO

et la présence de racines contribuant également à la diminution de la densité dans l’horizon de surface de la parcelle de maïs et dans la parcelle sous jachère. A la Loma, la décroissance de la densité apparente peut être reliée à l’augmentation de la teneur en allophanes (voir paragraphe 3.4 et Tableau Annexe A) avec la profondeur. D’après Nanzyo et al. (1993a), les allophanes sont les matériaux amorphes qui contribuent à la faible densité apparente du sol. Les cendres fraîches ont une densité supérieure à 1.5 g cm-3, mais cette valeur décroît avec la dégradation, et se développe une structure poreuse grâce à la présence des composés amorphes et de la MO.

Courbe de rétention en eau / Réseau poral A Maré, la courbe de rétention en eau a été obtenue sur le terrain (grâce à l’installation de

tensiomètres et de TDR aux mêmes profondeurs) et en laboratoire (plaques poreuses). Des échantillons intacts du sol de la Loma ont été soumis aux méthodes du bac à sable et des plaques poreuses (les résultats définitifs sont en cours d’analyse).

D’après la courbe de rétention en eau (sur la Figure 13, l’échelle des ordonnées est tronquée),

la porosité des deux sols est importante, 65% sur Maré, et environ 75% à la Loma. Les sols de la Loma possèdent une forte microporosité (inférieure à 0.1 µm), d’environ 45%, alors qu’à Maré, elle varie entre 15 et 34% avec la profondeur. La forte microporosité des Andosols allophaniques est attribuée à la porosité intra et inter particulaire des allophanes (Nanzyo et al., 1993a). Parfitt et Henmi (1980) suggèrent que la feuille sphérique d’aluminosilicate de l’allophane est constituée de petits trous de diamètres de 0.3 à 0.5 nm qui permettent à la molécule d’eau de passer à travers (voir section 3.2, Figure 12). Les sols de Maré sont caractérisés par une forte mésoporosité (entre 0.1 et 50 µm), alors que la macroporosité est plus forte à la Loma qu’à Maré (Figure 13). La réserve en eau utile (calculée dans le chapitre 2, paragraphe 2.4) est très différente entre les deux sols, en conséquence des différences observées dans la distribution porale.

De plus, Prado et al. (2009) ont montré par analyses d’images de lames minces de la Loma

que la majorité des pores ont un rayon inférieur à 56 µm, et que ceux-ci ont tous une forme tubulaire qui peut être associée à la forme sphérique des allophanes. Les pores plus grands (de rayon supérieur à 564 µm) sont tous des pores d’assemblages, et ceux-ci diminuent avec la profondeur. En comparant les résultats d’analyse de lames minces de colonnes de sol et les courbes de sortie d’un traceur de l’eau à travers les mêmes colonnes, Prado et al. (2009) ont pu associer l’existence de flux préférentiels dans les horizons de profondeur de la Loma avec l’existence d’un plus grand nombre de pores de fissures. Ils ont également montré que le labour détruit les pores de formes bien définies pour les remplacer par des pores d’assemblage, et qu’il augmente la macroporosité. La comparaison entre les courbes de rétention en eau et la distribution des pores déterminée par traitement d’image est en cours.

D’après Basile et al. (2007), la comparaison des deux techniques (analyses d’images/mesures

de la courbe de rétention en eau) ne convient pas du tout pour les pores de rayon inférieur à 500 µm. Ce comportement n’est pas simplement dû au fait que la distribution porale déterminée en 2D par analyse d’images ne correspond pas tout à fait au volume en 3D. En effet, les sols volcaniques ont généralement une structure bien développée et uniforme et on peut donc supposer une certaine isotropie. La différence est surtout due aux principes physiques derrières les deux techniques. L’analyse d’images détecte la vraie géométrie du réseau poral alors que la courbe de rétention en eau détermine une distribution porale équivalente fonctionnelle, basée sur un modèle de distribution

49

de pores capillaires cylindriques. Pour les systèmes poreux complexes, la courbe de rétention en eau, qui décrit le système poreux en terme fonctionnel, donne une description plutôt mauvaise de l’architecture du réseau poral.

Figure 13 : distribution des pores (Mar-01 : 10cm, Mar-02 : 20-30cm, Mar-03 :40cm, Lom-01 : 20-35cm, Lom-02 : 80-95cm).

3.4. Propriétés minéralogiques

Les méthodes employées sont les suivantes : méthode diffractométrique, Mössbauer, microscopie électronique de transmission et spectroscopie à infra rouge et extractions à l’oxalate (Siox, Alox, Feox, extration des composés amorphes ou mal cristallisés), au sodium pyrophosphate (Fep, Alp, extraction des métaux de la matrice organique) et au citrate dithionite (Ald, Fed, extraction des oxydes de fer et d’aluminium bien cristallisés).

Les résultats montrent que la fraction cristalline des sols de Maré est dominée par la

gibbsite et la boehmite (oxydes d’aluminium, de 32 à 44 %) et la goethite (oxydes de fer, de 18 à 25% de la fraction fine). Quelques minéraux silicatés primaires comme le quartz et le feldspar restent dans le sol. Les extractions ont montré que la goethite est fortement substituée en Al, ce qui est généralement le cas pour des sols gibbseux fortement désilicatés (Becquer et al., 2001). Les concentrations de Si dans l’extrait à l’oxalate (Siox) sont plus importants que ceux extrait au citrate dithionite (Sid). Il a été démontré que l’extration à l’oxalate extrait préférentiellement la Si dans les phases amorphes ; du fait, ceci dit la concentration de Siox (Tableau Annexe A) sont trop faibles pour parler d’une présence d’allophane, en comparant cette teneur à celle des sols allophaniques (Dahlgren et al., 1993).

En comparaison, les extraits du sol de la Loma contiennent beaucoup plus de composés

amorphes (de 18 à 28% d’allophane) ; la relation Alox + ½ Feox > 2 définissant un sol allophanique selon Parfitt et Clayden (1991) est satisfaite. Les dissolutions sélectives montrent également la présence de formes cristallines et non cristallines de Fe. Le rapport faible Feox/Fed indique un degré élevé de cristallisation des oxydes de fer. En ce qui concerne les phases cristallines, les minéraux primaires les plus importants sont les plagioclases et la cristobalite. Le MET a mis en évidence des halloysites sphériques et tubulaires (Figure 14). Les halloysites dans les sols volcaniques résultent

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

0.45

0.5

0 500 1000 1500 2000 2500 3000 3500 4000

rayon de pores, µµµµm

poro

sité

, %

Mar-01

Mar-02

Mar-03

Lom-01

Lom-02

0

2

4

6

8

10

0 100 200 300 400 500

50

de la dégradation directe des feldspars (Chartes et van Reuler, 1985) et la re-silication de l’allophane (Wada, 1989). La spectroscopie Mössbauer a permis de confirmer la présence conjointe de composés contenant de Fe2+ et Fe3+, i.e. de pointer la présence spécifique d’hématite et de magnétite.

Figure 14 : Microscopie électronique à transmission (MET) de la fraction fine (<2 µm) du sol de la Loma sans MO : agrégats d’allophanes (à gauche), halloysite sphérique (au milieu) et halloysite tubulaire (à droite)

3.5. Propriétés chimiques pH, contenu en carbone Le sol de Maré est légèrement acide en surface avec une diminution du pH en profondeur à

Maré alors que le pH augmente avec la profondeur à la Loma de 5.5 à 6.5, en même temps que le taux de Siox. Dans les sols allophaniques, l’acidité est tamponnée par les groupes fonctionnels de la MO (groupes carboxyles) et de ceux des aluminosilicates non cristallins (SiOH, AlOH).

A Maré, le ∆pH obtenu par différence entre le pH du sol en présence d’une solution de KCl et

celui du sol en présence d’H2O (pHKCl – pHH2O Tableau annexe B) est négatif dans l’horizon de surface (-0.5), est de +0.2 en profondeur. Ceci indique une prédominance des sites d’échange cationique dans l’horizon de surface (les sites réactifs principalement négatifs), et éventuellement un effet d’échange anionique en profondeur (accroissement des sites positifs). Ceci va de pair avec un accroissement des oxydes de fer avec la profondeur. A la Loma, le ∆pH est négatif tout au long du profil. Ce résultat, combiné aux faibles valeurs d’aluminium extractible, montre que le sol contient un mélange de charges variables et permanentes, et plus spécifiquement des charges négatives.

A Maré, le contenu en carbone est de 7% dans l’horizon de surface et décroît rapidement en

profondeur alors qu’à la Loma, ce taux est stable tout au long du profil (5 %). La MO est bien humifiée sur Maré, avec un rapport C/N de 10, alors que ce rapport est de 20 à la Loma, indiquant un faible degré d’incorporation du N dans les composés humiques. De plus, le taux de MO est faible à la Loma en comparaison à d’autres Andosols allophaniques (voir chapitre 2, paragraphe 2.4). Nous supposons que le fort contenu en MO en surface sur Maré est en lien avec son adhésion sur les oxydes d’aluminium et de fer ; ce processus est reconnu pour retarder la dégradation de la MO, en plus de la faible mobilité des acides et des complexes humiques au pH du sol (Schnitzer, 1996). Egalement dans le cas des Andosols, l’accumulation du C pourrait être liée à la formation de complexes stables entre les substances humiques et les composés amorphes : ces complexes sont ainsi protégés de la dégradation microbienne (Wada, 1989). D’autres facteurs retardant la dégradation peuvent être liés à l’insolubilité des composés organo-minéraux ou à la faible disponibilité du phosphate, au pH faiblement acide, limitant ainsi l’activité bactérienne (Boudot et al., 1989).

51

Minéralisation, nitrification et dénitrification du sol A Maré, la nitrification a été estimée par le bilan en azote nitrique sur sol nu, et en mesurant

directement les pertes par lixiviation, la variation de stock d’azote nitrique dans le profil de sol. Les résultats donnent un taux de nitrification variant entre 0 et 1.46 mg NO3-N kg-1 j-1 selon les années.

A la Loma, la nitrification a été déterminée par Prado et al. (2009, soumis7) par incubation du

sol pendant 30 jours à 25°C à la teneur en eau à la capacité au champ (Matus and Maire, 2000) et la dénitrification par incubation des sols à saturation à 35°C, en mesurant la réactivité des enzymes en analysant le N2O (Duxbury and McConnaughey, 1986). Le taux de nitrification potentielle calculé est de 1 mg NO3-N kg-1 j-1 et la dénitrification potentielle de 0.5mg N2O-N kg-1j-1. Ces taux sont faibles : le faible taux de phosphore disponible et le fort taux d’Al extractible (qui est toxique) limitent la présence de micro-organismes (Etchevers et al., 1978). Parfitt and Salt (2001) expliquent le faible taux de minéralisation des sols avec des propriétés andiques par rapports à des sols non andiques par la présence de l’allophane et des ions Al3

+. Propriétés de charge de surface

Les sols dérivés de matériaux volcaniques sont caractérisés par la présence d’une charge superficielle variable, c’est-à-dire, ce sont des sols dont la charge de surface dépend du pH et de la force ionique de la solution. Les composés à charge variable des sols dérivés de matériaux volcaniques les plus importants sont l’allophane, l’imogolite, les oxydes de fer non cristallins comme la ferrihydrite, les oxydes de fer et d’aluminium, les argiles et la MO (Parfitt et al., 1983; Parfitt et al., 1988; Childs et al., 1991).

Le sol de Maré comporte des charges variables, puisque sa fraction fine est principalement

constituée d’oxydes et d’hydroxydes de fer et d’aluminium. Le sol de la Loma contient en plus des allophanes et d’autres composés amorphes (imogolite, ferrihydrite).

A Maré, les charges de surface ont été évaluées grâce à la méthode d’adsorption d’ammonium

(Gillman and Sumpter, 1986) en faisant varier le pH, et à la Loma, en suivant la méthode de Zelazny et al. (1996) à pH du sol. Ces méthodes impliquent de remplacer les ions natifs du sol par des ions connus (généralement Ca++ et Cl-) dont la quantité absorbée est mesurée par différence entre la quantité appliquée et l’excès en solution. Ce remplacement se fait soit à différents pH fixés (méthode appliquée aux sols de Maré) soit au pH du sol (méthode appliquée au sol de la Loma).

La capacité d’échange cationique (CEC) du sol de Maré varie entre 20 et 36 cmolc kg-1 en

surface (0-12cm) et de 0 à 3.8 cmolc kg-1 en profondeur (30-60cm), et la capacité d’échange anionique (CEA) varie entre 0.6 et 3.0 en surface et 2.4 et 4.3 en profondeur, pour des valeurs de pH variant entre 4 et 7. Une relation entre l’adsorption des ions et le pH a été ajustée selon les modèles de Okamura et Wada (1983) (voir Figure 15).

7 Prado B., Duwig C., Hidalgo C., Gaudet J.P., Etchevers Barra J., Vauclin M., 2009. Preferential flow and nitrate fate

in a Mexican Andosol. Agricultural Water Management, soumis.

52

Figure 15: Sol de Maré : variations de la CEC et de la CEA avec le pH pour les horizon de surface (1B, 12-35 cm) et de profondeur (2B, 36-60 cm). Mesures et modélisation (tirée de Becquer et al., 2001).

Au pH du sol à la Loma, la CEC est pratiquement constante tout au long du profil (22 cmolc

kg-1, de 0 à 110 cm, Tableau Annexe A) et la CEA varie de 1.9 à 2.8 cmolc kg-1 avec la profondeur (Prado et al., 2007).

Pour les deux sols, la CEA est plus forte dans les horizons de profondeur qu’en surface (Figure

15). Ceci s’explique par la prédominance de la MO (et des composés silicatés amorphes pour le sol de la Loma) en surface (CEC élevée) et de la prédominance des oxydes (et des allophanes pour le sol de la Loma) positivement chargés en profondeur. Ces effets sont renforcés par le fait qu’une partie des allophanes et des oxydes sont recouverts par la matière organique et ne sont plus disponibles pour un échange anionique car selon Parfitt (1992), les groupes carboxyle de la MO peuvent réagir avec les allophanes et les oxydes de fer pour former des complexes de sphère interne masquant ainsi les charges positives de la surface des minéraux. A la Loma, cette augmentation de la CEA est également liée à l’augmentation de l’allophane avec la profondeur (Tableau annexe A).

Les calculs montrent que pour le sol de Maré, la CEC et la MO sont fortement corrélées (r2 =

0.96) tout au long du profil. Ceci s’explique par le fait que la composition des deux horizons est sensiblement identiques (donnée dans le tableau annexe A). Pour le sol de la Loma la CEC et la MO sont pratiquement constantes tout au long du profil. Le rapport CEC/CO peut donc s’exprimer par une simple valeur dans les deux systèmes étudiés. Nous supposons également que dans le sol de la Loma, les allophanes ont une contribution significative à la CEC. En effet, Nanzyo et al. (1993a) rapportent que dans les sols allophaniques avec un taux faible de MO, les groupes silanols contribuent plus à la CEC que les carboxyls de la MO.

Rétention des anions Cl- et NO3

- A la Loma, les mesures électrophorétiques (suivant la méthode adaptée aux sols volcaniques

par Escudey et Galindo (1983), en utilisant soit KNO3, soit KCl comme solution) montrent une forte dépendance du point isoélectrique (PIE, défini par le pH où le potentiel zeta est nul) en fonction de la force ionique (Figure 16). En fond KNO3, le PIE varie de 4 à 7 en en passant d’une force ionique de 10-2 à 10-4 dans les deux horizons (Figures 16a et 16c). Cet effet est lié à l’échange des protons par l’ion potassium, le nitrate étant un ion inerte en terme d’échange (la rétention constatée par ailleurs se fait par effet électrostatique et non pas par substitution de ions OH-). En

53

présence de Cl-, cette variation du PIE en fonction du pH diminue. Le Cl- est plus réactif que le NO3

- et est notamment connu pour complexer le fer. Il y a donc réel échange anionique i.e. libération de ions OH- suite à l’adsorption du Cl-. Globalement, l’effet du shift acide par effet d’échange cationique par le K+ et l’effet du shift basique par effet d’échange anionique par le Cl- se compensent, ce qui a pour conséquence le rapprochement des trois courbes dans les Figures 16b et 16d par rapport aux trois courbes KNO3. On observe que ce rapprochement est plus important dans l’horizon profond (80cm, Figure 16d). On sait que cet horizon contient légèrement plus d’allophane qui lui est positivement chargé dû à la présence d’Al et de Fe. Il y a donc cohérence entre ce constat et les mesures électrophorétiques.

(a) KNO3 (b) KCl

-30

-20

-10

0

10

20

30

0 2 4 6 8 10pH

Pot

entie

l Zet

a (m

V)

I=

I=

I=

10-2

10-3

10-4

-30

-20

-10

0

10

20

30

0 2 4 6 8 10pH

Pot

entie

l Zet

a (m

V)

I=

I=

I=

10-2

10-3

10-4

(c) KNO3 (d) KCl

-30

-20

-10

0

10

20

30

0 2 4 6 8 10pH

Pot

entie

l Zet

a (m

V)

)

I=

I=

I=

10-2

10-3

10-4

-30

-20

-10

0

10

20

30

0 2 4 6 8 10pH

Pot

entie

l Zet

a (m

V)

)

I=

I=

I=

10-2

10-3

10-4

Figure 16 : Sol de la Loma : courbes du potentiel Zeta en fonction du pH pour KNO3 (a) et KCl (b) à la profondeur de 30 cm pour différentes forces ioniques et pour KNO3 (c) et KCl (d) à la profondeur de 80 cm (tirée de Prado, 2006).

Ces résultats ont été complémentés par des analyses d’adsorption du NO3

- et du Cl- en batch. On rappelle qu’on observe en batch les effets combinés d’adsorption du NO3

- en sphère externe et de l’adsorption du NO3

- sur des sites échangeurs i.e. entre autres les sites ayant libérés des ions OH- en solution.

En essai batch, on note une sorption légèrement plus élevée du Cl- (Kd de 0.37, 0.57 et 0.47 L kg-1 pour les niveaux 0-30, 30-60 et 60-90 cm respectivement) que du NO3

- (Kd de 0.33, 0 .28 et 0.48 L kg-1 pour les niveaux 0-30, 30-60 et 60-90 cm respectivement). Des essais batch ont également été effectués avec le sol sans MO (lavage au H2O2). Le Kd du NO3

- dans le sol sans MO est supérieur à celui dans le sol avec MO. En éliminant la MO, les allophanes et les oxydes de fer préalablement recouvert par la MO, sont plus réactifs. C’est donc la présence d’allophane et d’oxydes de fer qui expliquerait l’augmentation du Kd du nitrate, après avoir éliminé la MO du sol. Cette augmentation du Kd est du à l’augmentation du PIE suite à l’élimination de la MO (le PIE de la MO est très acide). La surface devenue chargée, retient beaucoup mieux le NO3

- par adsorption de sphère externe.

54

(a) N-NO3- (b) Cl-

0

2

4

6

8

10

0 5 10 15 20

Equilibrium concentration (mM)

Ads

orbe

d co

ncen

trat

ion

((m

M k

g-1)

P1

P2

P3

Kd=0.33; r2=0.94; pH=5.8

Kd=0.28; r2=0.98; pH=6.0

Kd=0.48; r2=0.96; pH=5.9

0

2

4

6

8

10

12

0 5 10 15 20

Equilibrium concentration (mM)

Ads

orbe

d co

ncen

trat

ion

(

(mM

kg

-1)

P1

P2

P3

Kd=0.37; r2=0.94; pH=5.8

Kd=0.57; r2=0.93; pH=6.0

Kd=0.47; r2=0.95; pH=5.9

Figure 17 : isothermes d’adsorption de N-NO3

- (a) et Cl- (b) (tirée de Prado, 2006). Sorption du phosphore et des pesticides Le phosphate forme des complexes de sphère interne avec les allophanes (Nanzyo et al.,

1993b). A Maré comme à la Loma, la rétention du phosphate, obtenue suivant la méthode de Blakemore (1987), est supérieure à 80% dans tout le profil. Cette méthode implique d’ajouter une quantité connue de P (phosphate de potassium à 1 mg mL-1) à l’échantillon de sol et d’agiter à un pH de 4.6. La quantité de phosphore retenue par l’échantillon est déterminée par différence, par la mesure de la concentration de P restant en solution, et exprimée en % de la quantité apportée. Ce résultat est d’ailleurs une des caractéristiques des Andosols, comme notifié dans la classification de la FAO (1998). Cette forte rétention est probablement en lien avec le contenu élevé d’oxydes de fer et d’aluminium dans le cas de Ferralsol, et de composés amorphes en Al et Fe (extrait à l’oxalate) dans l’Andosol. Il en résulte une faible disponibilité du phosphore ce qui est un des principal facteur limitant la croissance des plantes dans ces sols. La sorption du phosphate dans ces sols diffère d’une simple réaction d’échange d’ions car : il a été démontré que le phosphate insoluble précipite sous forme de Ca-, Fe- et Al- phosphates, ou s’adsorbe en sphère interne sur les oxydes de Al et Fe (Beauchemin et al., 2003). Ces modes de fixation ne sont pas ou peu réversibles, il est difficile de ‘désorber’ le phosphate. Dubus et Becquer (2001) ont obtenu les isothermes d’adsorption des sols de Maré par essais batch : les courbes d’adsorption suivent un formalisme de type Langmuir, avec une capacité de sorption maximale variant de 6400 à 9250 mg P kg-1. Cette valeur très élevée est très probablement liée à un processus de précipitation du phosphate avec un cation majeur, qui pourrait être le Ca++. Une forte corrélation négative (r2 = 0.73) entre la sorption du phosphore et la MO a été notée, avec pour conséquence une meilleure disponibilité du phosphore dans l’horizon de surface. Il est probable que la MO couvre les minéraux contenant du Ca, Fe, Al limitant ainsi le processus de précipitation de phases insolubles et/ou d’adsorption e sphère interne, et augmentant du fait la disponibilité et la mobilité du phosphate. Ils ont également trouvé une désorption du phosphate très faible, ce qui corrobore l’hypothèse que la majorité du phosphate est lié sous forme de précipités insolubles et/ou sous forme de complexes en sphères interne.

Le sol de la Loma n’a pas fait l’objet de détermination particulière à part le calcul de Alox + ½

Feox dont la valeur varie entre 6.6 à 7.9 % avec la profondeur. D’après Nanzyo et al. (1993b) il existe une bonne corrélation entre Alox + ½ Feox et le pourcentage de rétention du phosphate. Pour une valeur de Alox + ½ Feox > 2% dans les Andosols allophaniques, la rétention est supérieure à 80%, ce qui est corroboré par la détermination de la rétention du phosphate par la méthode de Blakemore et al. (1987).

55

La sorption de l’atrazine et du 2,4-D ont été étudié dans l’Andosol de la Loma, mais leur comportement n’a pas été évalué dans le sol de Maré. L’atrazine est une base faible et le 2,4-D est un acide. Bien que l’utilisation de l’atrazine soit interdite en Europe car c’est un herbicide classifié comme perturbateur endocrinien, il est encore autorisé au Mexique, et est même le troisème pesticide le plus employé dans le pays (INE, 2000).

Ces deux herbicides sont fortement retenus dans l’Andosol (Figure 18). Les Kd déterminés par

batch varient de 14.4 L kg-1 pour l’horizon A2 (20-35 cm) à 8.03 L kg-1 pour l’horizon A5 (80-105 cm) pour l’atrazine (Duwig et al., 2008b), et est de 11.32 L kg-1 pour le 2,4-D pour l’horizon A2 (20-35 cm, Müller and Duwig, 2007).

En accord avec Baskaran et al. (1996a) et Stolpe et Kuzila (2002), Müller et Duwig (2007)

ont trouvé que la sorption du 2,4-D acide est corrélée positivement à la MO, à la CEC, au taux d’allophane et de ferrihydrite avec des coefficients de détermination de 0.89, 0.80, 0.67 et 0.72 respectivement. Comme le pH des solutions ajoutés pour la détermination des isothermes d’adsorption du 2,4-D (4<pH<6) est supérieur à la valeur du pKa du 2,4-D (2.8), celui-ci se trouve essentiellement sous forme anionique. La sorption des anions peut survenir sur les minéraux chargés du sol, sur les composés hydroxyles de fer et d’aluminium. Ces composés ont une capacité de sorption importante pour les phenoxy-herbicides (Spadotto and Hornsby, 2003). Hiradate et al. (2007) ont étudié la sorption du 2,4-D dans un Andosol japonais. Alors que le sol brut contient beaucoup de Corg (7.2 %), le retrait de la MO a peu d’effet sur la sorption du 2,4-D. Les hydroxyles actifs de surface, qui sont attachés aux hydroxydes actifs et libres de Fe et de Al et sur les complexes humus/métal ont été identifiés comme étant les principaux groupes fonctionnels pour la sorption du 2,4-D.

Füleky and Konda (2007) ont cherché quelles propriétés du sol pour les sols volcaniques

étaient responsables de la différence dans la sorption de l’atrazine entre les différents sols. La corrélation la plus effective est celle entre la MO et la quantité d’atrazine sorbée. Une corrélation négative avec le pH et positive avec la CEC montre que la molécule d’atrazine peut se sorber sur les surfaces chargées positivement ainsi que négativement. Ils ont également trouvé que la sorption de l’atrazine par les sols volcaniques est plus élevée que la moyenne des sols agricoles européens, ce qui peut être relié par les propriétés andiques comme le contenu élevé en MO ainsi que la dépendance de la CEC au pH.

0.0

1.0

2.0

3.0

4.0

5.0

6.0

7.0

8.0

0 0.2 0.4 0.6 0.8 1Ceq (mg L -1)

Sor

bed

conc

entra

tion

(m

g kg

-1)

A2-atrazine

3A-atrazine

A2-2,4-D

Puissance (A2-

Figure 18: isothermes d’adsorption de l’atrazine et du 2,4-D dans deux horizons de l’Andosol de la Loma. Données expérimentales et ajustement de l’équation de Freundlich (tirée de Duwig et al., 2008b). A2 est l’horizon de 20 à 35 cm et 3A de 80 à 105 cm.

56

En conclusion, l’Andosol de la Loma retient fortement les deux herbicides étudiées, grâce à deux de ces caractéristiques andiques, la présence de composés amorphes et un fort taux de matière organique. En l’absence de flux préférentiels, ces herbicides sont très peu mobiles. Des expériences préliminaires d’incubation donnent une demi-vie de l’atrazine dans l’Andosol de 15 jours, et le taux de dégradation obtenu par modélisation inverse des courbes d’élusion à la sortie de colonnes de sol remanié (chapitre 4) varie entre 10 et 39 jours. Ces demi-vies restent inférieures à celles trouvées dans la base de données FOOTPRINT (2007, 2009) avec une demi-vie moyenne de 75 jours. Donc, l’atrazine de l’Andosol est peu mobile et se dégrade rapidement. 3.6. Conséquences quant à la classification

Les deux sols ont été classifiés suivant les recommandations de la FAO (1998). Le sol de Maré est le résultat d’une altération intense et est fortement désilicaté. La

minéralogie est dominée par les oxydes de fer et d’aluminium. La texture limoneuse, la forte microaggrégation, la faible CEC effective entre et le ∆pH supérieur à 0.1 le situe dans l’ordre des Geric Ferralsols. Il présente également quelques caractéristiques andiques, comme la densité apparente inférieure à 0.9 Mg m-3, une rétention du phosphate supérieure à 70%, mais les valeurs de Al ox + ½ Feox inférieures à 1.5% sont trop faibles pour utiliser le terme Andic.

Le sol de la Loma présente des propriétés andiques jusqu’à 110 cm de profondeur : masse

volumique sèche inférieure à 0.9 Mg m-3, rétention du phosphate > 70%, contenu en verres volcaniques (matériaux amorphes) dans la fraction fine < 10% et Alox + ½ Feox > 2%. Le sol de la Loma est donc classifié comme un Pachic Andosol. 3.7. Effet des pratiques agricoles sur l’horizon superficiel

A Maré, il a été noté une diminution de la macroporosité grossière et de la porosité totale sous parcelles cultivées, et une augmentation de la densité apparente de l’horizon de surface. Le travail mécanisé semble compacter l’horizon de profondeur, ce qui constitue une limitation forte à la croissance des racines. De plus, la mise en culture des parcelles se fait généralement avec un bulldozer pour arracher les racines des arbres, ce qui a pour conséquence d’enlever la couche arable du sol.

A la Loma, en comparant l’horizon supérieur avec les horizons inférieurs, ce qui est possible vu la similarité de leur composition chimique et minéralogique, on observe une diminution de la capacité de rétention en eau, une diminution du pH du sol et de la teneur en carbone organique, et la perte de cohésion de la structure en surface. La mise à nu de l’horizon de surface lors de la saison sèche fait perdre certaines caractéristiques andiques du sol de manière irréversible. On observe aussi une augmentation de la macroporosité en surface avec le labour, mais une diminution de la connectivité des pores. 3.8. Conclusions

Nous sommes en présence de deux sols dérivés de cendres et de ponces volcaniques, mais à des stades complètement différents de dégradation. Le sol de Maré est un Ferralsol, très dégradé, alors que le sol de la Loma est un Andosol, beaucoup plus jeune avec un stade de dégradation moins avancé. La minéralogie du sol de Maré est dominée par les oxydes de fer et d’aluminium alors que l’Andosol de la Loma est caractérisé par 20% d’allophane. Leur minéralogie particulière et une forte teneur en carbone organique leur confèrent des propriétés uniques : une densité apparente faible, la présence de charges variables, conduisant à une capacité d’échange anionique

57

importante, une forte rétention du phosphate et des pesticides, la capacité de retarder le nitrate, une minéralisation et une nitrification de la matière organique lente due à la formation de complexes très stables entre les oxydes et/ou les allophanes et la matière organique. La porosité inter et intra particulaire des allophanes permet à l’Andosol de la Loma d’avoir une forte réserve en eau utile, contrairement au sol de Maré. Par contre, le sol de la Loma semble beaucoup plus fragile face aux pratiques agricoles, et surtout la mise à nu de l’horizon de surface.

58

4. Le transfert et la sorption des produits agrochimiques dans les sols d’origine volcanique : études dynamiques en conditions contrôlées

4.1. Introduction

Les expériences en batch sont couramment utilisées pour évaluer la capacité d’échange des sols. Ce sont des expériences réalisées en conditions statiques, où la solution est mise en présence du sol avec un ratio sol:solution généralement de 0.1, en agitant le mélange durant plusieurs heures, voire plusieurs jours. Bürgisser et al. (1993) ont observé que l’abrasion de la surface des grains du sol et la désintégration des agrégats durant l’agitation peuvent conduire à une capacité de sorption plus forte qu’elle ne l’est en milieu naturel. Wong et al. (1990) rapportent également que l’agitation et la dispersion des particules peuvent exposer des surfaces chargées qui ne le sont pas naturellement. Le rapport sol:solution est bien plus faible que sur le terrain. Hoyoux-Roche et Jamet (1988) ont trouvé que le coefficient d’adsorption Kd des pesticides, déterminé par batch était stable pour un rapport sol:solution supérieur ou égal à 0.8, ce qui est bien plus élevé que le rapport généralement utilisé dans ce type d’expérience. Un rapport sol:solution faible signifie également que la proportion d’eau libre dans les pores est bien plus importante que l’eau contenue dans la double couche, où le nitrate par exemple, peut s’accumuler. De plus, le temps de contact entre le sol et la solution est généralement plus grand dans les déterminations en batch, ce qui permet aux processus de dissolution et de précipitation de prendre place, conduisant à un déplacement de l’équilibre entre la solution du sol et la phase solide. Estrella et al. (1993) affirment également que les conditions de détermination de la sorption d’un élément en essai batch différent complètement des conditions de flux. Les coefficients de sorption calculés à un temps fixé ne prennent pas en compte la sorption dépendante du temps qui peut avoir lieu en conditions de flux. Qafoku et al. (2000) postulent qu’il vaut mieux déterminer les paramètres de sorption et de transport dans les mêmes conditions en dynamique.

La réalisation d’expériences contrôlées en colonnes se rapprochent des conditions de terrain puisqu’il est possible d’avoir le même rapport sol:solution. Elles permettent également de découpler les processus et de jouer sur les conditions initiales et aux limites pour en étudier les conséquences, ce qui n’est pas possible sur le terrain. De plus, on peut s’affranchir de la variabilité spatiale du sol et de la variabilité temporelle du climat.

Ainsi, et toujours dans l’optique de se rapprocher des conditions in situ, nous avons mis en

place plusieurs expériences en colonnes de sol, afin de déterminer, dans les mêmes conditions, les paramètres de transfert de l’eau et de transport du soluté considéré. Suivant les possibilités expérimentales et les objectifs de l’étude, les colonnes étaient soit intactes soit remaniées, sous conditions de régime permanent ou transitoire (voir annexe C pour des photos des dispositifs). Cela nous également permis de comparer l’effet de la destruction de la structure naturelle du sol sur les valeurs des paramètres de transfert et de sorption. Dans la totalité des études menées en colonnes de sol, une approche modélisatrice mécaniste déterministe a été utilisée pour la description du transport, et pour l’obtention des paramètres par modélisation inverse. Les approches 1D verticales sont bien adaptées aux expériences en colonnes de sol, car elles prédisent des flux, qui sont mesurables en sortie des colonnes. La pertinence des modélisations sera discutée avec la présentation des résultats.

4.2. Eléments de théorie : transport de solutés dans la zone non saturée du sol

L’étude du transfert de contaminants dans les sols est rendue complexe par les hétérogénéités du milieu, qu’elles soient chimiques (composition non uniforme des constituants du sol), ou physiques (géométrie non uniforme de l’espace poral). Les phénomènes de non équilibre chimique et physique sont étudiés depuis plus de 30 ans mais leur description mathématique a été laborieuse

59

due à la complexité et à la variabilité des sols. On peut trouver dans Addiscott et Wagenet (1985), Vauclin (1994) et Duwig et Vauclin (2003) par exemple, la description de différentes méthodes expérimentales et d’approches modélisatrices à plusieurs échelles pour étudier et modéliser le transport de contaminants dans la zone non saturée du sol. L’annexe D précise les modèles et les hypothèses qui ont été utilisés dans le cadre de ce travail.

4.3. Traceurs de l’eau

Dans toutes les expériences en colonne réalisées, un traceur de l’eau a été employé afin d’obtenir les paramètres hydro dispersifs indépendamment des paramètres de sorption ou de retard. Les traceurs les plus couramment employés sont le Cl- ou le Br-, car de tels anions sont généralement inertes et présents dans les sols naturels à de très faibles concentrations, et leur détermination est peu coûteuse. Dans les sols à charge variable, ils sont retardés et ne peuvent donc pas être utilisés comme traceur. Nous avons donc utilisé de l’eau marquée à l’oxygène 18, dont la détermination est coûteuse, mais qui est un meilleur traceur de la molécule de l’eau. Dans certains cas, le Br- a été utilisé, pour diminuer les coûts d’analyse, mais son retard a été au préalablement déterminé par comparaison des courbes de sortie entre le H2

18O et le Br-. Nous avons également testé le tritium, lors de l’utilisation de pesticides marqués au carbone

14. Dans Müller et Duwig (2007), la modélisation inverse des courbes de sortie du tritium donne un facteur retard supérieur à un pour les trois sols allophaniques étudiés. D’autres études rapportent des facteurs retard supérieurs à 1 pour le tritium (Jacobsen et al., 1992 ; Logsdon et al., 2002 ; Kjaergaard et al., 2004). Le retard du tritium est expliqué par de l’échange isotopique entre le tritium et les hydroxyles des cristaux. Le tritium de la molécule d’eau peut s’échanger avec de l’hydrogène naturellement présent dans le sol. L’isotope entre majoritairement dans des liaisons faibles d’hydrogène, qui sont caractéristiques des bio-polymères et des substances organiques. L’isotope peut aussi se fixer sur les argiles et d’autres minéraux hydratés. Le retard du tritium a été plus particulièrement observé dans les vermiculite et certains types de montmorillonites. Ceci pourrait expliquer pourquoi dans nos expériences, le tritium est retardé dans les sols allophaniques, qui contient des groupes OH très réactifs et de fort taux de matière organique capable de retenir l’eau. 4.4. Expériences en régime permanent

A la Loma, plusieurs expériences en colonnes de sol intact et remanié, toujours en régime permanent (saturé ou non) ont été mises en place pour étudier le transport de l’eau, du nitrate, de l’atrazine et du 2,4-D. Ces expériences n’ont pas eu lieu sur le sol de Maré. La Figure 19 illustre le dispositif expérimental utilisé. Toutes les expériences ont été effectuées en utilisant une solution reconstituée de la solution du sol (analysée sur le terrain), pour avoir une composition géochimique et une force ionique stable de la solution de base. Cette solution est celle injectée au départ pour établir le régime permanent, ainsi que celle utilisée pour préparer les solutions de nitrate et chlorure injectées en créneaux. Il est important de contrôler la force ionique de la solution pour les sols à charge variable, car les propriétés d’échange varient avec la force ionique.

60

Figure 19 : dispositif expérimental (Prado, 2006) pour les études en colonne du sol de la Loma

En colonnes remaniées

Les expériences en colonnes intactes sont bien sûr préférables, puisque gardant intacte la structure du sol, mais leur utilisation n’est pas toujours possible, car il est nécessaire de réduire au maximum le transport de la colonne après son excavation. L’étude du transport en colonne de sol remanié peut aussi avoir l’intérêt de mettre en avant l’effet des propriétés chimiques du sol sur le mouvement du soluté, par rapport aux propriétés hydrodynamiques. Les sols sont tamisés humides à 2 mm afin de ne pas modifier de façon irréversible la structure et les propriétés de charge de l’Andosol, puis compactés par tranche de quelques cm, dans une colonne de 2.5 cm de diamètre et de 25 cm de long, à la densité apparente du terrain.

Prado et al. (2006) et Prado et al. (2009, soumis8) ont étudié le transport du nitrate dans

l’Andosol de la Loma en colonne remaniée, en régime permanent, pour différentes valeurs de flux, proche de la saturation (vitesse de pores variant de 0.36 à 2.91 cm h-1) et pour différentes valeurs de concentrations initiales en nitrate (4 à 16 mM de N). La Figure 20a illustre les courbes de sortie obtenues entre 5 et 21 cm de profondeur pour le NO3

- et H218O et le Tableau 3 donne les paramètres

de transport optimisés par CXTFIT.

a) Packed column b) Intact column

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0 1 2 3 4

Pore volume

C/Co

Simulated

Observed

Simulated

Observed

H218O

H218O

N-NO3-

N-NO3-

0.0

0.1

0.2

0.3

0.4

0.5

0.6

0 1 2 3 4Pore volume

C/Co

Simulated

Observed

Simulated

Observed

H2180

H2180

N-NO3-

N-NO3-

Figure 20 : (a) Courbes de sortie expérimentales et simulées de NO3

- et H218O pour une colonne

remaniée entre 5 to 21 cm et (b) une colonne intacte entre 80 to 96 cm (Prado et al., 2009 soumis8) Les deux courbes de la figure 20a sont symétriques, et sont donc considérées comme

représentatives d’un équilibre physique et chimique du transport de l’eau et du nitrate, et la courbe du nitrate est retardée par rapport à celle du traceur de l’eau, avec des facteurs de retard R variant de

8 Prado B., Duwig C., Hidalgo C., Gaudet J.P., Etchevers Barra J., Vauclin M., 2009. Preferential flow and nitrate fate in a Mexican Andosol. Agricultural Water Management, soumis

61

1.07 à 1.29. Ces facteurs ne varient pas avec la concentration initiale en nitrate, montrant que l’adsorption est linéaire. On peut donc facilement calculer le coefficient de distribution (Kd) pour comparer celui-ci à ceux obtenus par expérience en batch (chapitre 3). Les coefficients Kd obtenus en batch sont systématiquement supérieurs (de 8 à 14%) à ceux obtenus en colonnes de sol remanié, ce qui était attendu comme expliqué dans le paragraphe introductif. Selon Maraqa (2001), une déviation inférieure à 20% peut être considérée comme une variation faible. MacIntyre et al. (1991) expliquent qu’une isotherme linéaire et une sorption instantanée indiquent que le temps de contact entre le sol et le soluté en colonne est suffisant et donc que des coefficients Kd comparables sont généralement obtenus par les deux techniques.

Müller et Duwig (2007) ont étudié le transport du 2,4-D (acide et esther, Figure 21b) et

Raymundo et al. (2009) le transport de l’atrazine , en colonne de sol remanié de la Loma (Figure 21a). Comme pour le nitrate, les courbes du traceur de l’eau ont servi pour obtenir la dispersivité du sol, en fixant le retard à 1 pour l’H2

18O et en l’optimisant dans le cas du tritium. Seuls les paramètres β (coefficient de partition), ω (coefficient de transfert de masse) et µ1 et µ2 (coefficients de dégradation) ont donc dû être optimisés pour les courbes de sortie des herbicides (voir équations A.17 et A.18 en annexe). Dans le cas du 2,4-D, la dégradation a été supposée négligeable car le temps d’expérience était d’un jour au maximum, alors que pour l’atrazine, il était de plusieurs jours. Ele a donc été évaluée par expériences d’incubation et par modélisation inverse.

a) H2

18O et atrazine b) 2,4-D acetic acid (○)and 2,4-D ethylhexyl ester (●)

Figure 21 : courbes de sortie expérimentales et simulées de a) H2

18O et atrazine dans l’Andosol remanié A5 80-95 cm (en haut) Br- et atrazine dans l’Andosol remanié A2 20-35 cm (en bas) ; b) 2,4-D acetic acid and 2,4-D ethylhexyl ester dans l’Andosol remanié A2 20-35 cm (tiré de Raymundo et al., 2009, et Müller et Duwig, 2007).

Comme indiqué par les fortes valeurs du coefficient Kd en batch (voir chapitre 3), le retard de l’atrazine et du 2,4-D en colonnes de sol remanié est important dans l’Andosol de la Loma, avec un retard de l’atrazine plus important dans l’horizon A2, et un retard plus important du 2,4-D ethylhexyl ester que du 2,4-D acid. Le 2,4-D dans sa forme ester subit premièrement une sorption hydrophobique sur la MO, qui peut être décrite par un isotherme linéaire (Brusseau and Rao, 1989a). Les isothermes non linéaires observées démontrent que des réactions de sorption plus spécifiques ont aussi eu lieu. Les esters du 2,4-D peuvent se transformer rapidement par hydrolyse en la forme acide et alcool (Wilson et al., 1997). Durant les expériences en batch et en colonne, des phénomènes de sorption spécifique ont dû avoir lieu, et ils ont stabilisé la forme ester et empêché une hydrolyse complète. Ces résultats montrent que la formulation de l’herbicide a un impact significatif sur le transport et la sorption de celui-ci, et ce immédiatement après l’application à la surface du sol.

62

Les courbes de sortie (Figure 21) des herbicides sont visiblement dissymétriques avec une traînée, indiquant un non équilibre chimique (les courbes de percée symétriques du traceur de l’eau ne semblent pas indiquer de cinétique physique), et la dissymétrie est plus prononcée pour le 2,4-D ethylhexyl ester que pour le 2,4-D acide. La traînée est caractéristique des sorptions avec cinétique, et/ou de non-linéarité des isothermes d’adsorption. Pour les deux herbicides (atrazine et 2,4-D), la fraction des sites de sorption en équilibre instantanée (fraction f) varie de 0.41 à 0.55 pour les sols allophaniques, indiquant qu’environ la moitié des sites sont facilement accessibles. Ceci peut être expliqué par l’accès rapide de l’herbicide à des sites de la MO ou à des sites localisés sur les surfaces des allophanes ou des oxydes. Le Tableau 3 donne les valeurs de Kd pour l’atrazine et le 2,4-D.

Tableau 3 : Paramètres de transport optimisés avec CXTFIT à partir des courbes expérimentales de l’atrazine, du 2,4-D et du nitrate en colonnes de sol remanié.

Les essais en colonnes de sol remanié, que ce soit pour le nitrate, l’atrazine ou le 2,4-D ont été

effectués dans des conditions similaires de densité apparente, de flux et de teneur en eau (sauf pour le nitrate où les valeurs de teneur en eau doivent être vérifiées). Par contre, la dispersivité, obtenue par optimisation des courbes d’élusion simulée par CXTFIT aux courbes expérimentales, diffèrent grandement d’un type d’expérience à l’autre. Ceci pose le problème de l’unicité de la solution lors de l’optimisation des paramètres.

En colonnes intactes Pour l’étude en colonne intacte se pose la question du protocole d’extraction et de leur

conditionnement. Dans la littérature, le protocole d’excavation des colonnes de sol diffère grandement d’une étude à l’autre. On trouve par exemple :

- des colonnes excavées en poussant manuellement un cylindre d’acier dans le sol et en enlevant le sol autour (Kjaergaard et al., 2004) ;

- des cylindres d’acier, recouverts d’huile végétale, poussés dans le sol (Logdson et al., 2002) ;

- pour éviter les effets de bord, Vanderborght et al. (2002) a sorti les colonnes de sol du cylindre et les a encastrés dans de la résine ;

- un block de sol a été sculpté et mis dans une boite de taille légèrement supérieure. Le block de sol a été recouvert de film plastique puis l’espace entre le sol et la boite a été remplie de mousse polyuréthane (Morris et Mooney, 2004).

Les premières colonnes intactes que nous avons prélevées pour la thèse de Prado (2006) ont

été excavées en poussant un cylindre dans le sol (Figure 20b, pour le nitrate, Tableau 4). Pour l’étude des pesticides (thèse de Raymundo, 2008, Tableau 4), nous avons sculpté les colonnes à l’instar de Morris et Mooney (2004) mais comme nous avions des colonnes plus petites (9 cm de diamètre), nous avons rempli l’espace entre les cylindres et le sol par de la cire liquide chaude. Des essais ont été effectués et ont confirmé que la cire ne pénétrait pas dans le sol. Au vu des résultats des paramètres hydro-dispersifs (dispersivité et fraction d’eau mobile plus importantes, Tableau 4) et des images obtenues par tomographie à rayon X sur la première série de colonnes, il se peut que la structure du sol ait été modifiée lors de l’excavation de la colonne. Cette observation peut

horizon-solute average stdevaveragestdev average stdev average stdev average stdev average stdevA2-atrazine 0.80 0.00 0.65 0.08 0.013 0.001 0.022 0.000 12.00 1.13 0.14 0.03A5-atrazine 0.71 0.13 0.72 0.13 0.019 0.001 0.025 0.001 8.63 1.36 0.16 0.02A2-2,4-D 0.75 0.64 0.047 0.030 7.57 0.09A1 nitrate 0.68 0.03 0.95 0.06 0.028 0.005 1.520 1.000 0.20 0.03 0.73 n.c.A2 nitrate 0.71 0.01 0.93 0.05 0.023 0.008 1.520 1.000 0.09 0.01 0.77 n.c.A5 nitrate 0.64 0.01 0.93 0.02 0.018 0.004 1.520 1.000 0.42 0.02 0.68 n.c.

λcm

Co

mg cm-2

Kd

L kg-1ρ

g cm-3ν

cm min-1θ

cm3 cm-3

63

remettre en cause les résultats obtenus dans plusieurs articles qui disent travailler sur des sols intacts. Mais l’objectif, ici, n’est pas ne reproduire (à l’identique) le sol en laboratoire, mais d’observer si le fait de s’en rapprocher fait émerger des mécanismes qui n’ont pas été identifiés précédemment, notamment avec l’usage de sols remaniés. Tableau 4 : paramètres de transport optimisés par CXTFIT à partir des courbes de sorties du nitrate et de l’atrazine des colonnes intactes. Les résultats sur l’atrazine sont en cours d’analyses, ce sont donc des résultats préliminaires.

En comparant les paramètres hydro-dispersifs et de sorption entre les essais batch, les essais en colonnes de sol remanié (Tableau 3) et intact (Tableau 4), on observe les points suivants : - Une réduction drastique de la dispersivité en colonne de sol remanié, due au tamisage et

compactage du sol (10 fois pour l’horizon A2) ; - La disparition de la fraction d’eau immobile en colonne de sol remanié (par exemple, Figure

20a et Figure 20b), ce qui signifie (de manière non exclusive) soit la disparition des chemins préférentiels par tamisage, soit une stabilisation du front d’infiltration ;

- Une légère sur-évaluation du coefficient Kd : les coefficients Kd obtenus en essais batch et en colonne de sol remanié sont en général légèrement supérieurs aux coefficients Kd des colonnes intactes, surtout pour l’horizon de surface. Dans l’horizon A5 intact, alors qu’on obtient une fraction d’eau immobile (10 à 27%), les Kd du nitrate et de l’atrazine entre les différentes méthodes sont comparables. Ceci montre que la présence d’eau immobile influe peu sur la sorption du nitrate ou de l’atrazine.

Comparaisons avec la littérature

Les Kd du nitrate obtenus pour le sol de la Loma (voir Tableaux 3 et 4) et les facteurs retard R correspondants (1.1 en surface et 1.3 en profondeur) obtenus en batch, colonnes de sol remanié et intact sont comparables entre les différentes techniques employées mais en moyenne inférieurs aux Kd et R reportés dans la littérature sur les Andosols. Katou et al. (1996) ont obtenu des facteurs retard de 1.2 à 2.2 pour le nitrate dans un Andosol ayant 80% de composés amorphes, ce qui est beaucoup élevé que le taux d’amorphes dans l’Andosol de la Loma. Feder and Findeling (2007) et Payet et al. (2009) obtiennent également des Kd et R plus élevés que notre étude, pour un Andic Cambisol (CEC = 9 cmolc kg-1, pH = 6, taux d’allophane de 7.8%). Seuls Sansoulet et al. (2007) ont trouvé des Kd et R comparables aux nôtres, dans un Humic Andosol (taux d’allophane de 19.4% dans l’horizon B). Dans notre sol, le facteur retard est lié au taux d’allophane qui augmente avec la profondeur, et qui contient la majorité des charges positives de ce sol, les oxydes de fer ne contribuant que pour une faible partie à l’AEC.

Le retard de l’atrazine varie entre 8 et 16 selon les horizons et les méthodes employées (voir

Tableau 3 et 4). Ce retard équivaut à un Koc entre 155 et 222 L kg-1. Ces résultats sont légèrement supérieures aux valeurs reportées dans la base de données Footprint (2007, 2008), qui reporte une valeur moyenne de 100 L kg-1 pour le Koc de l’atrazine. Baskaran et al. (1996) ont étudié 32 sols allophaniques et non allophaniques et ont trouvé des valeurs de Kd entre 1.16 et 7.70 L kg-1 pour les sols classifiés Typic Hapludand avec des contenus en carbone organique variant entre 1.9 et 16.7%, ce qui correspond à nos valeurs de Kd pour le taux de carbone organique correspondant. Les courbes de sortie de l’atrazine sont visiblement en non équilibre chimique (cinétique dans la

f

-horizon-solute average stdevaveragestdev average stdev average stdev average stdev average stdevA2-atrazine 0.56 0.02 0.67 0.01 0.011 0.005 0.014 0.009 8.38 1.99 1.41 0.15 1.00A5-atrazine 0.59 0.04 0.71 0.03 0.011 0.006 0.019 0.007 9.00 2.01 0.49 0.01 0.90A1 nitrate 0.75 0.10 0.73 0.09 0.024 0.022 0.720 0.470 0.15 0.04 2.40 nc 0.75A2 nitrate 0.70 0.09 0.79 0.03 0.013 0.008 0.720 0.470 0.25 0.05 2.30 nc 0.72A5 nitrate 0.65 0.06 0.83 0.06 0.013 0.010 0.720 0.470 0.39 0.07 1.70 nc 0.73

ρ ν Co Kdθg cm-3 cm min-1 mg cm-2 L kg-1cm3 cm-3

λcm

64

rétention), comme indiquées par les courbes d’élusion des colonnes de sol intact, en accord avec les résultats de Baskaran et al. (1996b) également pour des sols allophaniques.

4.5. Expériences en régime transitoire

Tubes de Perroux Le transfert du nitrate dans le sol de Maré n’a pas été étudié en colonnes verticales, mais en

colonnes horizontales, avec du sol remanié et en régime transitoire, en utilisant les tubes de Perroux (Duwig et al., 1999 ; Duwig et al., 2003 ; voir le dispositif en annexe C).

Le transport de l’eau est décrit mathématiquement par l’équation de Richards, et le front de

l’eau moyen correspond au mouvement d’un piston (Bond, 1990). En supposant que l’adsorption des solutés réactifs suit une isotherme linéaire, Watson & Jones (1981) ont trouvé une solution analytique approximative à l’équation de convection dispersion, après l’apport d’une solution injectée en créneau. Cette solution a été ajustée aux données expérimentales de concentration totale de nitrate dans chaque tranche de la colonne à la fin de l’expérience afin de déterminer le retard du nitrate (Figure 22). Le sol de surface (0-20 cm) retarde le transfert du nitrate, en suivant une isotherme linéaire (facteur de retard de 1.1). Dans les horizons profonds (> 20 cm), les facteurs de retard sont plus importants, et augmentent de 1.15 à 2.05 alors que la concentration injectée diminue de 71.4 à 0.4 mM, indicatifs d’une isotherme non linéaire. L’horizon de surface et les horizons plus profonds contiennent tout deux des quantités importantes d’aluminium et de fer alors que l’horizon de surface est beaucoup plus riche en matière organique, qui contrebalance les charges positives des oxydes. En effet, les groupes organiques déplacent les liaisons des molécules d’eau sur les sites chargés positivement des oxydes, réduisant les sites disponibles pour la sorption des anions comme le nitrate (Parfitt, 1992).

Figure 22 : Concentrations en nitrate dans le sol, mesurées et simulées, vs. la variable de Boltzman (fonction de la distance et du temps), pour l’horizon 20-40cm, pour différentes concentrations appliquées C1. La ligne pointillée verticale indique la position du front d’eau. La concentration initiale dans le sol est de 0.68 mM.

65

Les charges positives de surface dans les horizons profonds ont été considérées comme étant

de deux types : une partie avec une forte affinité pour le nitrate lorsque la concentration est faible, et l’autre partie avec une faible affinité pour le nitrate à fortes concentrations. Ce type d’isotherme est identique à ceux trouvés pour les métaux lourds et les oxyanions. La capacité d’échange anionique qui découle de ces résultats pour le Ferralsol de Maré est supérieure à celles trouvées en général pour les sols à charge variable (Qafoku and Sumner, 2001).

Lysimètres remaniés et intacts Le retard du nitrate a également été mesuré dans le Ferralsol de Maré dans un lysimètre de

30 cm de diamètre (voir photos Annexe C), où le sol a été compacté après tamisage, avec ou sans plantes. Le nitrate a été appliqué à la surface du sol sec, avant d’appliquer une intensité de pluie constante proche de la conductivité hydraulique à saturation du sol. La conductivité électrique du sol mesurée par des sondes TDR installées à différentes profondeurs dans le lysimètre a servi à calculer le facteur de retard (Vogeler et al., 2000), en mesurant la différence de temps d’apparition entre les pics des courbes de conductivité électrique, et en supposant que le sol se comporte comme un sol de Green et Ampt (Kutilek et Nielsen, 1994). Des retards pour le nitrate entre 1.2 et 1.7 ont été obtenus, ce qui correspond aux retards évalués par d’autres techniques (Duwig et al., 2003). Par contre, la méthode basée sur la mesure de la conductivité électrique n’est pas adaptée à l’estimation des vitesses du front d’eau, à cause de l’hétérogénéité du front d’humectation. Des lysimètres de sol intact de l’Andosol de la Loma prélevés selon la méthode de Morris et Mooney (2004) ont été utilisés pour visualiser l’infiltration d’un colorant fluorescent (pyranine) appliqué par un infiltromètre à disque en dépression à h = -10 mm (Duwig et al., 2008a). Des faces ont été coupées verticalement dans le lysimètre : la distribution du colorant observée était très hétérogène avec des pénétrations en forme de doigts (« fingered penetration »). Une calibration indépendante entre la concentration et les caractéristiques de la couleur, obtenue par analyses d’images nous a permis d’obtenir la carte de distribution de la concentration en 2D (Figure 23). Le colorant est fortement retenu dans le sol et suit un isotherme de Langmuir : le front moyen du colorant n’atteint que quelques cm en dessous de la surface du lysimètre, mais des chemins préférentiels ont fait qu’il atteigne le bas de la colonne en certains endroits.

Figure 23 : zoom sur le profil à 13.3 cm du bord montrant les variations locales de concentrations du colorant, le long d’un chemin de flux préférentiel

66

4.6. Conclusions

On est présence de deux sols qui retardent le nitrate (facteur de retard de 1.1 à 2.05 pour le Ferralsol et 1.1 à 1.5 pour l’Andosol). L’Andosol de la Loma adsorbe également fortement les herbicides (facteurs de retard entre 8.0 et 15.8 pour l’atrazine et 9.9 pour le 2,4-D) en suivant des isothermes non linéaires et une adsorption non instantanée (transport en non équilibre chimique). De plus, la structure intacte de l’Andosol conduit à des chemins de flux préférentiels, essentiellement dans les horizons de profondeur.

L’utilisation de plusieurs dispositifs expérimentaux et de conditions expérimentales

différentes permet de tirer plusieurs conclusions dignes d’intérêt : - La comparaison des résultats en colonnes de sol remanié et intact montre clairement que la

destruction de la structure naturelle du sol a un effet sur le transfert hydro-dispersif, mais aussi sur les propriétés de sorption ou de retard. On soupçonne également que le type d’expériences et la manière d’excaver les colonnes de sol intact influent également sur les propriétés de transfert.

- Dans les gammes de flux étudiés, il n’y a pas d’effet significatif de la vitesse de l’eau sur les propriétés de sorption, et comme les régimes permanents et transitoires n’ont pas été étudiés sur les mêmes sols, il est de difficile de conclure sur l’effet du régime d’écoulement sur la sorption. Ce point serait très intéressant à regarder, car c’est une caractéristique importante du climat tropical, où les intensités de pluie sont très variables. Les Kd obtenus par essais batch et par expériences en colonnes de sol sont comparables, mais seules les expériences en colonnes de sol ont permis de mettre en évidence l’existence d’une cinétique de sorption pour les herbicides.

- L’isotherme d’adsorption du nitrate dans l’Andosol et le sol de surface du Ferralsol est linéaire, par contre l’adsorption du nitrate dans l’horizon profond du Ferralsol, ainsi que celle des herbicides dans l’Andosol varient avec la concentration.

- La formulation de l’herbicide a également une incidence sur le transfert : le 2,4-D ester, présent dans la formulation du Pasture Kleen est adsorbé plus fortement et avec une cinétique significative par rapport au 2,4-D acide. Ce dernier est généralement l’ingrédient actif étudié dans la plupart de la littérature sur le 2,4-D, alors que ce n’est pas l’ingrédient actif présent dans la formulation de l’herbicide commercialisé.

Dans ces travaux en colonnes de sol, nous avons initié des études d’analyse d’images pour la

compréhension et la description mathématique du réseau poral du sol et du transfert de solutés. Pour l’instant, la méthode utilisée est destructive : visualisation de l’infiltration d’un colorant (Duwig et al., 2008a) ou analyse du réseau poral des lames minces (Prado et al., 2009). Le traitement d’image nous a permis d’obtenir des paramètres indépendamment. Il faudrait maintenant pouvoir les introduire dans des modèles de transfert et comparer les simulations numériques avec celles des modèles plus classiques (CDE). L’idée est aussi d’utiliser des méthodes non destructives, comme la tomographie à rayon X (voir perspectives).

67

5. Conclusion générale

Ces travaux, à l’interface entre l’hydrologie, la physique et la chimie du sol visent à une meilleure compréhension et représentation des processus impliqués dans le transfert de l’eau et des produits agrochimiques dans les sols d’origine volcanique. Les différentes approches expérimentales développées ont permis de mieux comprendre les principaux mécanismes impliqués et de se questionner sur l’influence des dispositifs expérimentaux sur les résultats obtenus. L’étude à différentes échelles a permis d’affiner la description multi-échelle des processus, avec la contrainte de rester toujours proche des conditions de terrain. Le découplage des processus grâce à l’expérimentation en laboratoire m’a aidé à comprendre quels sont les processus prépondérants, et a permis d’expliquer les résultats obtenus in situ et de paramétrer les différents processus pour la modélisation des transferts.

L’approche par modélisation est indispensable pour répondre dans des délais acceptables avec

des coûts raisonnables aux attentes de notre société, compte tenu de la très grande diversité des polluants et des milieux récepteurs, mais cette modélisation doit être fondée sur une bonne connaissance des sols et des processus prépondérants. Enfin, ces travaux ont mis en évidence l’intérêt de coupler observations de terrain – expérimentation – approche théorique – modélisation dans la compréhension des systèmes naturels et la prédiction de leur évolution dans le temps et en fonction des conditions du milieu.

Les principales conclusions sont les suivantes :

- les conditions climatiques « extrêmes », avec des intensités de pluies importantes et une forte pluviométrie font que les risques de contaminations des ressources en eau sont importants ;

- ces risques sont aggravés par le fait que les deux sols étudiés ont une forte porosité et une conductivité hydraulique importante ; de plus, le sol de Maré est peu profond, et l’horizon de profondeur de l’Andosol contient des chemins de flux préférentiels,

- les caractéristiques minéralogiques et chimiques uniques (présence d’oxydes de fer et d’aluminium dans le Ferralsol et de composés allophaniques dans l’Andosol) font qu’ils retardent le mouvement du nitrate et l’Andosol adsorbe fortement les herbicides ;

Les pratiques agricoles peuvent être adaptées pour utiliser aux mieux ces propriétés uniques et diminuer le risque de dégradation des ressources. Voici quelques recommandations, tirées des résultats de nos études :

- maintien d’une couverture végétale minimale sur les Andosols pour éviter une dégradation rapide des propriétés andiques (forte rétention en eau et des nutriments par les composés amorphes, agrégation), et la perte de nutriments en surface par le séchage, le ruissellement et l’érosion ;

- fractionnement de l’apport d’engrais, pour éviter la lixiviation rapide de ceux-ci lors des pluies intenses, et apport lorsque les racines sont suffisamment développées. La présence de racines dans les horizons de profondeur, permet d’exploiter les caractéristiques de rétention plus importantes de ces horizons, afin d’augmenter le temps de contact entre les racines et les nutriments, et contrebalancer la présence de chemins préférentiels. De plus l’horizon de profondeur du Ferralsol de Maré a une capacité d’échange anionique plus importante pour des concentrations de la solution du sol plus faible, d’où l’intérêt du fractionnement de l’engrais, pour diminuer la concentration du nitrate dans la solution du sol.

- la quantité d’engrais apportée doit être calculée en fonction des nutriments présents naturellement et initialement dans le sol, et des processus de minéralisation et nitrification ;

68

sur les deux sites, une partie du nitrate lixivié au bas de la zone racinaire vient de la minéralisation ;

- un labour minimum des Andosols détruit les chemins de flux préférentiels et peut donc diminuer la vitesse d’arrivée des contaminants dans les horizons plus profonds, mais cause une déstructuration du sol (également observée sur le Ferralsol de Maré) qui le rend plus vulnérable aux processus d’érosion.

- Les herbicides sont fortement retardés dans les Andosols étudiés, et présentent un taux de dégradation rapide. Il ne présentent donc à priori pas de risque majeur pour les eaux souterraines. Mais la présence de flux préférentiels dans les horizons profonds peut rendre ces polluants plus mobiles.

Ces travaux résultent bien évidemment d’un travail collaboratif avec les chercheurs,

notamment mexicains (voir publications) et avec les étudiants dont j’ai encadré les travaux de thèse (Blanca Prado et Elias Raymundo). A travers ces recherches, je pense avoir répondu à l’attente de nos partenaires, tant au niveau scientifique, que au niveau du transfert de connaissance par la formation de techniciens et d’étudiants mexicains ou néo-calédoniens.

Afin de parvenir à une étude complète sur le transfert dans les sols volcaniques, certains aspects des recherches présentées ci-dessus devraient être approfondis ou même abordés. Le transport n’a été étudié qu’en régime permanent. Cependant, pour se rapprocher des conditions de terrain, il conviendra d’étudier l’effet du flux (régime transitoire, ou au moins arrêt du flux pendant un temps fixé) sur la rétention des produits agrochimiques. Lors de l’étude du transfert dans l’Andosol, nous avons toujours essayé de travailler avec des solutions dont la force ionique était identique à celle in situ, mais constante. Du fait de cycles d’évapotranspiration et de drainage sur le terrain, et des intrants, les caractéristiques géochimiques de la solution peuvent évoluer, ce qui a des conséquences sur les capacités d’échange ionique des Andosols, et donc sur la rétention des produits agrochimiques. Certains points sont abordés dans les perspectives.

69

EEEE. . . .

PERSPECTIVESPERSPECTIVESPERSPECTIVESPERSPECTIVES 2009200920092009----2020202013131313

70

Perspectives

1. Préambule

La qualité de l’eau et le devenir des polluants dans l’environnement sont devenus des thèmes de préoccupations majeurs pour notre société. Les activités industrielles, l'accroissement de la population mondiale, les rejets d'eau domestiques non traités, les pollutions d'origine agricole (nitrate, phytosanitaires), la salinisation relative à l'irrigation, ont entraîné une dégradation progressive de la qualité des eaux, soit liée à la pollution de l'eau elle-même, soit à la dégradation des contenants à travers lesquels circule l'eau (atmosphère, cours d'eau, sols, aquifères). Aujourd’hui, environ 1,5 milliard de personnes n'ont pas accès à l'eau potable et ce chiffre va en augmentant rapidement. Une meilleure connaissance scientifique des impacts et des comportements des polluants dans les différents milieux récepteurs et l’amélioration des moyens analytiques nous ont fait prendre conscience des risques de toxicité pour l’homme des produits comme les produits pharmaceutiques, ou des mélanges de contaminants à des concentrations considérées au-dessous des limites de toxicité (pollutions multiples ou mixtes, expositions très longues même à faibles doses…). Les pays industrialisés utilisent de grandes quantités de produits pharmaceutiques, et leur présence a été signalée dans les usines de traitements des eaux et dans les eaux superficielles (Lindberg et al., 2005), alors que ces produits ne sont généralement pas analysés de façon courante dans les eaux. Les Etats-Unis et la Communauté Européenne (US Department of Health and Human Services, 1998; EMEA, 2006) ont émis des directives sur les risques de ces produits pour la santé humaine. Au Mexique, l’étude de ces produits dans les eaux est très récente et ne concerne à notre connaissance que ‘el Valle del Mezquital’, une région proche de Mexico où les eaux usées de la ville de Mexico sont utilisées pour l’irrigation sans aucun traitement.

La compréhension du transport en milieux poreux a donc de nombreuses applications dans

les domaines des sciences de l’environnement, l’agronomie, la pédologie et l’hydrologie. Les sols naturels présentent des structures hétérogènes, variables spatialement et temporellement, et soumises à des forçages climatiques et anthropiques qui affectent le transfert de matière et d’énergie. Pourtant, la plupart des équations classiques de transport supposent que le milieu est homogène. Les approches macroscopiques sont souvent monodimensionnelles, ne considèrent pas la structure du sol et utilisent des paramètres empiriques. Les modèles mécanistes à l’échelle du pore souffrent de l’impossibilité d’effectuer des mesures directes et indépendantes pour valider les hypothèses sous-jacentes. C’est en partant de ces deux constats que je vais poursuivre mes travaux de recherche sur le transfert de contaminants dans les sols. Ils visent d’une part à améliorer les relations entre architecture/structure du sol et transfert, et d’autre part à mieux comprendre le devenir des « nouveaux » contaminants dans le sol et en surface. Je reste bien sûr attachée à la problématique « Recherche pour le Développement », puisque la plupart de mes projets en cours ou prévus seront menés dans le cadre d’un partenariat avec nos collègues mexicains et sur la demande de nos collaborateurs.

2. Cadre général

J’ai eu la « chance » de rédiger mon HDR en même temps que la rédaction du rapport d’activité et du rapport de prospective 2011-2014 du LTHE, ce qui m’a permis de mettre en phase mes questions et intérêts scientifiques à ceux du laboratoire et de l’équipe TRANSPORE (TRANsferts en milieux POreux et REactions). Tout en gardant les mêmes contours thématiques qu’au cours du quadriennal précédent, l’équipe aura une inflexion forte vers la qualité environnementale des milieux poreux en phase avec une demande sociétale de plus en plus

71

prégnante. Mon thème de recherche se retrouve également en phase avec certaines priorités affichées dans les documents de la prospective de l’INSU-SIC (2007) et de l’INEE (2009).

Les activités de l’équipe se font en grande partie grâce à l’existence de plateaux (par exemple, MOME-PMPN, PS2E, ..) organisés au sein du pôle Envirhonalp, qui permettent aussi de renforcer les collaborations entre les laboratoires impliqués dans le domaine environnemental grenoblois. Ce cadre de travail me permettra de développer dans de bonnes conditions de support scientifique et analytique les questions scientifiques décrites ci-dessous.

3. Questions scientifiques

Mes travaux vont donc concerner la compréhension et la modélisation du transfert de contaminants dans les sols à travers les quatre questions principales suivantes:

- Comment mieux prendre en compte la structure/architecture des milieux poreux, et quelle est l’influence des écoulements préférentiels qui en résultent sur les flux, les temps de séjour, les profondeurs de migration de l’eau, et des substances chimiques ?

- Comment varie cette architecture dans le temps et dans l’espace en fonction des conditions

initiales et aux limites (teneur en eau initiale du sol, régime transitoire ou permanent, effet de l’ambiance géochimique de la solution) ?

- Quelle est l’influence des processus gouvernant le devenir et l’impact de pollutions organiques ou

inorganiques, notamment les produits pharmaceutiques ? Cette thématique sera développée en collaboration avec deux (géo)chimistes qui intégreront le LTHE et la soumission d’un projet au CONACYT au Mexique.

- quel est l’effet de l’échelle spatio-temporelle d’observation sur les questions précédentes ? Cette question sera développée en étudiant et observant les transferts à différentes échelles (micropores, colonnes, lysimètres, parcelles).

4. Méthodologies mises en œuvre et objets étudiés

Ces quatre axes de recherche seront développés dans le cadre de plusieurs programmes en cours ou projets soumis à financement.

4.1. « Méthodes non invasives pour le suivi du transport de matière en milieu poreux » Il est reconnu que la mesure et la quantification exacte de la structure du sol et des chemins

préférentiels posent un défi important. Le réseau poral, plus spécifiquement la distribution des tailles de pores et leur connectivité contrôlent les propriétés hydrodynamiques des sols. Il est donc important d’étudier le transport en sol intact, dans un volume suffisamment important pour intégrer des chemins préférentiels comme les fractures, le réseau racinaire, les pores dus à l’activité faunique. De plus, la dynamique des flux préférentiels (ou le transport non-darcien) et leurs relations avec la structure du sol ne sont pas totalement comprises. Il est nécessaire d’approfondir les recherches sur les relations temporelles entre structure du sol et dynamique des flux. Je me propose de mettre en place un système d’étude polyvalent du transport de matière (eau, solutés, bactéries, particules) en colonnes de sol intact (ou mini-lysimètres) afin de pouvoir obtenir et étudier les courbes d’élution à la sortie, et ceci en fonction des conditions initiales et aux limites (en condition saturée, non saturée, en régime de flux permanent ou transitoire). L’idée est de combiner la détermination indirecte des paramètres de transport par l’étude de la courbe d’élution à l’observation directe du mouvement de l’eau ou d’un traceur par tomographie à rayon X ainsi que par inversion de mesures électriques. La tomographie à rayon X est un instrument utilisé depuis une

72

dizaine années en science du sol pour observer de manière non invasive la structure du sol. La plupart des mesures en tomographie servent à l’étude statique en 3D de la structure du sol, mais la répétition de la mesure en utilisant un traceur visible durant une certaine période permet une étude en dynamique du transport ainsi que de l’évolution de la structure avec le temps et la teneur en eau (milieu déformable). La mise en place de plusieurs sondes électriques ou TDR dans un mini-lysimètre permettra d’appliquer la méthode d’inversion de la permittivité électrique pour obtenir la teneur en eau (collaboration avec l’équipe Hydrogéophysique du LTHE). Ces données seront comparées à celles acquises par le tomographe à rayon X.

Ce projet sera développé au sein du LTHE, en collaboration avec la UNAM au Mexique et

l’Université d’Auckland en Nouvelle-Zélande. J’ai prévu d’encadrer une étudiante en M2R en collaboration avec la UNAM au Mexique sur le sujet à partir de février 2010 : « étude du transport d’un traceur de l’eau dans des colonnes intactes d’un Vertisol de la Valle du Mezquital (voir projet ci-dessous) et modélisation du transport en utilisant la structure 3D du sol visualisée par tomographie à rayon X ». Elle a déjà émis son intention de poursuivre ce travail dans le cadre d’une thèse. De plus, une proposition de sujet thèse portant sur ce thème a été soumise au programme « Beca Chile », financé par le gouvernement chilien.

4.2. Application à l’évolution spatio-temporelle de l’architecture des Andosols et des sols irrigués par des eaux usées

Dans la suite de mes travaux actuels, la méthodologie décrite ci-dessus sera appliquée aux

Andosols mexicains, présentés dans ce document. Cela permettra de relier les paramètres hydro dispersifs obtenus par les courbes de caractérisation hydrologique avec les résultats relatifs à l’analyse d’images en 2D et en 3D. L’emploi d’un traceur visible par le tomographe servira également à visualiser et à décrire les chemins de flux préférentiels.

Les sols de la Vallée du Mezquital (Leptosols et Vertisols) sont soumis à irrigation avec les

eaux usées, provenant de la ville de Mexico. La ville de Mexico manque d’eau, et le problème est devenu crucial en 2009 : en janvier, le niveau des barrages qui alimentent Mexico en eau a atteint un minimum historique, et la distribution a été rationnée cette année pendant cinq mois, dans cette capitale de 20 millions d'habitants, a annoncé, la Commission mexicaine de l'Eau (Conagua)9. Des nouvelles ressources sont constamment recherchés, et l’idée d’utiliser les eaux des aquifères de la Valle du Mezquital a fait son apparition10, car c’est certainement le seul aquifère de la région de la ville de Mexico dont le niveau augmente. L’aquifère est largement contaminé, et la description du transport de l’eau et des contaminants dans les sols soumis à l’irrigation est donc hautement prioritaire. Cette description passe premièrement par une bonne compréhension du transport de l’eau, et notamment de l’évolution de la structure du sol sous irrigation à saturation, et par des eaux chargées en sodium et en détergents, entre autres. La tomographie X permettra de visualiser l’évolution de la structure du sol, avant et après l’application de lames d’eau à la composition géochimique connue, à une colonne de sol. L’étude en colonne de sol permettra d’étudier l’effet du type d’irrigation (par inondation, ou par aspersion, par exemple), et des caractéristiques géochimiques de la solution (teneur en sodium, détergents, carbone organique, pH). Il est également prévu de comparer plusieurs sols, notamment ayant des contenus en argile et en matière organique différent. Ce thème sera abordé dans le cadre d’un projet soumis au CONACYT (voir paragraphe 4.3) sur le devenir des produits pharmaceutiques dans ces mêmes sols.

9 Joëlle Stolz, « A Mexico, l'eau devient une denrée rare », dans Le Monde du 13-01-2009, mis en ligne le 12-01-2009 10 Acuíferos del Mezquital alternativa de agua para la Ciudad de México, academia Mexicana de Ciencias, Boletín AMC/44/08, México, D.F., jueves 24 de abril de 2008 http://www.comunicacion.amc.edu.mx/comunicados/acuiferos-del-mezquital-alternativa-de-agua-para-la-ciudad-de-mexico/

73

4.3. « Evaluation du risque de contamination du sol et des eaux souterraines par les produits pharmaceutiques dans les sols irrigués avec des eaux usées dans la Vallée du Mezquital au Mexique » Siemens et al. (2008) ont estimé la quantité théorique de produits pharmaceutiques libérés

dans l’environnement (eau et sols) en se basant sur les données de consommation de produits pharmaceutiques et sur leurs concentrations dans les eaux résiduelles, en réalisant des analyses à différents points du système d’irrigation de la Vallée du Mezquital, utilisant les eaux usées (sans traitement) en provenance de la ville de Mexico. Les auteurs ont trouvé en quantités importantes 11 produits, parmi lesquels se trouvent la clarithromycine, le metoprolol, le naproxène, et l’ibuprofène, qui sont des anti-inflammatoires, des antibiotiques, des médicaments contre l’hypertension.

Les produits pharmaceutiques en solution s’ionisent et peuvent donc être retenus dans les sols, selon les propriétés du milieu. La rétention de ce type de produits dans les sols, et spécialement les antibiotiques, peut représenter un risque pour la santé humaine dû au transfert de résistances entre bactéries du sol et bactéries pathogènes humaines, ce qui impliquerait que les traitements médicamenteux deviennent inefficaces. Thiele-Bruhn (2003), dans une revue sur le devenir des antibiotiques dans les sols, rappellent que l’adsorption des produits antibiotiques sur les sites d’échange minéraux ou organiques est principalement due au transfert de charge et à l’interaction des ions plutôt qu’à de la partition hydrophobique. La sorption est fortement influencée par le pH du milieu et celle-ci gouverne la mobilité et le transport des antibiotiques. De plus, les antibiotiques fortement adsorbés peuvent se déplacer rapidement par les macropores ou les chemins préférentiels. La biotransformation est une façon efficace pour la dégradation et inactivation effective des antibiotiques, bien que certains produits de dégradation soient encore actifs.

Les conditions de transport, de dégradation et d’accumulation des produits pharmaceutiques sont donc des informations importantes pour les microbiologistes et les médecins afin de mieux connaître le transfert des gènes de resistance entre les microorganismes. Suite à ces constats, nous nous proposons d’identifier les chemins de transferts des produits pharmaceutiques à travers le sol, d’évaluer la capacité du sol à les retenir et les temps caractéristiques d’incorporation de ces produits à la matrice du sol, notamment dû aux processus de bioturbation. Des suivis seront fait sur le terrain (mesure du drainage grâce à l’installation de TDR/tensiomètres, et analyse de la solution du sol collectées dans des bougies poreuses) seront effectués régulièrement, ainsi que des études en colonnes de sol pour obtenir les paramètres hydrodispersifs, de sorption, ainsi que l’impact des flux préférentiels sur le transport des produits pharmaceutiques. Ces études seront complémentés par la détermination des isothermes de sorption en batch. Les expériences en batch serviront également pour analyser la fraction du sol responsable de la rétention du produit pharmaceutique, analyses que seront corroborés par la spectroscopie infrarouge et la diffraction X des différentes fractions du sol avec le produit étudié. Les transformations des produits pharmaceutiques et leur produits de dégradation, ainsi que les conditions du milieu (par exemple, type de population bactérienne, humidité) favorisant ou défavorisant leur dégradation et inactivation seront étudiés.

Ce projet a été soumis à financement au CONACYT, et sera mené en collaboration avec la UNAM au Mexique. Ces études se feront essentiellement sur le terrain et laboratoire, en utilisant les méthodes explicitées dans ce rapport, ainsi que les techniques de tomographie en colonnes intactes, qui seront d’abord développées grâce au projet soumis au LTHE. L’interaction des produits pharmaceutiques avec le sol durant leur transport sera étudié en collaboration avec Marie-Christine Morel, chimiste qui intègre le LTHE en fin d’année.

4.4. Transport de nutriments dans des bassins versants agricoles d’altitude

Bassin du Cointzio, Mexique En parallèle aux études plus fondamentales et en conditions contrôlées sur la compréhension

des processus impliqués dans le transport de contaminants dans les sols, je continue à participer à

74

des projets sur le terrain, ce qui me permet d’étudier l’effet de l’échelle spatio-temporelle d’observation sur les différents processus, en combinant suivi sur le terrain, et mesures en conditions contrôlées, selon les processus étudiés.

Le bassin du Cointzio, alimenté par la rivière Rio Grande est localisé dans la ceinture

volcanique transversale au Mexique. Dans les dernières années, la qualité des écosystèmes aquatiques a reçu une attention croissante, car le changement d’usage des sols (sols volcaniques dégradés) modifie la qualité des eaux du lac de réservoir de Cointzio, qui apporte 30% de l’eau potable consommée dans la région de Morelia.

Deux programmes débutés en 2007 ont pour objectifs de :

- déterminer les conditions de formation des écoulements de surface, et cartographier les sources potentielles de matière en suspension dans le sous-bassin versant de Cointzio, - améliorer la connaissance des mécanismes de transport des sédiments et des contaminants associés à l’échelle du bassin versant dans le cas particulier de terrains volcaniques de montagne, - préciser les conditions de dépôt et de re-mobilisation des sédiments en conditions naturelles dans la retenue de Cointzio, - établir un état de référence de la qualité de l’eau et des sources de contamination dans le Rio Grande de Morelia.

Une quinzaine de points dans les rivières du bassin sont suivis mensuellement pour l’analyse

des concentrations en sédiments, et en nutriments (C, N, P). Ces points sont situés dans des sous bassins aux occupations des sols contrastées (cultivés, forestiers, fortement urbanisés ou non). Ce suivi est complété par une étude hebdomadaire de la qualité de l’eau de la retenue de Cointzio (nutriments, chlorophylle, paramètres physico-chimiques), en plusieurs points du lac et à plusieurs profondeurs. Le but est de modéliser le comportement biogéochimique du lac et du bassin de Cointzio afin de proposer des scénarios de gestion du bassin pour une meilleure qualité des eaux pour la ville de Morelia. J’interviens essentiellement sur le compartiment « sol », et sur l’établissement d’une base de données de qualité des eaux dans ces recherches

Ces études sont financées par un programme européen DESIRE « Desertification Mitigation and Remediation of Land » (collaboration Mexique, Chili, Europe, etc..), et un programme ANR STREAMS « Sediments Transport and Erosion Across Mountain s », (collaboration LSCE, EDF, Hydrowide) qui sont coordonnés par l’équipe RIVER du LTHE.

Site du Lautarêt, France Dans les bassins versants de montagne, l’augmentation des dépositions de nitrate

atmosphérique (NO3atm) est maintenant bien connue grâce aux mesures effectuées dans les carottes de glace prélevées dans les Alpes et l’inventaire des émissions des pays européens. Cette augmentation modifie fondamentalement la disponibilité de l’azote et impacte la végétation, les sols, les ressources en eau et les écosystèmes. L’impact des dépositions de NO3atm reste peu étudié, spécialement quand ces dépositions sont combinées avec une évolution de l’usage des sols, dans les bassins versants d’altitude. De nouveaux outils sont nécessaires pour mieux prédire les conséquences à long terme de ces évolutions. L’hydrogéochimie de l’azote sera étudiée, en utilisant des techniques isotopiques combinant la signature en 17O, 18O, et 15N dans les aérosols, les précipitations, les sols et les eaux de surface afin d’établir le bilan en azote du bassin. Des études in situ établiront l’interaction entre l’évolution de l’usage des sols et la déposition de l’azote. Je participerai plus spécifiquement à la partie 2 du projet qui s’intéresse à tracer le NO3atm dans les eaux de lixiviation, et relier ces signatures aux caractéristiques du terrain (usage du sol, diversité de la végétation, mésotopographie). Est-ce que la signature isotopique peut être un indicateur de l’état de saturation de l’azote dans un sol, ou un écosystème donné ? Est-ce qu’elle est relié aux flux dans le sols d’autre nutriments ? Les flux d’eau et de nutriments dans les sols seront étudiés grâce à l’utilisation de 80 lysimètres installés sur 30 sites, et l’installation de TDR/tensiomètres et de

75

bougies poreuses. Ces sites sont suivis par le LECA depuis 2003 pour le cycle de l’azote et la diversité de la végétation au Lautaret, qui fait partie de la Zone Atelier Alpes (inclus dans les « Long Term Environmental Research » (LTER) network).

Un projet coordonné par le LECA (OSUG Grenoble) et intitulé NATEAU « Impact des dépôts d’azote atmosphérique dans les bassins versants agricoles d’altitude » a été soumis pour financement à l’ANR. Le LTHE et le LGGE également de l’OSUG à Grenoble collaborent à ce projet. Il sera intégré dans le plateau PMPN (Plateau Milieu Poreux et Naturels) de Envirhonalp.

76

Références bibliographiques A2EP. 1993. Elaboration d'un schéma d'exploitation et de protection de la ressource en eau sur

Maré et Lifou. Agence pour l'Eau et l'Environnment du Pacifique. Addiscott, T.M., and R.J. Wagenet. 1985. Concepts of solute leaching in soils - a review of

modeling approaches. Journal of Soil Science 36:411-424. Basile, A., A. Coppola, R. De Mascellis, G. Mele, and F. Terribile. 2007. A comparative analysis of

the pore system in volcanic soils by means of water-retention measurements and image analysis, p. 493-513, In O. Arnalds, et al., eds. Soils of volcanic regions in Europe. Springer, Berlin.

Baskaran, S., N.S. Bolan, A. Rahman, and R.W. Tillman. 1996a. Pesticide sorption by allophanic and non-allophanic soils of New Zealand. New Zealand Journal of Agricultural Research 39:297-310.

Baskaran, S., N.S. Bolan, A. Rahman, and R.W. Tillman. 1996b. Non-equilibrium sorption during the movement of pesticides in soils. Pesticide Science 46:333-343.

Beauchemin S., D. Hesterberg, J. Chou, M. Beauchemin, R.R. Simard, and D.E. Sayers. 2003. Speciation of phosphorus in phosphorus-enriched agricultural soils using X-ray absorption near-edge structure spectroscopy and chemical fractionation. Journal of Environmental Quality 32(5):1809-1819.

Becquer, T., E. Bourdon, and W. Nigotte. 1993. Contribution à l'étude des sols de Maré: les sols des Centres d'Appuis au Développement Agricole de Tawaïnèdre et de Taodé. Caractérisation morphologique et cartographie des épaisseurs de sols, Nouméa, ORSTOM. Conv. Sic. Vie: Agropédologie, n°17, 37p.

Becquer, T., J. Petard, C. Duwig, E. Bourdon, R. Moreau, and A.J. Herbillon. 2001. Mineralogical, chemical and charge properties of Geric Ferralsols from New Caledonia. Geoderma 103:291-306.

Blakemore, L.C., P.L. Searle, and B.K. Daly. 1987. Procedures for soil analysis, Technical paper n°9, Wageningen, Netherlands: International Soil Reference and Information Centre.

Blanchard, F. 1990. Ile de Maré (Archipel des Loyauté): gestion et modélisation de la lentille d'eau douce. BRGM.

Bogdanov, I., D. Huaman, J.F. Thovert, P. Genthon, and P.M. Adler. 2007. A model for fracturation in the Loyalty Islands. Comptes Rendus Geoscience 339:840-848.

Bond, W.J., and I.R. Phillips. 1990. Ion-transport during unsteady water-flow in an unsaturated clay soil. Soil Science Society of America Journal 54:636-645.

Boudot, J.P., A.B.H. Brahim, R. Steiman, and F. Seiglemurandi. 1989. Biodegradation of synthetic organo-metallic complexes of iron and aluminium with selected metal to carbon ratios. Soil Biology & Biochemistry 21:961-966.

Brusseau, M.L., and P.S.C. Rao. 1989. Sorption nonideality during organic contaminant transport in porous-media. Critical Reviews in Environmental Control 19:33-99.

Buurman, P., K. deBoer, and T. Pape. 1997. Laser diffraction grain-size characteristics of andisols in perhumid Costa Rica: The aggregate size of allophane. Geoderma 78:71-91.

Buytaert, W., J. Deckers, G. Dercon, B. De Bievre, J. Poesen, and G. Govers. 2002. Impact of land use changes on the hydrological properties of volcanic ash soils in South Ecuador. Soil Use and Management 18:94-100.

Bürgisser, C.S., M. Cernik, M. Borkovec, and H. Sticher. 1993. Determination of nonlinear adsorption-isotherms from columns experiments - An alternative to batch studies. Environmental Science & Technology 27:943-948.

Chesworth, W. 2008. Encyclopedia of Soil Science, Series: Encyclopedia of Earth Sciences Series, W. Chesworth (eds), 902 p.

Childs, C.W., N. Matsue, and N. Yoshinaga. 1991. Ferrihydrite in volcanic ash soils of Japan. Soil Science and Plant Nutrition 37:299-311.

77

Clothier, B.E., M.B. Kirkham, and J.E. McLean. 1992. In situ measurement of the effective transport volume for solute moving through soil. Soil Science Society of America Journal 56:733-736.

Clothier, B.E., I. Vogeler, S.R. Green, and D.R. Scotter. 1998. Transport in unsaturated soil: aggregates, macropores, and exchange, p. 273-295, In M. H. Selim and L. Ma, eds. Physical nonequilibrium in soils. Modeling and application. Ann Arbor Press, Inc., Chelsea, Michigan.

Colletta, M. 2005. Comportement hydrologique d’un petit bassin versant agricole. La Loma, Etat de Mexico, Mexique, Mémoire de stage de deuxième année de l’INAPG, Paris Grignon.

Dahlgren, R., S. Shoji, and M. Nanzyo. 1993. Mineralogical characteristics of volcanic ash soils, p. 101-143, In S. Shoji, et al., eds. Volcanic ash soils. Genesis, properties and utilization, Vol. Development in Soil Science 21. Elsevier, Amsterdam.

Dahlgren, R.A., M. Saigusa, and F.C. Ugolini. 2004. The nature, properties and management of volcanic soils. Advances in Agronomy, Vol 82 82:113-182.

de Smedt, F., and P.J. Wierenga. 1984. Solute transfer through columns of glass-beads. Water Resources Research 20:225-232.

Dixon, J.B., and D.G. Schulze. 2002. Soil mineralogy with environmental applications SSSA, Madison, WI.

Dubus, I.G., and T. Becquer. 2001. Phosphorus sorption and desorption in oxide-rich Ferralsols of New Caledonia. Australian Journal of Soil Research 39:403-414.

Duwig, C. 1998. Etude des transferts d'eau et de nitrate dans les sols ferrallitiques de Maré (Nouvelle-Calédonie): risques de pollution des lentilles d'eau douce, Doctorat de l'Université Joseph Fourier, Grenoble 1.

Duwig, C., T. Becquer, B.E. Clothier, and M. Vauclin. 1998. Nitrate leaching through oxisols of the Loyalty Islands (New Caledonia) under intensified agricultural practices. Geoderma 84:29-43.

Duwig, C., T. Becquer, B.E. Clothier, and M. Vauclin. 1999. A simple dynamic method to estimate anion retention in an unsaturated soil. Comptes Rendus de l'Académie des Sciences, Série II Fascicule a - Sciences de la Terre et des Planètes 328:759-764.

Duwig, C., T. Becquer, I. Vogeler, M. Vauclin, and B.E. Clothier. 2000. Water dynamics and nutrient leaching through a cropped ferralsol in the Loyalty Islands (New Caledonia). Journal of Environmental Quality 29:1010-1019.

Duwig, C., and M. Vauclin. 2003. An overview of measurement techniques and modelling approaches of agrochemicals transport in the vadoze zone. Innovative Soil-Plant Systems for Sustainable Agricultural Practices:444-456.

Duwig, C., T. Becquer, L. Charlet, and B.E. Clothier. 2003a. Estimation of nitrate retention in a Ferralsol by a transient-flow method. European Journal of Soil Science 54:505-515.

Duwig, C., B. Normand, M. Vauclin, G. Vachaud, S.R. Green, and T. Becquer. 2003b. Evaluation of the WAVE model for predicting nitrate leaching for two contrasted soil and climate conditions. Vadose Zone Journal 2:76-89.

Duwig, C., P. Delmas, K. Muller, B. Prado, K. Ren, H. Morin, and A. Woodward. 2008a. Quantifying fluorescent tracer distribution in allophanic soils to image solute transport. European Journal of Soil Science 59:94-102.

Duwig, C., B. Prado, E. Raymundo, J. Lopez, C. Hidalgo, K. Müller, and J. Etchevers. 2008b. Herbicide fate in an allophanic soil: comparison of several experimental and data analysis techniques, EGU General Assembly 2008, 14-18 April, Vienna, Vol. 10, EGU2008-A-04906.

Duxbury, J.M., and P.K. McConnaughey. 1986. Effect of fertilizer source on denitrification and nitrous-oxide emissions in a maize field. Soil Science Society of America Journal 50:644-648.

Egashira, K., and S. Aomine. 1974. Effects if drying and heating on the surface area of allophane and imogolite. Clay Science 4:231-242.

78

EMEA. 2006. Guideline on the environmental risk assessment of medicinal products for human use (European Agency for the Evaluation of Medicinal Products), EMEA/CHMP/QWP/4447/00. London, 12p.

Escudey, M., and G. Galindo. 1983. Effects of iron coatings on electrophoretic mobility and dispersion of allophane. Journal of Colloid and Interface Science 93:78-83.

Escudey, M., J.E. Forster, and G. Galindo. 2004. Relevance of organic matter in some chemical and physical characteristics of volcanic ash-derived Soils. Communications in Soil Science and Plant Analysis 35 (5&6):781–797.

Estrella, M.R., M.L. Brusseau, R.S. Maier, I.L. Pepper, P.J. Wierenga, and R.M. Miller. 1993. Biodegradation, sorption, and transport of 2,4-dichlorophenoxyacetic acid in saturated and unsaturated soils. Applied and Environmental Microbiology 59:4266-4273.

Etchevers, J., L. Longeri, and J. Venegas. 1978. Nitrificación en suelos de Ñube, Concepción y Biobío. Agricultura Técnica 38:89-93.

FAO, ISRIC, and ISSS. 1998. World Reference Base for Soil Resources (WRB). World Soil Resources. FAO, Rome.

Feder, F., and A. Findeling. 2007. Retention and leaching of nitrate and chloride in an andic soil after pig manure amendment. European Journal of Soil Science 58:393-404.

Flury, M. 1996. Experimental evidence of transport of pesticides through field soils - A review. Journal of Environmental Quality 25:25-45.

FOCUS. 2001. FOCUS Surface water scenarios in the EU evaluation process under 91/414/EEC. Report of the Focus Working Group on Surface Water Scenarios. EC Document Reference SANCO/4802/2001 rev.2 [Online].

FOOTPRINT (2007, 2008). The FOOTPRINT Pesticide Properties DataBase. Database collated by the University of Hertfordshire as part of the EU-funded FOOTPRINT project (FP6-SSP-022704). En http://www.eu-footprint.org/ppdb.html.

Füleky, G., and L.N. Konda. 2007. Pesticide sorption of European volcanic soils, p. 539-544, In O. Arnalds, et al., eds. Soils of volcanic regions in Europe. Springer, Berlin.

Gillman, G.P., and E.A. Sumpter. 1986. Surface-charge characteristics and lime requirements of soil derived from basaltic, granitic, and metamorphic rocks in high-rainfall tropical Queensland. Australian Journal of Soil Research 24:173-192.

Harsch, J. 2000. Poorly crystalline aluminosilicate clays, In M. E. Sumner, ed. Handbook of soil science.

Hiradate, S., A. Furubayashi, N. Uchida, and Y. Fujii. 2007. Adsorption of 2,4-dichlorophenoxyacetic acid by an Andosol. Journal of Environmental Quality 36:101-109.

Hodder, A.P.W., B.E. Green, and D.J. Lowe. 1990. A 2-stage model for the formation of clay-minerals from tephra-derived volcanic glass. Clay Minerals 25:313-327.

Hoyoux-Roche, D., and P. Jamet. 1988. Rôle des conditions expérimentales sur l'équilibre d'adsorption d'un pesticide, p. 185-189, In P. Jamet, ed. Methodological aspects of the study of pesticide behaviour in soil. INRA, Versailles.

Katou, H., B.E. Clothier, and S.R. Green. 1996. Anion Transport Involving Competitive Adsorption during Transient Water Flow in an Andisol. Soil Science Society of America Journal 60:1368 – 1375.

INE. 2000. Características de peligrosidad ambiental de los plaguicidas, Instituto Nacional de Ecología , D.F., México.

ISEE. 2003. L'agriculture en Province des Iles Loyauté, Directeur de publication: DANVAR, Nouméa. http://www.isee.nc/recensagricol/telechargement/rga_iles.pdf.

Jacobsen, O.H., F.J. Leij, and M.T. Vangenuchten. 1992. Parameter determination for chloride and tritium transport in undisturbed lysimeters during steady flow. Nordic Hydrology 23:89-104.

Jaffré, T., and J.M. Veillon. 1987. Etude des jachères et premiers aperçus sur la succession secondaire à Lifou (Iles Loyauté, Nouvelle-Calédonie), Nouméa, ORSTOM. Rapp. Sci. et Tech.: Sic. Vie: Botanique, n°1, 28p.

Jarvis, N.J. 2007. A review of non-equilibrium water flow and solute transport in soil macropores: principles, controlling factors and consequences for water quality. European Journal of Soil Science 58:523-546.

79

Jauzein, M., C. Andre, R. Margrita, M. Sardin, and D. Schweich. 1989. A flexible computer code for modeling transport in porous-media - IMPACT. Geoderma 44:95-113.

Johnson, G.R., K. Gupta, D.K. Putz, Q. Hu, and M.L. Brusseau. 2003. The effect of local-scale physical heterogeneity and nonlinear, rate-limited sorption/desorption on contaminant transport in porous media. Journal of Contaminant Hydrology 64:35-58.

Kimble, J.M., C.L. Ping, M.E. Sumner, and L.P. Wilding. 2000. Andisols, p. E209-E223, In M. E. Sumner, ed. Handbook of Soil Science. CRC Press, Boca Raton.

Kjaergaard, C., T.G. Poulsen, P. Moldrup, and L.W. de Jonge. 2004. Colloid mobilization and transport in undisturbed soil columns. I. Pore structure characterization and tritium transport. Vadose Zone Journal 3:413-423.

Kutilek, M., and D.R. Nielsen. 1994. Soil Hydrology Catena Verlag, Cremlingen-Destedt, Germany.

Latham, M. 1980. Les oxydisols dans quelques milieux insulaires du Pacifique Sud. Cahiers ORSTOM, série Pédologie XVIII:305- 316.

Latham, M., and P. Mercky. 1983. Etude des sols des Iles Loyauté. Carte pédologique et carte d'aptitude culturale et forestière à 1:200 000., Paris.

Leamy, M.L. 1884. Andisols of the world, p. 368-387 Congreso internacional de Suelos Volcanicos, Vol. Serie informes 13. Secretariado de Publicaciones, Universidad de la Laguna.

Leij, F.J., and N. Toride. 1998. Analytical solutions for nonequilibrium transport models, p. 117-156, In H. M. Selim and L. Ma, eds. Physical nonequilibrium in soils. Modeling and application. Ann Arbor Press, Chelsea, Michigan.

Leij, F.J., and M.T. van Genuchten. 2000. Solute transport, p. A183-A227, In M. E. Sumner, ed. Handbook of soil science. CRC Press, London.

Lindberg, R.H., P. Wennberg, M.I. Johansson, M. Tysklind, and B.A.V. Andersson. 2005. Screening of human antibiotic substances and determination of weekly mass flows in five sewage treatment plants in Sweden. Environmental Science & Technology 39:3421-3429.

Logsdon, S.D., K.E. Keller, and T.B. Moorman. 2002. Measured and predicted solute leaching from multiple undisturbed soil columns. Soil Science Society of America Journal 66:686-695.

Ma, L., and H.M. Selim. 1998. Coupling of retention processes to physical nonequilibrium models, p. 83-116, In M. H. Selim and L. Ma, eds. Physical nonequilibrium in soils. Modeling and application. Ann Arbor Press, Inc., Chelsea, Michigan.

Ma, L.W., and H.M. Selim. 1994. Predicting the transport of atrazine in soils - 2nd order and multireaction approaches. Water Resources Research 30:3489-3498.

Macintyre, W.G., T.B. Stauffer, and C.P. Antworth. 1991. A comparison of sorption coefficients determined by batch, columns and box methods on a low organic-carbon aquifer material. Ground Water 29:908-913.

Maraqa, M.A. 2001. Effects of fundamental differences between batch and miscible displacement techniques on sorption distribution coefficient. Environmental Geology 41:219-228.

Martins, J. 2008. Microbiologie, transfert réactif et impact des micropolluants dans les sols, rapport HDR, Université Joseph Fourier, Grenoble.

Matus, F., and V. Maire. 2000. Interaction between soil organic matter, soil texture and the mineralization rates of carbon and nitrogen. Agricultura Técnica 60:112-126.

Morris, C., and S.J. Mooney. 2004. A high-resolution system for the quantification of preferential flow in undisturbed soil using observations of tracers. Geoderma 118:133-143.

Müller, K., and C. Duwig. 2007. Transport and sorption of 2,4-dichlorophenoxyacetic acid in allophanic soils. Soil Science 172:333-348.

Nanzyo, M., S. Shoji, and R. Dahlgren. 1993a. Physical Characteristics of Volcanic Ash Soils, p. 189-207 Volcanic Ash Soils - Genesis, Properties and Utilization, Vol. Volume 21. Elsevier.

Nanzyo, M., R. Dahlgren, and S. Shoji. 1993b. Chemical Characteristics of Volcanic Ash Soils, p. 145-187 Volcanic Ash Soils - Genesis, Properties and Utilization, Vol. Volume 21. Elsevier.

Okamura, Y., and K. Wada. 1983. Electric charge characteristics of horizons of Ando (B) and red-yellow B soils and weathered pumices. Journal of Soil Science 34:287-295.

80

Olvera, V. 1990. Estudio de la eutroficacion del embalse Valle de Bravo, México., Master Thesis, UNAM, Mexico, 87 p.

Olvera-Viascán, V., L. Bravo-Inclan, and J. Sánchez-Chavez. 1998. Aquatic ecology and management assessment in Valle de Bravo reservoir and its watershed. Aquatic Ecosystem Health and Management 1:277-290.

Parfitt, R.L. 1990. Allophane in New Zealand—a review. Australian Journal of Soil Research 28:343–360.

Parfitt, R.L. 1992. Surface charge in some New Zealand soils measured at typical ionic strength. Australian Journal of Soil Research 30:331-341.

Parfitt, R.L., and T. Henmi. 1980. Structure of some allophanes from New Zealand. Clay and Clay Minerals 28:285-294.

Parfitt, R.L., and J.M. Kimble. 1989. Conditions for formation of allophane in soils. Soil Science Society of America Journal 53:971-977.

Parfitt, R.L., and B. Clayden. 1991. Andisols- The development of a new order in soil taxonomy. Geoderma 49:181-198.

Parfitt, R.L., and G.J. Salt. 2001. Carbon and nitrogen mineralisation in sand, silt, and clay fractions of soils under maize and pasture. Australian Journal of Soil Research 39:361-371.

Parfitt, R.L., M. Russell, and G.E. Orbell. 1983. Weathering sequence of soils from volcanic ash involving allophane and halloysite, New Zealand. Geoderma 29:41-57.

Parfitt, R.L., C.W. Childs, and D.N. Eden. 1988. Ferrihydrite and allophane in 4 Andepts from Hawaii and implications for their classification. Geoderma 41:223-241.

Parkhurst, D.L., and C.A.J. Capelo. 1999. User's guide to PhreecqC (Version 2). A computer programm for speciation batch-reaction, one dimensional transport and inverse geochemical calculations. Water Resources Investigations Report:99-4259.

Payet, N., A. Findeling, J.L. Chopart, F. Feder, E. Nicolini, H. Saint Macary, and M. Vauclin. 2009. Modelling the fate of nitrogen following pig slurry application on a tropical cropped acid soil on the island of Réunion (France). Agriculture, Ecosystems and Environment in press.

Prado, B. 2006. Etude du mouvement de l'eau et du transfert réactif du nitrate dans les sols volcaniques du bassin versant élémentaire de la Loma, Mexique, Doctorat de l'Université Joseph Fourier, Grenoble 1.

Prado, B., C. Duwig, M. Escudey, and M. Esteves. 2006. Nitrate sorption in a mexican allophanic andisol using intact and packed columns. Communications in Soil Science and Plant Analysis 37:2911-2925.

Prado, B., C. Duwig, C. Hidalgo, D. Gomez, H. Yee, C. Prat, M. Esteves, and J.D. Etchevers. 2007. Characterization, functioning and classification of two volcanic soil profiles under different land uses in Central Mexico. Geoderma 139:300-313.

Prado, B., C. Duwig, J. Marquez, P. Delmas, P. Morales, J. James, and J. Etchevers. 2009. Image processing-based sudy of soil porosity and its effect on water movement though Andosol intact columns. Agricultural Water Management 96:1377-1386.

Qafoku, N.P., M.E. Sumner, and D.E. Radcliffe. 2000. Anion transport in columns of variable charge subsoils: Nitrate and chloride. Journal of Environmental Quality 29:484-493.

Qafoku, N.P., and M.E. Sumner. 2001. Retention and transport of calcium nitrate in variable charge subsoils. Soil Science 166:297–307.

Raymundo, E. 2008. Parámetros de transporte de atrazina en un Andosol y Vertisol de México, Doctorado del Colegio de Postgraduados, Montecillo, Mexique.

Raymundo, E., I. Nikolskii, C. Duwig, B. Prado, C. Hidalgo, F.G. Reyes, and B. Figueroa. 2009. Transporte de atrazina en un Andosol y un Vertisol disturbados de Mexico. Interciencia 34:330-337.

Rojas, F. 2004. Cuantificación y caracterización del escurrimiento superficial y erosión hídrica utilizando variables pluviométricas en una microcuenca de la cuenca Valle de Bravo, Thesis de Maestria de la Universidad Autónoma del Estado de México.

Roulier, S., and N.J. Jarvis. 2003. Analysis of inverse procedures for estimating parameters controlling macropore flow and solute transport in dual-permeability model MACRO. Vadose Zone Journal 2:349-357.

81

Sansoulet, J., Y.M. Cabidoche, and P. Cattan. 2007. Adsorption and transport of nitrate and potassium in an andosol under banana (Guadeloupe, French West Indies). European Journal of Soil Science 58:478-489.

Schnitzer, M. 1986. Binding of humic substances by soil mineral colloïds, p. 77-101, In P. M. Huang and M. Schnitzer, eds. Interactions of soil minerals with natural organics and microbes. SSSA, Special Publication N°17, Madison WI.

Segalen, P. 1985. Les sols ferralitiques et leur répartition géographique. Tomes 2 et 3, ORSTOM, collection Etude et Thèses, Paris, France.

SEMARNAT. 2003. Plan de Ordenamiento Ecologico Regional de la Subcuenca Valle de Bravo-Amanalco.

Shoji, S., M. Nanzyo, and R.A. Dahlgren, (eds.) 1993. Volcanic Ash Soils. Genesis, properties and utilization, pp. 1-288. Elsevier, Amsterdam.

Siemens, J., G. Huschek, C. Siebe, and M. Kaupenjohann. 2008. Concentrations and mobility of human pharmaceuticals in the world's largest wastewater irrigation system, Mexico City-Mezquital Valley. Water Research 42:2124-2134.

Spadotto, C.A., and A.G. Hornsby. 2003. Soil sorption of acidic pesticides: Modeling pH effects. Journal of Environmental Quality 32:949-956.

Sparks, D.L. 2001. Foreword, In H. M. Selim and D. L. Sparks, eds. Physical and chemical processes of water and solute transport/retention in soils, Vol. 56. Soil Science Society of America, Inc., Madison.

Soil Survey Staff. 1999. Soil taxonomy. A basic system of soil classification for making and interpreting soil surveys, Agriculture Handbook 436. Department of Agriculture, Natural Resources Conservation Service. Washington D.C.

Stolpe, N.B., and M.S. Kuzila. 2002. Relative mobility of atrazine, 2,4-D and dicamba in volcanic soils of south-central Chile. Soil Science 167:338-345.

Sumner, M.E., and M.J. McLaughlin. 1996. Adverse Impacts of Agriculture on Soil, Water and Food Quality, p. 125-182, In R. Naidu, et al., eds. Contaminants and the Soil Environment in the Australasia-Pacific Region, 20-24 Feb., Adelaide, Australia. Kluver Academic Publishers, Dordrecht, Boston.

Tercinier, G. 1971. Contribution à la connaissance des phénomènes de bauxitisation et d'allitisation. Les sols de karst d'atolls surélevés du Sud-Ouest Pacifique. Cahiers ORSTOM, série pédologie IX:307-334.

Thiele-Bruhn, S. 2003. Pharmaceutical antibiotic compounds in soils - a review. Journal of Plant Nutrition and Soil Science 166 (2):145 - 167.

Thouvenel, F. 2009. Détermination de la courbe de rétention en eau d’un Andosol mexicain et calibration du modèle WAVE, Stage de Master 1, Université Joseph Fourier.

Toride, N., F.J. Leij, and M.T. van Genuchten. 1999. The CXTFIT code for estimating transport parameters from laboratory or field tracer experiments, Version 2.1. U.S. Salinity Laboratory Research Report 137, Riverside, California.

USDHHS. 1998. Guidance for industry environmental assessment of human drug and biologics applications, US Departement of Health and Services, Food and Drug Administration, Center for Drug and Research, Center for Biologics Evaluation and Research. July, CMC6, revision 1.

Vachaud, G., M. Vauclin, and J. Colombani. 1981. Water-budget in Southern Tunisia. 1. Experimental characterization of water movementin the non-saturated zone. Journal of Hydrology 49:31-52.

van Genuchten, M.T., and P.J. Wierenga. 1976. Mas-transfer studies in sorbing porous-media. 1. Analytical solutions. Soil Science Society of America Journal 40:473-480.

van Genuchten, M.T., and F.N. Dalton. 1986. Models for simulating salt movement in aggregated fields soils. Geoderma 38:165-183.

van Genuchten, M.T., and R.J. Wagenet. 1989. 2-site 2-region models for pesticide transport and degradation - Theoritical development and analytical solutions. Soil Science Society of America Journal 53:1303-1310.

82

Van Ranst, E., S.R. Utami, and J. Shamshuddin. 2002. Andisols on volcanic ash from Java Island, Indonesia: Physico-chemical properties and classification. Soil Science 167:68-79.

Van Wambeke, A., P. Hastings, and M. Tolomeco. 1986. Newhall Simulation Model, Department of Agronomy, Cornell University, Ithaca, N.Y.

Vanclooster, M., P. Viaene, J. Diels, and J. Feyen. 1995. A deterministic evaluation analysis applied to an integrated soil-crop model. Ecological Modelling 81:183-195.

Vanderborght, J., P. Gahwiller, and H. Fluhler. 2002. Identification of transport processes in soil cores using fluorescent tracers. Soil Science Society of America Journal 66:774-787.

Vauclin, M. 1994. Modélisation du transport de solutés dans la zone non saturée du sol. Revue des Sciences de l'eau 7:81-102.

Viramontes, D., M. Esteves, L. Descroix, C. Duwig, F. Rojas-Rojas, A. Gutierrez, and B. de Leon-Mojarro. 2008. Quantification of runoff and erosion in an experimental andosol watershed in Valle de Bravo. Ingenieria Hidraulica En Mexico 23:89-103.

Vogeler, I., C. Duwig, B.E. Clothier, and S.R. Green. 2000. A simple approach to determine reactive solute transport using time domain reflectometry. Soil Science Society of America Journal 64:12-18.

Wada, K. 1989. Allophane and imogolite, p. 1051-1087, In D. J.B. and S. B. Weeds, eds. Minerals in soil environment. Soil Sci. Soc. Am. Book, Madison WI.

Wada, K., and D.J. Greenland. 1970. Selective dissolution and differential infrared spectroscopy for characterization of amorphous constituents in soil clays. Clay Minerals 8:241-254.

Watson, K.K., and M.J. Jones. 1981. Estimation of hydrodynamic dispersion in a find sand using an aproximate analytical solution. Australian Journal of Soil Research 19:265-273.

Wehrer, M., and K.U. Totsche. 2005. Determination of effective release rates of polycyclic aromatic hydrocarbons and dissolved organic carbon by column outflow experiments. European Journal of Soil Science 56:803-813.

Wilson, R.D., J. Geronimo, and J.A. Armbruster. 1997. 2,4-D dissipation in field soils after applications of 2,4-D dimethylamine salt and 2,4-D 2-ethylhexyl ester. Environmental Toxicology and Chemistry 16:1239-1246.

Wong, M.T.F., R. Hughes, and D.L. Rowell. 1990. Retarded leaching of nitrate in acid soils from the Tropics - measurement of the effective anion exchange capacity. Journal of Soil Science 41:655-663.

Zelazny, L.W., L. He, and A. Vanwormhoudt. 1996. Charge analysis of soils and anion exchange, p. 1231-1253, In D. L. Sparks, ed. Methods of soil analysis, Part 3, Chemical methods. SSSA, Madison, WI.

83

Annexe A : Principales caractéristiques physiques, minéralogiques et chimiques du Ferralsol de Maré et de l’Andosol de la Loma

Prof horizon porosity ρd Sand Silt Clay pH pH CEC Ca2+ Mg2+ K+ Na+ EA CO Pret Feox Alox Siox Feox/Fed cm cm3 cm-3 g cm-3 % H2O KCl cmolc kg-1 g kg-1 % Maré 0-12 Ap 0.75 2.4 35.3 62.4 6.8 6.3 28.8 17.8 10.5 0.3 0.2 0.05 72.3 82.2 0.4 1.1 0.01 0.0032 35-60 Bo2 0.70 0.5 13.9 85.6 5.8 6.0 1.7 0.7 0.8 0.0 0.1 0.02 10.1 93.8 0.4 0.5 0.01 0.0025 La Loma

0-15 Ap 0.61 0.76 29 62 9 5.5 5.1 22.3 7.8 0.5 0.4 0.3 0.08 54 >90 1.9 5.7 3.1 0.8 15-20 A1 0.71 0.64 45 50 5 6.1 5.8 23.0 18.2 1.2 0.5 0.3 0.07 53 >90 1.2 6.8 3.8 0.4 20-45 A2 0.71 0.53 23 66 11 6.2 5.8 20.0 19.3 1.6 0.1 0.1 0.05 56 >90 1 6.4 3.7 0.4 45-65 2A1 0.72 0.51 25 66 9 6.3 5.9 24.0 20.1 1.4 >10-4 0.1 0.05 53 >90 1.6 6.7 4.3 0.7 65-85 2A2 0.72 0.49 26 63 11 6.3 5.9 23.1 16.1 0.9 >10-4 >10-4 0.05 47 >90 1.6 6.0 4.3 0.5 85-110 A5 0.72 0.48 22 68 10 6.5 6.0 23.6 16.0 1.1 0.2 0.1 0.05 51 >90 1.3 7.2 4.6 0.5

Allophane : 6 × Siox (Parfitt, 1990) Ferrihydrite : 1.7 × Feox (Childs et al., 1991)

84

Annexe B : photos de sols Du Ferralsol de Maré, Nouvelle-Calédonie (jusqu’à 60 cm). On peut observer les remontées de calcaire corallien.

De l’Andosol de la Loma, Mexique (jusqu’à 110 cm)

85

Annexe C : Photos des différents dispositifs d’étude en colonne de sol

a

Dans le sens des aiguilles d’une montre : colonne de sol remanié, lysimètre remanié sous simulateur de pluie, tube de Perroux, lysimètre intact sous infiltromètre à disque, colonne de sol intact

86

Annexe D : Approche modélisatrice mécaniste déterministe 1D

L’équation de convection-dispersion

( ) ( )

∑+∂

∂−

∂∂

∂∂=

∂∂

+∂

∂i

irr

ssr

z

qC

z

CD

zt

S

t

C ϕθρθ (A.1)

où q est la densité volumique de flux d’eau donnée par la loi de Darcy (m s-1), Ds le coefficient de dispersion hydrodynamique (m2s-1) et Cr la concentration résidente de soluté (kg ou mol m-3), et Ss la concentration de la partie du soluté adsorbé (mol kg-1).

Le modèle de non équilibre physique MIM (Mobile Immobile Model) Cette classe de modèle a été proposée pour représenter l’hétérogénéité du flux durant le

transport de solutés dans les sols. Les hypothèses classiques sont que le sol est isotrope non déformable et qu’il n’y a qu’une dimension de flux. L’eau dans le sol est divisée en deux fractions : une mobile et une immobile. Les mécanismes de rétention des solutés dans les deux phases sont supposés être identiques. Le transfert de solutés entre les deux phases est décrit empiriquement par de la cinétique du premier ordre, avec un coefficient de transfert (α). En supposant un volume représentatif de sol V (cm3) avec une masse totale de sol P (g), l’équation de continuité pour le transport de soluté 1D peut s’écrire de la façon suivante (Ma and Selim, 1998) :

( )z

J

tV

M

∆∆−=

∆

∆ (A.2)

où J est le flux de soluté (µg cm-2 h-1) dans la direction z, M est la masse totale de solutés dans le volume V (µg) et t est le temps (h). J s’exprime par :

mm

m Cqz

CDJ +

∂∂

−= (A.3)

où q est le flux d’eau dans la région mobile (cm h-1), Cm est la concentration du soluté dans l’eau mobile (µg mL-1), et Dm est le coefficient de dispersion dans l’eau mobile (cm2 h-1). La masse totale de soluté M peut être écrite comme :

immimm NNMMM +++= (A.4)

où Mm et Mim sont les quantités de solutés adsorbés (µg) sur le sol en contact avec les zones mobile et immobile respectivement. Nm et Nim sont les quantités de solutés présents en équilibre (µg) dans les zones mobiles et immobiles, respectivement. En supposant V, q et Dm constants (régime permanent) l’équation de continuité devient

∂∂

∆∆+

∆∆

−=

∆∆

+∆

∆+

∆∆

+∆

∆z

CD

zz

Cq

t

N

t

N

t

M

t

M

Vm

mmimmimm1

(A.5)

En prenant Wm et Wim les volumes d’eau associés aux phases mobile et immobile (cm3)

respectivement, on a donc : mmm CWN = ; et imimim CWN = (A.6)

Cm et Cim ont une dimension de masse par volume (µg mL-1).

Soit,

87

t

C

t

C

t

N

t

N

Vim

imm

mimm

∆∆

+∆

∆−=

∆∆

+∆

∆ θθ1 (A.7)

où θm et θim sont les teneurs en eau mobile et immobile (cm3 cm-3) avec θ = θm + θim. Quand ∆ → 0, on arrive à l’équation de convection dispersion (CDE) généralisée :

2

21

z

C

t

Cq

t

C

t

C

t

M

t

M

Vmmim

imm

mimm

∂∂

+∂

∂−=

∂∂

+∂

∂+

∂∂

+∂

∂ θθ (A.8)

La masse totale de sol (P) est divisé en Pm, masse de sol en contact avec l’eau mobile, et

enPim, masse de sol en contact avec l’eau immobile. La concentration sorbée Sm sur Pm et celle sorbée Sim sur Pim (unités µg µg-1), peuvent être définies comme :

mmm PSM = et imimim PSM = (A.9)

donc,

( )2

2

1z

C

t

Cq

t

C

t

C

t

Sf

t

Sf mmim

imm

mimm

∂∂

+∂

∂−=

∂∂

+∂

∂+

∂∂

−+∂∂ θθρρ (A.10)

où f = Pm/P et ρ = P/V. L’équation (A.9) a été utilisée par van Genuchten et Wierenga (1976) où le bilan de masse était exprimé comme :

( )[ ]immimimmm SfSfCCM −+++= 1ρθθ (A.11)

L’équation (A.9) est associée à l’équation de transfert de masse du premier ordre entre les

phases mobiles et immobiles :

( )immimim

im CCt

S

t

C−=

∂∂

+∂

∂ αρθ (A.12)

où α est un coefficient de transfert de masse (h-1). L’application du modèle MIM nécessite deux paramètres additionnels en comparaison

avec l’équation CDE traditionnelle (équation A.1). Ce sont f, la fraction d’eau mobile (θm/θ) (supposée ici égale au rapport Pm/P) et α, le coefficient de transfert de masse entre les phases mobiles et immobiles. Ces paramètres sont difficiles à estimer pour la plupart des applications, et sont souvent optimisés. On peut trouver un exemple de détermination indépendante de f dans Clothier et al. (1992) par exemple, en employant un traceur de l’eau dans un infiltromètre à disque. Le deuxième paramètre, α, peut être dérivé de la géométrie du sol (van Genuchten et Dalton, 1986). D’autres études montrent que ce paramètre dépend des conditions expérimentales, et qu’il augmente avec la vitesse du flux (De Smedt et Wierenga, 1984).

Modèles de rétention chimique – Modèles à multi réaction non linéaire (MRM) Cette classe de modèles est fondée sur l’hypothèse que les affinités de sorption varient selon les constituants du sol, et sont représentées par un système de réactions simultanées et/ou consécutives et concurrentes. La Figure 16 tirée de Ma et Selim (1998) montre un diagramme schématique d’un modèle général de non équilibre chimique, où C est la concentration du soluté dans la solution du sol (µg mL-1) et Se est la quantité de soluté retenue par la matrice du sol (µg g-1 de sol), et est en équilibre avec C. Les phases sorbées S1 et S2 sont en contact direct avec C et sont

88

gouvernées par deux types de réactions de cinétique (mg g-1 de sol), S3 est une adsorption consécutive à S2 (µg g-1 de sol), et Sirr se réfère à la quantité de soluté adsorbée sur les sites d’adsorption irréversible (µg g-1 de sol). Les paramètres associés sont les suivants : Kd est la constante d’équilibre, k1 et k2 sont les coefficients de sorption et de désorption associés à S1 ; k3 et k4 sont les coefficients associés à S2, ; k5 et k6 sont associés à S3. kirr est le coefficient d’adsorption irréversible.

Figure 16 : Diagramme schématique du concept de non équilibre chimique (Ma and Selim, 1998) Les modèles à multi réaction supposent que les sites d’adsorption sur la matrice du sol sont illimités. D’après Ma et Selim (1994), les réactions chimiques gouvernant les différents mécanismes d’adsorption sont les suivantes :

nde CKS

=ρθ

(A.13)

1211 SkCkt

S n −

=∂

∂ρθ

(A.14)

36252432 SkSkSkCkt

S n +−−

=∂

∂ρθ

(A.15)

36253 SkSkt

S−=

∂∂

(A.16)

Ckt

Sirr

irr =∂

∂ (A.17)

où n est l’ordre de la réaction (sans dimension), θ est la teneur en eau volumétrique (cm3 cm-3), et ρ est la densité apparente du sol (g cm-3). Le coefficient de distribution à l’équilibre Kd est en cm3 g-1

et k1 à k6 sont en h-1. Les équations ci-dessus proposent un modèle dont seulement certains processus peuvent être sélectionnés selon l’étude ou les données disponibles. Par exemple, le programme CXTFIT (Toride et al., 1999) qui a été mis en œuvre dans ce travail pour modéliser les données expérimentales des expériences en colonnes en condition de régime permanent, utilise simplement des réactions du premier ordre avec adsorption linéaire (n = 1), et divisent les sites d’adsorption en deux types : des sites qui sont équilibre (adsorption instantanée, type 1) et des sites où l’adsorption suit une cinétique

S1

C

Sirr

Se S2 S3 Kd

k1 k2

k3

k4 k6

k5

kirr

89

de premier ordre (type 2). Ce modèle est appelé modèle de non équilibre chimique à deux sites (van Genuchten et Wagenet, 1989).

Couplage des modèles de non équilibre chimique et non équilibre physique Il est possible de coupler les modèles décrits ci-dessus, MIM et MRM. Ma et Selim (1998) donnent la formulation mathématique et proposent également d’autres couplages. D’autres types de modèles incluant le non équilibre chimique et physique sont les modèles d’échanges compétitifs d’ions, les modèles à fonction de transfert, les modèles à deux domaines ou domaines multiples de flux. L’utilisation de ces modèles soulèvent deux types de difficultés : l’identification des réactions chimiques et leurs éventuelles cinétiques, ainsi que le chiffrage des paramètres associées.

On présente seulement ici l’équation adimensionnelle utilisée dans CXTFIT (Toride et al., 1999) qui a une formulation mathématique et une solution analytique formellement identique, que ce soit pour le non équilibre chimique ou physique. Le transfert entre les zones à l’équilibre (réaction chimique instantanée et/ou eau mobile) et les zones en non équilibre (cinétique de premier ordre et/ou eau immobile) est décrit par un processus de premier ordre.

( ) )(1

111121

21

21 ZCCC

Z

C

Z

C

PT

CR λµωβ +−−+

∂∂

−∂∂

=∂

∂ (A.18)

( ) ( ) ( )ZCCCT

CR 22221

21 λµωβ +−−=∂∂

− (A.19)

où β est un coefficient de partition (relié à f), R est le coefficient retard (relié à Kd), C1 est la concentration adimensionnelle dans la région mobile (MIM) ou dans la phase liquide (modèle à deux sites), C2 est la concentration dans la région immobile ou la concentration adsorbée par les sites de type 2, T est le temps et Z la distance (adimensionnels), P est le nombre de Peclet, ω est le coefficient de transfert de masse (relié à α ou aux k1 à k6), µ et λ sont des terme de dégradation de premier ordre et de production d’ordre 0 respectivement. Les indices 1 et 2 se réfèrent aux phases où les processus sont à l’équilibre et ne sont pas à équilibre respectivement. Les formules reliant les paramètres dimensionnels et adimensionnels peuvent être consultées dans Toride et al. (1993) ou Leij et Toride (1998) par exemple. Dans certaines des expériences en colonnes de sol que nous avons menées sur les sols du Mexique, nous étions visiblement en présence de non équilibre chimique et physique (par exemple, transport de l’atrazine dans un Vertisol). En utilisant le programme CXTFIT, nous avons fait le choix de privilégier le processus dominant, c’est-à-dire généralement le non équilibre physique, en ayant conscience du fait que les paramètres adimensionnels obtenus par modélisation inverse (ou par optimisation) reflètent le couplage des deux processus. Il est également important de bien définir les conditions initiales et aux limites, ainsi que les termes de production. En général, pour des solutés appliqués pendant un temps limité à l’entrée de la colonne de sol, en régime d’écoulement de l’eau permanent, et dont les concentrations sont analysées en sortie, on impose une condition de flux à l’entrée et à la sortie (eau et soluté). Lorsqu’il y a plusieurs espèces chimiques en présence, les descriptions précédentes peuvent s’avérer insuffisantes, notamment lorsque d’autres facteurs, tels que le pH ou l’existence de complexants ou d’autres polluants en compétition, deviennent importants (Martins, 2008). Dans ces cas, une modélisation de type géochimique est nécessaire avec résolution des réactions et équilibres chimiques comme par exemples les modèles IMPACT (Jauzein et al., 1989) ou PhreeqC (Parkhurst et Appelo, 1999).

SELECTION D’ARTICLES

Duwig C., Becquer T., Clothier B.E., Vauclin M., 1998. Nitrate leaching through oxisols of the Loyalty Islands (New Caledonia) under intensified agricultural practices. Geoderma 84: 29-43.

Duwig C., Becquer T., Vogeler I., Vauclin M., Clothier, B.E., 2000. Water dynamics and nutrient leaching through a cropped Ferralsol in the Loyalty Islands (New Caledonia). Journal of Environmental Quality 29: 1010-1019.

Vogeler I., Duwig C., Clothier B.E., Green S.R., 2000. Time Domain Reflectometry and Solute Transport : Measurements and Modelling. Soil Science Society of America Journal 64: 12-18.

Becquer T., Pétard J., Duwig C., Bourdon E., Moreau R., Herbillon A.J., 2001. Mineralogical, chemical and surface properties of Geric Ferralsols of New Caledonia. Geoderma 103: 291-306.

Duwig C., Becquer T., Charlet L., Clothier B.E., 2003. Nitrate retention in a variable charge soil from the Loyalty Islands, New Caledonia. European Journal of Soil Science 54: 505-515.

Duwig C., Normand, B., Vauclin, M., Green, S.R., Becquer, T., Vachaud, G., 2003. Evaluation of the WAVE-model on two contrasted soil and climate conditions. Vadoze Zone Journal 2:76-89.

Prado B., Duwig C. , Escudey M., Esteves M., 2006. Nitrate sorption in a mexican allophanic andisol using intact and packed columns. Communications in Soil Science and Plant Analysis 37 (15-20): 2911-2925.

Prado B., Duwig C., Hidalgo C., Gómez D., Prat C., Etchevers J. D., Esteves M., 2007. Characterization, classification and functioning of two profiles under different land uses in a volcanic sequence in Central Mexico. Geoderma 139: 300-313.

Müller, K., Duwig C., 2007. The transport and sorption of 2,4-D in allophanic soils. Soil Science 72(5):333-348.

Duwig C., Delmas P., Müller K., Prado B., Morin H., Ren, K., 2008. Quantifying fluorescent tracer distribution in allophanic soils to image solute transport. European Journal of Soil Science 59, 94-102.

Prado B., Duwig C., Márquez J., Delmas P., Morales P., James J., Etchevers J., 2009. Image Processing-based study of soil porosity and its effect on water movement through Andosol intact columns. Agricultural Water Management, 96 (10), 1377-1386.

Ž .Geoderma 84 1998 29–43

Nitrate leaching through oxisols of the Loyaltyž /Islands New Caledonia under intensified

agricultural practices

C. Duwig a,c,), T. Becquer a, B.E. Clothier b, M. Vauclin c

a ORSTOM, Laboratoire d’Agropedologie, B.P. A5, 98848 Noumea Cedex, New Caledonia´ ´b EnÕironment Group, HortResearch, P.B. 11-030, Palmerston North, New Zealand

c (Laboratoire d’etude des Transferts en Hydrologie et EnÕironnement CNRS-UMR 5514, INPG,´)UJF , B.P. 53, 38041 Grenoble Cedex 9, France

Accepted 23 July 1997

Abstract

Ž .For the uplifted coral atolls of the Loyalty Islands New Caledonia , the prime source ofpotable water is the freshwater lenses that underlie the islands. The recent adoption of more-inten-sive agricultural practices, particularly the use of nitrogeneous fertilizers, may, however, representa threat for these fragile Pacific ecosystems. To assess the risk posed by nitrate leaching,experiments have been conducted on the permeable oxisols of the island of Mare, using both´cropped and bare soil sites. Drainage below the root zone was found to be very important, about50% of the rainfall, even on the cropped site. The soils are thin and permeable, and the frequenttropical storms have high rainfall intensities. Nitrate fertilizers thus have potential to be leached, inlarge amounts, even up to 100% of the nitrate supply, especially if fertilizers are not suppliedaccording to weather conditions and in concert with the plant’s ability to extract them. q 1998Elsevier Science B.V. All rights reserved.

Keywords: coral atoll; hydraulic conductivity; fertilizer; groundwater

1. Introduction

The adverse effects of intensified agricultural practices on soil and waterŽ .quality are well documented Sumner and McLaughin, 1996 . Pollution of our

) Corresponding author. Tel.: q64-6-3568080; Fax: q64-6-3546731; E-mail:[email protected]

0016-7061r98r$19.00 q 1998 Elsevier Science B.V. All rights reserved.Ž .PII S0016-7061 97 00119-5

( )C. Duwig et al.rGeoderma 84 1998 29–4330

reserves of drinking water by nitrate leaching from the root zone is a majorpublic concern of increasing intensity. On the other hand, the world’s populationgrowth demands that there should be an increase in food production as well as inother agricultural products. The ability to use new arable or pastoral land islimited. Therefore, the use of fertilizers, mainly nitrogen, is often the first meansadopted to achieve an improvement in the productivity of agriculture, so as to

Ž .increase the return from the land Ange, 1992 .´As in many developing countries, the islands of the South Pacific must

increase their agricultural production because their population is likely toŽ .increase some twofold over the next 30 years WRI, 1990 . Traditional agricul-

ture is still the most important source of food for the people of the PacificŽ .Brookfield, 1989 , and agriculture is generally the main source of national

Žincome. Furthermore, the cash-cropping sector is increasing in size Naidu et al.,.1991 . However, more and more young people are leaving their rural villages to

go to urban areas. Intensification of agricultural practices means that less labouris required relative to traditional practices. Thus, while there is an increasingdemand for the potable water by urban dwellers, at the same time in rural areasmore agrochemicals are being applied to the soils as agricultural practicesbecome more intensive. Leaching of these agrochemicals from the root zonemight compromise the quality of the water being demanded by the city.

The Loyalty Islands in New Caledonia, are no exception to these conflicts.Because of economic and demographic pressure, farmers are using more inten-sive agricultural practices, rather than traditional procedures. The increased useof nitrogen fertilizers could, however, pose a risk to this fragile ecosystem.Indeed, the freshwater lenses that underlie these uplifted coral atolls oftenconstitute the sole source of drinking water. The primary aim of this currentstudy is to quantify nitrate leaching under intensified agricultural practices. Thisresearch stresses the need to understand drainage from the root zone so as todetermine the optimal fertilization rate. In this way, the amount of nitrateleaching beyond the bottom of the profile can be reduced, without unacceptablyreducing crop production.

2. Materials and methods

2.1. Experimental site

Ž .The study was carried out on the island of Mare New Caledonia , the´southernmost island of the Loyalty Islands. Like some other volcanic islands inthe Pacific Ocean, Mare is an uplifted coral atoll which is build upon an´underlying volcanic structure. Above the basalt substratum, the coral rock isfrom 50 to 100 m deep and displays many fractures whose complex structure isnot yet well known. The freshwater lenses are subterranean, and float upon the

Žunderlying salt water. The soil of the study site is an oxidic ferrallitic soil type

( )C. Duwig et al.rGeoderma 84 1998 29–43 31

.Acorthox, USDA, 1975 . This type of soil represents just 19% of the entiresurface of the island, but nearly 60% of the cultivated soils. The soil derivesfrom altered volcanic ejecta and ash, and is relatively thin, ranging from nothingto one meter deep. Across the experimental field site on average the depth is 0.4

Žm. The soil primarily comprises iron and aluminium oxides Latham and.Mercky, 1983 , mainly as gibbsite, boehmite and goethite with a very low level

of silicates. The main physico-chemical characteristics of the soil for varioussubplots at the Tawaınedre site are presented in Table 1. Here the code Bush¨ `refers to a profile under bushfallow, and was collected before commencement offield experimentation, while Corn and Grass are profiles of two plots of theexperimental field, collected under corn and grassland, after two years ofcultivation.

2.2. Field experimentation

Three plots of 400 m2 were studied. They were ploughed at the beginning ofJanuary 1995. The bare soil plot was kept bare by the application of herbicide.

Ž .Another plot was sown with corn Zea mays, cv. Hycorn 90 at 50,000 plantsy1 Žha on the same date. The last plot was a two-year old grassland Rhode grass,

.Chloris gayaha, cv. Callide . On January 11, 1995, all plots received 104 kg Nhay1 as ammonium nitrate. The results presented here relate only to the first halfof 1995, between January and June, during the growth cycle of the corn. Thisperiod corresponds to the wet season under this tropical climate, during whichtime leaching is most likely to occur.

2.3. Measurements

The purpose of this study was to monitor both the drainage from the rootzone and the leaching of nitrogen under the different plots.

Each plot was instrumented for soil moisture and soil solution concentrationŽmeasurements. Eight waveguides for water content measurement by TDR Time

Domain Reflectometry; TRASE, Soilmoisture Equipment Corporation, Santa.Barbara, CA were installed horizontally to measure the water content at 10, 20,

30 and 40 cm. Until the month of April, when the TDR became available, soilmoisture measurements were made gravimetrically. Nearby, eight tensiometerswere installed to record the hydraulic head of water in the soil at 10, 20, 30, 40and one at 60 cm where the soil was deep enough. Eight suction cups werelocated at both 10 and 40 cm depth. These allowed measurement of the nitrogenconcentration of the soil solution.

Measurements of soil moisture and tensiometers were recorded every twodays just after a rainfall, and every week during drying periods. Collection ofsoil solution samples were made weekly when the soil was wet enough.Temperature, rainfall and micrometeorological data were obtained every hourwith an automatic data acquisition system. These data allowed estimation of the


Tab

le1

Ž.

Ž.

Ž.

Sel

ecte

dpr

oper

ties

ofso

ilpr

ofil

esun

der

bush

fall

owB

ush

,co

rnC

orn

,an

dgr

assl

and

Gra

ssy

1Ž

.Ž.

Ž.

Ž.

Pro

file

sD

epth

Bul

kP

arti

cle

pHO

rgan

icE

xcha

ngea

ble

base

scm

olkg

aC

EC

aT

otal

elem

ents

bŽ

.cm

dens

ity

dens

ity

CN

atpH

7H

OK

Cl

Ca

Mg

Na

KS

iOF

eO

Al

O2

22

32

3y

3y

3y

1y

1y

1Ž

.Ž

.Ž

.Ž

.Ž

.M

gm

Mg

mg

kgg

kgcm

olkg

Ž.

Ž.

Ž.

%%

%

Bus

h-1

0–

150.

62—

6.8

6.3

76.8

06.

7715

.79

10.0

80.

260.

2826

.93

nd21

.05

37.6

2B

ush-

215

–35

0.73

—6.

36.

224

.02

2.10

4.32

3.33

0.18

0.08

10.7

11.

0624

.51

42.9

5B

ush-

335

–60

0.80

—5.

86.

08.

440.

950.

280.

300.

080.

013.

481.

4625

.14

44.0

0C

orn-

10

–30

0.73

2.60

6.6

6.2

78.6

06.

0616

.01

11.6

00.

280.

7333

.06

——

—C

orn-

230

–40

0.82

2.75

6.5

6.1

50.8

54.

198.

748.

330.

200.

3024

.08

——

—C

orn-

340

–60

0.87

—5.

75.

916

.65

1.36

1.21

1.29

0.06

0.12

9.79

——

—G

rass

-10

–15

—2.

606.

66.

375

.57

5.62

14.0

910

.05

0.23

0.29

29.3

0—

——

Gra

ss-2

15–

25—

2.85

6.3

6.1

39.2

03.

255.

244.

830.

200.

1117

.00

——

—G

rass

-325

–40

——

5.8

6.2

11.2

41.

120.

680.

530.

140.

084.

44—

——

Ž.

Ž.

aT

ucke

r,19

54;

bD

iges

tion

wit

hpe

rchl

oric

acid

;—

sno

data

.


Ž .potential evapotranspiration ETP with the use of the Penman–Monteith methodŽ .Brunel, 1994 .

2.4. Methods

The relationship between soil hydraulic conductivity, K , and water content,u , was determined by two different and complementary methods.

The first method involved the use of the ‘zero flux plane’ approach andŽ . 3 y3provided values of K u in the unsaturated range of uf0.3 cm cm . As

Ž .described by Vachaud et al. 1978 , this method is based on an analysis of thesoil moisture and soil hydraulic head profiles during periods of drainage in theabsence of rainfall, when there are no plants, or plants just having a veryshallow depth of rooting.

ŽThe use of a tension disc infiltrometer Clothier and White, 1981; Ankeny et.al., 1991 allowed determination of the soil’s unsaturated hydraulic conductivity

in the region close to saturation. Measurements were made at three differentŽ .suctions 0.5, 5 and 15 cm of water . These measurements were carried at a

depth of approximately 40 cm in the profile, in the third horizon at the base ofthe root zone.

Ž .With the K u relationship formed by cobbling together these two data sets,Ž .and from the measured hydraulic information, the drainage D m at a depth of

40 cm was calculated from Darcy’s law. Here the cumulative amount ofdrainage can be found using:

D HDsqPD tsyK u P PD t 1Ž . Ž .

D zŽ y1. Ž .where q is the mean volumetric flux density m day during D t, K u is the

Ž y1.hydraulic conductivity m day corresponding to the measured water contentat 40 cm, and D HrD z is the hydraulic head gradient measured at this depth.

Nitrate and ammonium concentrations were measured in the soil solution.Ž y2 .The rate of nitrogen leaching L kg m below the root zone was thusN

obtained from the relationship:

L sDPC 2Ž .N

Ž . Žwhere D m is the drainage at depth 40 cm, as calculated above, and C kgy3.NO –N m is the NO –N concentration measured by suction cups at this3 3

Ž .depth Kengni et al., 1994 .

3. Results

3.1. Hydraulic conductiÕity

Determination of hydraulic conductivity was carried on every plot, below theroot zone. Because there are not enough data to characterize individually each

Žsite, one curve was fitted to all the data using a power law. Others Vachaud et


Fig. 1. The fitted relationship between measurements of soil hydraulic conductivity and thevolumetric water content.

.al., 1981; Poss and Saragoni, 1992 have suggested this approach. In our case,the soil across the entire site is likely to be quite uniform at this depth. Theresults are presented in Fig. 1.

On this figure, two groups of measured points can be distinguished dependingŽ .on the method used to obtain the K u data. As a matter of fact, the

infiltrometer gives values near the saturation and the ‘zero flux plane method’ inŽ .the range of the water content mostly found on the field see Fig. 2 . We can

note the high sensitivity of the hydraulic conductivity to water content, and moreimportantly this soil is found to be very permeable at saturation. The two data‘clouds’ provide good upper and lower limits to the values of hydraulicconductivity in the range of the water content that is important for drainageŽ . Ž .Figs. 1 and 2 . The fitted line thus allows interpolation so that Eq. 1 can beused to compute the water flux from measurements of the water content andhydraulic head gradient.

3.2. Soil water content and hydraulic head

As is shown in Fig. 2, the response of the water content, even at 40 cm depthis significant to any important rainfall input. Such a significant response isexpected for a soil having such high values of hydraulic conductivity nearsaturation. However, the temporal variation at 40 cm would not be as large as inthe upper horizons. Both these traits lend evidence to our proposed use ofDarcy’s law at this depth. The fluctuations are muted, and there is a consistencybetween the various plots, as might be expected due to the greater uniformity at40 cm.

Likewise, as is shown in Fig. 3, the hydraulic head gradient in this permeablesoil also responds quickly to any rainfall input. On bare soil, apart from after the


Fig. 2. The water content at 40 cm with time on the different plots.

first important rainfall followed by a 3-week dry period, the hydraulic headgradient is mostly negative. This indicates that the soil is often subject todrainage, and consequently the possibility for nitrate leaching would alwaysexist.

The behaviour of the bare soil plot and the corn plot were similar, at least upuntil one month after sowing. However, after that, root extraction started to beimportant at the cropped site. The upwards hydraulic head gradient was some-times very high, perhaps because of root dry down of the surface soil, with theprospect of capillarity bringing water upwards from below the root zone.

On the grassland plot, the gradient was almost always negative, being similarto the bare soil plot. However, after the first rainfall event, the gradient herebecame positive sooner than in the two other plots. So even with an active, butshallow-rooted pasture, water can be lost rapidly after any significant rainfallevent.


Fig. 3. The hydraulic head gradient at 40 cm during several rainfall events.

Fig. 4. Calculated drainage at 40 cm for each plot.


3.3. Drainage flux

To calculate the drainage, it was necessary to establish on a daily basis thewater flux density by Darcy’s law. But because of the dynamic behaviour of thissoil, this poses some problems with respect to water content determination. Toovercome this, it was necessary that the water content values were interpolatedbetween two measurements far apart using tensiometer data. For this, a water-re-tention curve was used having being obtained from simultaneous TDR-tensiome-ter measurements.

The drainage flux at each site is shown in Fig. 4. Drainage was at a maximumthe day after rainfall, but it then diminished rapidly, and became negligible afterone week. Such peaked behaviour is to be expected for a soil possessing such a

Ž . Ž .steep K u relationship Fig. 1 . Following the first and intense rain, some 170mm in 8 h, the total drainage was almost 70–78% of the rainfall. For thefollowing rains, the drainage fraction was much less. These values are nonethe-less high because of the high value of the hydraulic conductivity when the soil is

Table 2Ž a.Comparison between ETR estimated from ETP ETR and calculated by the water balance

Ž b.ETR . NO –N contents in soil solution and standard deviations3

Periods of drainage

Date: 29r01–08r02 24r02–13r03 03r04–14r04 15r04–5r05

Ž .P mm 179.4 104.8 65.9 127.0Ž .ETP mm 67 82 49 85a Ž .ETR mm 31 52 21 54

Bare soilŽ .DS mm 10 30 10 10

Ž .D mm 141 25 34 62b Ž .ETR mm 29 50 22 55

y3Ž .NO –N g m 32.5"29.5 24.5"20.7 14.9"15.6 17.4"20.53

CornŽ .DS mm 5 27 10 10

Ž .D mm 141 37 38 62b Ž .ETR mm 33 41 18 55

y3Ž .NO –N g m 72"39.8 44.1"41.8 12.4"25.2 16.4"16.93

GrasslandŽ .DS mm 10 38 10 10

Ž .D mm 132 35 22 63b Ž .ETR mm 37 31 29 54

y3Ž .NO –N g m 10.2"14.2 0.1"0.2 0.0 0.04"0.083

a Ž .ETR calculated by Chopart and Siband 1988 method.b ETR estimated by the water balance.


close to saturation, as happens during and just after such heavy tropicaldownpours that result from convective weather systems.

3.4. Water balance

A water balance was only calculated over periods when the hydraulic headgradient was negative, that is to say during each drainage period. There was alack of data during drying periods to allow this. Integration of the law of massconservation between the surface and 40 cm deep leads to:

ETRsPyRyDSyD 3Ž .

Ž . Ž .where ETR mm is the actual evapotranspiration, P mm the amount ofŽ . Ž .rainfall, R mm the surface runoff, DS mm the water storage variation and D

Ž .mm the drainage. All these parameters were determined during drainageperiods. Surface runoff did not occur as the highest rainfall intensity recordedwas 50 mm hy1, while the hydraulic conductivity near saturation was about 100mm hy1. The water storage was calculated using water contents in the wholeprofile.

This calculation of ETR was compared with the estimation of ETR using theŽ . Žpotential evapotranspiration ETP by Penman–Monteith Chopart and Siband,

.1988 . This last estimation as well as the other components of the water balanceare presented in Table 2.

Fig. 5. Amount of NO –N leached at 40 cm for various periods.3


On the bare soil plot, the two values of ETR were very close, while they weredifferent for the two other plots. This can be explained by the estimation of ETRusing ETP. Indeed, this estimation of ETR does not include the effect of a plantand its transpiration but only the state of soil water content.

3.5. Nitrogen leaching

Cumulative drainage was also calculated over each drainage period. Anaverage value of the nitrogen content in the soil solution at 40 cm was

Ž .determined for each period Table 2 so that the amount of nitrogen lost belowŽ . Ž .that depth could be calculated using Eq. 2 Fig. 5 . The ammonium content

was found to be almost negligible in the soil of each plot, presumably becauseof a rapid nitrification of ammonium to nitrate in the moist, warm, tropical soils.

In the drainage period after the first rain, just 20 days after the fertilization,the loss of nitrogen was high on all plots. On the bare plot, some 44% of thenitrogen was lost, relative to the total nitrogen supplied. The loss was 98% forthe corn and 13% for the grass. Under grassland, the rate of nitrogen leachingthen became negligible, whereas losses continued for the other plots.

4. Discussion

4.1. Water fluxes

Our proposed use of Darcy’s law is confirmed by the similarity between thetwo different estimations of ETR for the bare soil plot. The Chopart and SibandŽ . Ž .1988 method was assessed in Chopart and Vauclin 1990 and they also founda good agreement between the two estimations of ETR.

On all plots, the soil water content at 40 cm was maintained within a narrowband, being between 0.25 and 0.35 cm3 cmy3. This was much less thanvariations of water content near the surface. This small variation made easier the

Ž .Darcy calculation of drainage, as K u was known in this range. Rains wererelatively heavy and regular during the study. The temporal evolution of thewater content within the profile became different for each plot during the secondpart of the cycle studied. Here the rainfall was less intense, albeit still regular,and the plants’ consumption was more significant.

ŽDuring the first part of the cycle of the corn’s growth for the first 70 days. Ž .after sowing , the hydraulic head gradients Fig. 3 under the young corn and for

the bare soil were basically similar. The water consumption by the young cornwas not significant at this stage. Under grassland, the active roots of thewell-developed grass took up the water from the top soil, and thereby decreasedits water content. The remoistening of this upper horizon by water drawn frombelow created a highly positive, upward hydraulic head gradient. During the


second part of this study, the consumption of water by the growing corn becamegreater and the gradient also became strongly positive by the end of the study.On the other hand, the bare soil and the grassland maintained a similar patternbecause the mature grass had reached the end of its growth cycle and wasprobably becoming senescent. In summary then, the variation of the hydraulichead gradient was found to be rapid yet critical in establishing the direction andrate of water flow. Compared to the sandy and permeable soil studied by

Ž .Vachaud et al. 1978 in Senegal, where the gradient reached the value of 3 afterthe dry season of 7 months, this ferrallitic soil of Mare under a tropical climate´displays similar behaviour, but with an even quicker temporal response. How-

Ž .ever, for the soil studied by Kengni et al. 1994 under temperate conditions, aglacial terrace soil near Grenoble, France, the variation in the hydraulic headgradient was far less important.

Finally, during drainage periods, cumulative amounts of drainage calculatedfor each plot are very similar, probably because in this humid climate, theamount of water consumed by plants is small compared to the drainage.Drainage was found to be about 50% of the rainfall. For the first event, atropical storm, 55% of the total drainage was lost within 2 days, 80% in 3 daysand drainage became negligible within 7 days of the rain. Overall the totaldrainage was about 64% of this 170 mm rainfall event. This indicates that onMare water is rapidly lost in significant amounts. In this case, the soil is not able´to retain the water and the plant can derive little benefit from such heavy andintense rains. Given this dominance of precipitation, the behaviour of the baresoil and the cropped site were not that different. However, this would not be thecase when a heavy rain occurred after a long dry period, for the cropped sitewould now be drier than the bare soil because of root water extraction. In such acase, the drainage could be less on the cropped site.

4.2. Nitrogen leaching

At the beginning of the study, the high nitrogen levels found under the cornprobably reflected the low level of uptake by the young corn plants. Undergrassland, the applied nitrogen had all disappeared within the first month,presumably because of rapid uptake or immobilization. However, for the initialperiod of the first 70 days after sowing, the amount of nitrogen leached undercorn was high compared to that from bare soil. Indeed, the two plots had beenploughed at the same time and the young corn was not yet effective at uptake.Another corn plot was also studied, and although detailed results are notpresented here. We found the amount of nitrogen leached during this first periodwas less high, being 4.2 g my2. This difference between the two plots, some 5.9g my2, may be due to spatial variability, and more likely to the error inestimating drainage or measuring the nitrogen content. Indeed, the standard

Ž .deviations in nitrogen concentration Table 2 are high. Furthermore, measure-


ments were made every week during the rainfall period, as recommended byŽ .some authors Poss, 1991; Kengni, 1993 . In Mare where fluxes are rapid in this´

permeable soil, more frequent measurements may be required. Furthermore,high variability in results might also result from variability of the soil, theheterogeneity of fertilizer supply, and the local variability in the location ofroots in this case of corn sown in rows.

Furthermore, the leaching losses can be higher than the nitrogen fertilizerinput. So, the mineralization of organic matter can even contribute to theleaching. In the case of corn, for the whole growing period, 133 kg N hay1 waslost by drainage and we found that approximately 50 to 100 kg N hay1 was

Žconsumed by plants. Compared to the amount of nitrogen supplied 104 kg Ny1.ha , the mineralization on this cultivated plot is relatively weak, between 0.6

and 1% of the native organic nitrogen contained in the 0–0.3 m layer of this plotŽ .Table 1 . A higher mineralization rate, up to 1.5%, could have been expected in

Ž .this aerated and fine-textured soil Schepers and Meisinger, 1994 . Boudot et al.Ž .1988 found that gross mineralization rate for N was inversely related tocontents of amorphous Al and allophanic constituents. Indeed, the amorphous Alin this soil is high, and that could explain the accumulation of organic matter inthe 0–0.3 m layer.

5. Conclusions

The tensiometer-TDR method of hydraulic characterization, coupled with theinfiltrometer, provided us with a good measure of the hydraulic properties ofthis permeable tropical oxisol. These characteristics suggest that this thin, verypermeable soil, and a tropical climate with high rainfall intensities, conspire tocreate an agricultural ecosystem that has the potential to pollute underlyinggroundwater due to significant leaching. We found high rates of nitrogen loss,especially on bare soil, and on plots where the plant cover was not yet active, orwhere native nitrate levels were high. The nitrogen contents recorded in the soilsolution were between 0 and 72 g my3 of NO –N, being almost always higher3

Ž y3that the maximum level suggested by WHO for drinking water 12 g m of.NO –N . The exception was under grassland one month after the fertilizer3

application. So fertilizer supply needs to be carefully tailored to match theplants’ ability to extract it. In our case, the ammonium content was found to benegligible. Hence, there appears to be a rapid nitrification of ammonium intonitrate in this warm and moist tropical soil.

This experiment has already led to changes in the fertilizer programme, withan emphasis now on more frequent applications of small amounts. In the future,the frequency of experimental measurements will be likewise increased in orderto find out exactly what happens after a tropical rainfall and to determine moreprecisely the water balance. Some additional analyses, such as the amount of


nitrate taken up by the plants, and the amount of mineralization, will lead toeven better estimates of the nitrogen balance of this fragile tropical agro-ecosys-tem.

Acknowledgements

This research was supported by the ‘‘Contrat de Developpement entre la´Province des Iles et l’ORSTOM pour l’etude des risques de degradation de la´ ´fertilite des sols et de pollution des lentilles d’eau douce’’ and by the French´Ministry of University Education and Research for the Operation no. 6DEF. Theauthors thank the analysis laboratory staff from ORSTOM–Noumea for the´chemical analyses and CIRAD for their collaboration in the field.

References

Ange, A.L., 1992. Environmental risks of fertiliser use. Overuse or misuse. FADINAP regional´seminar fertilization and environment, 7–11 september 1992, Chiang Mai, Thailand, 57 pp.

Ankeny, M.D., Ahmed, M., Kaspar, T., Horton, R., 1991. Simple field method for determiningunsaturated hydraulic conductivity. Soil Sci. Soc. Am. J. 55, 467–470.

Boudot, J.P., Brahim, A.B.H., Chone, T., 1988. Dependence of carbon and nitrogen mineralizationrates upon amorphous metallic constituents and allophanes in Highland soils. Geoderma 42,245–260.

Brookfield, H.C., 1989. The human context of sustainable smallholder development in the Pacific.In: IBSRAM Proc., Bangkok, 8, 189–204.

Brunel, J.P., 1994. L’evaporation sous differents climats du Sud-Ouest Pacifique. Etudes energe-´ ´ ´ ´tique et isotopique. These Doct., Univ. Paris-Sud, ORSTOM Ed., 426 pp.`

Chopart, J.L., Siband, P., 1988. PROBE: programme de bilan de l’eau. Mem. Trav. IRAT 17, 76´pp.

Chopart, J.L., Vauclin, M., 1990. Water balance estimation model: field test and sensitivityanalysis. Soil Sci. Soc. Am. J. 54, 1377–1384.

Clothier, B.E., White, I., 1981. Measurement of sorptivity and soil water diffusivity in the field.Soil Sci. Soc. Am. J. 45, 241–245.

Kengni, L., 1993. Mesure in-situ des pertes d’eau et d’azote sous culture de maıs irriguee.¨ ´Ž . Ž .Application a la plaine de la Bievre Isere . These Doct., Univ. Joseph Fourier UJF , Grenoble` ` ` `

1, 200 pp.Kengni, L., Vachaud, G., Thony, J.L., Laty, R., Garino, B., Casabianca, H., Jame, P., Viscogliosi,

R., 1994. Field measurements of water and nitrogen losses under irrigated maize. J. Hydrol.162, 23–46.

Latham, M., Mercky, P., 1983. Etudes des sols des ıles Loyaute. Carte pedologique et carteˆ ´ ´d’aptitude culturale et forestiere a 1:200 000. ORSTOM, Notice Explic. 99.` `

Naidu, R., Haynes, R.J., Gawandar, J.S., Morrison, R.J., Fitzpatrick, R.W., 1991. Chemical andmineralogical properties and soil solution composition of acid soils from the South Pacific

Ž .Islands. In: Wright, R.J. et al. Eds. , Plant–Soil Interactions at Low pH. Kluwer AcademicPublishers, pp. 4–53.

Poss, R., 1991. Transferts de l’eau et des mineraux dans les terres de Barre du Togo. Conse-´quences agronomiques. These Doct., Univ. Paris VI, 291 pp.`


Poss, R., Saragoni, H., 1992. Leaching of nitrate, calcium and magnesium under maize cultivationon an oxisol in Togo. Fert. Res. 33, 123–133.

Schepers, J.P., Meisinger, J.J., 1994. Field indicator of nitrogen mineralization. In: Soil Testing:Prospects for improving Nutrient Recommendations. Soil Sci. Soc. Am. Spec. Publ. 40, 31–47.

Sumner, M.E., McLaughin, M.J., 1996. Adverse impacts of agriculture on soil, water and foodŽ .quality. In: Naidu, R., Oliver, D.P., Rogers, S., McLaughin, J.M. Eds. , Contaminants and the

Soil Environment in the Australasia–Pacific Region, Adelaide, Australia, 18–23 February1996. Kluwer Academic Publishers, pp. 125–181.

Tucker, B.M., 1954. The determination of exchangeable calcium and magnesium in carbonatesoils. Aust. J. Agric. Res. 5, 706–715.

USDA, 1975. Soil Taxonomy, a basic system of soil classification for making and interpreting soilsurvey. U.S. Dep. Agric. A and B, 436 pp.

Vachaud, G., Dancette, C., Sonko, S., Thony, J.L., 1978. Methodes de caracterisation hydrody-´ ´namique in situ d’un sol non sature. Application a deux types de sol du Senegal en vue de la´ ` ´ ´determination des termes du bilan hydrique. Ann. Agron. 29, 1–36.´

Vachaud, G., Vauclin, M., Colombani, J., 1981. Bilan hydrique dans le sud tunisien. 1.Caracterisation experimentale des transferts dans la zone non saturee. J. Hydrol. 49, 31–52.´ ´ ´

Ž .WRI World Resource Institute , 1990. 1990–91 World Resources. A Guide to the GlobalEnvironment. Oxford Univ. Press, New York.

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

Water dynamics and nutrient leaching through a cropped ferralsol in the loyal...Celine Duwig; Thierry Becquer; Iris Vogeler; Michel Vauclin; Brent E ClothierJournal of Environmental Quality; May/Jun 2000; 29, 3; ProQuest Science Journalspg. 1010

Reproduced with permission of the copyright owner. Further reproduction prohibited without permission.

A Simple Approach to Determine Reactive Solute TransportUsing Time Domain Reflectometry

Iris Vogeler,* Celine Duwig, Brent E. Clothier, and Steven R. Green

ABSTRACT of anion adsorption depends on the exchange capacityof the soil, which is generally determined from batchTime domain reflectometry (TDR) possesses potential for de-experiments. Limitations of batch techniques includetermining solute-transport parameters, such as dispersion coefficients

and retardation factors for reactive solutes. We developed a simple breakdown of soil aggregates, disturbance of flow path-method based on peak-to-peak measurements of water and solute ways, and the use of soil-to-solution ratios that are muchvelocities through the soil using TDR. The method was tested by smaller than in natural soil systems. This results in inap-carrying out unsaturated leaching experiments in the laboratory on propriate estimates of the exchange characteristicstwo soil columns packed with a South Pacific soil from Mare, which (Bond and Phillips, 1990). Alternatively, exchange char-is a ferrasol with variable surface charge. One column was left bare acteristics can be inferred from leaching experimentsand the other was planted with mustard. Pulses of CaBr2 and Ca(NO3)2 on undisturbed soil columns using fitting procedures towere applied to the surface of either wet or dry soil and then leached

the concentration of the effluent (Vogeler et al., 1997a).by water from a rainfall simulator applied at a steady rate of betweenHowever, such leaching experiments are laborious and30 and 45 mm h21. Water and solute transport were monitored bycannot be used for in situ measurements in the field.collecting the effluent. Contemporaneous in situ measurements of

the water content and electrical conductivity were made using TDR. The objective of this study was to develop a simpleTransport parameters for the convection–dispersion equation, with a method by which anion retardation of soils can be in-linear adsorption isotherm, were obtained from the flux concentration ferred from TDR measurements of water content andand the solute resident concentrations measured by TDR. Anion bulk soil electrical conductivity. We present some mea-retardations between 1.2 and 1.7, and dispersivities between 1 and 9 surements of bromide and nitrate movement through amm, were found. Retardations also were calculated using our simple ferrasol. The soil has a variable surface charge that ad-approach based on TDR-measured water and solute front velocities.

sorbs anions and therefore retards their movementThese used TDR measurements of soil water content and bulk soilthrough the soil. We studied anion transport through aelectrical conductivity with time, and were similar to those obtainedbare soil column and a column growing mustard. Thefrom the effluent. The agreement suggests TDR could be a valuablemustard was used to study the impact of the presencein situ technique for obtaining the parameters relating to reactive

solute transport through soil. of roots on the velocity on solute movement. Nitrateand bromide transport were measured by collecting theeffluent exiting at the base of the column and monitoringthe change in water content and electrical conductivityLack of adequate instrumentation limits in situ mea-as measured by TDR probes installed at various depthssurements of transport processes of water andwithin the soil column. These measurements are com-chemical movement in the unsaturated zone. Within thepared to the results generated from a numerical solutionlast decade TDR has become widely used for measuringof the convection–dispersion equation (CDE), in whichsoil water content. Now TDR is seen as a means bysoil water transport is predicted using Richards’s equa-which the changing concentration of electrolyte in thetion. Parameters describing chemical transport obtainedsoil solution can also be observed. The ability to takein situ from TDR-measured peak-to-peak velocities ofsuch measurements continuously and automatically, inwater and solutes fronts are compared with those ob-a nondestructive way, makes TDR a valuable tool fortained from the flux concentration in the effluent. Al-observing solute transport in situ. So far applicationthough the TDR technique was used to obtain the trans-of this technique to monitor solute transport has beenport parameters for repacked soil columns in thelimited to soils in which anions such as nitrate are consid-laboratory, the approach should also be suitable for inered to be nonreactive solutes (Kachanoski et al., 1992;situ measurements in the field.Vanclooster et al., 1993; Mallants et al., 1994). However,

many soils around the Pacific region, and elsewhere inTHEORYvolcanic regions, carry variable surface charge and are

known to adsorb anions. Therefore these soils need to The Transport Modelbe managed differently, because adsorption critically

One-dimensional transient water flow into a uniform unsat-controls the depth and pattern of leaching. The degreeurated soil can be described by Richards’s equation. Assumingthat root water uptake is negligible, this equation can be writ-ten asI. Vogeler, B.E. Clothier, and S.R. Green, Environment and Risk

Management Group, Hort Research Inst., Private Bag 11-030, Palmer-ston North, New Zealand; and C. Duwig, ORSTOM, Laboratoire ]u

]t5

]

]z 3Dw(u)]u

]z4 2dKw

du

]u

]z[1]

d’Agropedologie, B.P. A5, 98848 Noumea, Nouvelle Caledonie, andLaboratoire d’etude des Tranferts en Hydrologie et Environment,

where u is volumetric water content (m3 m23), Dw is soil waterB.P.53, 38041 Grenoble Cedex 9, France. Received 5 Aug. 1998. *Cor-responding author ([email protected]).

Abbreviations: CDE, convection–dispersion equation; PV, pore vol-umes; TDR, time domain reflectometry.Published in Soil Sci. Soc. Am. J. 64:12–18 (2000).

12

VOGELER ET AL.: TIME DOMAIN REFLECTOMETRY AND SOLUTE TRANSPORT 13

diffusivity (m2 s21), Kw is hydraulic conductivity (m s21), t is with the constants for the Dw(u) function g 5 3.847 3 1025

and b 5 12, sorptivity S 5 2.4 mm s21/2 for the transient case,time (s), and z is depth (m) (Kutılek and Nielsen, 1994).For the purpose of modeling the water flow we assume that and the saturated water content us 5 0.69 m3 m23.diffusivity can be described using an exponential function(Brutsaert, 1979), and that conductivity can be described usinga power law function (Quadri et al., 1994). The appropriate Time Domain Reflectometryinitial and boundary conditions for unsaturated flow into a

Here we describe a simple method to determine the soil’ssoil column under steady rainfall aresolute transport properties based on TDR measurements ofsoil water content and bulk soil electrical conductivity. For au 5 ui(z) t 5 0 0 # z # l,

qw 5 q0 z 5 0 t . 0 [2] Green–Ampt soil (that is, a soil possessing a Dirac-d diffusivityfunction Dw[u]), the invading water enters the soil as a rectan-

where ui is initial water content (m3 m23), qw is water flux gular wet front and rides atop the initial water content ui.density (m s21), l is column length (m), and q0 is the constant Thus, the wet front, zf (m), at any time is located atflux imposed at the surface (m s21).

The convection–dispersion equation (CDE) for one-dimen- zf 5I

(u0 2 ui)[8]

sional transport of reactive anions under transient conditionsis

where I is the cumulative depth of water infiltrated (m) andu0 is the final water content (m3 m23) (Clothier, 1998). The](uCr)

]t1

](qSA)]t

5]

]z 1uDs]Cr

]z 2 2](qw Cr)

]z[3] velocity of the wet front nf is therefore given by

where Cr is solute concentration in the resident soil solutionnf 5

]zf

]t5

i(u0 2 ui)

[9](mol m23), SA is the amount of solute adsorbed (mol kg21),and r is the bulk density of the soil (kg m23) (Kutılek and

where i is the infiltration rate (m h21). By monitoring u(t)uzNielsen, 1994). We assume that Ds, the diffusion dispersionwith TDR probes installed horizontally at sequential depthscoefficient (m2 s21), is given byz it should be possible to measure nf as the wet front passes

Ds 5 an 1 t Dm [4] the probes. Frequent measurements using an automated sys-tem are therefore required.where a is dispersivity [m], n is average pore water velocity of

We use a similar approach to analyze the solute movement.qw/u (m h21), t is the tortuosity factor, and Dm is the molecularAssuming the soil water is fully mobile, and that solute disper-diffusion coefficient in a free solution (10210 m2 s21).sion and diffusion can be ignored, the solute front, becauseFor the purpose of modeling solute transport we assumeof this complete invasion of the wetted pore space, will bethat the soil is initially free of the solute of interest. A solutelocated at a depth ofpulse with a concentration C0 was applied to the soil surface

over a very short time interval, 0 , t , ti. This was followedsf 5

Iu0

[10]by a continuous application of solute-free water at a steadywater flow. Thus, for the solute the appropriate initial andboundary conditions are (van Genuchten and Wierenga, 1986) The peak concentration will also be at the depth sf for a pulse

application of solute, even if dispersion and diffusion occur.Cr 5 0 0 # z # l t 5 0The solute velocity is given by

2uDs]Cr

]z1 qwCr 5 q0C0 z 5 0; 0 , t , ti

ns 5]sf

]t5

iu0

[11]2uDs

]Cr

]z1 qwCr 5 0 z 5 0; t . ti [5]

Using the TDR to monitor changes in the bulk soil electricalFor the lower boundary condition it was assumed that the conductivity should also allow measurement of ns. It followssoil column was part of an effectively semi-infinite system, as thatsuggested by van Genuchten and Wierenga (1986, p. 1034).The adsorption of anions by the soil was modeled using a ns 5 nf

(u0 2 ui)u0

[12]simple linear isotherm of the form

SA 5 KD Cr [6] Both the wet front velocity and solute front velocity canbe inferred simply from peak-to-peak measurements of u(t)uzwith distribution coefficient KD (L kg21) taken to be constant.and s(t)uz as measured by TDR. If the measured solute veloc-For a linear isotherm, any anion adsorption retards the soluteity, n*s , is smaller than the ns calculated using Eq. [12], thenfront by the factor R defined asanion adsorption must have occurred. This anion adsorptionmust be related to a change in anion adsorption capacity with

R 5 1 1qKD

u[7] change in soil solution concentration, as TDR would not detect

solute retardation due to anion exchange. For simplicity weassume that this change in anion adsorption capacity with soilAs part of our modeling procedures, Eq. [1] through [7]

were solved numerically using a fully implicit Newton– solution concentration is linear, and thus Eq. [7] applies. Theretardation will simply be given by ns/n*s . For an excludedRaphson iteration for the water flow equation and a time-

centered Crank–Nicholson scheme for the solute flow (Green, anion, the measured solute velocity will likewise be greaterthan the predicted one. This analysis requires that there be1997). Soil hydraulic properties were determined from one-

dimensional, free-water adsorption experiments, using hori- complete invasion of the wetted pore space by the invadingsolute. Should this not be the case, then the denominator inzontal sectionable columns (Duwig, 1998). The value found

for the saturated conductivity was Ks 5 1.56 3 1022 mm s21, Eq. [10]–[12] will be the mobile water fraction um. The tech-

14 SOIL SCI. SOC. AM. J., VOL. 64, JANUARY–FEBRUARY 2000

nique will then also require a sample of the resident concentra- Time Domain Reflectometry Watertion so that um can be determined (Clothier et al., 1992). Content Calibration

This simple approach of obtaining the retardation and anionThe determination of the water content from the TDR-exchange parameters was tested under controlled conditions

measured dielectric constant was based on a third-order poly-on repacked soils in the laboratory, which are known to havenomial equation fitted to calibration measurements carriedno immobile water. The approach will be compared to resultsout with the same soil material (Duwig, 1998). The equationobtained from breakthrough curves.is given as

u 5 20.17 1 6.3 3 1022 ε 2 1.7MATERIALS AND METHODS3 1023 ε2 1 1.6 3 1025 ε3 [17]

Use of Time Domain ReflectometryThis equation deviates by up to 0.2 m3 m23 from the curve

The TDR technique for measuring soil water content (u) suggested by Topp et al. (1980) for mineral soils. The deviationand solute resident concentration (Cr) is based on the measure- is probably due to the combination of high organic matterment of the soil’s dielectric constant (e) and bulk soil electrical content and the low bulk density of the soil (Jacobsen andconductivity (s). The dielectric constant is calculated from Schønning, 1993).Topp et al. (1980) as

Laboratory Soil Column Transport Experimentsε 5 3ct2lt

42

5 3 la

lt mp4

2

[13]Oven-dried soil was sieved and packed in columns to the

field bulk density of 0.8 Mg m23. Two soil columns were used,where c is the propagation velocity of an electromagnetic waveone with a diameter of 300 mm and length of 280 mm, thein free space (3 3 108 m s21), t is travel time (s), lt is the realother with a diameter of 300 mm and length of 300 mm. Onelength of the transmission line (m), la is the apparent lengthhad a bare soil surface and the other had mustard growing on(m) as measured by a cable tester, and np is the relative velocityit. The columns were placed on inverted tension infiltrometerssetting of the instrument. (Magesan et al., 1995) to ensure unsaturated flow at the base,Following the thin-sample theory of Giese and Tiemann yet also allow regular sampling of the effluent. Time domain(1975), the electrical conductivity of the soil s can be described reflectometry probes were installed at depths of 30, 130, andby Topp et al. (1988) as 230 mm into the bare soil column and at 50, 150 and 250 mminto the mustard column. Three-wire TDR probes, 150 mm

s 51

120plt

Z0

Zu12V0

Vf

2 12 [14] long, with a wire diameter of 2 mm, and a spacing of 12.5mm, were used. The probes were connected via a multiplexer(similar in design to that of Heimovaara and Bouten [1990]),where Z0 is the characteristic impedance of the probe (V), Zuto the Tektronix cable tester (1502C, Tektronix, Beaverton,is the characteristic impedance of the TDR system (50V), V0OR). A laptop computer controled the settings of the TDRthe voltage of the incident step, and Vf the final reflectedand also recorded and analyzed the waveforms using softwarevoltage. The probe impedance Z0 was calculated using Toppdeveloped in the laboratory, based on curve-fitting algorithmset al. (1988):described by Baker and Allmaras (1990). Measurements ofboth u and s were taken every 5 min at the beginning and every

Z0 5 60 ln12sd 2 [15] 30 min after the fourth hour. A rainfall simulator (Vogeler et

al., 1997b) was used to apply the water at a steady rate toeach column. The experimental setup is shown in Fig. 1.where s is the rod spacing (m) and d the rod diameter (m).

We assume the soil to be initially solute free with an initialbulk soil electrical conductivity of si. If a steady state water Mustard Experimentflux, i, has already been established, and a pulse with a total

In the mustard column, the roots had invaded the entiremass of M (mol m22) is applied to the soil surface, thencolumn length after 3 wk. To study the effect of the initialwater content at the soil surface prior to solute application,M 5 i#

∞

0Cr (t)dt 5 i#

∞

0Cf (t)dt

the mustard was used to dry the soil down to a water content of5 ia #

∞

0(s(t) 2 si)dt [16] 0.25 m3 m23. A pulse of CaBr2, equivalent to a nitrate–nitrogen

application of 100 kg N ha21, was then sprayed onto the drysoil surface. The column was subsequently leached with fourwhere Cf is the flux concentration as measured in the effluentpore volumes (PV) of distilled water applied via the rainfall(mol m23), Cr is the resident concentration (mol m23) as mea-simulator. To mimic tropical rainfall intensities, a water fluxsured by TDR at a depth z, and a is an empirical constantdensity of initially 50 mm h21 was used. Because ponding onthat provides an integrally correct interpretation of the con-the surface occurred, the intensity was decreased to 46 mmductivity measurements.h21. After the 4 PV a concentrated pulse of Ca(NO3)2 at thesame concentration was sprayed onto the soil surface whileSoil Materialthe steady-state rainfall was maintained.

The soil material used was a Geric Ferrasol from Mare(Loyalty Islands, New Caledonia), derived from volcanic

Bare Soil Experimentejecta and ash. Details of chemical and mineralogical proper-ties of the soil are given in Duwig (1998) and Duwig et al. The bare soil column was first leached with distilled water(1998). Only material from the upper 20 cm of a cultivated using the rainfall simulator. Initially, the water was applied atarea was used. The soil is variably charged, relatively rich in a water flux density of 40 mm h21. Because ponding on theorganic matter (about 13%), and primarily composed of Al bare soil column occurred at this rate, the rate was dropped

to 36 mm h21. When steady state flow was reached, a bromideand Fe oxides.


Fig. 1. Experimental setup with sprinkler reservoir (S), pressure head regulator (P), and water reservoir (R). The pressure potential h1 (5 h2)controls the water head (H) in the sprinkler reservoir and the pressure potential h0 controls the pressure potential at the base of the soil column.

pulse, again an anion equivalent to 100 kg N ha21, was sprayed Time Domain Reflectometry Measurementson the soil surface and leached with 4 PV of water. This was of Water Contentfollowed by a nitrate pulse, again leached under the same

Figure 3a shows the measured water contents at vari-steady-state water flow. The effluent samples from both exper-ous depths with time, as measured by TDR during theiments were analyzed for NO2

3 and Br2.infiltration of water into the bare soil column. Figure 3bshows the same, but following an application of bromide

RESULTS AND DISCUSSION onto mustard. The early part of the u(t) signal of the

Effluent ConcentrationsThe flux concentrations of bromide and nitrate mea-

sured in the effluent from the bare soil and the mustardcolumn are shown in Fig. 2 as a function of cumulativeinfiltration Q (mm). Note that the bromide pulse in themustard column was applied to a dry soil surface. Alsoshown are fitted numerical solutions of the water andsolute flow equations (Eq. [1] through [7]). Dispersivi-ties ranging from 3 to 9 mm were found (Table 1), givingKD values ranging from 0.11 to 0.4 L kg21. This impliesR values of 1.2 to 1.5 (Eq. [7]), which are within the1.4–1.7 range of R values found by Katou et al. (1996)

Fig. 2. Measured and predicted (using the CDE model) breakthroughfor their andisol from Japan. The effect of plant roots curves of bromide (d and solid line) and nitrate (h and brokenand the initial water content of the soil surface on anion line) for (a) bare soil column and (b) column with mustard. Cf

denotes the flux concentration.movement and retardation seems negligible.


Table 1. Column data and model parameters obtained from the convection–dispersion equation†. The hyphen indicates a missing valueand the semicolon separates replicates.

a (mm) Rnf n*s

BTC- TDR- TDR- TDR- BTC- MRqw u fitted fitted peak-peak fitted fitted n*r nf n*s ns TDR

mm h21 m3 m23 mm h21 %Bare soil: bromide pulse on wet soil, steady-state water flow36.4 0.663 3 2 1.4 1.3 1.3 91 50 38 55 105

1 1.4 79 66 38 104Bare soil: nitrate pulse on wet soil, steady-state water flow29.7 0.663 4 3 1.4 1.2 1.2 32 45 116

1 1.3 1.3 34 131Mustard: bromide pulse on dry soil, transient water flow43.6 0.625 4 – 1.9 – 1.5 181 115 55 105 172

194 125 –Mustard: nitrate pulse on wet soil, steady-state water flow44.5 0.625 9 1 1.4 1.7 1.4 52 71 94

2 2.4 1.6 30 70

† qw 5 water flux density, u 5 volumetric water content, a 5 dispersivity, R 5 retardation factor, nr 5 water front velocity, ns 5 solute front velocity,MR 5 mass recovery, BTC 5 breakthrough curve, TDR 5 time domain reflectometry.

deepest probe was eliminated due to water falling out- TDR measurements of the water content in Fig. 3b,where there is a slow rise in u to the final water content.side the core onto the external connector of the TDR

probe. Water front velocity through the soil could thus A certain degree of hydrophobicity, which seems to bewidespread in this soil, might be the overall cause. Thebe calculated from the wet front arrival times at the

various depths. For the bare soil we determined veloci- wetting front is thus not moving as a rectangle as pre-scribed by the Green and Ampt assumption, but ratherties n*f of 91 and 79 mm h21, while under mustard veloci-

ties were 181 and 194 mm h21 (Table 1). is affected by the soil structure and texture and plantleaves and roots. This, however, did not affect soluteFrom Eq. [9] we calculated wet front velocities (nf)

for the bare soil of 65 and 64 mm h21, based on an initial transport as the solute front lags behind the water frontdue to the initial water content. Furthermore, pondingflow rate i of 40 mm h21. For the mustard column we

calculated water front velocities nf of 128 and 135 mm of water occurred only at the beginning of the experi-ment, before the solute was applied.h21, based on an initial i of 50 mm h21. The measured

and calculated velocities are quite different. This mightTime Domain Reflectometrybe due to preferential water flow or nonuniform wetting

and Solute Transportof the soil, which could have occurred because of waterponding on the soil surface, or a nonuniform water con- The TDR-measured electrical conductivity of thetent in the horizontal plane (caused by nonuniform wa- bulk soil (s) following the various pulses of bromide

and nitrate is shown in Fig. 4a and 4b. From these peak-ter uptake). This nonuniform wetting can be seen in the

Fig. 3. Time domain reflectometry–measured water content for (a) bare soil column and (b) column with mustard.


Fig. 4. Time domain reflectometry–measured bulk soil electrical con- Fig. 5. Bromide (d) and nitrate (h) concentrations, obtained fromductivity for (a) bare soil column and (b) column with mustard, time domain reflectometry–measured bulk soil electrical conduc-following a pulse of bromide (d) and nitrate (h). Also shown are tivities for (a) bare soil column and (b) a column with mustard.the times for one pore volume for each probe and for bromide Also shown are the predictions using the convection–dispersion(solid lines) and nitrate (broken lines). equation for the upper two TDR probes for bromide (solid line)

and nitrate (broken line).

to-peak velocities (n*s ) were calculated (Table 1). Forsoil does not seem to behave like an ideal Green-Amptthe transient case only the peaks of the upper two probessoil. It is worthwhile to explore the impact that nonidealcould be used, because the lower TDR probe gave inex-behavior might have on this technique, because in theplicable measurements of s. This confirms the problemfield such simplicity is unlikely to be encountered.caused by local heterogenities around TDR probes.

Predicted solute velocities (ns) using Eq. [12] were inall cases higher compared to n*s . Because we know that Time Domain Reflectometry and Modelingall the soil water is mobile, this disparity suggests an of Solute Transportincrease in anion adsorption capacity with increasing

For modeling solute transport from TDR measure-soil solution concentration that effectively retards thements, the measured bulk soil electrical conductivitiesdownward movement of bromide and nitrate. R values(s) of the upper probe were converted into concentra-ranged from 1.3 to 2.4. The highest R value from thetions using Eq. [16]. The values of a found for eachpeak-to-peak measurements following the nitrate pulsepulse application were then used to convert measured son the mustard column is probably due to a misinterpre-values of the other two TDR probes into concentrations.tation of the unusual third peak. A slight increase in

Mass recoveries for steady-state water flow cases cal-water content measured by the lower two TDR probesculated from these concentrations ranged from 94 tomight also have caused a delayed increase in s. Apart116% for the middle TDR probes and from 70 to 130%from the transient flow case, all other R values arefor the lower TDR probes. The mass recoveries for thesimilar to those obtained from the flux concentrationeffluent ranged between 97 and 107%. Reasons for thein the effluent. Implicit in the use of this TDR approachpoorer recoveries for the lower TDR probes, as well asfor obtaining retardation factors during invasion of wa-the secondary peaks observed in the measurements ofter into a dry soil is the assumption that the soil behavess for the lower TDR probes, are not clear. For thelike a Green-Ampt soil with a rectangular wetting pro-transient flow case, again only the upper two TDRfile of complete invasion. However, as shown by the

u( t)|z measurements (Fig. 3) and discussed above, this probes were used, but a recovery of 172% for the second


REFERENCESprobe was obtained. This is probably again due to non-uniform wetting. If the water content still changes while Baker, J.M., and R.R. Allmaras. 1990. System for automating and

multiplexing soil moisture measurement by time-domain reflec-the solute front passes the TDR probe, the assumptiontometry. Soil Sci. Soc. Am. J. 54:1–6.of a unique value of a (Eq. [16]) does not hold. Thus,

Bond, W.J., and I.R. Phillips. 1990. Cation exchange isotherms ob-for modeling purposes, only the two upper probes were tained with batch and miscible displacement techniques. Soil Sci.used and the measurements for the transient flow case Soc. Am. J. 54:722–728.

Brutsaert, W. 1979. Universal constants for scaling the exponentialwere not used at all.soil water diffusivity. Water Resour. Res. 15:4812483.The converted TDR-measurements are shown in Fig.

Clothier, B.E., M.B. Kirkham, and J.E. McLean. 1992. In situ measure-5a and 5b. Also shown for the steady-state flow cases ment of the effective transport volume for solute movementare fitted numerical solutions for Eq. 1 through 7. Dis- through soil. Soil Sci. Soc. Am. J. 56:733–736.

Clothier, B.E. 2000. Infiltration. In K.A. Smith and C.E. Mullins (ed.)persivities range from 1 to 3 mm and are slightly lowerSoil analysis. Marcel Dekker, New York (in press).than those obtained from the effluent. Similar observa-

Duwig, C. 1998. Etude des transferts d’eau et de nitrate dans les solstions were made by Hart and Lowery (1998), when TDR ferrallitiques de Mare (Nouvelle-Caledonie): risques de pollutionwas used to measure transport of a bromide tracer. des lentilles d’eau douce. Ph.D. diss. Universite Joseph Fourier,

Grenoble, France.They found that their simulated dispersion, using theDuwig, C., T. Becquer, B.E. Clothier, and M. Vauclin. 1998. NitrateLEACHM model, was greater than that measured by

leaching through oxisols of the Loyalty Islands (New Caledonia)TDR. They argued that the low electrolyte concentra- under intensified agricultural practices. Geoderma 84:29–43.tion in combination with the low moisture content of Giese, K., and R. Tiemann. 1975. Determination of the complex per-

mittivity from thin sample time domain reflectometry. Improvedtheir sandy soil might have biased the TDR measure-analysis of the step response waveform. Adv. Mol. Relax. Pro-ments. However, in our study, electrolyte concentra-cesses 7:45–59.tions were high enough for detection by TDR, as Green, S.R. 1997. THETA-1D: A user’s manual for the THETA-1D

evidenced by the relatively good mass recoveries. Retar- model to compute the 1-D measurement of water and chemicalinto uniform soil. p. 43. In HortResearch Internal Report IR97-dation factors between 1.2 and 1.7 were found for the66. HortResearch, Palmerston North, New Zealand.steady-state flow cases. These are again similar to those

Hart, G.L., and B. Lowery. 1998. Measuring instantaneous solute fluxobtained from the effluent, indicating that TDR is a and loading with time domain reflectometry. Soil Sci. Soc. Am.promising tool for obtaining solute transport parameters J. 62:23–35.

Heimovaara, T.J., and W. Bouten. 1990. A computer-controlled 36-in situ.channel time domain reflectometry system for monitoring soil wa-ter contents. Water Resour. Res. 26:2311–2316.

CONCLUSIONS Jacobsen, O.H., and P. Schjønning. 1993. A laboratory calibration oftime domain reflectometry for soil water measurements includingOur study has demonstrated that TDR is a useful tool effects of bulk density and texture. J. Hydr. 151:147–157.

for monitoring the transport of not only inert solutes, Kachanoski, R.G., E. Pringle, and A. Ward. 1992. Field measurementof solute travel times using time domain reflectometry. Soil Sci.but also reactive solutes provided that the soil inhibitsSoc. Am. J. 56:47–52.an increase in anion adsorption capacity with increasing

Katou, H., B.E. Clothier, and S.R. Green. 1996. Anion transportsoil solution concentration. A simple theory was used involving competitive adsorption during transient water flow in anto obtain retardation factors from peak-to-peak mea- Andisol. Soil Sci. Soc. Am. J. 60:1368–1375.surements of the bulk soil electrical conductivity ob- Kutılek, M., and D.R. Nielsen. 1994. Soil hydrology. Catena Verlag,

Cremlingen-Destedt, Germany.tained by TDR. This method was not quite accurate forMallants, D., M. Vanclooster, M. Meddahi, and J. Feyen. 1994. Esti-estimating water front velocities because of nonuniform

mating solute transport in undisturbed soil columns using timewetting of the soil. However, it did not affect the TDR- domain reflectometry. J. Cont. Hydr. 17:91–109.estimated solute transport parameters, which were com- Magesan, G.N., I. Vogeler, D.R. Scotter, and B.E. Clothier. 1995.

Solute movement through two unsaturated soils. Aust. J. Soilparable to those obtained from breakthrough curvesRes. 33:585–596.and by directly fitting the TDR measurements to the

Quadri, M.B., B.E. Clothier, R. Angulu-Jaramillo, M. Vauclin, andCDE. This means that the proposed TDR method S.R. Green. 1994. Axisymmetric transport of water and soluteshould be a simple and nondestructive way for determin- underneath a disk permeameter: Experiments and numerical

model. Soil Sci. Soc. Am. J. 58:696–703.ing solute transport parameters of not only conservative,Topp, G.C., J.L. Davis, and A.P. Annan. 1980. Electromagnetic deter-but also reactive chemicals such as nitrate in variably

mination of soil water content: Measurement in coaxial transmis-charged soils. sion lines. Water Resour. Res. 16:574–582.The TDR technique for obtaining retardation factors Topp, G.C., M. Yanuka, W.D. Zebchuk, and S. Zeglin. 1988. Determi-

nation of electrical conductivity using time domain reflectometry:was demonstrated under controlled conditions in theSoil and water experiments in coaxial lines. Water Resour. Res.laboratory and should now be tested on undisturbed24:9452952.soil columns. If immobile water fractions need to be Vanclooster, M., D. Mallants, J. Diels, and J. Feyen. 1993. Determining

taken into account, as is common in the field, then it local-scale solute transport parameters using time domain reflec-tometry (TDR). J. Hydr. 148:93–107.may also be necessary to take an independent measure

van Genuchten, M. Th., and P.J. Wierenga. 1986. Solute dispersionof the resident concentration.coefficients and retardation factors. In A. Klute (ed.) Methods ofsoil analysis. Part I: Physical and mineralogical methods. 2nd ed.

ACKNOWLEDGMENTS Agron. No. 9. ASA, Madison, WI.Vogeler, I., D.R. Scotter, B.E. Clothier, and R.W. Tillman. 1997a.This research was supported by ORSTOM (French Re- Cation transport during unsaturated flow through two intact soils.

search Institute for Development in Cooperation) and the Eur. J. Soil Sci. 48:401–410.New Zealand Foundation for Research, Science, and Technol- Vogeler, I., D.R. Scotter, S.R. Green, and B.E. Clothier. 1997b. Soluteogy program entitled “Soil Processes and Environmental Pro- movement through undisturbed soil columns under pasture during

unsaturated flow. Aust. J. Soil Res. 35:1153–1163.tection.”

Ž .Geoderma 103 2001 291–306www.elsevier.comrlocatergeoderma

Mineralogical, chemical and charge properties ofGeric Ferralsols from New Caledonia

Thierry Becquera,b,), Jean Petardc, Celine Duwigb,´ ´Emmanuel Bourdonb, Roland Moreauc, Adrien Jules Herbillond

a IRD, cro Centre de Pedologie Biologique 17, rue Notre-Dame des PauÕres,´UPR 6831 du CNRS, Associee a l’UniÕersite Henri Poincare, Nancy I, BP 5,´ ` ´ ´

54501 VandoeuÕre les Nancy, France`b IRD, Laboratoire d’Agropedologie, BP A5, 98848 Noumea cedex, New Caledonia´ ´

c IRD, Laboratoire d’Etude du Comportement des Sols CultiÕes, BP 5045,´34032 Montpellier cedex 1, France

d Unite des Sciences du Sol, UniÕersite Catholique de LouÕain, Place Croix du Sud 2r10,´ ´B1348, LouÕain la NeuÕe, Belgium

Received 29 September 2000; received in revised form 15 January 2001;accepted 8 February 2001

Abstract

The mineralogical, chemical and surface charge properties of Geric Ferralsols of NewCaledonia were examined. These soils, which corresponded to two soil mantles formed either onultramafic rocks or volcanic ejecta and ashes, were respectively dominated by iron and aluminiumoxides. The electric charge characteristics were studied by measuring retention of Ca2q and Cly

Ž .at different pH values ranging from 3 to 7. The cation exchange capacity CEC increased withT

soil organic carbon and pH and varied from 0 to 35 cmol kgy1 soil. The anion exchange capacitycŽ . y1AEC reached 4.25 cmol kg soil in Bo horizons at pH 4. The magnitude of the CEC andc T

AEC variations was modelled according to CECs10a1pH)10B1 and AECsy10a2pH

)10B2.TŽParametersa and a were low for samples with high organic carbon from 0.10 to 0.19 and from1 2

) Corresponding author. IRD, cro Centre de Pedologie Biologique 17, rue Notre-Dame des´Pauvres, UPR 6831 du CNRS, Associee a l’Universite Henri Poincare, Nancy I, BP 5, 54501´ ` ´ ´Vandoeuvre les Nancy, France. Tel.:q33-3-83-51-84-56; fax:q33-3-83-57-65-23.`

Ž .E-mail address: [email protected] T. Becquer .

0016-7061r01r$ - see front matterq2001 Elsevier Science B.V. All rights reserved.Ž .PII: S0016-7061 01 00045-3

( )T. Becquer et al.rGeoderma 103 2001 291–306292

.y0.44 toy0.66, respectively and could increase to 1.25 andy0.39, respectively, when organiccarbon content decreased. The parametersa and a could be also useful to regroup soil types1 2

according to the nature of the properties that can be manipulated for their management.q2001Elsevier Science B.V. All rights reserved.

Keywords: Iron and aluminium oxides; Organic matter; pH-dependent charge; Ion adsorption;Cation exchange capacity; Anion exchange capacity

1. Introduction

ŽHighly weathered Ferralsols occur extensively in New Caledonia Latham et.al., 1978 . Two types of soil very rich in iron and aluminium oxides are present.

The first type is derived from ultramafic rocks and represents a third of theŽ .surface of the main island Grande Terre . The second one, derived from

volcanic ejecta and ashes, covers 19% of the surface area of the Loyalty Islands.These soils are generally deficient in most plant nutrients. In the Grande

Terre, the lack of land and the increasing demand for food and vegetables leadto the use of these rather infertile soils. They had never been cultivated beforethe beginning of the 1980s, when vegetable and fruit started to be produced. Inthe Loyalty Islands, Ferralsols are the main cultivated soils because they are thedeepest. The agricultural production was mainly traditional, with long fallowperiods that maintained the soil fertility. However, the development of acash-cropping sector since a few years is leading to a breakdown in thetraditional strategies of soil fertility maintenance.

Appropriate management of these soils requires detailed information on theirproperties. Little is known about their mineralogical and physico-chemical

Žcharacteristics. A few mineralogical studies Tercinier, 1971; Schwertmann and.Latham, 1986 indicates that iron and aluminium oxides and oxyhydroxides are

their main constituents, while silicates are present in small amounts. A pH-de-Ž . Ž .pendent cation exchange capacity CEC and anion exchange capacity AEC

result from the presence of Fe and Al oxides and organic matter. Variablecharges in these soils should have a major incidence on the bioavailability and

Ž .movement of nutrients as well as toxic metals Ni, Cr . . . . Phosphorus is highlyŽ .sorbed on the two types of soils Dubus et al., 1998 and nitrate sorption also

Ž .occurs on the soils of the Loyalty Islands Duwig et al., 1999 . The high levelsŽof bioavailable Ni found in some soils of the Grande Terre Becquer et al.,

. Ž .1995 can reach toxic levels for crops L’Huillier and Edighoffer, 1996 .Changes in management practices can affect environmental factors, such as

pH or organic matter content, and the sorption characteristics of the soils. TheŽ .aims of the study were therefore: i to characterise the mineralogy and the

Ž .chemistry of these soils, and ii to study the influence of organic matter, Fe andAl oxides, and pH on their charge characteristics.

( )T. Becquer et al.rGeoderma 103 2001 291–306 293


2.1. Study area and soils

Two sets of soil profiles corresponding to the two types of oxide-rich soils ofŽ .New Caledonia Fig. 1 were studied.

Ž X .The first set, located in Ouenarou E 166844 –S 2288’ in the south of the´main island, came from a highly weathered mantle derived from peridotite.Three different profiles under natural vegetation situated along a soil topose-

Ž . Ž .quence were selected Table 1 . They correspond to a piedmont soil OUE 1 , aŽ .colluvio-alluvial soil OUE 3 and an alluvio-colluvial soil with temporary

Ž .reducing conditions OUE 4 , respectively. OUE 1 was a very compact soil witha loamy sand texture and a poorly developed fine sub-angular polyhedralstructure. Profiles OUE 3 and OUE 4 were constituted by colluvio-alluvialmaterials with variable texture occurring in successive horizons of irregularthickness. OUE 4 contained a lot of ferro-manganese coatings suggesting theexistence of reducing conditions at the base of the toposequence. These soilswere poorly structured except in the topsoil where roots and organic matter werequite abundant.

ŽFig. 1. Map of the two types of highly weathered Ferralsols in New Caledonia in dark grey on.peridotite and in light grey on uplifted coral atolls and location of the study areas.


Table 1Soil profile descriptions

Soil Horizon Depth Color Texture Structure Cohesion Roots BoundaryŽ . Ž .profile cm moist

OUE 1OUE 1-1 A 0–5 10 R 3r3 LAS sgsrfcr 1 1 sOUE 1-2 ABo 5–20 2.5 YR 3r4 LAS fsaprfap 3 3 sOUE 1-3 Bo1 20–35 10 R 3r2 LS fap 4 4 dOUE 1-4 Bo2 35–65 10 R 3r2 LAS fap 5 5 dOUE 1-5 Bo3 65–100 10 R 3r4 LSa fap 4 5

OUE 3OUE 3-1 Ah 0–4 5 YR 3r4 La fcrrfap 1 1 sOUE 3-2 ABo 4–15 2.5 YR 4r4 La fsaprfcr 2 1 sOUE 3-3 Bo 15–30 2.5 YR 3r6 LSa m 3 3 sOUE 3-4 2Bo1 30–55 5 YR 3r2 Lsa m 3 3 sOUE 3-5 2Bo2 55–80 2.5 YR 4r6 Lsa m 1 4 sOUE 3-6 3Bo 80–100 2.5 YR 3r2 LSa m 3 4

OUE 4OUE 4-1 Ah 0–4 2.5 YR 3r4 La fcrrfap 1 1 sOUE 4-2 A 4–12 2.5 YR 4r6 A fsaprfcr 2 1 sOUE 4-3 2Bog 12–42 2.5 YR 3r6 Sa sgs 3 3 sOUE 4-4 3Bog 42–70 2.5 YR 4r6 S1 sgs 1 3 sOUE 4-5 4Boc 70–90 10 YR 3r2 S sgs 1 5

MAR 3MAR 3-1 Ap 0–12 7.5 YR 3r2 L fcrrsgs 1 1 sMAR 3-2 Bo1 12–35 5 YR 3r3 L mic 1 2 dMAR 3-3 Bo2 35–70 5 YR 3r4 L mic 3 3 s

R

MAR 5MAR 5-1 Ap 0–12 7.5 YR 3r4 L fcrrsgs 1 1 sMAR 5-2 Bo1 12–25 5 YR 3r4 L mic 1 2 dMAR 5-3 Bo2 25–40 5 YR 3r4 L mic 3 4 dMAR 5-4 Bo3 40–60 5 YR 3r4 L mic 1 3 s

R

Ž .Horizon: according to FAO 1994 .Color: according to the Munsell color chart.

Ž . Ž . Ž .Texture: S sssandy; L lssilty; A a sclayed.Structure: fcrs fine crumb; fsaps fine subangular polyhedral; faps fine angular polyhedral;msmassive; micsmicroaggregated; sgsssingle-grain structure.Cohesion: 1s very loose; 2s loose; 3squite compact; 4scompact; 5s very compact.Roots: 1sabundant; 2squite abundant medium and fine; 3spresent; 4s few; 5s very few.Boundary: dsdiffuse; sssharp.

The second set of profiles was located in the Loyalty Islands. These islandsare uplifted coral atolls built upon an underlying volcanic structure. The soils

Ž .derive probably from volcanic ejecta and ashes Tercinier, 1971 deposited on


the coral rocks. The soil studied belonged to the deepest type, with depthsranging from 0.3 to 1 m. Two profiles located on the island of Mare, at´

Ž X X. Ž . Ž X X.Tawaınedre E 16884 –S 21830 MAR 3 and Taode E 167855 –S 21827¨ ` ´Ž .MAR 5 , were selected. The first site was covered by a regularly burnt bushfallow with 1–4 m shrubs; the second one was an older fallow with re-growingforest of 10–15 m. These profiles were located on the level surface of the

Žformer lagoon. They were characterised by little horizon differentiation Table.1 . The surface horizons showed a fine crumb to sub-angular polyhedral

structure related to their high organic matter content, whereas the deepest oneswere micro-aggregated. Root density decreased sharply with depth and withdecreasing carbon content and increasing soil cohesion.

Two samples were selected and analysed in each profile. The first oneŽ .corresponded to a surface or sub-surface horizon A generally rich in organic

matter and the second one was a Bo horizon poorer in organic matter.

2.2. Soil analysis

The soil samples were dried at 408C and sieved at 2 mm. Particle size wasŽ . Ž .determined by wet sieving 2000–50mm and pipetting -50 mm following

treatment with 35% H O and dispersion through overnight shaking with2 2Ž .Na-resin Bartoli et al., 1991 . The pH was measured in water and in 1 M KCl

Ž .suspensions using 1:2.5 wrv soil:solution ratio. Total organic carbon andnitrogen were analysed by dry combustion in a Carlo Erba 1108 chromatograph.

Ž . Ž .Phosphate retention P was determined according to Blakemore et al. 1987 .retŽ . Ž .Effective cation exchange capacity ECEC , exchangeable bases EB and

Ž .exchangeable acidity EA were assessed according to the methods described byŽ .Rouiller et al. 1994 . Total elements were measured by atomic absorption

Ž . Ž .spectrometry AAS Varian AA300 after a nitro-perchloric acid attack.Selective dissolutions of aluminium and iron compounds were carried out

using two chemical reactants. The oxalate extraction was performed by shaking2 g of soil in 80 ml of a 0.2 M ammonium oxalate buffered at pH 3 solution in

Ž .the dark at 208C for 4 h McKeague and Day, 1966 . The dithionite–citrate–bi-Ž . Ž .carbonate DCB extraction is a modification of the Holmgren’s 1967 method.

Two sets of extractions were performed, the first one with only citrate–bi-Ž . Ž . Žcarbonate CB , the second one with the addition of dithionite CBD Jeanroy. Ž .et al., 1991 . Centrifuge tubes containing 75 mg of soil with or without 250 mg

Ž .of dithionite and 25 ml of citrate–bicarbonate solution CB were placed in anend-over-end shaker at 258C for 5 days. Then, the samples were centrifuged at7300=g for 10 min and the supernatant analysed for Fe, Al and Si by ICP. TheFe, Al and Si extracted by oxalate were referred to as Fe , Al , and Si ando o o

those extracted by DCB as Fe , Al , and Si . The aluminium substitution ratiod d d

in iron oxides was calculated from the composition of the DCB minus CBŽ .extracts according to Jeanroy et al. 1991 .


Ž . Ž . Ž .X-ray diffraction XRD analyses were performed i on total powder, ii onsiltqsand and clay fractions after deferrification. The clay samples were alsoX-rayed after solvation with ethylene glycol or heating at 5508C.

The AEC and the CEC were determined according to the method describedŽ .by Gillman and Sumpter 1986 . The soil was first saturated with 0.1 M CaCl2

and washed with 0.002 M CaCl to remove the excess of salt. Then, 0.002 M2

CaCl solutions with pH ranging from 3 to 7 were added to the soil in a 1:52Ž .wrv soil:solution ratio and the suspension was gently shaken in an horizontalroller–shaker. After a week of shaking at 258C, the pH was recorded and thesuspension centrifuged at 2000=g. The supernatant solution was retained forCa2q, Al 3q and Cly analysis and the tubes were weighed to estimate thevolume of entrained solution. Ca2q, Al 3q and Cly were extracted with a 1 MNH NO solution and analysed after centrifugation. Ca2q and Al3q were4 3

measured by ICP and Cly was determined by titrimetry. The basic exchangeŽ . 2qcapacity CEC referred only to the amount of Ca adsorbed, the totalB

Ž . 2q 3qexchange capacity CEC to the sum of Ca and Al adsorbed and the anionTŽ . yexchange capacity AEC to Cl adsorbed.

The relationship between ion adsorption and pH were fitted according to theŽ . Ž .models of Wada and Okamura 1980 and Okamura and Wada 1983 :

log CEC sa pHqb logCqcT 1 1 1

log yAEC sa pHqb logCqcŽ . 2 2 2

where pH is the pH of the soil suspension,C is the concentration of theelectrolyte anda , a , b , b , c , c are constants adjusted for each soil.1 2 1 2 1 2

Ž y1.However, in our experiments, only one concentrationCs0.002 mol lwas used and the term of the model related to the concentration was integratedinto a constant termB, where Bsb logCqc. The equations became:

log CEC sa pHqB or, CEC s10a1pH)10B1 1Ž .T 1 1 T

log yAEC sa pHqB or, AECsy10a2pH)10B2 2Ž . Ž .2 2

3. Results and discussion

3.1. Chemical properties

The main chemical properties of the soils are shown in Table 2.The soils from Ouenarou were acidic with the lowest pH in the topsoils. The´

Ž .DpH pH –pH were close to zero or slightly positive in the upperKCl H O2

horizons but they could be over one unit in the deep ones. PositiveDpHŽ .indicates that positive charges prevail in these horizons Parfitt, 1980 . The

organic carbon content decreased sharply with depth. The CrN ratios were close

()

T.B

ecqueret

al.rG

eoderma

1032001

291–

306297

Table 2Main physical and chemical properties of the soils

Horizon Depth of Particle size distribution pH DpH Organic matter P EB EA ECECretŽ .sampling % 2q 2q q qCL FS CS S H O KCl C N Ca Mg K Na2Ž .cm y1 y1 y1 y1 y1 y1Ž . Ž . Ž . Ž . Ž . Ž . Ž . Ž . Ž . Ž .% % % % g kg g kg cmol kg cmol kg cmol kg cmol kgc c c c

Ouenarou´OUE 1-1 0–5 52.7 27.3 6.9 13.0 4.6 5.1 0.47 36.2 1.62 54.7 0.5 0.4 0.1 0.1 0.05 1.7OUE 1-4 38–57 18.9 35.8 27.4 17.9 4.8 5.8 1.06 2.4 0.16 93.8 0.1 0.1 0.0 0.0 0.04 0.2OUE 3-1 0–4 40.0 45.6 9.0 5.4 5.1 5.1 y0.01 76.3 2.37 50.5 9.9 2.4 0.5 0.2 0.07 14.4OUE 3-4 40–51 42.2 40.8 12.9 4.1 5.0 6.1 1.10 11.9 0.71 69.1 0.2 0.6 0.1 0.0 0.03 1.1OUE 4-2 4–9 52.9 34.6 6.5 5.9 4.6 4.7 0.10 25.4 1.35 71.0 0.2 0.5 0.1 0.0 0.03 1.4OUE 4-3 26–39 39.4 36.9 16.5 7.2 5.0 6.2 1.24 9.6 0.51 78.1 0.1 0.2 0.0 0.0 0.05 0.5

MareMAR 3-1 0–12 62.4 26.5 8.8 2.4 6.8 6.3 y0.43 72.3 6.84 82.2 17.8 10.5 0.3 0.2 0.05 28.8MAR 3-3 35–60 85.6 8.8 5.1 0.5 5.8 6.0 0.17 10.1 0.90 93.8 0.7 0.8 0.0 0.1 0.02 1.7MAR 5-1 0–12 60.9 32.4 4.8 1.9 6.3 5.9 y0.37 117.3 8.75 75.2 23.6 11.5 0.4 0.3 0.1 36.0MAR 5-4 40–50 83.1 12.3 3.8 0.9 5.4 5.6 0.27 9.8 0.86 91.6 1.0 1.2 0.0 0.1 0.02 2.4

Ž . Ž . Ž . Ž .CLs clay - 2 mm ; FSs fine silt 2–20mm ; CSs coarse silt 20–50mm ; Ss sand 50–2000mm3q qŽ .P s phosphate retention; EBs exchangeable bases; EAs exchangeable acidity Al qH ; ECECs effective cation exchange capacity measured in 0.5 M NH Clret 4

Mineral fine earth composition

LOI residual SiO Al O Fe O TiO MnO NiO Cr O CoO CaO MgO K O Na O TRB K K2 2 3 2 3 2 2 2 3 2 2 r iy1 y1Ž . Ž .g kg cmol kgc

Ouenarou´OUE 1-1 174 37 22 107 623 2 6 8 31 1 0 4 0 0 18.0 0.05 0.20OUE 1-4 145 31 17 103 648 3 9 9 30 1 0 3 0 0 17.3 0.04 0.16OUE 3-1 166 85 91 70 539 1 8 12 27 1 1 8 0 0 41.6 0.26 1.25OUE 3-4 138 144 75 72 521 2 7 11 28 1 0 8 0 0 41.1 0.22 0.99OUE 4-2 197 64 73 63 563 1 7 12 20 1 0 4 0 0 19.6 0.21 1.12OUE 4-3 146 87 42 95 579 2 9 10 32 1 0 6 0 0 31.1 0.11 0.43

MareMAR 3-1 329 19 8 376 210 11 13 n.d. n.d. n.d. 1.2 0.6 0 0 75.7 0.02 0.02MAR 3-3 248 6 15 440 251 8 8 n.d. n.d. n.d. 0.2 0.3 0 0 19.2 0.02 0.03MAR 5-1 391 6 10 321 186 11 8 n.d. n.d. n.d. 1.8 0.5 0 0 92.1 0.02 0.03MAR 5-4 252 9 13 427 241 13 6 n.d. n.d. n.d. 0.6 0.3 0 0 35.9 0.02 0.03

n.d.s not determined.LOIs loss on ignition.TRBs total reserve in bases;K sSiO rAl O q Fe O ; K sSiO rAl O .r 2 2 3 2 3 i 2 2 3


to 20, indicating a low degree of N incorporation in the humic components. TheECEC were very low except in the upper horizon of OUE 3.

The soils from Mare were slightly acid in the topsoil with a decrease of pH´with depth. TheDpH were negative in the surface horizons but were over 0.1 inthe deeper ones. The organic matter was well humified with a CrN ratio closeto 10. Carbon contents reached 12% in the upper horizons and decreased sharplywith depth. These high amount of organic matter in surface horizons could beattributed to the strong binding of aluminium and iron compounds by organicsubstances and to the low mobility of the humic acid or humate complexes at

Ž . Ž .soil pH Schnitzer, 1986 . The phosphate retentions P exceeded 75% andret

90% in the A and Bo horizons, respectively. In these soils, the CEC is highlyTŽ 2 .correlated to organic carbonr s0.96 and exchangeable bases are dominated

by calcium and magnesium.

3.2. Mineralogy

The oxidic nature of these soils was evident from the total elements analysisŽ .of the fine earth Table 2 , from the extraction by DCB and from the XRD

Ž .results Table 3 .In the samples from Ouenarou, total Fe O content was always higher than´ 2 3

50% of the mineral fine earth composition and could reach 65% in OUE 1Ž .Table 2 . The mineralogy was dominated by goethite, whereas hematite and

Žmaybe magnetite which was not distinguishable from chromite with XRD as. Ž .there are both spinels were present at lower levels Table 3 . The dark red

Ž .colour of some horizons Table 1 was rather unexpected as hematite contentŽ .was low. The amounts of Fe and Al were very low Table 3 indicating thato o

Fe and Al oxides were mainly crystalline. Generally, more than 95% of the totalŽ .iron content of the samples were dissolved by the DCB extraction Table 3 . The

y1 ŽAl-substitution ratios of iron oxides varied from 0.1 to 0.17 mol mol Table.3 . In our study, we used a 5-day DCB extraction as the dissolution rates were

Ž . Ž .low results not shown . Jeanroy et al. 1991 explained that the low DCBsolubility of different iron oxide-rich tropical soils is related to the high levels ofAl substitution. Moreover, chromium, which was also present in the Fe-oxides

Ž .in these soils result not shown , reduces their dissolution rate much moreŽ .Cornell and Schwertmann, 1996 .

The amounts of silica were very low. Total SiO contents were close to 2%2Ž . Žand 4–9% in the piedmont OUE 1 and alluvio-colluvial soils OUE 3 and

.OUE 4 , respectively. The XRD analysis, after deferification, showed littledifferences between samples. The main silicate mineral was talc and traces of

Ž . Ž .quartz and chlorite OUE 3 and OUE 4 or mica OUE 1 were also detectedŽ .Table 3 . The extent of the weathering of the soils can be estimated by theK r

w Ž . xand K ratios K sSiO r Al O qFe O ; K sSiO rAl O and by thei r 2 2 3 2 3 i 2 2 3Ž . Ž .total reserve in bases TRB Table 2 . These parameters showed that the

()

T.B

ecqueret

al.rG

eoderma

1032001

291–

306299

Table 3Selective extractions of Al, Fe and Si and mineralogy of the soil samples

3Oxalate extraction DCB extraction Fe)10 r Fe r Al substition Mineralogyo dy1Ž .Fe Fe mol mold totAl Fe Si Al Fe Sio o o d d d

y1 y1 y1 y1 y1 y1Ž . Ž . Ž . Ž . Ž . Ž .g kg g kg g kg g kg g kg g kg

Ouenarou´OUE 1-1 0.97 2.11 0.03 43 420 4.1 5.0 0.96 0.171 Go, He, Ta, Sp, QOUE 1-4 1.87 1.57 0.03 47 447 4.0 3.5 0.99 0.179 Go, He, Ta, Sp, Q, MiOUE 3-1 1.44 3.08 0.22 20 327 6.4 9.4 0.87 0.105 Go, He, Ta, Sp, Ch, QOUE 3-4 1.13 1.9 0.27 26 347 6.1 5.5 0.95 0.131 Go, He, Ta, Sp, Ch, QOUE 4-2 1.03 7.31 0.24 32 427 6.3 17.1 1.08 0.132 Go, He, Ta, Sp, Ch, QOUE 4-3 1.35 2.97 0.20 37 393 5.4 7.6 0.97 0.160 Go, He, Ta, Sp, Ch, Q

MareMAR 3-1 11.34 4.41 0.11 45 138 0.0 32.0 0.94 0.371 Gb, Bo, Go, Q, Ma, Ch, FeMAR 3-3 4.59 4.07 0.08 46 162 0.0 25.1 0.92 0.359 Gb, Go, Bo, Q, Ma, FeMAR 5-1 12.43 3.76 0.04 46 118 0.0 31.9 0.91 0.401 Gb, Go, Bo, Q, Ma, Ch, FeMAR 5-4 7.21 3.15 0.02 50 162 0.0 19.4 0.96 0.375 Gb, Go, Bo, Q, Ma, Ka, Fe

Bo, boehmite; Ch, chlorite; Fe, feldspar; Gb, gibbsite; Go, goethite; He, hematite; Ka, kaolinite; Ma, magnetite; Mi, mica; Q, quartz; Sp, spinelle;Ta, talc.


Ž .piedmont soil OUE 1 is the most weathered one. In the colluvio-alluvial soilsŽ .OUE 3 and OUE 4 talc was likely more abundant and the TRB, which wasclose to 30–40 cmol kgy1 in Bo horizons, was mainly due to magnesium.c

However, Si varied from 4 to 6 g kgy1, whereas Si extracted with CB wasdŽ .close to zero except for OUE 3-1 . Si represented 40–50% of the total Sid

content for OUE 1 and 15–30% for OUE 3 and OUE 4. Therefore, a substantialpart of Si seemed to be associated with iron oxides.

In the soils of the Loyalty Islands, the crystalline fraction is mostly present asgibbsite and boehmite for aluminium oxides and goethite for iron oxides. These

Ž .results are in agreement with those of Tercinier 1971 . A few primary silicatessuch as quartz and feldspar remained in the soil. As for Ouenarou soils, a 5-day´

Žextraction with DCB was necessary to solubilize up to 90% of total Fe Table. y13 . The amount of Al was 45–50 g kg , corresponding to Al-substitutiond

ratios of the iron oxide fractions of about 0.35 mol moly1. This high Al-substitu-tion in goethite, which is generally observed in highly desilicified, gibbsitic soilsŽ .Cornell and Schwertmann, 1996 , explained the slow dissolution rates of theiron oxides. The intense weathering of this soil led to total SiO contents close2

Ž .to 1% and thus to very lowK and K ratios Table 2 . The TRB in the deepestr i

horizons is relatively high probably due to carbonate particles contaminationfrom the underlying coral rocks. Chemical extractions of Si showed thatSi )Si . This result might indicate the presence of allophanic products, but theo d

levels of Si were too low to account for the presence of substantial amount ofo

allophane.

3.3. Surface charge properties

The ion adsorption method has been used extensively to determine theŽnegative and positive surface charges of soils with pH-dependent charges Wada

and Okamura, 1980; Okamura and Wada, 1983; Wada and Wada, 1985; Gillman.and Sumpter, 1986; Parfitt, 1992; Van Ranst et al., 1998 .

The CEC varied between 0.3–36 and 0–3.8 cmol kgy1 in A and BoT cŽ .horizons, respectively, in the range of pH 4–7 Fig. 2 . It was mainly related to

the organic matter content, as the mineralogical composition of A and Bohorizons were similar. The dissociation of carboxyl groups of the organic matter

Žis strongly pH-dependent and increases progressively with pH Tate and Theng,.1980 . However, some differences exist between the two types of soils. For

similar levels of organic carbon, the CEC was higher in the soil from Mare´T

than in those from Ouenarou. The CEC per unit of organic carbon calculated at´ T

pH 7 from A horizons was 281 cmol kgy1 C in Ouenarou soils, whereas in´c

Marean soils it was 360 cmol kgy1 C. This could be related to differences of´ c

organic matter qualities as expected from their different CrN ratios. The oxidiccompounds could also have consequences on the CEC . OUE 1–4 was veryT

Ž y1.poor in organic carbon 2 g kg and exhibited a particular shape of its


Ž . Ž .Fig. 2. Variation of CEC and AEC with pH for the surface 1A, 1B and deep 2A, 2B horizonsTŽ . Ž .of Ouenarou A and Mare B . Measurements and modelling.´ ´

Ž .CEC –pH curve Fig. 2A . The CEC was nil below pH 5.5 and increasedT T

sharply above. An increase of pH causes a decrease of the electric potentials ofŽ .oxides and the increase of the dissociation of bivalent cations Barrow, 1987 .

These two phenomena are favourable to cation sorption on the surface of theŽoxides and result in a sudden increase in adsorption with increasing pH Barrow,

.1987 .The CEC was close to CEC except at pH below 4. Aluminium oxides wereB T

solubilised at lower pH during the equilibrium phase with 0.002 M CaCl .2

During the subsequent extraction with 1 M NH NO , leading to an increase of4 3

1–2 pH units, the Al3q previously solubilised was either sorbed on negativesurface charge, if present, or precipitated as Al-oxides leading to some artefacts.Therefore, the CEC below pH 4 were not taken into account.T


Ž .The AEC was lower in A than in Bo horizons Fig. 2 with values rangingbetween 0.63–2.99 and 2.38–4.25 cmol kgy1 at pH 4, respectively. Thec

differences between samples could be related to the types of oxides, to theparticle size distribution as well as to the organic matter content, the latter beingprobably the most important. A close association of Fe and Al compounds withorganic anions at the surface of the oxides could reduce the sites available forCly sorption and therefore decrease the AEC. The values found here were

Žhigher than those observed in other studies on Ferralsols Okamura and Wada,.1983; Gillman and Sumpter, 1986; Parfitt, 1992; Van Ranst et al., 1998 and are

Žof the same magnitude as those generally found on Andosols Okamura and.Wada, 1983; Parfitt, 1992 . In those studies, the soils exhibit permanent negative

charges and the negatively charged surfaces repel anions from the double layerŽ .Gillman and Sumpter, 1986 . Therefore, AEC does not represent the totalamount of positive surface charges. In our conditions, the permanent negativecharges were practically nil and the AEC measured was equal to the totalpositive surface charge of the soil.

Ž . Ž .CEC and AEC were modelled according to Eqs. 1 and 2 , respectively,T

and the values of adjustable parametersa , B , a , B and the coefficient of1 1 2 2Ž 2.determination r are given in Table 4. The agreement between the calculated

Ž 2 .and measured values was goodr s0.971 to 1.000 . The values ofa and a1 2

describe the effect of pH on the variation of charge of the soils. The values ofa1

were low in organic matter-rich samples; they increased when carbon contentsŽ .decrease. Thea values 0.10–0.19 obtained for the samples with C)5%1

Ž .were of the same magnitude as those reported by Wada and Okamura 1980 ,Ž . Ž .Okamura and Wada 1983 and Wada and Wada 1985 . They reported values in

Table 42 Ž . Ž .Coefficients a , a , B , B and coefficients of determinationr for regression Eqs. 1 and 21 2 1 2

Negative charge Positive charge2 2a B r a B r1 1 2 2

Ouenarou´OUE 1-1 0.391 y1.647 0.996 y0.597 2.312 0.996OUE 1-4 1.246 y7.786 0.999 y0.389 2.177 0.996OUE 3-1 0.189 y0.004 0.987 y0.721 2.814 0.996OUE 3-4 0.421 y2.430 0.989 y0.440 2.304 0.993OUE 4-2 0.420 y2.189 1.000 y0.571 2.662 0.998OUE 4-3 0.508 y3.182 0.990 y0.412 2.331 0.995

MareMAR 3-1 0.116 0.631 0.971 y0.547 2.665 0.996MAR 3-3 0.502 y2.909 0.988 y0.473 2.560 0.993MAR 5-1 0.100 0.896 0.987 y0.553 2.447 0.983MAR 5-4 0.409 y1.997 0.988 y0.655 3.131 1.000


the range of 0.11–0.34 for Andosols and 0.04–0.23 for Alfisols, Oxisols andUltisols. The different kinds of functional groups of the organic matter andror

Ž .the different mineral compounds clay minerals, oxides, allophane . . . present inthose soils led to a slight increase in CEC with pH. For the samples in theT

deepest horizon, the narrow range of points of zero charge of Fe and Al oxides,ranging from 7 to 9.5, led to a sharp increase of the CEC with pH. In the sameT

way, the absence of repulsion of anions by permanent negative charge at low pHas described above led to a sharp increase of the AEC. Thea values ranged2

from y0.55 to y0.72 and fromy0.39 to y0.66 in A and Bo horizons,respectively. These values were lower than those reported by Okamura and

Ž .Wada 1983 , which are betweeny0.05 andy0.36.

3.4. Soil classification

Ž .The two types of soils were classified according to FAO 1998 . Theyresulted from an intense weathering and were highly desilicified. Their mineral-ogy was dominated by iron and aluminium oxides. Different field observationsand chemical characteristics, such as their loamy to loamy–sandy texture, theirstrong microaggregation, and their low ECEC led to classify these soils asFerralsols. However, the TRB was sometimes over 25 cmol kgy1 in the Boc

horizons. For the Loyalty Islands, this could be related to traces of calciumcarbonate originating from the underlying coral rocks. At Ouenarou, the collu-´

Ž .vio-alluvial soils OUE 3 and OUE 4 have been rejuvenated by the input ofŽ .traces of Mg-rich phyllosilicates talc and chlorite . Accordingly, the TRB of

their Bo horizons are somewhat larger than 25 cmol kgy1 and are similar inc

this respect to those found in other ferralsols deriving from Mg-bearing parentŽ .materials Herbillon, 1986 .

According to theirDpH higher than 0.1 and to their low ECEC in the Bohorizons, the soils were classified as Geric Ferralsols.

The soils from Mare indicated some Andic properties such as a bulk density´y3 Ž .less than 0.9 Mg m result not shown and a phosphate retention of more than

70% in A and Bo horizons. However, the Alq0.5Fe values close to 1.5% ino oŽ .the A horizons or 1% in the Bo horizons Table 3 were too low to use the

Andic qualifier. The Al oxide content above 25% of the fine earth fraction led tothe use of the Gibbsic qualifier. The Humic qualifier could also be used as therewas an average carbon content of 2.6–3.8% in the fine earth fraction on thewhole profile. Therefore, the soil of Mare can be classified as a Geri-Gibbsic´

Ž .Ferralsol Humic .For Ouenarou soils, none of the qualifiers of the Ferralsols could be used to´

described these soils except the Geric one. For OUE 4, reducing conditionscould occur leading to some iron and manganese oxide coatings. However, thegleyic colour pattern was too weak to use this qualifier. Therefore, these soilswere simply classified as Geric Ferralsol.


Ž .These soils were previously classified by Latham 1980–1981 according totheir dominant oxide types: theAferritiquesB soils, rich in Fe-oxides, developedon the ultramafic rocks; theAallitiquesB soils, rich in Al and Fe-oxides, on the

Ž .coral rocks. According to FAO 1998 , the Gibbsic qualifier can be used for thesoils from the Loyalty Islands. However, the Ferric qualifier cannot be used forthe soils on ultramafic rocks as segregation of iron has not taken place in thesesoils. In order to take into account the high level of iron oxide and in agreementwith the regional classification system in use in New Caledonia, a Ferriticqualifier could be suggested for Ferralsols with more than 25% of iron oxide inthe fine earth fraction. In this way, the two types of Ferralsols could be easilyseparated in the FAO classification scheme on the basis of their differences inoxide composition.

3.5. Agronomic and enÕironmental consequences

Geric Ferralsols with high levels of Fe and Al oxides are known to adsorbanions. On the soil of New Caledonia, a very high fixation of phosphorus occursŽ . ŽDubus et al., 1998 and leads to yield limitations for crops L’Huillier et al.,

.1998 . The adsorption of non-specifically adsorbed anions such as nitrate is alsoŽ .observed on the soil of the Loyalty Islands Duwig et al., 1999 . The reduction

of the movement of nitrate through the soil limits the contamination of underly-Ž .ing groundwater Duwig et al., 2000 .

The two types of soils varied mainly according to their surface horizons. Inthe soil from Mare, the organic matter content was larger and better humified.´CEC and AEC varied to a lesser extent in this soil as shown by the lowerT

values of parametersa and a . The CEC that was mainly related to the1 2 T

organic matter was greater and saturated by Ca and Mg owing to the underlyingŽ .coral rocks. In the soils from Ouenarou, exchangeable trace metals Ni, Mn . . . ,´

which are often pointed out to explain the infertility of soils derived fromŽ . Žultramafic rocks Proctor and Woodell, 1975 , are also present Becquer et al.,

.1995 . In both soil types, exchangeable potassium was very low. Thus, thefertility of Mare Ferralsol is certainly higher than that of Ouenarou Ferralsol.´ ´

The CEC and AEC act on the retention of major nutrient cations or anionsT

and on the availability of toxic metals like Ni, Cr, Mn and Co. Both CEC andAEC are affected by factors such as the pH, the soil solution concentration andthe nature of the surface charge of the different compounds of the soils. As thedevelopment of negatively charged sites will enhance cation retention on the one

Žhand and anion leaching on the other and the opposite for positively charged. Ž .surfaces , management strategies inputs of fertilizers, organic matter or lime . . .

could have major agronomic and environmental consequences. The shapes ofthe curves giving the surface charge variations with pH, or the parametersa1

and a describing these shapes, could be useful tools to separate different soil2

types into management groups.


References

Barrow, N.J., 1987. Reactions with variable-charge soils. Developments in Plant and SoilSciences, vol. 31, Martinus Nijhoff Publishers, Dordrecht, The Netherland.

Bartoli, F., Burtin, G., Herbillon, A.J., 1991. Disaggregation and clay dispersion of Oxisols: Naresin, a recommended methodology. Geoderma 49, 301–317.

Becquer, T., Bourdon, E., Petard, J., 1995. Disponibilite du nickel le long d’une toposequence de´ ´ ´sols developpes sur roches ultramafiques de Nouvelle-Caledonie. C. R. Acad. Sci. Paris, Serie´ ´ ´ ´IIa 321, 585–592.

Blakemore, L.C., Searle, P.L., Daly, B.K., 1987. Methods for chemical analysis of soils NewZealand Soil Bureau, Scientific Report no. 80. New Zealand Society of Soil Science, LowerHutt, New Zealand.

Cornell, R.M., Schwertmann, U., 1996. The iron oxides. Structure, properties, reactions, occur-rence and uses. VCH Publishers, Weinheim, Germany.

Dubus, I., Becquer, T., Bonzon, B., Calvet, R., 1998. Phosphorus sorption in two Geric Ferralsolsof New Caledonia. XVIe Congres Mondial de Science du Sol, Montpellier, France.`

Duwig, C., Becquer, T., Clothier, B.E., Vauclin, M., 1999. A simple dynamic method to estimateanion retention in an unsaturated soil. C. R. Acad. Sci. Paris, Ser. IIa 328, 759–764.

Duwig, C., Becquer, T., Vogeler, I., Vauclin, M., Clothier, B., 2000. Water dynamics and nutrientŽ .leaching through a cropped Ferralsol in the Loyalty Islands New Caledonia . J. Environ. Qual.

29, 1010–1019.FAO, 1994. Directives pour la Description des Sols. 3eme edn. FAO, Rome.` ´FAO, 1998. The world reference base for soil resources. World Soil Ressources Report no. 84.

ISSSrAISSrIBGrISRICrFAO, Rome.Gillman, G.P., Sumpter, E.A., 1986. Surface charge characteristics and lime requirements of soil

derived from basaltic, granitic, and metamorphic rocks in high-rainfall tropical Queensland.Aust. J. Soil Res. 24, 173–192.

Herbillon, A.J., 1986. Chemical estimation of weatherable minerals present in the diagnosticŽ .horizons of low activity clay soils. In: Beinroth, F.H., Camargo, M.N., Eswaran, H. Eds. ,

Proceedings of the Eighth International Soil Classification Workshop: Classification, Charac-terization and Utilization of Oxisols, Part I: EMBRAPA, Rio de Janeiro, Brazil, pp. 39–48.

Holmgren, G.G.S., 1967. A rapid citrate–dithionite extractable iron procedure. Soil Sci. Soc. Am.Proc. 31, 210–211.

Jeanroy, E., Rajot, J.L., Pillon, P., Herbillon, A.J., 1991. Differential dissolution of hematite andgoethite in dithionite and its implication on soil yellowing. Geoderma 50, 79–94.

Latham, M., 1980–1981. Les oxydisols dans quelques milieux insulaires du Pacifique sud:caracterisation, conditions de formation, fertilite, classification. Cah. ORSTOM, Ser. Pedol.´ ´XVIII 3–4, 305–316.

Latham, M., Quantin, P., Aubert, G., 1978. Etude des Sols de la Nouvelle-Caledonie. ORSTOM,´Paris.

L’Huillier, L., Edighoffer, S., 1996. Extractability of nickel and its concentration in cultivatedplants in Ni rich ultramafic soils of New Caledonia. Plant Soil 186, 255–264.

L’Huillier, L., Dupont, S., Bonzon, B., Dubus, I., Becquer, T., Bourdon, E., 1998. Carence etfixation du phosphore dans les sols ferrallitiques ferritiques du sud de la Nouvelle-Caledonie.´XVIe Congres Mondial de Science du Sol, Montpellier, France.`

McKeague, J.J., Day, J.A., 1966. Dithionite and oxalate-extractable Fe and Al as aids indifferentiating various classes of soils. Can. J. Soil Sci. 46, 13–22.

Ž .Okamura, Y., Wada, K., 1983. Electric charge characteristics of horizons of Ando B andred-yellow B soils and weathered pumices. J. Soil Sci. 34, 287–295.

Ž .Parfitt, R.L., 1980. Chemical properties of variable charge soils. In: Theng, B.K.G. Ed. , Soils


with variable charge. New Zealand Society of Soil Science, Lower Hutt, New Zealand, pp.167–194.

Parfitt, R.L., 1992. Surface charge in some New Zealand soils measured at typical ionic strength.Aust. J. Soil Res. 30, 331–341.

Proctor, J., Woodell, S.R.J., 1975. The ecology of serpentine soils. Adv. Ecol. Res. 9, 255–366.Rouiller, J., Souchier, B., Bruckert, S., Feller, C., Toutain, F., Vedy, J.C., 1994. Methodes´ ´

Ž .d’analyses des sols. In: Bonneau, M., Souchier, B. Eds. , Pedologie: 2. Constituants et´Proprietes du Sol. Masson, Paris, pp. 619–652.´ ´

Schnitzer, M., 1986. Binding of humic substances by soil mineral colloids. In: Huang, P.M.,Ž .Schnitzer, M. Eds. , Interactions of Soil Minerals with Natural Organics and Microbes. SSSA,

Madison, WI, USA, pp. 77–101, Special Publication No. 17.Schwertmann, U., Latham, M., 1986. Properties of iron oxides in some New Caledonian oxisols.

Geoderma 39, 105–123.Tate, K.R., Theng, B.K.G., 1980. Organic matter and its interactions with inorganic soil

Ž .constituents. In: Theng, B.K.G. Ed. , Soils with variable charge. New Zealand Society of SoilScience, Lower Hutt, New Zealand, pp. 225–249.

Tercinier, G., 1971. Contribution a la connaissance des phenomenes de bauxitisation et d’allitisa-` ´ `tion. Les sols de karst d’atolls sureleves du Sud-Ouest Pacifique. Cah. ORSTOM, Ser. Pedol.´ ´IX 3, 307–334.

Van Ranst, E.V., Shamshuddin, J., Baert, G., Dzwowa, P.K., 1998. Charge characteristics inrelation to free iron and organic matter of soils from Bambouto, Westren Cameroon. Eur. J.Soil Sci. 49, 243–252.

Wada, K., Okamura, Y., 1980. Electric charge characteristics of Ando A1 and buried A1 horizonsoils. J. Soil Sci. 31, 307–314.

Wada, S.-I., Wada, K., 1985. Charge characteristics and exchangeable cation status of KoreanUltisols and Alfisols and Thai Ultisols and Oxisols. J. Soil Sci. 36, 21–29.

Estimation of nitrate retention in a Ferralsol by atransient-flowmethod

C. DUWIGa, T. BECQUER

b, L. CHARLETc & B. E. CLOTHIER

d

aIRD/Laboratoire d’etude des Transferts en Hydrologie et Environnement, LTHE-UMR (CNRS, INPG, IRD, UJF), BP 53 X, 38041

Grenoble, France, bIRD/Embrapa Cerrados, CP 7091, 71619-970 Brasilia – DF, Brazil, cLaboratoire de Geophysique Interne et

Tectonophysique (UJF-CNRS), OSUG, BP 53, 38041 Grenoble, France, and dHortResearch, PB 11-030, Palmerston North,

New Zealand

Summary

Anion retention is important in highly weathered soils that contain large amounts of iron and aluminium

oxides with surfaces of variable charge. Sorption mechanisms retard anionic solute transfer through these

soils. We determined the retardation factor for nitrate in highly weathered Ferralsols from New

Caledonia from dynamic experiments using a transient-flow method, and we evaluated the effect of soil

solution concentration and organic matter content. A simple method with sectionable tubes was used to

determine the nitrate isotherm during non-steady-state water flow under unsaturated conditions. The

topsoil retarded the movement of nitrate, and the sorption followed a linear isotherm. In subsoils,

retardation factors were larger and increased from 1.15 to 2.05 at soil pH as the NO3–-N concentration

of the input solution decreased from 71.43 to 0.35mM, indicative of a non-linear isotherm. Positive

surface charge sites were considered to be of two types: one with strong affinity for nitrate at small

concentrations and one with weak affinity for adsorption of nitrate at larger concentrations. This type of

isotherm with high- and low-energy sites is similar to those found for oxyanions and heavy metals. The

related anion exchange capacity was larger than that usually observed in soils of variable charge. Not all

exchange sites were detected with our method, and some sites were obviously not available for nitrate

retention.

Introduction

Leaching of chemicals through the vadose zone to the ground-

water represents an important threat to public health because

of possible contamination of drinking water. For this reason,

we must know more about transport in soil to develop strat-

egies for managing the environment in a sustainable state.

Solute transport can be retarded by sorption of solutes on to

the soil’s matrix. Retention of contaminants is of particular

importance because it retards the transport of contaminants

from the surface to groundwater, thereby allowing more time

for transformation and dissipation of these contaminants by

biotic and chemical processes. Anion retention can be espe-

cially important in highly weathered tropical soils, which have

substantial anion exchange capacity (AEC). In subtropical and

tropical regions, highly weathered soils such as Oxisols and

Ultisols are widely distributed. Because of the abundant rain-

fall and high temperatures there, these regions are potentially

among the most important for agriculture in the world. But

the abundant rainfall also leads to potentially large losses of

nutrients from the root zone. Anionic nutrients such as nitrate

do, however, sorb on tropical soils, and this may lead to the

retardation of nitrate compared with the movement of water

(Wong et al., 1990; Bellini et al., 1996; Katou et al., 1996;

Qafoku & Sumner, 2001). Sorption must be taken into account

in predicting the fate of nitrate through such soils.

Oxisols and Ultisols carry variable charges. The common

feature of these highly weathered soils is the abundance, in the

clay fraction, of minerals with amphoteric surfaces, such as

iron and aluminium oxides and hydrous oxides. Qafoku et al.

(2000) related the surface-charge characteristics to the relative

content, and to the interaction and surface reactivity of the

mineralogical constituents. Subsoil usually retains anions more

readily than does topsoil, because the AEC is larger in the

subsoil. According to Parfitt (1992), carboxylate groups of

organic matter can react with allophane and iron oxides to

Paper given at the Michel Rieu Memorial Colloquium, 8–10 October

2001, in Paris.

Correspondence: C. Duwig. E-mail: [email protected]

Received 7 November 2001; revised version accepted 5 August 2002

European Journal of Soil Science, September 2003, 54, 505–515

# 2003 Blackwell Publishing Ltd 505

form inner-sphere complexes, and these shield the positive

charge on the mineral surfaces, thereby decreasing the positive

charge in surface horizons. Anion retention depends on the

surface charge, which in turn depends on the pH, and the ionic

strength of the soil solution (Charlet & Sposito, 1987). The

AEC and the retardation factor are positively correlated with

electrolyte concentrations at fixed pH (Ishiguro et al., 1992;

Qafoku & Sumner, 2001). Bellini et al. (1996), studying a

highly weathered acid soil, found that for concentrations in

excess of the equilibrium concentration that would saturate the

AEC, retardation decreased when the concentration of the

electrolyte increased. All these findings accord with the

double-layer theory: as the electrolyte concentration increases,

so does the surface positive charge on oxides. However, once all

the positive charge sites are saturated, any further increase in

electrolyte concentration will just lead to a compression of the

double layer (Rietra et al., 2000).

Batch experiments have often been used to assess the

exchange capacity of soils. Burgisser et al. (1993) observed

that the surface abrasion and disintegration of particles during

shaking can lead to an apparently larger sorption capacity.

Wong et al. (1990) reported that shaking and dispersion can

expose charged surfaces which would not be exposed natu-

rally. The soil:solution ratio is also larger in the field. Hoyoux-

Roche & Jamet (1988) found that the distribution coefficient,

Kd, determined by batch experiments for pesticide adsorption

was stable for a soil:solution ratio less than 0.8, which is much

smaller than is usual in batch studies. A small soil:solution

ratio also means that the proportion of free water in pores is

larger than water in the double layer where the nitrate accu-

mulates. Also, the time of contact between the soil and the

solution is longer in batch determination, and this allows

dissolution and precipitation to occur, leading to an equilib-

rium between the soil solution and the solid phase.

A field study conducted on Mare Island, New Caledonia

(Duwig et al., 1998, 2000) showed the water transport through

the soil to be very rapid, due to the soil’s large porosity and to

the rapid precipitation. Under such conditions, the chemical

equilibrium of the soil solution with the mineral phases is not

reached, and exchange reactions could be the dominant pro-

cesses in controlling the composition of the soil solution. In a

previous paper, Duwig et al. (1999) described a simple method

using sectionable soil tubes to measure the retardation of

anions during transient flow under unsaturated conditions.

This was validated against nitrate breakthrough curves

obtained for those conditions. Clothier et al. (1988), Bond &

Phillips (1990) and Katou et al. (1996) had previously used the

so-called Perroux-tube technique to study cation and anion

movement. We have used these transient experimental condi-

tions, which mimic more closely those found in the field, to

determine the entire nitrate isotherm across the different hori-

zons of the soil profile with contrasting organic matter con-

tents. We have characterized nitrate retention in a highly

weathered soil from New Caledonia and analysed the effect

of the nitrate concentration in the input solution, as well as the

effect of organic matter content on this retention and on the

AEC. We report the results below.

Theory and analytical methods

Duwig et al. (1999) described the theory in detail, and here we

summarize only the important concepts. The theory applies to

the transport of water and reactive solute during transient

wetting of soil in a tube, at an initial volumetric water content

of �n. The transport is considered to be horizontal, one-dimen-

sional and transient (time-dependent).

Water

Water transport is described by Richards’ equation under the

conditions described above (Smiles et al., 1978; Duwig et al.,

1999). Smiles et al. (1978) and Bond (1986) have shown that

when water is absorbed by a horizontal unsaturated soil col-

umn, provided that there is no preferential flow of water, the

average position of the front of a non-reactive solute corres-

ponds to the piston front of the water. A plane of separation,

x*, which identifies the front of the invading water, can be

calculated from the water content profile y(x) (Smiles & Philip,

1978):

ð�s�n

xd� ¼ðx�0

�dx; ð1Þ

where x is the horizontal distance (m) from the proximal end

of the soil column, i.e. the end where the invading solution

enters the soil column, y is the water content (m3m�3), yn is

the initial water content in the entire column, ys is the water

content at the proximal end of the soil column (m3m�3), and

x* (m) is the plane of separation.

After an elapsed time t (s), the sorptivity S (m s�1/2), which

is a measure of the capillary uptake of water by the soil and a

function of the initial and final water content, can be calcu-

lated as (Philip, 1969):

S ¼ð�s�n

xd�.t1=2: ð2Þ

Solute

The solute transport can be described by the one-dimensional

convection–dispersion equation (CDE) for reactive solute

under transient conditions. For short duration horizontal

experiments of 0.5–1 hours, where the water flow is fast

enough, molecular diffusion can be ignored in relation to

dispersion. Assuming a linear isotherm for reactive solutes,

Watson & Jones (1981) found an approximate solution to

the CDE for the normalized resident concentration C (–)

after a step input of solution of concentration C1:

506 C. Duwig et al.

# 2003 Blackwell Publishing Ltd, European Journal of Soil Science, 54, 505–515

C ¼ Cr � C0

C1 � C0¼ 1

2erfc

�B � �0B� �

t1=4

2 �0 �0B� �1=2

" #; ð3Þ

where Cr is the resident solute concentration, C0 is the initial

concentration in the soil solution, and �B, �0B and �0 are given

by

�B ¼ x� ffiffi

tp;

�0

B ¼ ��B=R; and ��B ¼ x�=ffiffitp

ð4Þ

and

�0 ¼ �=R;

where � is the water dispersivity (m), R (–) is the retardation

factor for the reactive solute, and thus �0 is the reactive solute

dispersivity (m). The quantity �B is the so-called Boltzmann

variable (Philip, 1969), which is a similarity scale used as the

independent variable. Thus, �*B is the plane of separation in

terms of the Boltzmann variable and �0B the retarded centroid

position of the reactive solute front, again in terms of the

Boltzmann similarity variable.

In our experiments, the soil column was cut into slices at a

given time t, and the total concentration of nitrate in the soil,

M (mol kg�1), rather than the resident concentration, was

measured in each soil sample (Figures 2–6). For a reactive

solute species, the total concentration per unit mass of soil is

the sum of the contents in the adsorbed phase (q, molc kg�1)

and in the aqueous solution:

M ¼ �

�Cr þ q; ð5Þ

where � is the soil’s bulk density (kgm�3).

For a linear isotherm with the form

q ¼ Kd Cr;

R ¼ 1þ ð�Kd=�Þ; ð6Þ

where Kd is the distribution coefficient (l mol�1).

Using Equations (5) and (6), we canwrite Equation (3) now as

M ¼ C0 þ C1 � C0ð Þ½ � ��

R1

2erfc

�B � �0B� �

t1=4

2 �0 �0B� �1=2

" #: ð7Þ

In Equation (7), C0 and R are unknowns, which we deter-

mined for each soil horizon by fitting Equation (7) to the

measured points (M, �B), using a least-squares minimiza-

tion method (Solver software, Microsoft Excel 2000). Here,

�¼1.5mm was found to be an appropriate value for water

movement in this soil – see Duwig et al. (1999).

We can obtain the adsorption isotherm by plotting the

adsorbed concentration, q, against the equilibrium concentra-

tion, Ceq. The equilibrium between sorbed and aqueous anions

was assumed to have been reached at the proximal end of the

tube by time t. With Meq being the value of M at the proximal

end of the tube, Ceq and q can be calculated from Equations

(5) and (6) as follows:

Ceq ¼Meq

R

�

�s

and

q ¼Meq ��s�Ceq: ð8Þ

Materials and methods

Soil

The soil used was from Mare, which is in the Loyalty Islands

(New Caledonia). These islands are uplifted coral atolls built

upon an underlying volcanic structure. The soil probably

derives from the weathering of volcanic ejecta and ashes

deposited on the coral, and has a depth ranging from 0.3 to

1m. It is a Geric Ferralsol in the FAO classification. Four

horizons were sampled: one surface horizon from a site rich in

humus H1 (0–10 cm), and the three other horizons from a soil

under cultivation (H2: 0–20, H3: 20–40, and H4: 50–60 cm).

We chose H1 for its larger organic content, yet its mineralogy

is similar to the other horizons. We could thus compare four

horizons with contrasting organic matter content. Table 1 lists

Table 1 Main chemical characteristics of the four soil horizons studied

Organic matter Exchangeable cations Total elements, as oxides

C N Ca2þ Mg2þ Kþ Naþ CECa SiO2 Al2O3 Fe2O3 TiO2 MnO2 CaO

Horizon Depth /cm /g kg�1 /cmolc kg�1 /cmolc kg

�1 /g kg�1

H1b 0–10 54.7 4.85 13.3 7.9 0.14 0.14 18.3 9.54 366 182 16.6 7.25 14.9

H2 0–20 35.9 3.5 11.4 4.1 0.07 0.10 15.0 5.47 400 202 17.1 10.3 9.6

H3 20–40 17.6 2.1 6.2 2.1 0.02 0.04 10.4 5.48 413 218 18.2 9.62 5.0

H4 50–60 14.4 1.3 7.5 3.3 0.02 0.04 10.0 5.53 415 216 17.7 6.38 14.1

aCation exchange capacity.bH1 was taken from a different location from H2 to H4.

Nitrate retention in a Ferralsol 507


the main chemical properties of each soil horizon. Becquer

et al. (2001) present other mineralogical and chemical proper-

ties of this soil.

The crystalline fraction is present mostly as gibbsite and

boehmite for aluminium oxides, and goethite for iron oxides.

The intense weathering of the soil has led to a very small total

SiO2 content, close to 1%. The amount of oxalate-extractable

Si was too small to account for the presence of substantial

amount of allophane (Becquer et al., 2001).

As shown in Table 1, the soil is slightly acidic or neutral,

and the deepest horizon has the largest pH. This latter might

be due to the presence of carbonate particles from the under-

lying coral rock. The organic matter is well humified, with a

C/N ratio more than 10. The carbon content reaches 5.5% in

the upper horizon and decreases sharply with increasing depth.

This large amount of organic matter in the surface horizon

can be attributed to the strong binding of humic substances

to the aluminium and iron oxides, and to the immobility of

the humic acid or humate complexes at the soil’s pH. The

cation exchange capacity (CEC) is related mainly to the

organic matter content, and thus varies with pH. It reaches

18 cmolc kg�1 at soil pH (6.9) in the surface horizon and

decreases sharply with depth (2.5 cmolc kg�1 at pH6.7 for the

subsoil). Exchangeable bases are dominated by calcium and

magnesium.

The surface positive charges of the soil are created by the

protonation of iron and aluminium oxide hydroxyl sites. The

AEC at the soil’s pH (Table 2), calculated from the data in

Becquer et al. (2001), was determined by the method of

Gillman & Sumpter (1986) at several fixed pH values with an

equilibrium concentration of 0.002M CaCl2, and a relationship

between it and pH was established. Calculated values of AEC

at soil pH (6.9–7.2) vary from 0.04 to 0.14 cmolc kg�1 for H1

and H4, respectively. The surface horizon has smaller positive

charge than deeper horizons. At pH5, which is close to the

usual pH of variable-charge soils, AEC varied between 0.48

and 1.57 cmolc kg�1 for H1 and H4, respectively. These values

are larger than those common on soils of variable charge

(Qafoku et al., 2000; Qafoku & Sumner, 2001). Part of the

positive charge is balanced by permanent negative charge,

with iron oxides often being present between clay layers

(Tandy et al., 1990). In our study, the soil has no significant

permanent negative charge, and the measured AEC is equal to

the total positive surface charge of the soil.

Horizontal nitrate transport experiments

The soil samples were moistened with distilled water to attain

a volumetric water content of �n¼ 0.16 (standard error of 0.02

calculated on six samples) for the H1 horizon, and �n¼ 0.13

(standard error of 0.01 calculated on six samples) for the three

others. The soil was then packed to a fixed bulk density � for

each horizon (Table 3), in a sectionable transparent acrylic

tube, 25 cm long with a 2-cm internal diameter. This device,

the so-called Perroux tube, is similar to those used by Smiles

et al. (1978). All the tubes were packed in a similar way for the

experiments with different input concentrations so as to study

the movement of nitrate, independent of that of the water. The

initial water content was presumed to be constant along the

tube. One-dimensional, horizontal, absorption experiments

were carried out using KNO3 solutions with NO3–-N concen-

trations of 71.43, 35.71, 7.14, 3.57, 0.71 and 0.36mM. Only one

tube was built per horizon and concentration. The invading

solution was supplied from a Mariotte bottle to the proximal

end of the tube. The position of the wetting front in the soil

was recorded frequently. When the front reached the penulti-

mate section of the tube at time t (see Table 4), the experiment

was terminated and the tubes were sectioned rapidly. The

lengths of the samples ranged from 0.42 to 2.57 cm. Some of

the longest samples were divided into two parts, one for deter-

mining water content and one for analysis of nitrate concen-

tration. The value of �s was taken as the water content at the

proximal end of the tube, and �n, the initial water content, was

found as the water content of the last sample beyond the reach

of the invading solution. The other samples were analysed only

for total nitrate concentration. A 2-g fresh sample of the

sampled soil was extracted with 10ml of a 2MKCl solution,

and the nitrate concentration in the supernatant solution was

measured with an Auto Analyser (Technicon). As the concen-

tration of the KCl greatly exceeded the incident concentration

of KNO3, we can assume that the nitrate ions were totally

desorbed from the soil surfaces. Each total nitrate concentra-

tion was related to the sample position in the soil column (x),

and by calculating the Boltzmann variable for each sample

Table 2 Anion exchange capacity (AEC) obtained from the horizontal

imbibition experiments, compared with the AEC calculated from

Becquer et al. (2001)

Horizon Soil pH

Imbibition experiment

AEC /cmolc kg�1

Calculated

AEC /cmolc kg�1

H1 6.9 > 0.64 0.04

H2 6.9 > 0.59 –

H3 6.7 3.2 –

H4 7.2 4.3 0.14

Table 3 Means of the water contents, �n and �s, and bulk density, �,

and standard errors (in parentheses) for the water imbibition experi-

ments

Horizon �n /m3m�3 �s /m3m�3 � /g cm�3

0–10 cm 0.16 (0.02) 0.67 (0.02) 0.76 (0.02)

0–20 cm 0.13 (0.01) 0.63 (0.03) 0.79 (0.04)

20–40 cm 0.13 (0.01) 0.62 (0.03) 0.81 (0.03)

50–60 cm 0.13 (0.01) 0.60 (0.02) 0.91 (0.02)

508 C. Duwig et al.


using the time t for each experiment (Table 4) we could draw

the M(�B) profiles.

Results

Water movement

Figure 1 shows the water content profiles, as functions of

actual distance, x, for the various input concentrations, and

for H1 and H4 horizons. Tables 3 and 4 list characteristics

related to water movement. The initial and final water content

and the bulk density were kept constant. However, the time t

required for the absorption to be complete varied from one

experiment to the other (Table 4). Thus the values of �*B and

the sorptivity, S, varied. This variation could be explained by a

slight hydrophobicity of the horizons rich in organic matter.

The difficulty of packing the soil homogeneously in the tube

could also induce some variation.

Also shown in Figure 1 is a simulated water content profile,

for the 7.14mM input concentration, computed numerically by

solving Richards’ equation using the model THETA-1D

(Green, 1997). The hydraulic properties derive from the

0

0.2

0.4

0.6

0 5 10 15 20 25

Distance /cm

H4: horizon 50–60 cm

x *

0

0.2

0.4

0.6

0 5 10 15 20 25

71.4

35.7

7.14

3.57

0.71

0.36H1: horizon 0–10 cm

x *

Vol

umet

ric w

ater

con

tent

/m3 m

–3

Concentrations /mM

Figure 1 Water content, �, plotted against the

distance along the tube, x, for horizons H1 and

H4 and for the different input NO3–-N

concentrations C1 in mM. Simulations (solid

lines) are shown for C1¼ 7.14mM with the

associated plane of separation, x* (vertical

dashed lines).

Table 4 Duration of the experiment, t, sorptivity, S, and the centroid of the invading water front, �*B, for each horizon and input concentration

Concentration

t /s S /mm s�1/2 �*B /mm s�1/2

C1 /mM H1 H2 H3 H4 H1 H2 H3 H4 H1 H2 H3 H4

71.43 3420 1730 1100 2220 1.34 2.26 2.86 1.70 2.072 3.687 4.649 2.827

35.71 1800 1500 1200 1815 2.15 2.25 2.68 1.84 2.384 3.833 4.468 3.145

7.14 3120 1500 1560 2340 1.67 2.56 2.26 1.64 2.522 3.974 3.826 2.925

3.57 3780 2580 1700 3360 1.32 1.79 1.93 1.29 1.981 3.037 3.517 2.259

0.71 2700 1770 2220 3600 1.87 2.02 1.95 1.44 2.787 3.540 3.318 2.458

0.36 4260 1560 1440 3900 1.41 2.46 2.55 1.30 2.074 3.911 3.890 2.248



measured sorptivity of each experiment. This model

reproduces the measured water content profiles.

Nitrate movement

Figures 2–5 show the measured total nitrate concentration

profiles, here in terms of the Boltzmann variable �B, for each

soil horizon. The profiles simulated by Equation (7) are also

shown. We obtained these by optimizing C0 and R, as

described above. After the initial fitting, the values of C0

from all the tubes were averaged for each horizon, and Equa-

tion (7) was fitted again to obtain R. Before the fitting, a first

approximation for C0 was made by inspection of Figures 2–5,

from which it can be seen when the inflowing solution con-

centration differs from the initial concentration. In horizon

H1, the imbibitions with nitrate concentrations equal to or

less than 3.57mM show that the incident solution dilutes the

nitrate initially present in the soil. The initial concentration of

nitrate must then exceed 3.57mM, but be less than 7.14mM

(Figure 2). Similarly, the initial concentration ranges from 0.71

to 3.57mM for the horizon H2, between 0.36 and 0.71mM for

H3, and is less than 0.36mM for H4 (Figures 3–5).

The inferred values of C0 and R are presented on Figures 2–

5. We found a small retardation, with 1.09 and 1.10 for R for

the two top horizons H1 and H2, respectively. This retardation

is constant whatever the input concentrations of NO3–-N. In

the two deeper horizons, the retardation increases with

increasing depth. In the deepest horizon, 50–60 cm, R can be

as large as 2.52 for the input concentration of 0.36mM. These

values of retardation depend on the input concentration,

which indicates a non-linear isotherm. However, in developing

Equation (3), fromwhich our Equation (7) is derived,Watson&

Jones (1981) assumed a linear sorption isotherm. Application

of Equation (7) is therefore subject to the assumption that the

nitrate sorption isotherm is piece-wise linear for each input

concentration used.

To describe the whole isotherm over the range of input

concentrations used, we plotted the points on a double loga-

rithmic graph for horizons H3 and H4 (Figure 6). The equilib-

rium and adsorbed concentrations were calculated with

Equation (8). Graphs for the surface horizons are not shown

00 1 2 3 4 5

100

50

C1 = 71.43 mM

00 1 2 3 4 5

10

20C1 = 35.71 mM

00 1 2 3 4 5

5

10C1 = 7.14 mM

00 1 2 3 4 5

5

6

3

C1 = 3.57 mM

00 1 2 3 4

6

3

C1 = 0.71 mM

00 1 2 3 4 5

6

3

C1 = 0.36 mM

Tot

al c

once

ntra

tion

of N

O3-

N /m

mol

kg–1

Boltzmann variable /m s–1/2

C

Figure 2 Measured and simulated (solid lines) total NO3–-N concentration, M, against the Boltzmann variable, �B, in the H1 horizon for the input

nitrate concentrations C1. The vertical dashed lines indicate the location of the invading water front �B*. R ¼1.09 and C0¼ 6.57mM.

510 C. Duwig et al.


as the isotherm was found to be linear. For the two deeper

horizons, the isotherms exhibit similar behaviour, with

two linear isotherms apparently separated by a plateau. The

isotherm indicates a range of sites with strong affinity for

NO3–-N at concentrations <14.29mM for horizon H3, and

about 1.43mM for H4. For H4, this relation is indistinct,

and more experiments are needed for the very small concen-

trations (< 0.36mM). A second set of sites with weak affinity

for nitrate appears to start at solution concentrations

>14.29mM for the horizon H3, and at 1.43mM for H4.

The sum of the sorption sites can be given by two Langmuir

curves:

q ¼ K1 qmax1 Cr

1þ K1 Crþ K2 qmax2 Cr

1þ K2 Cr; ð9Þ

where K1 and K2 are the equilibrium constants (l mol�1), and

qmax1 and qmax2 are the amount of adsorbent (molc kg�1).

The following values were found for H3: K1¼ 3.50� 102

l mol�1 and qmax1¼ 1.43� 10�3molc kg�1 and K2¼ 2.8 l mol�1

and qmax2¼ 3.21� 10�2molc kg�1. For H4, these values are:

K1¼ 1.12� 104 l mol�1 and qmax1¼ 2.86� 10�4molc kg�1 and

K2¼ 5.6 l mol�1 and qmax2¼ 4.29� 10�2molc kg�1 (Figure 6).

Discussion

Methodology

Few studies have been made with dynamic methods to deter-

mine the retardation or adsorption of anions. Determinations

of isotherms are commonly obtained in the laboratory from

batch experiments. Another method for studying sorption

phenomena under dynamic conditions is the column method.

This technique usually involves steady-state water flow, and

the initial saturation of exchange sites with a single anion. The

isotherm is either fitted with a model, or determined independ-

ently (Wong et al., 1990; Bellini et al., 1996). In our experi-

ments with horizontal tubes, in order to be close to field

conditions the soil was not leached previously, so that inflow-

ing nitrate ions had to compete with other anions present in

the soil in its field state. In an Andisol using the same tech-

nique, Katou et al. (1996) found retardation factors for Cl– and

NO3– of 1.2 to 2.3, depending on the input concentration.

These values are similar to those we found.

For each horizon in our study, AEC was evaluated from the

largest adsorbed concentration, qmax, and compared with the

values of AEC obtained in batch experiments by Becquer et al.

00 1 2 3 4 5 6

80

40

C1 = 71.43 mM

00 1 2 3 4 5 6

40

20

C1 = 35.71 mM

00 1 2 3 4 5 6

5

10C1 = 7.14 mM

00 1 2 3 4 5 6

2

4C1 = 3.57 mM

00 1 2 3 4 5 6

4

2

C1 = 0.71 mM

00 1 2 3 4 5 6

4

2

C1 = 0.36 mMTot

al c

once

ntra

tion

of N

O3-

N /m

mol

kg–1



nitrate concentrations C1. The vertical dashed lines indicate the location of the invading water front �B*. R ¼1.10 and C0¼ 3.57mM.



(2001) by the method of Gillman & Sumpter (1986), and

computed according to the models of Wada & Okamura

(1980) (Table 2). Only two horizons are comparable (H1 and

H4), as they were sampled at the same depth. Values for AEC

determined in our experiments are much larger than those

obtained by the Gillman & Sumpter method. Neither method

displaced all the anions present. The AEC in the batch experi-

ments was determined with an equilibrium concentration of

2mM CaCl2. Chloride ions at fairly small concentrations will

desorb only weakly adsorbed anions, but not anions such as

sulphate or phosphate. The same weak competition might

have occurred in our experiments in which nitrate was intro-

duced at 0.36 to 71.43mM concentrations. The discrepancy

between the two methods could lie therefore (i) in the differ-

ences in the methods used, (ii) in the interactions between

electrolyte and pH of the solution, and (iii) in the interactions

between the solution and the solid phase. As reviewed by, for

example, Barrow (1987), all these conditions could affect the

anion sorption.

Effect of soil horizon

We found adsorption of nitrate by the topsoil to be weaker

than by the subsoil. However, the topsoil and subsoil both

contain large amounts of iron and aluminium (see Table 1).

The two surface horizons contain much organic matter, which

counterbalances the effect of positive charges of the oxides on

nitrate retention. Indeed, organic groups displace water

ligands at the positive-charged sites on oxide surfaces

(Marcano-Martinez & McBride, 1989). This would reduce the

sites available for anion sorption and therefore the AEC (Parfitt,

1992). As shown in Table 2, and in Becquer et al. (2001), the

AEC increases with increasing depth, together with decreasing

organicmatter content. The same relationship betweenAEC and

depth was observed on basaltic soils by Gillman & Sumpter

(1986) and on allophanic and oxidic soils by Parfitt (1992).

The surface complex formation between organic matter and

hydroxide surface sites could also account for the slower

mineralization of carbon and nitrogen and accumulation of

organic matter in these soils than in most Ferralsols.

00 1 2 3 4 5 6

R = 1.15

C1 = 71.43 mM80

40

00 1 2 3 4 5 6

R = 1.17

C1 = 35.71 mM40

20

00 1 2 3 4 5 6

R = 1.35

C1 = 7.14 mM

5

10

0

2

0 1 2 3 4 5 6

R = 1.4

C1 = 3.57 mM

4

00 1 2 3 4 5 6

R = 2.05

C1 = 0.71 mM

1

2

00 1 2 3 4 5 6

R = 1.45

C1 = 0.36 mM

0.5

1.0Tot

al c

once

ntra

tion

of N

O3-

N /m

mol

kg–1



nitrate concentrations C1. The vertical dashed lines indicate the location of the invading water front �B*. C0¼ 0.68mM.

512 C. Duwig et al.


Effect of input concentration

For subsoil samples, we found increasing retardation with

decreasing concentrations. Bellini et al. (1996) and Katou

et al. (1996) had made similar observations. Adsorption iso-

therms generally had a slope less than 1 (Figure 6), which is

typical of a Freundlich-type isotherm. This indicates a log-

normal distribution of exchange sites (Sposito, 1994). A closer

examination of the data indicates the presence of two distinct

types of sites, as observed for cations and oxyanions

(Dzombak & Morel, 1990). The high-energy sites sorb ions at

small concentrations, whereas the low-energy sites are involved

in the sorption phenomena once the high-energy sites are satur-

ated. According to the results obtained with cations and

oxyanions, about 1/50 of the sites are in the high-energy site

class. In this study, qmax1/qmax2¼ 0.04 and 0.007 for the

00 1 2 3 4 5

R = 1.3

40

00 1 2 3 4 5

R = 1.38

4

2

00 1 2 3 4 5

R = 1.9

1.0

0.5

00 1 2 3 4 5

R = 2.52

0.8

0.4

00 1 2 3 4 5

R = 1.3

40

20

C1 = 35.71 mM

00 1 2 3 4 5

R = 1.35

8

4

80

C1 = 7.14 mM

C1 = 71.43 mM C1 = 3.57 mM

C1 = 0.71 mM

C1 = 0.36 mM

Tot

al c

once

ntra

tion

of N

O3-

N /m

mol

kg–1



nitrate concentrations C1. The vertical dashed lines indicate the location of the invading water front �B*. C0¼ 0.36mM.


–1 0 1 2 3

2

1

0

–1


–1 0 1 2 3

2

1

0

–1

Log(equilibrium concentration)

Log(

adso

rbed

con

cent

ratio

n)

Figure 6 Nitrate adsorption isotherm for the two deeper horizons H3 and H4: adsorbed concentration, q (mmolc kg�1), against equilibrium

concentration, Ceq (mM). The solid line indicates the prediction using the two-site form of a Langmuir isotherm.



horizons H3 and H4, respectively, which is of the same order

of magnitude.

The AEC values were small. For H4, which has a specific

surface area of 102.7m2 g�1 (N2 adsorption BET method), the

AEC (taken from Table 2) is equivalent to 0.25 sites nm�2.

However, the horizon comprises 22% of Fe2O3 and 42% of

Al2O3, on a weight basis (Becquer et al., 2001). Typical values

for the site density of Fe2O3 and Al2O3 are in the range of 1 to

20 sites nm�2 (Koretsky et al., 1998), with an average of 2.4

sites nm�2 (Davis & Kent, 1990). Thus our transient experi-

ments have used only about one site in 10. Sites unavailable

for nitrate retention might be complexed to humic and fulvic

acids, or occupied by strongly bound anions such as sulphate

(Hingston et al., 1972).

The compound isotherm can easily be described mathemat-

ically and so can be used in models of nitrate transport. A

critical issue for the use of a reactive transport model would be

to assess the extent to which such model parameters, obtained

independently, can be used to describe behaviour in the field.

Other processes should also be included in the mathematical

description. As Wong et al. (1990) described, a pulse of nitrate

will move through field soil and change the electrolyte concen-

tration and pH as it moves. This needs to be better described

in models for predicting the fate of contaminants.

Consequences for agricultural practices

Our findings have important consequences for agricultural

practices. The topsoil is rich in organic matter, and therefore

adsorbs large amounts of cations. But it has a small AEC and

therefore does not retain anions such as nitrate. On the other

hand, the subsoil can retain large percentages of percolating

nitrate, especially at small concentrations. For example, with

an AEC of 1 cmolc kg�1, half of which is filled with NO3

–-N

and half with other anions, a horizon 1m thick with a bulk

density of 1 g cm�3 would hold 70 kg of exchangeable NO3–-N

ha�1. Our results also show the importance of split applica-

tions of fertilizers, designed to increase nitrate retention in the

deeper horizons, and therefore to reduce the risk of pollution

of the groundwater. Split applications would reduce the nitrate

concentration in the soil solution, so that nitrate would then

sorb on to the high-energy sites. This would reduce the risk of

groundwater contamination at times of heavy rain. More

effective nitrate retention was observed in the deep horizons.

However, for crop roots to reach these deeper horizons, and to

use the stored nitrate, other nutrients and organic matter need

to be introduced there, by deep ploughing, for example. It is

thus important to favour crops with deep roots and effective

nutrient uptake. Previous studies on the same soil showed the

efficacy of deep-rooted crops such as perennial grass in dimin-

ishing nitrate losses from the root zone (Duwig et al., 2000).

Since the surface charge varies with ionic strength of the solu-

tion, factors affecting the composition of the soil solution,

such as fertilizers or irrigation with saline water, will also affect

the retention of anions.

Conclusions

We determined the nitrate adsorption isotherm of a Ferralsol

from New Caledonia using an unsaturated, transient flow.

This method allowed us to determine nitrate sorption and

retardation under more realistic conditions than batch

methods, in particular at much smaller solution:soil ratios.

The values of anion exchange capacity (AEC) obtainedwith this

technique were much larger than those found in batch experi-

ments. Imbibition experiments were performed on four soil

horizons with various input concentrations of nitrate, and

this enabled us to determine nitrate adsorption isotherms. At

small input concentrations, we could estimate the native

nitrate concentration in these soils. Retardation in nitrate

transport is strongly influenced by soil characteristics such as

mineralogy and organic matter content, and the solution ionic

strength. For the two organic matter-rich topsoils, nitrate

adsorption was shown to be small and to follow a linear

isotherm. In contrast, the two subsoils had large anion sorp-

tion capacities and nitrate adsorption followed a two-site sorp-

tion isotherm. This is similar to that found with sorption of

oxyanions and heavy metals on iron oxides, with high-energy

sites at smaller NO3–-N concentrations (< 1.4–14.3mM) and

low-energy sites for larger concentrations. These findings rein-

force the need for appropriate agricultural management, such

as the split application of fertilizers and the use of deep-rooted

crops so as to reduce nitrate concentration in the soil solution

at the base of the root zone.

Acknowledgements

This research was supported by the ‘Contrat de Developpe-

ment entre la Province des ıles et l’IRD pour l’etude des

risques de degradation de la fertilite des sols et de pollution

des lentilles d’eau douce’ and by the French Ministry of Edu-

cation and Research for Operation no 6 DEF. We thank

Massey University soil chemistry laboratory for helping with

nitrate analyses.

References

Barrow, N.J. 1987. Reactions with Variable-Charge Soils.

Martinus-Nijhoff Publishers, Dordrecht.

Becquer, T., Petard, J., Duwig, C., Bourdon, E., Moreau, R. &

Herbillon, A.J. 2001. Mineralogical, chemical and charge properties

of Geric Ferralsols of New Caledonia. Geoderma, 103, 291–306.

Bellini, G., Sumner, M.E., Radcliffe, D.E. & Qafoku, N.P. 1996.

Anion transport through columns of highly weathered acid soil:

adsorption and retardation. Soil Science Society of America Journal,

60, 132–137.

514 C. Duwig et al.


Bond, W.J. 1986. Velocity-dependent hydrodynamic dispersion during

unsteady, unsaturated soil water flow: Experiments. Water

Resources Research, 22, 1881–1889.

Bond, W.J. & Phillips, I.R. 1990. Ion transport during unsteady water

flow in an unsaturated clay soil. Soil Science Society of America

Journal, 54, 636–645.

Burgisser, C.S., Cernik, M., Borkovec, M. & Sticher, H. 1993. Deter-

mination of non linear adsorption isotherms from column experi-

ments: an alternative to batch studies. Environmental Science and

Technology, 27, 943–948.

Charlet, L. & Sposito, G. 1987. Monovalent ion adsorption by an

Oxisol. Soil Science Society of America Journal, 51, 1155–1160.

Clothier, B.E., Sauer, T.J. & Green, S.R. 1988. The movement of

ammonium nitrate into unsaturated soil during unsteady absorp-

tion. Soil Science Society of America Journal, 52, 340–345.

Davis, J. & Kent, D.B. 1990. Surface complexation modeling in aqueous

geochemistry in mineral–water interface geochemistry. In: Reviews in

Mineralogy, Volume 23 (eds M.F. Hochella & A.F. White), pp.

177–260. Mineralogical Society of America, Washington, DC.

Duwig, C., Becquer, T., Clothier, B.E. & Vauclin, M. 1998. Nitrate

leaching through Oxisols of the Loyalty Islands (New Caledonia)

under intensified agricultural practices. Geoderma, 84, 29–43.

Duwig, C., Becquer, T., Clothier, B.E. & Vauclin, M. 1999. A simple

dynamic method to estimate anion retention in an unsaturated soil.

Comptes Rendus de l’Academie des Sciences de Paris, Sciences de la

Terre et des Planetes, 328, 759–764.

Duwig, C., Becquer, T., Vogeler, I., Vauclin, M. & Clothier, B.E. 2000.

Water dynamics and nutrient leaching through a cropped Ferralsol

in the Loyalty Islands (New Caledonia). Journal of Environmental

Quality, 29, 1010–1019.

Dzombak, D.A. & Morel, F.M.M. 1990. Surface Complexation Mod-

elling – Hydrous Ferric Oxide. John Wiley & Sons, New York.

Gillman, G.P. & Sumpter, E.A. 1986. Surface charge characteristics

and lime requirements of soils derived from basaltic, granitic, and

metamorphic rocks in high-rainfall tropical Queensland. Australian

Journal of Soil Research, 24, 173–192.

Green, S.R. 1997. A Users’ Manual for the THETA-ID Model to

Compute the Movement of Water and Chemical into Uniform Soil.

Report No IR97-66, HortResearch, Palmerston North, New Zealand.

Hingston, F.J., Posner, A.M. & Quirk, J.P. 1972. Anion adsorption by

goethite and gibbsite I. The role of the proton in determining

adsorption envelopes. Journal of Soil Science, 23, 177–192.

Hoyoux-Roche, D. & Jamet, P. 1988. Role des conditions experimen-

tales sur l’equilibre d’adsorption d’un pesticide. In: Methodological

Aspects of the Study of Pesticide Behaviour in Soil (ed. P. Jamet), pp.

185–189. INRA, Versailles.

Ishiguro, M., Song, K.-C. & Yuita, K. 1992. Ion transport in an

allophanic Andisol under the influence of variable charge. Soil

Science Society of America Journal, 56, 1789–1793.

Katou, H., Clothier, B.E. & Green, S.R. 1996. Anion transport invol-

ving competitive adsorption during transient water flow in an Andi-

sol. Soil Science Society of America Journal, 60, 1368–1375.

Koretsky, C.M., Sverjensky, D.A. & Sahai, N. 1998. A model of

surface site types of oxide and silicate minerals based on crystal

chemistry: implications for site types and densities, multi-site

adsorption, surface infrared spectroscopy, and dissolution kinetics.

American Journal of Science, 298, 349–438.

Marcano-Martinez, E. & McBride, M.B. 1989. Comparison of the

titration and ion adsorption methods for surface charge measure-

ment in Oxisols. Soil Science Society of America Journal, 53,

1040–1045.

Parfitt, R.L. 1992. Surface charge in some New Zealand soils meas-

ured at typical ionic strength. Australian Journal of Soil Research,

30, 331–341.

Philip, J.R. 1969. Theory of infiltration. Advances in Hydrosciences, 5,

215–296.

Qafoku, N.P. & Sumner, M.E. 2001. Retention and transport of

calcium nitrate in variable charge subsoils. Soil Science, 166,

297–307.

Qafoku, N.P., Sumner, M.E. & West, L.T. 2000. Mineralogy and

chemistry of some variable charge subsoils. Communications in

Soil Science and Plant Analysis, 31, 1051–1070.

Rietra, R.P.J.J., Hiemstra, T. & van Riemsdijk, W.H. 2000. Electrolyte

anion affinity and its effect on oxyanion adsorption on goethite.

Journal of Colloid and Interface Science, 229, 199–206.

Smiles, D.E. & Philip, J.R. 1978. Solute transport during adsorption of

water by soil: laboratory studies and their practical implications.

Soil Science Society of America Journal, 42, 537–544.

Smiles, D.E., Philip, J.R., Knight, J.H. & Elrick, D.E. 1978. Hydro-

dynamic dispersion during absorption of water by soil. Soil Science

Society of America Journal, 42, 229–234.

Sposito, G. 1994. The Surface Chemistry of Soils. Oxford University

Press, New York.

Tandy, J.C., Grimaldi, M., Grimaldi, C. & Tessier, D. 1990. Miner-

alogical and textural changes in French Guyana Oxisols and their

relation with microaggregation. In: Soil Micromorphology: A Basic

and Applied Science (ed. L.A. Douglas), pp. 191–198. Elsevier,

Amsterdam.

Wada, K. & Okamura, Y. 1980. Electric charge characteristics of

Ando A1 and buried A1 horizon soils. Journal of Soil Science, 31,

307–314.

Watson, K.K. & Jones, M.J. 1981. Estimation of hydrodynamic

dispersion in a fine sand using approximate analytical solution.

Australian Journal of Soil Research, 19, 265–273.

Wong, M.T.F., Hughes, R. & Rowell, D.L. 1990. Retarded leaching

of nitrate in acid soils from the tropics. Measurement of the

effective anion exchange capacity. Journal of Soil Science, 41,

655–663.



Evaluation of the WAVE Model for Predicting Nitrate Leaching for Two ContrastedSoil and Climate Conditions

Celine Duwig,* Beatrice Normand, Michel Vauclin, Georges Vachaud, Steven R. Green, Thierry Becquer

ABSTRACT vere problems of pollution resulting from intensive agri-culture. However, the modeling exercise is not a simpleThe integrated soil–crop–atmosphere model Water and Agrochem-

icals in the Vadose Environment (WAVE) (Vanclooster et al., 1995) one because the fate of N is determined by many physi-was evaluated for two contrasted sets of data. One was from the trop- cal, chemical, and biological processes that can varyical climate and ferrallitic soil conditions that exist in Mare Island, tremendously in time and space. Addiscott and Wagenetin New Caledonia. The other was from a glacial terrace under the (1985) noted that the many N models also differ mark-continental climate of La Cote Saint-Andre (Isere) in France. Water and edly, depending on the background and expertise of theNO3 concentrations and fluxes were monitored during three consecu-

developers, as well as the questions and problems thetive years at instrumented sites with different surface covers (maizemodels are trying to solve. Only a few holistic models[Zea mays L.] or bare soil) or amount of applied fertilizer. The com-that describe each process in detail with the same level ofprehensive set of measurements allowed us to evaluate the prediction

capabilities of the WAVE model. Several parameters were determined complexity have been published. These holistic models,independently, while others were adjusted on the basis of simulations while being more complex, still face limitations in termsfor the wettest year at Mare and an average hydrological year at La of parameterization and validation.Cote Saint-Andre. A stepwise approach was used to calibrate these WAVE, developed by Vanclooster et al. (1994), is aparameters by sequentially integrating each individual model compo- model that is both mechanistic and deterministic. Thisnent. A screening sensitivity analysis was performed to address the most

model was initially developed and evaluated under tem-critical parameters. The predictive ability of the model was evaluatedperate climate conditions (Vanclooster et al., 1995; Du-by comparing simulated and measured states variables and water andcheyne and Feyen, 1999; Meiresonne et al., 1999; Du-NO3 fluxes using two different years of data obtained at the samecheyne et al., 2001). Often, models are evaluated onlysites. For both sites, the model gave the best results for wet conditions,

which actually posed the most critical problems in terms of groundwa- by their developers at the site for which the model waster pollution under our specific conditions. However, the model was developed. According to Thorsen et al. (1998), a modelused beyond its capacity as both soils had specificities for which the cannot generally be validated, but must be tested undermodel was not designed. Overall, WAVE gave quite good predictions, all the conditions for which it will be used, that is, forbut further studies are needed to fully evaluate WAVE with its crop different soil, climate, and crop conditions. We chose togrowth model, SUCROS.

evaluate WAVE using two comprehensive sets of datafrom very different field and climate situations. One wasfrom the tropical climate and soil pedological conditionsOne consequence of the dramatic change in the that exist in the Loyalty Islands of New Caledonia. Theagricultural sector in the last several decades hasother was from the continental climate of La Cote Saint-been the intensive use of agrochemicals, which is notAndre (Isere) in France, on a glacial terrace.always in harmony with ecological constraints. Nitrogen

Following a description of the field experiments, weis a key crop nutrient. Any shortage of N results directlygive a brief overview of the WAVE model and describein reduced crop growth and loss of income to farmers.the calibration procedure used to determine suitable val-Hence farmers often increase their use of fertilizer toues for some of the unknown model parameters basedmaximize the growth of their crops. As a consequence,on data from one specific year. The predictive capacitythe mobility of some N compounds in the environmentof WAVE is then evaluated against measurements ac-has become a crucial component to be studied. Anyquired during two other years. The results of a sensitivityleaching of mobile NO3 beyond the root zone can be-analysis of the main parameters are also presented.come an unwanted contaminant in drinking water.

The need for modeling N fate in the soil–crop–atmosphere system is now widely accepted. Many mod- MATERIALS AND METHODSels of different types and for different applications (Ad- Field Experimentsdiscott and Wagenet, 1985; Wagenet and Hutson, 1989;

Two intensive experiments were conducted during threeBrusseau and Rao, 1990; Simunek and Suarez, 1993;consecutive years to study water and N transport in the soilSimunek et al., 1999, among many others) have beenunsaturated zone to establish a good fertilizer managementdeveloped either to improve our understanding of trans-practices that protect local groundwater resources from pollu-port processes in soils or to address the increasingly se-tion. The studies took place between 1991 and 1993 (Kengni,1993; Kengni et al., 1994; Normand, 1996; Normand et al.,C. Duwig, B. Normand, M. Vauclin, and G. Vachaud, Laboratoire

d’etude des Transferts en Hydrologie et Environnement (LTHE) 1997) at La Cote Saint-Andre, near Grenoble, France, and(CNRS, INPG, IRD, UJF), BP 53 X, 38041 Grenoble, France; S.R. between 1995 and 1997 (Duwig, 1998; Duwig et al., 1998, 2000)Green, HortResearch, PB 11-030, Palmerston North, New Zealand; on Mare (Loyalty Islands), New Caledonia.T. Becquer, IRD/Embrapa Cerrados, CP 7091, 71619-970 Brasilia-DF, Brazil. Received 26 Aug. 2002. Original Research Papers. V02-

Abbreviations: DAS, days after sowing; ETP, Penman–Monteith po-0039. *Corresponding author ([email protected]).tential evapotranspiration; GMP, Good Modeling Practice code;WAVE, Water and Agrochemicals in the Vadose Environment model.Published in Vadose Zone Journal 2:76–89 (2003).

76

www.vadosezonejournal.org 77

Table 1. Summary of agricultural practices at the Mare and La Cote Saint-Andre sites.

Number of irrigationSowing Harvest Fertilization Fertilization Rain between sowing applications and Total dry

Crop Year date date amount date and harvest cumulated amounts matter

kg N ha�1 mm Mg ha�1

MareRainfed corn 1995 20 Jan. 20 Apr. 104 11 Jan. 434 – 10

1996 18 Jan. 27 Apr. 104 (52 � 52) 15 Feb. � 14 Mar. 850 – 111997 15 Jan. No harvest 104 (52 � 52) 13 Feb. � 17 Mar. 300† – –

La Cote Saint-AndreIrrigated corn 1991 22 Apr. 05 Oct. 260 22 Apr. 360 6–216.6 mm 24

1992 23 Apr. 08 Oct. 160 (50 � 110) 23 Apr. � 16 June 516 4–106.4 mm 281993 20 Apr. 20 Oct. 180 (22 � 158) 20 Apr. � 09 June 892 3–117.4 mm 24

† Until the 20 Apr. 1997.

evapotranspiration was calculated from the mass conservationTreatment and Measurement Protocolsequation. The NO3–N storage in the root zone was derived

Different treatments were considered: bare soil without and from water contents and NO3 concentrations in the soil solu-with fertilization, and corn crop with fertilization. Each plot tion sampled by the suction cups. Finally, NO3–N leachingon Mare was 400 m2 and on La Cote Saint-Andre was 5000 from the root zone was calculated by multiplying the drainagem2. Ammonium-nitrate in granular form was used as N fertil- rate by the NO3 concentrations measured at the consideredizer (61% NH4–N and 39% NO3–N on Mare, and 50 and 50%, depth, assuming pure convective transport. Details of the vari-respectively, on La Cote Saint-Andre). Corn was rainfed at ous measurements and calculations can be found in severalMare (cv. Hycorn 90) and irrigated at La Cote Saint-Andre previous papers (Kengni et al., 1994; Normand et al., 1997;(cv. Furio). Table 1 presents a summary of the agriculture Duwig et al., 1998, 2000).practices at each site.

For both studies, similar measurement protocols were fol-lowed, which focused on obtaining estimates of water and SoilsNO3 fluxes under bare soil and corn. The various terms of

The soil at Mare is an oxidic ferrallitic soil (Anionic Acru-the water balance and N cycle were obtained from intensivedox Oxisol), which primarily comprises Al and Fe oxides. Soilmonitoring of the root zone of the crop. At La Cote Saint-depths range from 0 to 1 m, with an average of 0.4 m acrossAndre, a set of five tensiometers (at 15, 30, 50, 70, and 90 cmthe experimental field (1 ha).depth), six replicates of suction cups (30, 50, and 80 cm), and

The soil at La Cote Saint-Andre is a heterogeneous, shal-a neutron moisture meter (measurements every 10 cm fromlow, stony and sandy soil (Alfisol), representative of the glacial10 to 90 cm depth) were established in the soil to measurealluvial plain of La Bievre, France. Its chemical and physicalwater pressures, NO3 concentrations of the soil solution, andproperties have been described in other studies (Kengni, 1993;water contents, respectively. At Mare, two replicates of tensi-Kengni et al., 1994; Angulo-Jaramillo et al., 1997; Netto et al.,ometers (10, 20, 30, 40, and 50 cm) and six replicates of suction1999; Sauboua, 2001). The soil upper layer (0–30 cm) is acups (10 and 40 cm) were utilized as well, whereas two repli-loamy sand that consists of approximately 40% coarse materialcates of horizontal time domain reflectometry probes (at 10,and is reasonably rich in organic matter (2–3%). An important20, 30, and 40 cm) were employed for water content measure-point is the increase in percentage and size of gravel andment. Instrumentation for the water balance study was placedstones with depth. As a result, the effective root zone is notin 1 m2 of each plot, whereas suction cup measurements weremuch deeper than 0.8 m (Kengni et al., 1994). The main charac-taken over the entire agricultural field. Climate variables wereteristics of both soils are presented in Table 2.recorded at both sites. Some characteristics of the soil such

as the hydraulic conductivity, solute dispersivity parameters,and N transformation rates were determined in situ or via Climatecomplementary laboratory experiments. Other parameters

Mare island is situated at 21�30� South and 168�3� East inwere either taken from literature reviews or obtained by modelthe South Pacific. The climate is semitropical with a hot, wetcalibration. Because the time series of the hydraulic headseason between December and March, and a dry season be-gradient and the water content were measured with high tem-tween June and September. The average annual rainfall isporal resolution, drainage from the rooting zone could be

inferred from the measured data using Darcy’s Law. Actual 1641 mm. The average annual Penman–Monteith potential

Table 2. Selected properties of soils at the La Cote Saint-Andre and Mare sites.

Soil granulometry, %Organic

Site Horizon Coarse† A Lf Lg Sf Sg matter N C/N

cm % %Mare 0–15 0.0 35.9 36.2 4.3 8.9 1.9 13.1 6.06 13.0

15–30 0.0 35.9 29.9 6.9 18.5 3.1 5.7 4.19 12.130–50 0.0 46.8 32.9 5.8 9.3 2.3 3.3 1.36 12.2

La Cote 0–30 40.0 17.5 23.3 17.7 16.6 24.9 2.6 1.25 10.3Saint-Andre 30–60 71.6 18.9 22.3 15.3 15.7 27.7 1.6 0.89 9.1

60–90 69.3 13.9 17.7 8.6 13.8 46.1 0.7 0.46 7.4

† Coarse (�2 mm) expressed in percentage of the total (fine fraction � coarse material).‡ In percentage in weight of fine fraction: A � clay (�0.002 mm); Lf � fine silt (0.002–0.005 mm); Lg � coarse silt (0.005–0.02 mm); Sf � fine sand

(0.02–0.2 mm); Sg � coarse sand (0.2–2 mm).

78 VADOSE ZONE J., VOL. 2, FEBRUARY 2003

evapotranspiration (ETP) is 1341 mm, and the monthly tem- the integral of the root water uptake term from the soil surfaceto some depth z less than or equal to the rooting depth Lrperatures vary between 10 and 31�C.

The climate around La Cote Saint-Andre (45�24� North, such that the integral becomes equal to the potential rate. If theintegration over the complete rooting depth was insufficient to5�15� East) is of a continental type. The average annual rainfall

is 907 mm, and the average annual ETP is 869 mm. The months explain Tp, water stress was considered to occur and Ta is setequal to the integral of S(h,z) over the entire rooting depth.of July and August are relatively dry, with a monthly rainfall

rate below 70 mm. During those dry months, irrigation was This concept is written as follows:applied (high pressure gun). Irrigation rates and the time ofapplication were those used by farmers on conventional irriga- Ta � �

z�Lr

0

S(h,z)dz � Tp [5]tion practices of the region; they are given in Table 1. Monthlyaverage temperatures vary between �2 and 26�C.

where Lr is the rooting depth and Tp is the potential croptranspiration (L T�1).

MODELING Tp as well as the potential soil evaporation rate Ep of ahealthy crop are obtained by splitting the potential crop evapo-Model Descriptiontranspiration rate ETcrop (L T�1), using the leaf area index

The process-based WAVE model (Vanclooster et al., 1994) (LAI) as a division parameter:describes the one-dimensional transport and transformations

Ep � exp(�0.6LAI) ETcrop [6]of matter and energy in the soil, crop, and vadose zone envi-ronments. It combines the SWATNIT (Vereecken et al., 1991) Tp � ETcrop � Ep � CanStor [7]and SUCROS models (Van Keulen et al., 1982; Spitters et

CanStor( j) � min[Rainfall � irrigation,al., 1988).While a detailed description of the model was given by (CanStormax � CanStor( j � 1)] [8]

their developers (Vanclooster et al., 1994, 1995), we brieflywhere CanStor is the amount of water that has been inter-review here the different modules, focusing on those relevantcepted and is now released from the crop canopy (L T�1) and jto this study (i.e., water, heat, solute, and N aspects of theis the time step. ETcrop is calculated by multiplying the potentialmodel). Water, heat, and solute mass balance equations areevapotranspiration rate (ETo) of a reference surface with a cropsolved for each soil compartment specified by the user, usingcoefficient Kc. LAI, Kc, ETo, Lr, (h), and Smax(z) are modelfinite difference techniques.input parameters, as is the potential interception capacity Can-Stormax (L T�1).Water Flow

A crop growth module, SUCROS, is available within WAVE,Water movement in the unsaturated soil is modeled using and, if it is used, LAI and Lr are calculated by the model

the Richards equation: depending on photosynthesis and water and nutrient availabil-ity. However, we did not use the SUCROS module, since ourfield experiments did not provide us the necessary parameters,�

�h�h�t

��

�z�K() ��h�z

� 1�� S [1]and as such would have to rely on literature values for Dutchconditions (Vanclooster et al., 1994).where h is the soil pressure head (L), is volumetric water

content (L3 L�3), z is depth (L) defined as positive downwards,Heat Transportt is time (T), and S (T�1) is a sink term accounting for the crop

water uptake. This formulation requires knowledge of the water Heat transport was modeled with the one-dimensional Fou-retention (h) and hydraulic conductivity K() (L T�1) functions. rier transport equation. The soil thermal properties (conduc-

The van Genuchten (1980) (h) parametric expression is tivity and volumetric heat capacity) were calculated as sug-used in WAVE: gested by de Vries (1952).

(h) � r �s � r

[1 � (d|h|)n]m[2] Solute Transport

Solute transport was described by the convection–dispersionwhere r and s are the residual and saturated water content, equation with a linear reversible adsorption isotherm for reac-respectively; d (L�1), m, and n are fitting parameters, and tive solutes:where the Burdine (1953) condition (m � 1 � 2/n) is used.

We considered the Brooks and Corey (1964) K() expres- �(Cr)�t

� �Kd�Cr

�t�

�

�z�Ds�Cr

�z � ��(qCr)

�z �i

�i [9]sion among those available to model users:

where Cr is the resident concentration (M L�3) in the soil so-K() � Ks � � r

s � r�

2

�3

[3] lution, � the soil dry bulk density (M L�3), Kd the solute dis-tribution coefficient (L3 M�1), Ds the apparent dispersion co-

where Ks is the saturated hydraulic conductivity (L T�1) and efficient (L2 T�1), q the Darcian water flux (L T�1), and �i�i is a shape factor. All of the above parameters are input data. a solute sink term (M L�3 T�1) that includes crop uptake and

The crop water uptake rate S was modeled using the macro- transformations. Here we are mainly interested in nitrate andscopic sink model proposed by Hoogland et al. (1980). S is ammonium transport. Transformations refer to the N cycle.calculated from a maximum root water uptake rate as a func- Nitrogen uptake by the crop was described using a macro-tion of depth Smax(z) (T�1), and a dimensionless reduction scopic model, and the uptake rate was restricted to a potentialfunction (h) that accounts for water stress by reducing the level. The potential uptake rate depends on a maximum N up-maximum extraction rate according to: take rate, Nmax (M L�2), specified by the user, and is separated

into a convective and diffusive fraction. The convective frac-S(h,z) � (h)Smax(z) [4]tion is a function of water uptake and the total concentration ofnitrate and ammonium. The diffusive fraction, calculated onlyThe actual transpiration rate, Ta (L T�1), was calculated as


Table 3a. Summary of soil hydraulic and other parameters used in the WAVE model for the corn plot.

Site Layers � �r �s �d n Ks � �

cm Mg m�3 m3 m�3 cm�1 cm d�1 (mm)Mare 0–14 0.68 0.02 0.65 0.115 2.16 720 0.16 30

14–34 0.77 0.02 0.65 0.115 2.16 700 0.20 3034–52 0.88 0.02 0.65 0.115 2.16 600 0.20 3052–152† 0.80 0.0 0.03 0.030 2.50 90 0.50 30

La Cote 0–40 1.38 0 0.30 0.16† 2.12† 35† 0.12 100Saint-Andre 40–60 1.34 0 0.30 0.05 2.17 12 0.17 100

60–75 1.28 0 0.33 0.04 2.27 15 0.27 10075–150 1.30 0 0.33 0.15 2.22 15 0.22 100

† Calibrated value.

if the convective uptake rate is smaller than the potential level, thickness and grouped into three and four pedological soillayers for the Mare and La Cote Saint-Andre sites, respec-is a function of the root density depth profile RDENS (z),

the mean root radius RORAD (L), and the average distance tively, each having constant physical, chemical, and biologicalparameters (Table 2). We were interested in the quantities ofbetween the soil solution and the root surface D0 (L).

The apparent dispersion coefficient Ds is calculated as: water and NO3 leaving the base of the root zone at 40 cmdepth in Mare and at 80 cm depth in La Cote Saint-Andre.

Ds � �q

� De [10]Boundary Conditions

where � is the dispersivity (L) and De is the effective diffusion At Mare, a free drainage boundary condition at the verycoefficient (L2 T�1) given by: bottom of the soil profile was imposed by adding a fourth

layer going down to 1.50 m. This deeper layer, which mimickedDe �

Do aexp(b)

[11] the underlying coral rock below the root zone, was describedusing hypothetical flow and transport properties. At La CoteSaint-Andre, the fourth existing layer was extended down toin which Do is the molecular diffusion coefficient of the consid-1.50 m. The upper boundary condition was chosen to be aered solute in pure water (L2 T�1) and a and b are empiri-flux type with no ponding at the soil surface since the saturatedcal constants.hydraulic conductivity at both sites was higher than the rainfallWhen solute is applied at the soil surface (during a fertiliza-and irrigation intensities.tion or irrigation event), it is assumed to dissolve instantane-

The model used a variable time step, smaller than 1 d forously in the mass of water entering the profile during the daystrongly dynamic processes, such as the flow and transportof solute application (or the first day when infiltration occurs).processes and the solute transformations. Model input wasspecified on a daily basis, while the boundary conditions wereNitrogen Cycleassumed constant during a given day. This means, for example,The mineral N transformations (i.e., nitrification, denitrifi- that the daily precipitation is distributed equally over a day.cation, and volatilization) are described by means of first-

order kinetics. The corresponding rate constants (Knit, Kdenit,Water Flowand Kvol, respectively) are functions of soil temperature and

water content as proposed by Johnsson et al. (1987) and Ver- Water flow was first simulated for bare soil plots whereeecken et al. (1990). Mineralization of the N from the organic only the hydrodynamic characteristics of each soil layer andmatter is assumed to occur from three distinguishable soil the potential evaporation rate had to be estimated. For culti-organic matter fractions: litter, manure, and humus. The N vated plots, more processes had to be considered: plant tran-demand for the internal cycling of C and N in the three pools spiration, function of leaf area development, water uptake byis regulated by a constant C/N ratio identical for the soil plant roots, and rainfall interception by the crop canopy.biomass and the metabolization products. These three organic Hence, a number of parameters had to be determined bypools are characterized by degradation constants (T�1) called measurement or literature review.Klit, Kman, and Khum respectively, which are also functions of The soil water retention curves were determined using thesoil temperature and water content. The turnover efficiency, coupled measured water content and pressure heads at thefe, determines which fraction is decomposed into CO2, with same times and depths. The hydraulic conductivity was deter-the remainder being assimilated into another organic form. mined using the zero-flux plane method (Vachaud et al., 1978),The humification constant fh determines which fraction of the and a tension disk infiltrometer (Ankeny et al., 1991; Angulo-effectively turned-over C transfers to the humus pool. Jaramillo et al., 1996) for values near saturation. Values of

the fitting parameters of Eq. [2] and [3] are given in Table 3aInitial Conditions (note that of Eq. [3] was set equal to n � 2, except for the

The user has to specify an initial water content or pressureTable 3b. Summary of some of the WAVE parameters under cornhead value, as well as the initial concentration of solutes in

for the nitrogen cycle.each soil compartment. The initial concentrations of C and Nin the different organic matter pools must also be entered by Site Horizon Knit Kdenit Khum Klitthe user.

cm d�1

Mare 0–10 0.5 0.02 7 � 10�5 8 � 10�3

Model Parameters and Model Forcings 10–30 0.5 0 7 � 10�6 8 � 10�4

30–52 0 0 0 0Soil Profiles

La Cote 0–40 0.5 0 10�5 8 � 10�3

Saint-Andre 40–85 0.5 0 7 � 10�8 10�4The soil profile was numerically discretized into compart-85–150 0 0 0 0ments of 2 cm (in Mare) and 5 cm (in La Cote Saint-Andre)


Solute Movement

The diffusion coefficient Do and the empirical constants aand b in Eq. [11] were fixed at the following values: Do �120 mm2 d�1, a � 0.001 and b � 1.0 (Wagenet and Hutson,1989). The dispersivity � was set at 100 mm for La Cote Saint-Andre, estimated from breakthrough curves of inert solutesmeasured in a large undisturbed lysimeter near the experimen-tal plot (Schoen et al., 1999), and at 30 mm for the Mare siteafter calibration.

The NO3 was assumed to be more or less inert for the LaCote Saint-Andre soil (Kd � 10�9 m3 kg�1), but shown to bereactive for the Marean soil (Kd � 10�4 at the surface and4 � 10�4 m3 kg�1 for the deeper layers) as determined on soilcolumns (Duwig et al., 1999; Duwig et al., 2003). Ammoniumwas assumed to be inert at Mare but reactive at La Cote Saint-Andre (Kd � 1.5 � 10�3 m3 kg�1, Vereecken et al., 1991).

Nitrogen Cycle and Nitrate Uptake by PlantsFig. 1. Evolution in time (days after sowing, DAS) of LAI, Kc, and

Some of the N turnover parameters were selected fromrooting depth at both sites.literature values given in the user’s manual of WAVE. TheC turnover efficiency fe was fixed at 0.6 for Mare, as advisedsubsoil layer at Mare). Those two functions were determinedby McGill et al. (1981) for a well-aerated soil. For La Coteon bare soil and corn plots independently.Saint-Andre, fe was set equal to 0.3, which is the middle pointThe potential evapotranspiration ETo was calculated usingof the range [0.05; 0.6] advised by Vanclooster et al. (1994).the Penman–Monteith equation with daily values of wind speed,The humification constant fh is usually fixed at 0.2 for rapidsolar radiation, relative air humidity, and temperature mea-recycling (Johnsson et al., 1987 in Vanclooster et al., 1994).sured at the field sites. Kc values were taken from DoorenbosThe total soil organic matter was distributed across the litterand Pruitt (1977) with the dates of the different corn growth (5% for Mare and 2% for La Cote Saint-Andre, C/N ratio �stages as observed in the field (Fig. 1). LAI were measured 30) and the humus pool (C/N ratio � 10). The C/N ratio ofat La Cote Saint-Andre (Normand, 1996) and evaluated from the metabolized products and the soil biomass was measuredvalues given in Eik and Hanway (1966) for Mare (Fig. 1). to be 11 for Mare. For La Cote Saint-Andre, it was calculatedSince no values of CanStormax for corn were found in the lit- as the average of the C/N value for biomass in arable soils

erature, a value of 3 mm d�1 for small trees, as given by Rutter given in Bradbury et al. (1993) (C/N � 6.7) and the measuredand Morton (1977), was used. value of the organic matter at the soil surface (C/N � 10.3

The relationship between the water stress reduction factor for La Cote Saint-Andre; see Table 2).for root water uptake and the pressure head was considered The decomposition rate of the humus pool Khum was initiallyhyperbolic. The critical wet-end pressure head above which set at 7 � 10�5 d�1 for the entire soil profiles of both sites,water uptake is reduced was set equal to 100 cm, while the dry- but later calibrated since the simulated NO3 leaching wasend pressure head values were set equal to 1000 and 500 cm, for overestimated. The best fit for Mare was found to be 7 �the Mare and La Cote Saint-Andre sites, respectively. The 10�5 d�1 between 0 and 15 cm and 7 � 10�6 d�1 below 15 cm.maximum root water uptake function Smax(z) was determined These values are relatively small compared with those givenusing root length distribution observed in the field and taken by Desjardins et al. (1994) for an Ultisol in Brazil. However,from a literature review (Novak, 1987; Vanclooster et al., the biodegradation of organic matter decreases significantly1994). Smax(z) was set to 0.02 d�1 at the surface of the Mare site, when it is complexed by Al and Fe oxides (Boudot et al.,and assumed to decrease linearly to 0 at 40 cm depth. For La 1989). For La Cote Saint-Andre, Khum was 10�5 d�1 betweenCote Saint-Andre, Smax(z) values were 0.01 d�1 in the upper 0 and 30 cm and 7 �0�8 d�1 between 30 and 75 cm. Thelayer (0–20 cm) and 0.001 d�1 from 20 to 75 cm depth. decomposition rate of the litter pool Klit was calibrated to 8 �

10�3 d�1 in the surface layer and to a smaller value in thedeeper layer (8 � 10�4 d�1 for Mare and 10�4 d�1 for La CoteHeat TransportSaint-Andre). No manure was applied at either site.

As mentioned above, parameters to describe heat transport, Nitrification of mineral N at both sites was described by asuch as soil specific heat capacity and soil thermal conductivity, nitrification constant of Knit � 0.5 d�1 since ammonium waswere calculated within the model depending on soil bulk dry found to disappear rapidly from the soil solution. The denitrifi-density (� values given Table 2). The model assumes that the cation constant Kdenit was taken to be 0.02 d�1 for Mare, assoil surface temperature is equal to air temperature at 2 m given by Jabro et al. (1995). The denitrification activity of theabove the soil surface, which is calculated from the minimum soil from La Cote Saint-Andre was studied in the laboratoryand maximum temperature as specified by the user, while the and found to be relatively low compared with others soils,temperature of the bottom boundary condition was fixed at from 0.07 to 0.13 �g N g�1 h�1. Tables 3a and 3b summarize7�C. Since there were no soil temperature measurements for all the values of the parameters that were used in the modelthe Mare site, these conditions were kept constant. For La for the corn plot.Cote Saint-Andre, available measurements of air and soil sur- Considering N content measured in mature plants, the maxi-face temperature showed a 4�C difference. Thus the upper mum N root uptake (Nmax) was fixed at 70 kg N ha�1 for Mareboundary condition was defined as measured air temperature and 300 kg N ha�1 for La Cote Saint-Andre. Information aboutplus 4�C. For the bottom boundary condition, the soil profile root density were found in Durieux et al. (1994); densities atbeing extended to 1.50 m, the temperature was fixed at a the soil surface were fixed at 3000 and 3400 cm m�1 for Mare

and La Cote Saint-Andre, respectively. The model constrainsreasonable value of 7�C.


the root density to decrease exponentially with depth, and rates at the bottom and actual evapotranspiration at thethe corresponding decay coefficient was fixed at 0.0016 and top), but to look also at internal state variables, such0.002 mm�1 for Mare and La Cote Saint-Andre, respectively. as soil water contents, water pressure heads, and NO3D0 was set to 0.10 mm for Mare as suggested in the user’s concentrations at selected depths.manual of WAVE. For La Cote Saint-Andre, a much smallervalue of 0.001 provided the best model fit.

Water FlowAt a depth corresponding to the base of the corn rootModeling Approach

zone (40 cm at Mare and 80 cm at La Cote Saint-Andre),The model was calibrated step by step because of the inter-simulated and measured state variables ( and h for theactions among many of the N fate and transport processesbare soil with fertilizer and the corn plots) and fluxesinvolved (Normand, 1996). Nitrogen transformation parame-(drainage and actual evapotranspiration rates for theters were calibrated using bare-soil plot data without fertilizerbare and corn plots) for both the calibration and theamendment. Addition of fertilizer to the bare soil plot allowed

us to calibrate the convective and dispersive parameters for prediction periods are presented in Fig. 2a (Mare) and 2bsolute transport. Below we present only results for the bare (La Cote Saint-Andre). The statistical values (Eq. [12])soil plots with fertilizer. Finally, data from the cultivated plots are given in Table 4.were used to determine the remaining parameters related toN uptake by plant roots. This calibration procedure relies on Step 1: Model Calibrationthe assumption that the transport and transformation parame-ters do not vary from plot to plot. A trial and error method was For Mare, the only parameters related to water flowadopted, using statistical and graphical criteria for evaluating that were not measured were the hydraulic functions ofmodel performance. An intermediate sensitivity analysis was the coral rock layer below the root zone (between 50 andperformed to address the critical parameters in the system. 150 cm). A sensitivity analysis (Duwig, 1998) revealedThis calibration procedure was followed using 1992 mea- that these parameters did not have a large influence onsurements for La Cote Saint-Andre since the annual rainfall

cumulative drainage model output at 40 cm depth.was close to the average rainfall from 1952 to 1998 (953 vs.For La Cote Saint-Andre, water retention and hy-907 mm). For Mare, we used 1996 for calibration since that

draulic conductivity parameters were calibrated, sinceyear had the most data. The prediction capability of the modelthe experimental curves did not give much information atwas subsequently evaluated on the two remaining years, 1995

and 1997, for Mare, and 1991 and 1993 for La Cote Saint- the lower water contents. The calibration was performedAndre. in a trial and error approach defining as objective func-

To express differences between the simulated and observed tions both the measured time series of soil moisture andvalues in terms of statistical quantities, we used the modeling pressure head and the measured cumulative fluxes at theefficiency EF, the root mean square error RMSE, as defined boundaries. Parameters of the top layer were found toby Loague and Green (1991), as well as the bias B, as follows: be the most important (Normand, 1996). Among those,

the scale parameter d (Eq. [2]) was the most sensitiveto annual cumulative drainage at the base of the soil

EF ��n

i�1

�Oi � O�2 � �n

i�1

�Pi � Oi�2

�n

i�1

�Oi � O�2

[12a] profile and evapotranspiration (respectively, �5/�30%and �13/�77% for a change of �80% on the bare soil).Shape parameters (m, n, and ) were sensitive to timeseries of the pressure head, especially to the lowest valuesobtained during the crop season. The saturated hydrau-lic conductivity measured using a monodisk tension in-

RMSE �100O

�n

i�1

�Pi � Oi�2

n[12b] filtrometer (around 1.3 cm d�1 at a saturated water con-

tent of 0.3 cm3 cm�3) produced almost 190 mm of runoff,whereas no ponding water at all was observed at the soilsurface during the experiment. Ks was thus calibrated sothat no runoff occurred (see Table 3).

B �

�n

i�1

�Pi � Oi�

n[12c] For the bare soil plots of both sites, water contents

and drainage at depths corresponding to the base of thewhere Pi and Oi are the model calculated and observed values corn root zone (40 cm at Mare and 80 cm at La Coterespectively, n is the number of samples, and O is the mean Saint-Andre) were well simulated (Fig. 2a and 2b). Forof the observed data. These statistical quantities were calcu- both water content and pressure head, agreement be-lated on unsorted data, observed and predicted values being

tween measurements and simulations was also fair (EFcompared directly. EF, RMSE and B should be as close asvalues larger than 0.4, RMSE values lower than 50%,possible of 1, 0, and 0 respectively. A negative value of EFand B close to 0; see Table 4). However, at La Coteindicates that the model-predicted values are worse than sim-Saint-Andre, simulations were not so good in the topply using the observed mean.layer. The model overestimated water contents at 10cm depth, leading to a negative EF value. MeasurementsRESULTS AND DISCUSSION with a neutron probe are less accurate close to the soilsurface (between 0 and 15 cm depth). At 15 cm depth,Since the WAVE model is a mechanistic model, we

found it important to not only compare fluxes across pressure head showed larger differences than those fordeeper depths, as shown by the larger RMSE value.the boundaries of the soil profile (drainage and leaching


Fig. 2. Components of the soil water balance at (a) Mare and (b) La Cote Saint-Andre: measurements on bare soil with fertilizer (�) and oncorn plots (�); simulation on bare soil (thin line) and on corn plots (thicker line). DAS is days after sowing.

As explained above, some parameters describing Mare were simulated satisfactorily, except at the end ofthe cycle when soil dries out quicker than was simulatedevaporation and plant water uptake on the corn plot

had to be taken from the literature. Water contents on (Fig. 2a). The model stopped all root uptake at the


Table 4. Values of EF, RMSE, and B (Eq. 12) between observed and simulated values of h, �, and C for the bare soil and corn plotsat different depths. B is given in centimeters for the pressure head, in cubic centimeters per cubic centimeter for the water content,and in milligrams NO3–N per liter for the concentrations. n is the number of observations.

Bare soil Corn

EF, RMSE, B 1995‡ 1996† 1997‡ 1995‡ 1996† 1997‡

Mareh (cm) n � 25 n � 59 n � 69 n � 56 n � 56 n � 53

10 cm 0.7, 23, �16.7 0.8, 35, 30.6 0.5, 35, 46.0 0.6, 34, 13.2 0.3, 48, 68.9 �0, 356, �108320 cm 0.5, 23, �18.3 0.8, 29, 4.2 0.4, 30, �17.9 0.2, 44, 48.3 0.5, 41, 85.7 �0, 216, �75330 cm 0.9, 10, �0.2 0.9, 16, 12.7 0.2, 30, 27.4 0.5, 48, 63.3 0.5, 35, 59.8 �0, 208, �76840 cm 0.7, 15, 3.7 0.8, 22, 25.7 0.1, 27, 29.7 0.5, 49, 53.5 0.3, 32, 35.3 �0, 225, �79950 cm 0.0, 19, 27.7 0.8, 21, 22.7 �0, 28, 47.2 0.4, 43, 66.0 0.3, 33, 46.9 �0, 229, �797

� (cm3 cm�3) n � 29 n � 72 n � 68 n � 31 n � 68 n � 6110 cm 0.1, 13, 0.02 0.7, 11, �0.02 �0, 18, �0.05 0.7, 11, 0.02 0.2, 18, 0.04 0.5, 11, 0.0020 cm 0.3, 9, 0.01 0.7, 8, 0.009 �0, 11, �0.03 0.8, 7, �0.01 0.5, 9, �0.01 0.4, 15, 0.0230 cm 0.3, 11, 0.00 0.7, 6, �0.005 �0, 13, �0.05 0.7, 9, 0.008 0.4, 9, 0.02 0.1, 13, 0.0240 cm �0, 11, 0.02 0.4, 5, �0.002 �0, 6, �0.02 0.8, 7, 0.01 0.2, 8, 0.02 �0, 17, 0.05

C (mg N L�1) n � 6 n � 25 – n � 6 n � 25 –10 cm �0, 92, 27.4 �0, 98, 7.3 – �0, 78, �4.8 �0, 78, �7.9 –40 cm �0, 82, 16.8 �0, 64, 4.4 – �0, 75, 9.8 0.4, 57, 1.1 –

La Cote Saint-AndreEF, RMSE, B 1991‡ 1992† 1993‡ 1991‡ 1992† 1993‡h (cm) n � 109 n � 133 n � 158 n � 109 n � 134 n � 156

15 cm 0.8, 34, �4.4 0.5, 71, 19.1 0.4, 95, 22.8 0.4, 78, �28.5 �0, 322, �117 �0, 141, �10330 cm 0.8, 29, 0.5 0.7, 49, �3.4 0.6, 57, �12.1 0.3, 83, 85.0 0.7, 72, �11.4 0.8, 49, 5.250 cm 0.4, 26, 12.3 0.6, 26, 6.2 0.5, 24, �2.8 0.0, 91, 70.8 0.7, 55, 14.8 0.5, 61, 25.270 cm 0.0, 28, 13.1 0.5, 24, 6.5 0.3, 21, �0.7 0.0, 44, 4.1 0.7, 43, 10.0 0.6, 36, 3.390 cm �0, 28, 9.6 0.5, 21, 0.2 0.0, 30, �4.8 �0, 40, 2.9 0.6, 52, 19.5 0.6, 35, 4.4

� (cm3 cm�3) n � 31 n � 54 n � 65 n � 31 n � 51 n � 6210 cm �0, 13, �0.03 �0, 15, 0.03 �0, 37, 0.06 �0, 19, �0.04 �0, 14, 0.01 �0, 15, �0.0330 cm �0, 18, �0.05 0.6, 5, 0.00 0.5, 5, 0.00 �0, 19, �0.05 0.5, 14, �0.01 0.5, 10, �0.0150 cm �0, 7, 0.01 0.4, 4, 0.00 �0, 5, 0.00 �0, 6, �0.01 0.8, 8, 0.00 0.5, 8, �0.0170 cm �0, 7, 0.01 0.5, 5, 0.01 �0, 7, 0.00 �0, 9, �0.01 0.7, 8, 0.00 �0, 15, 0.0380 cm 0.3, 6, �0.01 0.8, 3, 0.00 0.0, 6, �0.01 �0, 16, �0.03 �0, 17, �0.04 0.1, 10, 0.01

C (mg N L�1) n � 28 n � 29 n � 34 n � 29 n � 26 n � 2930 cm 0.5, 62, �22.1 0.4, 54, 2.3 0.1, 84, 10.5 0.5, 52, 14.4 �0, 76, 0.2 0.3, 111, �2750 cm 0.6, 33, �8.7 0.5, 30, 1.9 0.4, 64, 4.6 0.7, 29, 3.8 0.2, 70, 3.2 0.4, 39, �8.580 cm 0.5, 44, �2.9 �0, 49, 3.3 �0, 56, 1.8 �0, 103, 31.1 0.6, 35, 0.5 �0, 56, �4.6

† Model calibration.‡ Model prediction.

harvest date (Table 1). However, the corn plants were tween measurements and simulations was better, as indi-cated by much larger EF values (0.6–0.7), even thoughleft uncut during more than 1 mo after grain harvest.a large deviation persisted (RMSE � 43%).These plants plus the weeds that colonized the plots

A sensitivity analysis of drainage for the unmeasuredmust have continued to consume water. This may ex-parameters Kc, LAI, Smax, and CanStormax for Mare is pre-plain why the EF values were lower and RMSE and Bsented in Fig. 3. The cumulative drainage was not verywere higher than those calculated for the bare soil plotsensitive to the root water uptake parameters. This is(Table 4). Indeed, parameters describing crop water

uptake were determined with less precision than thosedescribing water flow, thus leading to more imprecisesimulations of state variables and cumulated drainage.

For La Cote Saint-Andre, the maximum root wateruptake rate, Smax (Eq. [4]), was calibrated by trial anderror. Both the cumulative drainage at the base of theroot zone and the actual evapotranspiration were accu-rately simulated (Fig. 2b). Nevertheless, internal statevariables were not so well reproduced. In particular,from 100 to 130 days after sowing (DAS), calculatedpressure heads at depths deeper than 30 cm did notreach the very low observed values (�400 cm at 50 cmdepth and �300 cm at 70 cm depth), whereas the modelalso systematically underestimated water content at thebase of the root zone (80 cm), leading to negative EFand B values (Fig. 2b and Table 4). As observed for thebare soil plot, pressure heads at 15 depth were badly Fig. 3. Sensitivity of cumulative drainage (y-axis) toward parametersreproduced, with all three statistical indicators far from defining interception and root extraction of water by corn (x-axis),

in 1996 for Mare.the ideal values. At lower depths, the correlation be-


partly due to the fact that 1996 was a very wet season the EF values shown in Table 4. At 50 and 70 cm depth,the small RMSE and B values showed a strong correla-and water uptake by plant was low compared with other

components of the water balance. The most sensitive tion between measurements and the simulation, but farfrom the one-to-one line. For the corn plot, the modelparameter was Kc, which was also among the most diffi-

cult parameters to determine. Curiously, increasing LAI overestimated drainage during summer, while the evapo-transpiration rate was simulated very well, thus reflect-produced larger drainage rates. In fact, when increasing

the LAI and leaving all other parameters constant, soil ing a poor prediction of the change in soil water storage.At every depth, the EF values were either negative forevaporation decreased more than the added increase of

interception by canopy and plant transpiration. This sit- the water content or very low for the pressure head.Those poor results were expected, since the calibrateduation appears impossible because for a larger LAI,

CanStormax as well as Smax should also be larger and drain- water retention curve determined in 1992 did not fit themeasured sets of pressure heads and water contents inage would then decrease. Similar results were obtained

for La Cote Saint-Andre (results not shown). This sensi- 1991. This may have been due to a temporal change insoil hydraulic properties. This explanation is consistenttivity analysis highlights the uncertainty in model out-

puts when estimating parameters within a range of values. with findings by Angulo-Jaramillo et al. (1997), whoshowed that during the corn growing season the struc-ture of the fine fraction of soils at La Cote Saint-AndreStep 2: Model Predictionchanged from a well-interconnected microporous net-

The calibrated parameters were used to simulate the work to a poorly connected one. Furthermore, due tostate variables and fluxes in 1995 and 1997 (Mare) and tillage operations performed in spring a few weeks be-1991 and 1993 (La Cote Saint-Andre) with the appro- fore sowing, and to winter freezing, all the measurementpriate climatic inputs. devices were removed from the soil in December and

For Mare, the 1995 and 1997 cropping seasons were reinstalled after tillage, but not exactly at the same place.much drier than in 1996 (see Table 1). In 1995 at both The way we used WAVE for prediction purposesplots, the predicted pressure heads (not shown here) faced several difficulties:and water fluxes were very close to the measured values

1. There was no correlation between plant develop-(Fig. 2a), giving good values for the statistical indicatorsment and water and nutrient availability.for the pressure head (Table 4). However, EF values

2. Soil hydraulic properties varied from one year tofor the water content of the bare soil were less satisfying.another, and also spatially since the measurementThis may be explained by spatial variation in the reten-devices had to be removed every year before eachtion curve h() due to the fact that the instruments weretillage.removed at the end of each corn cycle before soil tillage.

A similar change would explain the poor simulation ofthe water content of the bare soil plot in 1997. Indeed, Nitrate MovementEF values were negative. However, notice that RMSE The measured NO3 collected with the solution sam-values are in the same range as those obtained for the plers at the base of the corn root zone (40 cm in Mare,calibration year. This shows that the correlation be- and 80 cm in La Cote Saint-Andre), the total amount oftween measurements and simulation is strong but with NO3 present in the root zone (expressed in kg N ha�1),a systematic bias. Thus, a unique statistical indicator and cumulative mass leaching are compared with themay not to be sufficient to evaluate the model simula- model simulations (Fig. 4a and 4b). Statistical indicatorstions. For the 1997 corn plot, water contents were well were calculated for NO3 concentration at various depthssimulated, whereas pressure heads were severely under- (Table 4). Nitrate storage and leaching were not mea-estimated (pressure heads being negative) by the model. sured in 1997 at Mare because there were very few datesIt did not rain for nearly 1 mo from DAS 30 to DAS when soil solution could be sampled in suction cups60. The soil then became very dry under the corn, and because of very severe drought conditions.the tensiometers stopped functioning properly for pres-sures lower than �700 cm. Furthermore, the parameters

Step 1: Model Calibrationdescribing crop growth were left unchanged from 1996to 1997. In 1997, since the climate was much drier, plants The first step in our calibration was to estimate somemust have been under water stress. of the parameters involved in the N cycle using measure-

For La Cote Saint-Andre (Fig. 2b), the 1993 season ments made on the bare soil plot without fertilization,was wetter than the calibrated one (1992) and the pre- especially those describing the soil production capacitydicted water content and terms of the water balance (nitrification rate constant and humus mineralizationwere once again very close to the measured values for rate; results not shown). The same parameters were thenboth plots. For the corn plot, the annual cumulative used to simulate the concentrations and leaching fromdrainage and evapotranspiration rates were well pre- the fertilized bare soil plot.dicted, except from DAS 15 to DAS 80. Figure 4a shows concentration, storage, and cumula-

The 1991 season was drier (see Table 1). For the bare tive leaching values vs. time for the Mare site in 1996.soil, simulations of evapotranspiration and drainage For the bare soil, all variables were well simulated, ex-were still very close to the observed values, whereas the cept for the concentrations after the second fertilizer

input. However, cumulative leaching was correctly sim-state variables were poorly predicted, as reflected by


Fig. 4. Components of the soil NO3–N balance at (a) Mare and (b) La Cote Saint-Andre: measurements on bare soil with fertilizer (�) and oncorn plots (�); simulation on bare soil (thin line) and on corn plots (thicker line). DAS is days after sowing. Error bars are estimates ofstandard deviations between six replicates of soil NO3–N concentration. Arrows indicate the time of fertilizer inputs.

ulated. The fertilizer was given just 10 d before a cyclone this may not have been the case because of the time-dependent dissolution of solid fertilizer and adsorptionthat produced 238 mm of rain in 4 d. The model assumed

that the fertilizer instantaneously dissolved in the soil of NO3 on soil particles. This adsorption is nonlinear(Duwig et al., 2003) and depends on NO3 concentration,solution and thus infiltrated with the water. In reality,


two features that were not considered by the model. for bare soil, since harvest residues were incorporatedinto the soil. Measured and simulated concentrationsThe measured NO3 peak was thus lower and more dis-

persed. EF values were lower than 0, and RMSE values and NO3 leaching at the base of the root zone are pre-sented on Fig. 4a and 4b for Mare and La Cote Saint-were quite large. The use of suction cups for measuring

soil solution NO3 concentrations leads to very dispersed Andre, respectively.Concentrations peaks for the Mare corn plot (Fig. 4a)data. Furthermore, the discrepancy between the second

simulated and measured peaks also produced poor EF were well predicted, suggesting the dispersivities werereasonable. Plant uptake (as measured by analysis ofvalues.

For the bare soil at La Cote Saint-Andre (Fig. 4b), total N in plant samples; data not shown) and cumulativeleaching were simulated quite well (�10 kg ha�1 differ-calibration was performed to best fit the beginning of the

concentration time series. Sensitive parameters were ence) considering the variability in measurements withsuction cups and in analyses of plant materials. How-initial contents in NH3 and NO3, nitrification, humus

and litter mineralization rates. Those rates drastically ever, simulated mineralization (35 kg N ha�1) and netnitrification (40 kg N ha�1) were somewhat lower fordecreased in the deeper part of the profile (depths

�30 cm), reflecting the fact that microbial activity was the relatively hot and humid climate conditions. Wedecided not to modify parameters related to these trans-not very effective due to a lack of available nutrients

and/or C. However, no parameter set could be found formations since there were already too many unknowns.The statistical indicators at 10 cm depth were quite poor,to accurately fit the data around DAS 60 where part of

the fertilizer input was not observed in the field. From presumably, as explained above, because soil solutionNO3 contents measured with suction cups are highlyDAS 53 to DAS 78, the model overestimated the ob-

served NO3 in the top of the soil profile by nearly 100 variable, especially near the soil surface. However, allthree statistical indicators values at 40 cm depth werekg N ha�1 (Fig. 4b). The apparent disappearance of the

NO3 after fertilization, and the associated slow release acceptable, which is quite satisfying.For the La Cote Saint-Andre corn plots (Fig. 4b), inin early spring, could be interpreted as an artifact caused

by heterogeneous flow in the soil and difficulty produc- the upper part of the profile (0–50 cm), the model under-estimated concentrations between the two fertilizer ap-ing representative soil water samples with the solution

samplers in the heterogeneous flow domain. The re- plications, resulting in less NO3 storage, as shown inFig. 4b. During this period, measured concentrationsmaining time series were quite well reproduced by the

model (positive EF values and small B values), espe- remained almost constant, which may be explained bya balance between crop uptake and production due tocially the soil NO3 content remaining at corn harvest

(DAS 168) and at the end of November (DAS 216) mineralization. Conversely, the decrease in NO3 storageafter the second fertilizer application was well repro-after autumn rainfall. Leaching was only influenced by

the concentration at 80 cm and the water percolation duced. During the growing season, the model generatedhigher plant uptake and higher soil NO3 productionrate, while the cumulative value was simulated quite

accurately. However, time series at 80 cm were less well rates, which were not in phase with the observed rates.On the water balance graph (Fig. 2b), it has already beenpredicted, as reflected in the negative EF values for the

concentration (Table 4). observed that the final cumulative evapotranspirationvalue was well reproduced, whereas the simulated timeTo simulate NO3 behavior in the crop plots, the lastseries did not always match the observed ones. In anystep was to take into account NO3 root uptake. Parame-case, the model accurately reproduced the low amountters describing crop NO3 uptake (both convective andof NO3 remaining in the soil at harvest resulting from adiffusive parts) were calibrated and initial concentra-reduction of NO3 fertilization. As the soil remained baretions of total N and C were modified relative to thoseafter corn harvest, this is an important point to considerwhen trying to limit groundwater pollution. Simulatedconcentrations at 80 cm depth were very close to themeasurements as shown by the statistical indicators inTable 4, whereas leaching (both time evolution and finalcumulative value) also was simulated accurately.

The calibration was completed by performing a sen-sitivity analysis. Figure 5 shows the sensitivity of thecumulative NO3 leaching to parameters defining theturnover of organic matter at Mare. The C turnoverefficiency, fe, was by far the most sensitive parameter.It was calibrated to its maximum possible value of 0.6,but a 30 % decrease in its value increased the leachingby nearly 100%. The degradation rate of the litter pool,Klit, appears also to be a significant parameter. Increas-ing this rate led to a decrease in NO3 leaching. In fact,the release of excess mineral N because of more rapiddegradation of the litter pool led to its immobilization.Fig. 5. Sensitivity of cumulative NO3 leaching (y-axis) toward organic

matter transformation parameters (x-axis), in 1996 for Mare. When Klit is increased by 150%, the immobilization dur-


ing the corn growth season is 140 kg ha�1, whereas it is ized corn, which reached 80% of the fertilized one, ascompared with 50% for the two other years) and largezero when Klit is decreased by 99%. The same trend was

observed for La Cote Saint-Andre. amounts of NO3–N observed in the bare soil (Fig. 4b).To best reproduce those large amounts of NO3–N storedin the soil during the 1991 corn cycle, the initial amountStep 2: Model Predictionof NH3 in the first 50 cm of soil was set equal to 168 kg

For the 1995 Mare NO3 concentration simulations, N ha�1 (instead of 5 kg N ha�1 in 1992). Calibration ofthe 1996 parameters were kept the same, except for the the initial NH4 content in the soil was relatively easylitter pool C/N ratio, which was increased from 20 to using a trial and error approach since it has a direct ef-30, considering that there would have been more plant fect on NO3 content. However, NH3 nitrification rateresidues because this plot was cultivated for the first was probably too high, and we probably also shouldtime in 1995. One can notice that in 1995, NO3 leaching have increased the stable organic N pool (humus andwas more pronounced under corn than under bare soil. litter) to better predict the beginning of the increaseThe initial pool of NO3 was much larger under corn in concentration. The resulting simulated cumulativethan under bare soil because of different previous agri- leaching appeared to be quite good considering the largecultural practices (Duwig et al., 2000). Thus, the initial variability in NO3 concentration observed for that plotconcentration under corn in the model was set larger (error bars in Fig. 4b). Conversely, for the corn plot, athan under bare soil. For both plots, simulated NO3 con- much lower amount of NO3 stored in the soil was mea-centrations at 40 cm and NO3 storage in the root zone sured the same year. Even with a very low initial contentoverestimated the measurements (Fig. 4a), as shown by of both NH3 and NO3, the simulations could not accu-the negative EF values and the large RMSE values. rately fit the measured data at 80 cm depth from DASThere are several explanations for this discrepancy: 60 to 120 where the concentrations remain constant at

a low level. As discussed for Mare, this discrepancy is1. The model is very sensitive to initial N concentra-probably linked to difficulties in estimating the parame-tion, which was not measured in 1995.ters dealing with crop development. Cumulative NO32. The model is even more sensitive to crop parame-leaching was thus overpredicted.ters, like the crop coefficient (Kc) or the decompo-

sition rate of the litter pool (Klit), which were notdetermined in the field, and probably did not ex- CONCLUSIONStrapolate well from a wet to a drier year.

Comprehensive data sets collected during three con-However, cumulative NO3 leaching was well simu- secutive years and for two contrasted field situations

lated. Overestimation of NO3 concentration in the root were used to evaluate the performance of the WAVEzone is thus balanced by other N transformations, which model under very different environmental conditions.leads to a good representation of leaching. The stepwise calibration approach used was found to

For the 1991 and 1993 La Cote Saint-Andre simula- be a valuable tool for evaluating the performance of thetions (Fig. 4b), the 1992 parameters were also kept the different model components. At both sites, the modelsame, except the initial contents in NH3 and NO3. These gave the best results for the wettest years (1992 andcontents had not been measured directly in the field and 1993 for La Cote Saint-Andre, 1996 for Mare), whichhence were adjusted to best fit the beginning of the con- actually posed the biggest problems in terms of ground-centration time series. In 1993, for both plots, EF values water pollution. The soil at both sites being very perme-calculated for the concentrations at 30 and 50 cm depth able, their hydraulic conductivities were found not towere satisfying. For the bare soil, as was observed for the be a very sensitive parameter. Under such wet condi-calibration year, simulated concentrations were over- tions and soil types, fluxes seemed to be mostly a func-estimated immediately after fertilizer application (DAS tion of climate. For dry years, predictions of NO3 con-50). But from mid September (�148 DAS), simulations centrations and fluxes were generally less accurate.were very close to measurements at every depth. Because Parameters related to the N cycle and plant NO3 uptakedrainage was very low during the summer (Fig. 2b), were more difficult to estimate directly in the field, whilethe underestimation had only a small impact on the some had to be either determined through independentsimulated leaching, which appears to be accurately pre- laboratory experiments or estimated from literature re-dicted. On the other hand, the model significantly un- views. Furthermore, the crop growth module SUCROSderestimated the concentrations at all depths of the corn was not used, since we did not have the necessary param-plot when the heavy rainfall events started (�140 DAS). eters. Thus, the prediction capability of WAVE wasThis period was crucial for water percolation, and thus limited due to the lack of interactions between cropthe model largely underestimated NO3 leaching. The growth, climate, fertilization, and soil variables.year before the experiments, large quantities of manure The model was used beyond its capacity in thatwere added. Unfortunately not enough information was WAVE was initially designed for European soil andavailable from the farmer to quantify those applications climatic conditions. The soil physical and chemical char-and initialize the different organic matter pools. Hence, acteristics at Mare are quite unique. The soil is verythe 1991 crop season was characterized by a particularly porous with a low bulk density, and the NO3 retentionlarge contribution of soil N to N plant nutrition (readily varies with pH and concentrations. These features are

not considered in WAVE. The soil at La Cote Saint-observable by the total dry matter yield of the unfertil-


Desjardins, T., F. Andreux, B. Volkoff, and C.C. Cerri. 1994. OrganicAndre is also somewhat unusual. It is very stony (withcarbon and 13C contents in soils and soil size-fraction, and theirporous stones), having hydraulic properties that cannotchanges due to deforestation and pasture installation in easternbe described using the parameterization in WAVE. Con- Amazonia. Geoderma 61:103–118.

sequently, the calibrated parameters should be viewed de Vries, D.A. 1952. The thermal conductivity of soil. Meded, Land-bouwhogeschool, Wageningen.as effective parameters, rather than as representing in

Doorenbos, J., and W.O. Pruitt. 1977. Crop water requirements. Irriga-situ processes, rendering the model semiempirical. Never-tion and drainage paper 24. FAO, Rome, Italy.theless, the calibrated model gave quite good results

Ducheyne, S., and J. Feyen. 1999. A procedure to reduce modeloverall. uncertainty by comparison with field data illustrated on a nitrogenThe evaluation of such a holistic model requires sev- simulation model. p. 457–466. In J. Feyen and K. Wiyo (ed.) Proc.

eral measured sets of time series, to implement the cali- Int. workshop on modeling of transport processes in soils at variousscales in time and space. 24–26 Nov. 1999. Leuven, Belgium.Wagen-bration and then proceed with predictions. Using theingen Pers., Wageningen, The Netherlands.same parameters to simulate several years of data may

Ducheyne, S., N. Schadeck, L. Vanongeval, H. Vandendriessche, andbe inaccurate because of temporal variability. For exam- J. Feyen. 2001. Assessment of the parameters of a mechanistic soil-ple, several parameters evolve from one crop cycle to crop-nitrogen simulation model using historic data of experimentalthe other, such as hydraulic properties that can vary field sites in Belgium. Agric. Water Manage. 51:53–78.

Durieux, R.P., E.J. Kamprath, W.A. Jackson, and R.H. Moll. 1994.with soil plowing or water and nutrient uptake by plantsRoot distribution of corn: The effect of nitrogen fertilization.that depend on water and nutrient availability in the soil.Agron. J. 86:958–962.Furthermore, measurements were not always obtained Duwig, C. 1998. Etude des transferts d’eau et de nitrate dans les sols

exactly at the same place. The devices had to be removed ferrallitiques de Mare (Nouvelle-Caledonie): Risques de pollutioneach year because of plowing or freezing periods (i.e., des lentilles d’eau douce. Ph.D. diss. Joseph Fourier Univ., Greno-

ble, France.at La Cote Saint-Andre). Comparisons between simula-Duwig, C., T. Becquer, L. Charlet, and B.E. Clothier. 2003. Nitratetions and measurements may then be problematic.

retention in a variable charge soil from the Loyalty Islands, NewWAVE was found to be a useful research tool for Caledonia. Eur. J. Soil Sci. In press.better understanding the various processes involved in Duwig, C., T. Becquer, B.E. Clothier, and M. Vauclin. 1998. NitrateNO3 leaching. The model was found to be robust enough leaching through oxisols of the Loyalty Islands (New Caledonia)

under intensified agricultural practices. Geoderma 84:29–43.to work for conditions for which it was not designed.Duwig, C., T. Becquer, B.E. Clothier, and M. Vauclin. 1999. A simpleCalibration of the model, especially the N cycle, could

dynamic method to estimate anion retention in an unsaturated soil.have been facilitated by using data recorded specially for C. R. Acad. Sci., Ser. IIa: Sci. Terre Planetes 328:759–764.this purpose. Also, additional field studies are needed Duwig, C., T. Becquer, I. Vogeler, M. Vauclin, and B.E. Clothier.to fully validate the SUCROS crop growth module of 2000. Water dynamics and nutrient leaching through a cropped

Ferralsol in the Loyalty Islands (New Caledonia). J. Environ.WAVE. Along with the model, Vanclooster et al. (2000)Qual. 29:1010–1019.proposed a code of Good Modeling Practice (GMP)

Eik, K., and J.J. Hanway. 1966. Leaf area in relation to yield of corn.whose main objective is to provide full transparency ofAgron. J. 58:16–18.

all steps used in the modeling process. This GMP should Hoogland, J.C., C. Belmans, and R.A. Feddes. 1980. Root waterbe used to improve both the quality of data sets and uptake model depending on soil water pressure and maximum

extraction rate. Acta Hortic. 119:123–136.the description of NO3 leaching processes in models.Jabro, J.D., J.D. Toth, Z. Dou, R.H. Fox, and D.D. Fritton. 1995.

Evaluation of nitrogen version of LEACHM for predicting nitrateREFERENCES leaching. Soil Sci. 160:209–217.

Johnsson, H., L. Bergstrom, P.E. Jansson, and K. Paustiaen. 1987.Addiscott, T.M., and R.J. Wagenet. 1985. Concepts of solute leachingSimulated nitrogen dynamics and losses in a layered field. Agric.in soils: A review of modelling approaches. J. Soil Sci. 36:411–424.Ecosyst. Environ. 18:333–356.Angulo-Jaramillo, R., J.P. Gaudet, J.L. Thony, and M. Vauclin. 1996.

Kengni, L. 1993. Mesure in-situ des pertes d’eau et d’azote sous cultureMeasurement of hydraulic properties and mobile water content ofde maıs irriguee. Application a la plaine de la Bievre (Isere). Ph.D.a field soil. Soil Sci. Soc. Am. J. 60:710–715.diss. Joseph Fourier Univ., Grenoble, France.Angulo-Jaramillo, R., F. Moreno, B.E. Clothier, J.L. Thony, G. Va-

Kengni, L., G. Vachaud, J.L. Thony, R. Laty, B. Garino, H. Casa-chaud, E. Fernandez-Boy, and J.A. Cayuela. 1997. Seasonal varia-bianca, P. Jame, and R. Viscogliosi. 1994. Field measurements oftion of hydraulic properties of soils measured using a tension diskwater and nitrogen losses under irrigated maize. J. Hydrol. (Am-infiltrometer. Soil Sci. Soc. Am. J. 61:27–32.sterdam) 162:23–46.Ankeny, M.D., M. Ahmed, T. Kaspar, and R. Horton. 1991. Simple

Loague, K., and R.E. Green. 1991. Statistical and graphical methodsfield method for determining unsaturated hydraulic conductivity.for evaluating solute transport models: Overview and application.Soil Sci. Soc. Am. J. 55:467–470.J. Contam. Hydrol. 7:51–73.Boudot, J.-P., A. Bel Hadj Brahim, R. Steiman, and F. Seigle-Murandi.

McGill, W.B., H.W. Hunt, R.G. Woodmansee, and J.O. Reuss. 1981.1989. Biodegradation of synthetic organo-metallic complexes ofPhoenix, a model of the dynamics of carbon and nitrogen in grass-iron and aluminium with selected metal to carbon ratios. Soil Bio-lands soils. Ecol. Bull. 33:49–115.chem. 21:961–966.

Meiresonne, L., N. Nadezhdina, J. Cermak, J. Van Slycken, and R.Bradbury, N.J., A.P. Whitmore, P.B.S. Hart, and D.S. Jenkinson. 1993.Ceulemans. 1999. Transpiration of a poplar stand: Model calibra-Modelling the fate of nitrogen in crop and soil in the years followingtion and validation by soil water and sap flow measurements.application of 15N-labelled fertilizer to winter wheat. J. Agric.p. 406–415. In J. Feyen and K. Wiyo (ed.) Proc. Int. workshop onSci. 121:363–379.modeling of transport processes in soils at various scales in timeBrooks, R.H., and A.T. Corey. 1964. Hydraulic properties of porousand space. 24–26 Nov. 1999. Leuven, Belgium. Wageningen Pers.,media. Hydrology paper 3. Colorado State University, Fort Col-Wageningen, The Netherlands.lins, CO.

Netto, A.M., R.A. Pieritz, and J.P. Gaudet. 1999. Field study on theBrusseau, M.L., and P.S. Rao. 1990. Modeling solute transport inlocal variability of soil water content and solute concentration. J.structured soils: A review. Geoderma 46:169–192.Hydrol. (Amsterdam) 215:23–37.Burdine, N.T. 1953. Relative permeability calculation from size distri-

butions data. Trans. AIME 198:71–78. Normand, B. 1996. Etude experimentale et modelisation du devenir de


l’azote dans le systeme sol–plante–atmosphere. Ph.D. diss. Joseph comparative validation of pesticide leaching models. J. Environ.Qual. 27:1183–1193.Fourier Univ., Grenoble, France.

Vachaud, G., C. Dancette, S. Sonko, and J.L. Thony. 1978. MethodesNormand, B., S. Recous, G. Vachaud, L. Kengni, and B. Garino. 1997.de caracterisation hydrodynamique in situ d’un sol non sature.Nitrogen-15 tracers combined with tensio-neutronic method to esti-Application a deux types de sol du Senegal en vue de la determina-mate the nitrogen balance of irrigated maize. Soil Sci. Soc. Am.tion des termes du bilan hydrique. Ann. Agron. 29:1–36.J. 61:1508–1518.

Vanclooster, M., J.J.T.I. Boesten, M. Trevisan, C.D. Brown, E. Capri,Novak, V. 1987. Estimation of soil-water extraction patterns by roots.O.M. Eklo, B. Gottesburen, V. Gouy, AMA. van der Linden. 2000.Agric. Water Manage. 12:271–278.A European test of pesticide-leaching models: methodology andRutter, A.J., and A.J. Morton. 1977. A predictive model of rainfallmajor recommendations. Agric. Water Manage. 44:1–29.interception in forests. J. Appl. Ecol. 14:567–588.

Vanclooster, M., P. Viaene, J. Diels, and J. Feyen. 1995. A determinis-Sauboua, E. 2001. Modelisation stochastique fonctionnelle du trans- tic evaluation analysis applied to an integrated soil–crop model.fert d’eau et d’azote sous culture de maıs. Application a l’evaluation Ecol. Modell. 81:83–195.de l’impact des pratiques agricoles en plaine de la Bievre. Ph.D. Vanclooster, M., P. Viaene, J. Diels, and K. Christianens. 1994.diss. Joseph Fourier Univ., Grenoble, France. WAVE: A mathematical model for simulating water and agro-

Schoen, R., J.P. Gaudet, and D.E. Elrick. 1999. Modelling of solute chemicals in the soil and vadose environment. Release 2.0. Refer-transport in a large undisturbed lysimeter, during steady-state water ences and user’s manual. Institute for Land and Water Manage-flux. J. Hydrol. (Amsterdam) 215:82–93. ment, Katholieke Universiteit Leuven, Belgium.

Simunek, J., and D.L. Suarez. 1993. The UNSATCHEM_2D code van Genuchten, M.Th. 1980. A closed-form equation for predictingthe hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am.for simulating two-dimensional variability-saturated water flow,J. 44:892–898.heat transport, carbon dioxide transport and solute transport with

van Keulen, H., F.W.T. Penning de Vries, and E.M. Drees. 1982. Amajor ion equilibrium and kinetic chemistry. Version 1.1. Researchsummary model for crop growth. p. 87–98. In F.W.T. Penning deRep. 128. U.S. Salinity Laboratory, USDA, ARS, Riverside, CA.Vries and H.H. van Laar (ed.) Simulation of crop growth and cropSimunek, J., M. Sejna, and M.Th. van Genuchten. 1999. The HY-production. PUDOC, Wageningen, The Netherlands.DRUS-2D Software package or simulating the one-dimensional

Vereecken, H., M. Vanclooster, and M. Swerts. 1990. A model formovement of water, heat and multiple solutes in variably-saturatedthe estimation of nitrogen leaching with regional applicability. p.media. IGWMC-TPS 53, Version 2.0. Int. Ground Water Modeling 250–263. In R. Merckx et al. (ed.) Fertilizer and the environment.

Center, Colorado School of Mines, Golden, CO. Leuven Academic Press, Belgium.Spitters, C.J.T., H. van Keulen, and D.W.G. van Kraailingen. 1988. Vereecken, H., M. Vanclooster, M. Swerts, and J. Diels. 1991. Simulat-

A simple but universal crop growth simulation model, SURCROS87. ing water and nitrogen behavior in soil cropped with winter wheat.In R. Rabbinge et al. (ed.) Simulation and systems management Fert. Res. 27:233–243.in crop protection. PUDOC, Wageningen, The Netherlands. Wagenet, R.J., and J.L. Hutson. 1989. Leaching estimation and chem-

Thorsen, M., P.R. Jorgensen, G. Felding, O.H. Jacobsen, N.H. Spliid, istry model (LEACHM). Version 2. Department of Agronomy,Cornell University, Ithaca, NY.and J.C. Refsgaard. 1998. Evaluation of a stepwise procedure for

POSTER PAPER

Nitrate Sorption in a Mexican AllophanicAndisol using Intact and Packed Columns

Blanca Prado and Celine Duwig

Institut de Recherche pour le Developpment, c/o Colegio de

Postgraduados, Laboratorio de Fertilidad de Suelo, Montecillo, Mexico

Mauricio Escudey

Departamento de Quımica de los Materiales, Facultad de Quımica y

Biologıa, Universidad de Santiago de Chile, Santiago, Chile

Michel Esteves

Institut de Recherche pour le Developpment, Laboratoire d’etude des

Transferts en Hydrologie et Environnement, Grenoble, France

Abstract: Contamination of groundwater by nitrate is a worldwide environmental

issue. A better knowledge of nitrate sorption characteristics by soils contributes to

efficient fertilizer use and prevents aquifer contamination. In volcanic soils, nitrate

sorption is induced by variable charges due to the presence of amorphous materials

and aluminum (Al) and iron (Fe) oxides. Anion transport in packed and intact

columns was investigated in a Mexican Allophanic Andisol, under different

permanent flow regimes in unsaturated conditions and several NO32-N and Br2 input

concentrations. In the packed columns, the NO32-N adsorption in the soil was

nonlinear. In the intact columns, the retardation coefficient variation was directly cor-

related to the increase of amorphous material with depth. The presence of preferential

flow in the intact columns significantly increased the mobility and velocity of nitrate

moving through the columns, whereas in the packed columns, NO32-N fate was only

affected by soil chemical composition and mineralogy.

Keywords: Allophane, physical nonequilibrium, retardation, volcanic soil

Received 28 January 2005, Accepted 7 May 2005

Address correspondence to Celine Duwig, Institut de Recherche pour le Developp-

ment, c/o Colegio de Postgraduados, Laboratorio de Fertilidad de Suelo, CP 56230,

Montecillo, Mexico. E-mail: [email protected]

Communications in Soil Science and Plant Analysis, 37: 2911–2925, 2006

Copyright # Taylor & Francis Group, LLC

ISSN 0010-3624 print/1532-2416 online

DOI: 10.1080/00103620600833017

2911

INTRODUCTION

Ion-exchange reactions in soils are processes dependant on soil properties,

which are influenced by local conditions such as ionic strength, soil–

solution concentration, pH, and water content of soil. These processes

govern the fate of dissolved nutrients and pesticides applied at the soil

surface in crop fields. The study of anions such as nitrate, which is both a

nutrient and a potential pollutant, and their transport in soil has intensified

in recent years because of increasing nitrate concentrations in drinking

water resources all over the world (Goodrich, Lykins, and Clark 1991;

Spalding and Exner 1993). Nitrogen (N) is an important plant nutrient, and

NO32-N is the most plant-available form. Nitrate naturally present in the

soil or applied as fertilizer is considered inert in most aerobics soils and

thus subject to leaching through the vadose zone.

In variable-charge volcanic soils, anion transport is retarded because of

electrostatic adsorption (Ishiguro, Song, and Yuita 1992), a fact that could

decrease risk of subsoil nitrate contamination. However, Katou, Clothier,

and Green (1996) found that nitrate-leaching depth in an Andisol increased

in the presence of Cl2, because of competitive adsorption. Similarly,

Ishiguro and Shoji (2002) studied the B horizon of a volcanic soil, showing

that nitrate retention time was reduced when sulfate was previously

adsorbed on the soil. Schalscha, Pratt, and Domecq (1973) demonstrated

that nitrate adsorption on Chilean volcanic soils decreased when the pH of

soil solution increased; additionally, the adsorption rate was lower in

surface horizons than subsoil. These results can be explained by the

increased amorphous material and the decreased organic matter with

increased depth. Kinjo and Pratt (1971a), in their study on nitrate adsorption

in acid soil from Mexico and South America, found a significant correlation

between the amount of nitrate adsorbed and the soil content of amorphous

material. At acid soil pH and low anion input concentrations, Qafoku,

Summer, and Radcliffe (2000) reported that nitrate adsorption was directly

related to nitrate concentration in the soil solution. A better understanding

of the adsorption processes during ion displacement in soil will contribute

to more efficient management of fertilizer applications to crops, thus

reducing loss of fertilizers in groundwater and increasing economic and

environmental benefit (Kinjo and Pratt 1971a).

The use of miscible-displacement or flow-through techniques in soil

columns allows the analysis of processes occurring between the liquid phase

and soil matrix. Batch technique is commonly used in adsorption studies.

However, the flow-through technique represents conditions closer to field,

including the soil–solution ratio, the immobility of the solid phase, and the

contact time between the soil and the solution (Miller, Summer, and Miller

1989). Considering the debate about the representativeness of packed versus

intact columns (Cassel, Genuchten, and Wierenga 1975; Camobreco et al. 1996)

to a natural system, results obtained from both types of columns were compared.

B. Prado et al.2912

The objective of this work was to study the transfer of reactive anions

through a Mexican volcanic soil. The retardation was estimated at different

soil depths, using intact and packed soil columns. The effects of the input

anion concentration and the flow rate were examined.

MATERIALS AND METHODS

Study Site and Soils Sampled

Soils were sampled from a small hydrologic catchment (53 ha), “La Loma”

(198 160N and 998 59 W; 2600–3000 m above sea level; in Amanalco de

Becerra, State of Mexico, Mexico), instrumented for water runoff, erosion,

and fertilizer transport monitoring. The climate at the site is tropical and

high altitude, with average annual precipitation of 1300 mm and annual

mean temperature of 10.78C. The soil in the catchment is a Typic

Hapludand (Soil Survey Staff 1999); indurated volcanic tuff (Tepetate) soils

are also present. The study was based on the first soil type, where the agricul-

tural practices and use of fertilizers are more intensive. Nitrogen fertilizer is

applied at an average rate of about 150 kg N/ha to maize, which represents

35% of the land in the catchment.

A soil profile was excavated under maize and each horizon was

sampled for physical, chemical, and mineralogical characteristics. Bulk

density was estimated from samples of known volume taken with a 5-cm

diameter steel core. The pH was measured in water suspension using a

1:2.5 (w/v) soil–solution ratio. Soil texture was determined by laser gran-

ulometry. Total organic carbon content (TOC) was analyzed by dry com-

bustion (TOC-5050A Shimadzu). Phosphate retention was determined

according to Blackmore, Searle, and Daly (1997). Cation and anion

exchange capacity were assessed according to Zelazny, He, and Vanworm-

houdt (1996) at soil pH. Selective dissolutions of iron (Fe), aluminium (Al),

and silica (Si) were carried out to characterize the poorly ordered

amorphous materials. Ammonium oxalate extraction at pH 3.0 was

conducted to extract the poorly crystalline active Fe, Al, and Si

compounds (Feo, Alo, and Sio respectively) (Blackmore, Searle, and

Daly 1997). The soils were extracted with sodium pyrophosphate to get

Fe and Al bound to organic matter (Fep and Alp) (Blackmore, Searle,

and Daly 1997). A dithionite-citrate dissolution procedure was used to

determine ‘free’ Al, Fe, and Mn oxides (Ald, Fed, and Mnd) (Mehra and

Jackson 1960). Concentrations of Fe, Al, Si, and Mn in solutions were

determined using an atomic absorption spectrophotometer (SpectrAA

220 Varian). The allophane and imogolite were calculated with the

relation Sio � 6 (Parfitt 1990), the ferrihydrite concentration with

the relation Feo � 1.7 (Parfitt 1998), and the crystalline Fe oxides with

the relation Fed – Feo (Parfitt 1998).

Nitrate Sorption in a Mexican Allophanic Andisol 2913

Column Experiments

Seven soil columns were studied. Four were excavated on the field at

different depths (C1 to C4 at 0.05, 0.3, 0.55, and 0.8 m respectively) by

inserting a 0.25-m long and 0.05-m inner diameter PVC tube into the

soil. Intact soil columns were stored at 48C until use. A soil sample was

taken at the same site between 0.2 and 0.6 m deep with a screw-type

auger. This soil was used to pack three columns (Cd1 to Cd3). The soil

sample was sieved moist at 2 mm before being packed into a 0.025-m-

diameter glass tube. The same procedure was followed for each of the

three packed columns to achieve a uniform and identical bulk density.

The field soil solution was analyzed and contained Ca2þ, Mg2þ, and K1þ

at concentrations of 0.307, 0.024, and 0.0486 mM respectively. A back-

ground solution containing the same concentrations of these cations with

Cl as accompanying cation was prepared. Columns were oriented vertically,

and a constant flow rate of the background solution under unsaturated

condition was passed through the columns using a peristaltic pump. This

solution was leached through the column for a minimum of two pore

volumes until reaching a steady state for the output flow rate and electrical

conductivity (EC). Half of a pore volume of KNO3, KBr, and 18O was

injected at concentrations of 6, 8, or 9 mM for NO32-N and Br2 and 34.1

(% of atomic excess) for 18O and pushed through by the background

solution. The 18O was used as a water tracer to calibrate the soil hydrody-

namic parameters of a numerical model. EC, pH, and temperature of the

leachate were monitored continuously. Samples were automatically

collected for chemical analysis. The experiments were terminated when

the leachate’s EC reached a level close to the background solution’s EC.

Water contents were determined by weighting the column at the

beginning and at the end of each run and by knowing the initial water

content. The pore water velocity was calculated as the input solution flux

over the volumetric water content. The leachate fractions were weighed to

check the flux at the column exit and were analyzed for NO32-N, Br2,

and Cl2 by capillary electrophoresis and for 18O by mass spectrometry.

The runs were conducted the same way in both packed and intact

columns. Two different flow rates were used, and the input concentration

was varied only into the packed columns. Table 1 presents the bulk

density, final water content (u), mean pore water velocity (n), pore

volume (Vo) and input concentration (Co) for each column.

Transport Parameters Determination

Convection–dispersion equation transport parameters were determined for the18O, NO3

2-N, and Br2 breakthrough curves (BTC) data using the CXTFIT 2.1

code (Toride, Leij, and van Genuchten 1999). The deterministic linear

B. Prado et al.2914

equilibrium adsorption model was used in inverse mode for flux concentration

in packed columns and the surface intact column to fit the dispersion coeffi-

cient D and the retardation R. This procedure assumed that all the water

was mobile and therefore at physical equilibrium. However, this hypothesis

proved incorrect on the intact columns, where the 18O BTC peak appeared

before one pore volume. The two-region nonequilibrium model for flux concen-

tration was fitted on the 18O BTC using R ¼ 1 to obtain D, the mobile water

fraction (um/u), and the dimensionless mobile–immobile region exchange

term (v). D and um/u were then kept constant, and the same two-region

model was fitted on the NO32-N and Br2 BTCs to obtain R, b (dimensionless

partitioning coefficient), and v. The use of CXTFIT assumes that adsorption

is linear.

RESULTS AND DISCUSSION

Soil Characteristics and Evidence for Anion Sorption

Main physical and chemical properties for each soil layer are given in Table 2.

Clay content, 27%, is dominated by allophane and imogolite. The average

content of allophane and imogolite varied from 18 to 28% of the clay

fraction depending on soil depth. The soil exhibits a high AEC, consistent

with previous studies on volcanic ash soils (Wong, Hughes, and Rowell

1990; Kinjo and Pratt 1971a). The TOC content is about 5.3% and did not

vary greatly with depth, as did the other soil characteristics except the

allophane and imogolite contents and the bulk density. As in other Andisols,

the soil is variably charged, a characteristic associated with the presence of

noncrystalline materials and high organic matter content (Nanzyo, Dahlgren,

and Shoji 1993).

Table 1. Column parameters: bulk density (r) input concentration (Co), final water

content (u), mean pore water velocity (n), and pore volume (Vo) for NO32-N and Br2

for a columns of a Typic Hapludand from Amanalco, Mexico

Columns

Depth

(cm)

r

(g cm23)

Co

(mM)

u

(cm3 cm23)

n

(cm min21)

Vo

(cm3)

Intact soil columns

C1 5–30 0.72 8 0.71 0.04 347

C2 30–55 0.61 8 0.69 0.05 338

C3 55–80 0.62 8 0.80 0.04 391

C4 80–105 0.63 8 0.72 0.04 352

Packed soil columns

Cd1 20–60 0.77 6 0.66 0.05 81

Cd2 20–60 0.75 6 0.70 0.033 86

Cd3 20–60 0.77 9 0.71 0.036 87


Table 2. Selected physical and chemical parameters of a Typic Hapludand from Amanalco, Mexico

Depth (cm) Horizon

ra

(g cm23)

OCb

(%)

AECc

(cmolc kg21)

CECd

(cmolc kg21) pH H2O

Allophane

and

imogolite (%)

Fe

oxides

(%)

Ferrihydrite

(%)

0–15 Ap 0.71 5.4 12.3 20.1 5.5 18.5 2.2 3.2

15–20 A 0.66 5.3 6.1 23.1 3.0 2.0

20–45 2A 0.49 5.6 11.6 23.4 6.2 22.5 2.7 1.8

45–65 2B 0.51 5.3 6.3 25.5 2.0 2.6

65–85 2Bw 0.49 4.7 10.3 17.9 6.3 26.0 3.0 2.7

85–110 2Bw2 0.50 5.1 6.5 27.7 2.3 2.1

105–140 11.2 21.1

ar: bulk density.bOC: organic carbon content.cAEC: anion exchange capacity.dCEC: cation exchange capacity.

B.

Pra

do

eta

l.2

91

6

Packed Columns

Table 1 presents the bulk density, final water content (u), mean pore water

velocity (n), pore volume (Vo), and input concentration (Co) for each

packed column. Figure 1 shows the Br2 and NO32-N BTCs for one packed

column (Cd3). As indicated before, curves are symmetrical. An equilibrium

convection–dispersion equation satisfactorily describes the experimental

results. Model parameters are presented in Table 3. The maximum C/Co

peak occurred in the column with the lower pore water velocity (Cd2). A

lower injection flux leads to a lower dispersion, which results in a narrower

curve with a higher peak. Because of the presence of variable charges in the

soil, the peak of NO32-N BTCs appeared after 1.25 pore volume, which

would be the position of the BTC peak for an inert tracer. The retardation

coefficient R varies with the concentration of input solution (Cd2 and Cd3,

Table 3), indicating a nonlinear adsorption of NO32-N in the soil columns.

This parameter decreases when the input concentration increases following

a Freundlich-type isotherm, similar to results obtained by Qafoku, Sumner,

and Redcliffe (2000). The authors reported that at low concentration

(,5 mM ), more NO32-N is adsorbed by the soil than in solution. Thus, the

maximum R value is obtained at low concentrations and decreases at higher

concentrations. There is a controversy concerning the mechanism of

NO32-N adsorption at low concentration in soil: Toner, Sparks, and Carski

(1989) proposed a totally reversible adsorption that is in fact a simple electro-

static retention, whereas according to Singh and Kanehiro (1969), the adsorp-

tion is the result of van der Waals interactions. Lately, Qafoku, Sumner, and

Radcliffe (2000) explained the adsorption as an overlapping or interpenetra-

tion of double layers around the positively charged Al polymers and nega-

tively charged silicate minerals. Each of these suggestions could imply a

different effect on retardation.

Figure 1. Measured (circle and cross) and simulated (line) BTCs for NO32-N and

Br2 for a packed soil column (Cd3) of a Typic Hapludand, Amanalco, Mexico. The

rectangle indicates the pulse.


Bromide adsorption was also studied, an anion not naturally present in

soil system and not subject to microbial degradation. A slightly higher

retardation coefficient was found for Br2 compared to NO32-N (not shown).

This result can either be due to a slight preference of the soil for Br2 over

NO32-N or to some bacterial degradation of the NO3

2-N anion during the

experiment duration. Degradation can be excluded because the mass

balances for both anions at conclusion were found to be close to 100%.

Additionally, the Darcy velocity flux is greater than the one indicated by

Corey, Nielsen, and Kirkham (1967) as being too rapid to allow

microorganism degradation of NO32-N during its transport through the soil

column.

The dispersivity parameter l was calculated using the equation l ¼ D/n,

assuming that the molecular diffusion was negligible compared to the convec-

tive dispersion, because in this study the pore volume velocity was high

(Kutılek and Nielsen 1994). The dispersion and also the dispersivity (n did

not vary greatly) were identical in all packed columns, which indicates that

the packing method was effective to obtain uniformity. According to

Magesan et al. (2003), the dispersivity is an indicator of how variable solute

mobility is in a soil.

Flux velocity or timescale can influence the transport and reactions of

anions in soils. If sorption process is truly at equilibrium, sorption and

transport parameters are invariant with flux velocity (Jardine et al. 1998).

As mentioned earlier, R and D values obtained in columns at different flux

velocity (Cd1 and Cd2) did not vary. It can be concluded that in the packed

columns, within the range of velocities and concentrations studied, the

sorption and transport processes are at equilibrium.

Intact Soil Columns

Table 1 presents the bulk density, final water content (u), mean pore water

velocity (n), pore volume (Vo), and input concentration (Co) for the intact

columns.

Table 3. Model fitted parameters

Columns

D

(cm2 min21)

l

(cm21) RNO32-N

KdNO32-N

(cm3 kg21) RBr2

KdBr2

(cm3 kg21)

C1 0.03 0.77 1.5 0.47 1.6 0.57

C4 0.23 5.5 2.2 1.37 2.3 1.49

Cd1 0.04 0.81 1.54 0.50 1.6 0.56

Cd2 0.03 0.82 1.56 0.52 1.6 0.56

Cd3 0.03 0.77 1.5 0.46 1.5 0.46

B. Prado et al.2918

Competitive Adsorption with Cl2

Figure 2a shows the variations of dimensionless Cl2 and NO32-N measured

concentrations at the bottom of column C1. As indicated in the method

section, Cl2 is the accompanying anion for all cations in the recomposed

soil solution. This solution is also used to prepare the pulse solution of Br2

and NO32-N. This means that Cl2 is constantly injected in the column at a

concentration of 1.5 mM. At the time of the pulse, Cl2 is in an equilibrium

state, the concentration in the leachate being equal to that in the input

solution. When the NO32-N and Br2 pulse is injected in the soil column,

Cl2 is supposed to be released into the leachate in higher proportion

(however, early samples were not analyzed), meaning that the Cl2 adsorption

level decreases. After the pulse, Cl2 concentration in the leachate decreases

(Figure 2), showing that in the absence of the two other anions, Cl2 is

adsorbed. Kinjo and Pratt (1971b) reported a slight preference for Cl2 over

Br2 and NO32-N in soils containing amorphous materials, whereas Harsh,

Chorover, and Nizeyimana (2002) pointed out that sorption of Cl2, Br2 and

NO32-N on allophane is not selective. In this study, NO3

2-N and Br2

Figure 2. Measured NO32-N and Cl2 BTCs for the intact soil columns: (a) in the top

horizon (5–30 cm, column C1) and (b) at 80–105 cm deep (column C4).


concentrations were 5.3 times higher than Cl2 concentration, meaning that

the latter cannot compete for the adsorption sites in the presence of the two

other concentrated anions. Figure 2b shows the same BTCs for column C4.

A similar reduction in Cl2 concentration can be noted but to a lesser extent,

which may be explained by the effect of dispersion being seven times

higher in C4 than in C1 (see Table 3). In Figure 2, it can also be observed

that the NO32-N BTC peak does not reach the level in C1 because of the

dispersion.

Physical Nonequilibrium

Figure 2 shows the variation of dimensionless measured concentrations

(C/Co) for Cl2 and NO32-N with pore volumes, at the bottom of the intact

soil columns C1 (a) and C4 (b). The C1 NO32-N BTC is symmetrical

(Figure 2a), which indicates that NO32-N movement is under physical and

chemical equilibrium. On the other hand, the C4 NO32-N BTC is asymmetrical

(Figure 2b), with an early peak and tailing at the end of the curve, indicative of

the presence of physical or chemical nonequilibrium.

Figure 3 shows the dimensionless measured and simulated NO32-N BTCs

of two intact columns, in the top horizon (5–30 cm, column C1) (a) and at

0.8 m deep (column C4) (b). The 18O (nonreactive tracer) BTC for C1 is

symmetrical, indicating that solute transport is under physical equilibrium.

However, in Figure 3b, the 18O BTC with an advanced peak and pronounced

tail indicates the presence of physical nonequilibrium between advection-

dominated pore domains and the soil matrix pores within the aggregates

(Jardine et al. 1998). Preferential flux or the presence of pores that do not par-

ticipate to the overall flux reduces the contact time between solute and soil

matrix and provokes an anticipated exit of the solute, inducing increasing

risks of groundwater contamination (Beven and Germann 1982). The

structure of C1 layer is the result of tillage, as this column was sampled

between 5 and 30 cm deep in a maize plot. Bejat et al. (2000) indicates that

tillage destroys the natural pore structure of surface soils, disrupting

macropore continuity and reducing the extent of bypass flow. In contrast to

C1, C4 was sampled at the bottom of the root zone between 0.8 and 1.1 m,

where the soil has never been disrupted and keeps its natural structure.

Adsorption and Relation with Soil Characteristics

On both graphs in Figure 3, a shift toward the right of the N-NO32 BTC

compared to the 18O one can be observed, indicative of nitrate retardation in

the soil. As in packed columns, Br2 was slightly more retarded than NO32-N

in intact soil columns (Table 3). The retardation coefficient increases signifi-

cantly with depth (Table 3), whereas AEC slightly decreases (Table 2). This

finding is contradictory to results obtained by Qafoku, Sumner, and

Radcliffe (2000), who found that on packed columns of variable charge

B. Prado et al.2920

subsoils, the retardation coefficient was positively correlated to the AEC. The

AEC determined through classical batch method [e.g., Zelazny, He, and

Vanwormhoudt (1996) in this study] is more related to the number of sites

where exchanges can occur and gives no information about the type of anion

interaction on these exchange sites. Duwig et al. (2003) determined the AEC

with both static and dynamic methods using the same anion and obtained

different values. They explained the discrepancy by differences in the exper-

imental conditions that can affect the adsorption. In this study, the AEC was

estimated through batch experiments with the anion Cl2. Cl2 is considered

to be a good tracer for NO32-N (McMahon and Thomas 1974). However, the

experimental conditions between batch and columns are not the same;

mostly the way the anion and the soil particles are in contact differs. In

addition to the difference in the anion and experimental conditions used, the

role of soil physical parameters is important when the retardation coefficient

through displacement experiments is determined, in comparison to static

methods.

Figure 3. Measured (circle and triangle) and simulated (line) BTCs for NO32-N and

18O in the intact soil columns: (a) in the top horizon (5–30 cm, column C1) and (b) at

80–105 cm deep (column C4).


The increase of R with depth can directly be related to the increase of

soil allophane and imogolite content, as reported by Kinjo and Pratt (1971a)

in their study of NO32-N adsorption in acidic soils from Mexico and South

America. In Table 1, it can be noted that the pH increases with depth as the

carbon (C) content decreases, whereas the retardation increases. As found in

many studies on anion transport in variable charge soils where soil conditions

are modified (Qafoku, Sumner, and Radcliffe 2000; Magesan et al. 2003), retar-

dation is negatively correlated with pH. However, while studying different soil

layers under natural conditions, this relationship is not so evident, because

other factors affect anion adsorption such as amorphous materials.

Dispersion

The dispersion was found to increase with depth. This phenomenon is linked to

the difference in soil structure from one soil layer to another. The soil has a loamy

structure; however the tillage, the C content, the amorphous material content, and

the water content modify the structure. It varies from granular of loose consist-

ency on the surface, to granular with subangular blocky moderately thick of

friable consistency at depth. The dispersivity coefficient was calculated as for

packed columns, and the values obtained are in the range of those reported by

Magesan et al. (2003). These authors studied solute movement through intact

columns of an allophanic soil under unsaturated flux.

Comparison between Packed and Intact Columns

The discrepancy found between the two columns type are linked to difference in

soil structure. The fine and loose structure of the top intact columns (C1) is hom-

ogenous, due to tillage practices, and comparable to the structure of packed

columns. It is confirmed by the fact that the dispersion of packed columns

(Cd1 to Cd3) where the soil was sieved and compacted to the same bulk density

was found to be similar to column C1. At a greater depth, the soil is more aggre-

gated and the deeper intact core dispersivity increases. Whereas solute transport in

packed columns was found to be uniform and under physical equilibrium, deep

intact soil core exhibited preferential flux, leading to accelerated solute

movement. Retardation coefficient obtained in packed columns was equal to

those obtained in intact soil cores from the surface horizon. This result can be

explained by two factors: similar soil characteristics because the soils were

sampled at the same depth and similar soil structure. In intact soil cores, retardation

increased with depth, due to variation in soil amorphous material contents.

Although the Br2 and NO32-N anions are more retarded at depth compared to

the water tracer, it must be noted that they will move through this deep layer

more quickly than an inert tracer in the same soil without preferential flux (see

Figure 3b). Consequently, nitrate present at this depth [which is below the root

zone (90 cm)] cannot be used by plants and will quickly move to the groundwater.

B. Prado et al.2922

CONCLUSIONS

The Andisol (Typic Hapludand) from the Mexican catchment “la Loma”

shows a NO32-N adsorption capacity that could retard its transfer toward the

aquifer. However, in deeper layers, the preferential flow is the dominant

process during the displacement experiments, a factor that accelerates nitrate

transfer toward groundwater resources. In the packed columns, where the con-

centration and flux density effect on the reactive anions was studied, the retar-

dation decreased when the input concentration increased, indicating that the

N-NO32 adsorption in the soil is nonlinear. The dispersivity was similar in

all packed columns, showing that the packing method was effective to

recreate comparable structure in each column. In the intact columns where

the variation of anion adsorption with depth was analyzed, the retardation

increased with depth. The variation of retardation coefficient was directly

correlated to the increase of amorphous materials with depth but not with

the anion exchange capacity. The dispersivity also increased with depth, due

to the change in soil structure between soil surface and the horizon at 1-m

depth. In both type of columns, Br2 was slightly more retarded than NO32-N.

The packed columns are useful to study the effect of the soil properties on

the anion adsorption; nevertheless, the intact columns describe better the

solute transport toward the aquifer as they parallel more closely the soil

structure and pore geometry found under field conditions. Results indicate

that extremely thoughtful consideration must be taken before extrapolating

fertilizer movement study results obtained from packed soils to the field

situation. Andisols, although having a fine and poorly developed structure

with mostly micropores, can display preferential flux and, as noted by Jarvis

(1998), this process must be considered more like a rule rather than an exception.

ACKNOWLEDGMENTS

Blanca Prado is indebted to Consejo Nacional de Ciencia y Tecnologia

(Mexico) and Societe Francaise d0Exportation des Resources Educatives

(France) for her PhD grant. The authors thank the ‘Laboratorio de Fertilidad

de Suelo’ of ‘Colegio de Postgraduados, Montecillo’ for the soil analyses.

The research was funded by the ‘Institut de Rercherche pour le Developpe-

ment’ (IRD), France.

REFERENCES

Bejat, L., Perfect, E., Quisenberry, V.L., and Haszler, G.R. (2000) Solute transport asrelated to soil structure in unsaturated intact soil blocks. Soil Science Society ofAmerica Journal, 64: 818–826.

Beven, K.J. and Germann, P.F. (1982) Macropores and water flows in soils. WaterResources Research, 18: 1311–1325.


Blakemore, L.C., Searle, P.L., and Daly, B.K. (1997) Procedures for Soil Analysis

(Technical Paper No. 9). International Soil Reference and Information Centre:

Wageningen, The Netherlands.

Camobreco, V.J., Richards, B.K., Steenhuis, T.S., Peverly, J.H., and McBride, M.B.

(1996) Movement of heavy metals through undisturbed and homogenized soil

columns. Soil Science, 161: 740–750.Cassel, D.K., Genuchten, M.T.V., and Wierenga, P.J. (1975) Predicting anion

movement in disturbed and undisturbed soils. Soil Sci. Soc. Am. P, 39: 1015–1919.

Corey, J.C., Nielsen, D.R., and Kirkham, D. (1967) Miscible displacement of nitrate

through soil columns. Soil Science Society of America Journal, 31: 497–501.

Duwig, C., Becquer, T., Charlet, L., and Clothier, B.E. (2003) Estimation of nitrate

retention by a Ferrasol by a transient-flow method. Eur. J. Soil Sci., 54: 505–515.

Goodrich, J.A., Lykins, B.W., and Clark, R.M. (1991) Drinking water from agricultu-

rally contaminated groundwater. Journal of Environmental Quality, 20: 707–717.

Harsh, J., Chorover, J., and Nizeyimana, E. (2002) Allophane and imogolite. In SoilMineralogy with Environmental Applications; Dixon, J.B. and Schulze, D.G.

(eds.); SSSA: Madison, Wisconsin, 291–322.

Ishiguro, M., Song, K.C., and Yuita, K. (1992) Ion transport in an allophanic Andisol

under the influence of variable charge. Soil Science Society of America Journal, 56:

1789–1793.

Ishiguro, M. and Shoji, S. (2002) Nitrate and sulphate transport in volcanic ash soil of A

and B horizons. In 17th World Congress of Soil Science; IUSS: Bangkok, Thailand;

Vol. V, 1629.

Jardine, P.M., O’Brien, R., Wilson, G.V., and Gwo, J.P. (1998) Experimental tech-niques for confirming and quantifying physical nonequilibrium processes in soils.

In Physical Nonequilibrium in Soils: Modeling and Application; Selim, H.M. and

Ma, L. (eds.); Ann Arbor Press: Chelsea, Michigan, 243–271.

Jarvis, N. (1998) Modeling the impact of preferential flow on nonpoint source

pollution. In Physical Nonequilibrium in Soils: Modeling and Application;

Selim, H.M. and Ma, L. (eds.); Ann Arbor Press: Chelsea, Michigan, 195–221.

Katou, H., Clothier, B.E., and Green, S.R. (1996) Anion transport involving competi-

tive adsorption during transient water flow in an Andisol. Soil Sciience Society of

America Journal, 60: 1368–1375.Kinjo, T. and Pratt, P.F. (1971a) Nitrate adsorption, I: In some acid soils of Mexico and

South America. Soil Science Society of America Proceedings, 35: 722–725.

Kinjo, T. and Pratt, P.F. (1971b) Nitrate adsorption, II: In competition with chloride,

sulfate, and phosphate. Soil Sci. Soc. Am. Proc., 35: 725–728.

Kutılek, M. and Nielsen, D.R. (1994) Soil Hydrology; Catena Verlag: Cremlingen-

Destedt, Germany.

Magesan, G.N., Vogeler, I., Clothier, B.E., Green, S.R., and Lee, R. (2003) Solute

movement through an allophanic soil. Journal of Environmental Quality, 32:

2325–2333.

McMahon, M.A. and Thomas, G.W. (1974) Chloride and tritiated water flow indisturbed and undisturbed soil cores. Soil Science Society of America Proceedings,

38: 727–732.

Mehra, O.P. and Jackson, M.L. (1960) Iron oxide removal from soil and clays by a

dithionite–citrate system buffered with sodium bicarbonate. Clays and Clay

Minerals, 7: 317–327.

Miller, D.M., Summer, M.E., and Miller, W.P. (1989) A comparison of batch and flow

generated anion adsorption isotherms. Soil Science Society of America Journal, 53:

373–380.

B. Prado et al.2924

Nanzyo, M., Dahlgren, R., and Shoji, S. (1993) Chemical characteristics of volcanicash soils. In Volcanic Ash Soils, Genesis, Properties and Utilization; Shoji, S.,Nanzyo, M., and Dahlgren, R. (eds.); Elsevier: Amsterdam, The Netherlands,145–187.

Parfitt, R.L. (1990) Allophane in New Zealand—a review. Australian Journal of SoilResearch, 28: 343–360.

Parfitt, R.L. and Childs, C.W. (1998) Estimation of forms of Fe and Al: A review andanalysis of constrating soils by dissolution and Moessbauer methods. Aust. J. SoilRes., 26: 121–144.

Qafoku, N.P., Sumner, M.E., and Radcliffe, D.E. (2000) Anion transport in columns ofvariable charge subsoils: Nitrate and chloride. Journal of Environmental Quality, 29:484–493.

Schalscha, E.B., Pratt, P.F., and Domecq, T.C. (1973) Nitrate adsorption by somevolcanic-ash soils of southern Chile. Soil Science Society of America Journal,117: 44–45.

Singh, B.P. and Kanehiro, Y. (1969) Adsorption of nitrate in amorphous and kaoliniticHawaiian soils. Soil Science Society of America Journal, 33: 681–683.

Soil Survey Staff (1999) Soil Taxonomy. A Basic System of Soil Classification forMaking and Interpreting Soil Surveys, 2nd Ed. (Agriculture Handbook 436), Depart-ment of Agriculture, Natural Resources Conservation Service: Washington, D.C.

Spalding, R.F. and Exner, M.E. (1993) Occurrence of nitrate in groundwater—areview. Journal of Environmental Quality, 22: 392–402.

Toner, C.V., Sparks, D.L., and Carski, T.H. (1989) Anion exchange chemistry ofmiddle Atlantic soils: Charge properties and nitrate retention kinetics. Soil ScienceSociety of America Journal, 53: 1061–1067.

Toride, N., Leij, F.J., and van Genuchten, M.T. (1999) The CXTFIT code for estimat-ing transport parameters from laboratory or field tracer experiments, version 2.1.Research Report 137, U.S. Salinity Laboratory: Riverside, California.

Wong, M.T.F., Hughes, R., and Rowell, D.L. (1990) Retarded leaching of nitrate inacid soils from the tropics. Measurement of the effective anion exchange capacity.Journal of Soil Science, 41: 655–663.

Zelazny, L.W., He, L., and Vanwormhoudt, A. (1996) Charge analysis of soils andanion exchange. In Methods of Soil Analysis, Part 3: Chemical Methods, 3rd edn.;Sparks, D.L. (ed.), SSSA: Madison, Wisconsin, 1231–1253.


07) 300–313www.elsevier.com/locate/geoderma

Geoderma 139 (20

Characterization, functioning and classification of two volcanicsoil profiles under different land uses in Central Mexico

B. Prado a,⁎, C. Duwig a, C. Hidalgo b, D. Gómez c, H. Yee e,C. Prat d, M. Esteves d, J.D. Etchevers b

a IRD, c/o Colegio de Postgraduados, Laboratorio de Fertilidad de Suelo, CP 56230 Montecillo, Méxicob Colegio de Postgraduados, Campus Montecillo, Laboratorio de Fertilidad de Suelo, CP 56230 Montecillo, México

c Departamento de Suelos, Universidad Autónoma de Chapingo, Chapingo, Méxicod IRD, Laboratoire d'étude des Transferts en Hydrologie et Environnement, BP 53, 38041 Grenoble Cedex 9, France

e Departamento de Física-ESFM-IPN Edif. 9, U.P. “ALM”, Col. Lindavista 07738, México, D.F. México

Received 8 March 2006; received in revised form 21 December 2006; accepted 11 February 2007Available online 30 March 2007

Abstract

Volcanic soils constitute an important resource for agriculture and forestry in Central Mexico, as well as in various world regions. They exhibitunique properties and high productive potential related to the amorphous materials they contain. The relationship between amorphous materials,soil characteristic and functioning, has not been well studied. The objectives of the present work were to assess the influence of land use(agricultural and forest), topography and other soil forming factors on physical, chemical and mineralogical characteristics and pedologicalprocesses responsible for soil genesis and soil classification of two volcanic soil profiles derived from andesitic parent material located 150 maway from each other within the same toposequences. The toposequence is located in the Trans-Mexican Volcanic Belt (TMVB), a highlypopulated region of Central Mexico that provides part of the water for Mexico City megapolis. A series of field and laboratory techniquesincluding physical, chemical, micromorphological, X-ray diffraction (XRD), transmission electron microscopy (TEM), infra-red analysis,Mössbauer spectroscopy were used.

The main factor affecting the present morphology of the soil profiles was the topography. The mineralogical features of the upper layers of themaize profile (Pachic Andosol), indicate redistribution of soil material from the upper part of the toposequence. The land use change favored thisredistribution. Deeper horizons of this profile were developed from volcanic ashes deposited in situ. hematite and ferrihydrite, considered markersof evolution in redistributed soil material were observed in this profile associated with allophane. The presence of hematite has been reported forthe first time in Mexican Andosols. The present characteristics of the forest profile (Dystric Cambisol) are mainly due to the pedological process ofthe volcanic ash layers remaining in situ after the redistribution and volcanic breccia. It was concluded that the forest profile evolved from anAndosol to an Inceptisol, which was evidenced by desaturation, loss of silica and organic carbon. In this profile the Fe minerals were associatedwith the presence of gibbsite and halloysite. The position in the toposequence and the physical and chemical characteristics of these profiles definetheir present functioning, such as losses by erosion and C dynamics.© 2007 Elsevier B.V. All rights reserved.

Keywords: Andosol; Cambisol; Land use; Allophanic soil; Gibbsitic soil; Trans-Mexican Volcanic Belt

1. Introduction

Soils derived from volcanic products cover about 0.84% ofthe world land surface (Takahashi and Shoji, 2002) and 1.2% of

⁎ Corresponding author. Tel./fax: +52 595 9511475, + 52 595 9520200x1237/1234 (Central Phones).

E-mail address: [email protected] (B. Prado).

0016-7061/$ - see front matter © 2007 Elsevier B.V. All rights reserved.doi:10.1016/j.geoderma.2007.02.008

Mexico's territory (INEGI, 1999), although they are generallylocated in regions of high population density (Leamy, 1984).Volcanic soils represent an important resource for agriculture invarious world regions due to their unique properties, placingthem among the most productive soils in the world. The pre-sence of amorphous materials in these soils gives them dis-tinctive properties, such as a low bulk density, high organicmatter content, high porosity and high water retention capacity,

mailto:[email protected]

http://dx.doi.org/10.1016/j.geoderma.2007.02.008

301B. Prado et al. / Geoderma 139 (2007) 300–313

all of them favourable for root development and plant growth(Nanzyo, 2002).

Most soils found along the TMVB are derived from volcanicmaterials (Aguilera, 1963). Numerous studies on soil classifica-tions were completed in this region (Aguilera, 1963; Cortés,1966; Valera, 1993). However, according to Alcalá (2003),these studies do not clearly explain the procedures followed toclassify the soils, and they should be updated according to themodifications made in international standards for soil classifi-cation (FAO-ISRIC and ISSS, 1998; Soil Survey Staff, 1999). Acomplete characterization of their physical, chemical andmineralogical properties is required for this purpose.

The effects of continuous changes in land use, vegetationtype, and management on the characteristics and pedogeneticsprocesses of the TMVB volcanic soils have not been studied indetail. The environmental effects on the site characteristics ofvolcanic soil were reported by Cruz and Campos (1997). Theseauthors studied six profiles on the southeast hillside of the Cofrede Perote volcano, each at different altitude and climate con-ditions. The authors concluded that tropical climate and the highsoil acidity intensify the weathering of original materials. Overthe last 20 years soil properties from the same region (Werner,1985; Hidalgo et al., 1986; Cruz-Huerta and Geissert, 2000;Campos-Cascaredo et al., 2001) have been studied. However,the mineralogy and alteration products of Andosols did notreceive much attention. Miehlich (1991) studied soils derivedfrom andesitic volcanic ash in the high-altitude areas of theSierra Nevada, in the central part of the TMVB, and reported thepresence of allophane and halloysite. In the same area Hidalgoet al. (1991) found imogolite and halloysite. Alvarado et al.(2001) reported gibbsita in Costa Rican soils. According toMizota and van Reeuwijk (1989), the presence of this mineralhas not been reported in Mexico, Ecuador, Nicaragua, El Sal-vador, countries with a dominant presence of volcanic soils.

Fig. 1. Location of

Iron oxides and oxyhydroxides are not frequently studied inthe volcanic soils of this region due to the scarce availability ofspecific equipments for its characterisation. Presently thetechniques used are of chemical nature, particularly selectivedissolution. Pedoenvironmental characteristics of volcanic soilsas soil temperature, soil water, soil pH, soil organic matter and Alactivity are related to forms of Fe in volcanic soils (Schwert-mann, 1988). According to this author, these pedoenvironmentalcharacteristics are used to obtain information about Fe forms inthe soil and the rate of Fe transformation during weathering.

The objectives of the present work were to assess the in-fluence of land use (agricultural and forest), topography, andother soil forming factors on the physical, chemical and miner-alogical characteristics and pedological processes responsiblefor the soil genesis and soil classification of two volcanic soilprofiles derived from andesitic parent material located very closeto each other.


The study site is located in Valle de Bravo, a watershed(63,473 km2) in the TMVB, on the western area of the State ofMexico (between 19°23′00″ and 19°05′30″ North, and 100°11′40″ and 99°52′00″ West) (Fig. 1). Valle de Bravo is anendorreic watershed part of the Balsas river hydrological region,within the Cutzamala river watershed that provides nearly 30%of the water for Mexico City megapolis. It is formed by six sub-watersheds: Amanalco, Molino Los Hoyos, Santa Mónica –Hierbabuena, San Diego – Gonzáles, El Carrizal, and LasFlores. In the Amanalco sub-watershed (Amanalco de Becerramunicipality), which provides 51% of the water to the Valle deBravo Lake, an elementary watershed, “La Loma”, is located,where the present study was conducted. La Loma has an area of53 ha, with an altitude ranging from 2500 to 3100 m, and is

the study area.

Fig. 2. Mean monthly climatic data at the study site.

Table 1Profiles site description

Maize profile Forest profile

Localization North 19°16′48.6″ 19°16′37.5″West 99°58′13.7″ 99°58′52.8″

Slope 28% 35%Erosion Laminar type Without erosionRelief Convex ConcavePermeability High MediumLand use Agricultural ForestField humidity Dry From 0 to 15 cm

Wet From 15 to 110 cm From 0 to 100 cm

302 B. Prado et al. / Geoderma 139 (2007) 300–313

representative of the different environmental conditions andland uses in the Amanalco watershed.

The watershed is characterised by the predominance ofvolcanic rocks from the Cenozoic Era. Soils are deep and derivefrom volcanic ashes and other volcanic materials. In someplaces, it is possible to observe deep dark and organic horizons,typical of paleosols, which are buried horizons that also comefrom volcanic ashes.

A tropical climate of altitude characterises La Loma, with amean annual temperature of 10.7 °C, a minimum monthlytemperature of 3.1 °C in February, and a maximum of 25.5 °C inApril (data available from 1962 to 1992 by the nearby Amanalcoweather station). The mean annual precipitation is 1300 mmwithmost of the rain falling between June and September, usually asstorm events of less than 1 hour duration (Fig. 2). The meanpotential annual evaporation is 1243 mm. The soil moistureregime was estimated by the Newhall model (Wambeke et al.,1986), using the same climatic data. Results show that the soilmoisture regime is udic.

The original vegetation of both sites was forest. The mainpresent land uses are forest (46%) and agriculture (37%, mainlymaize and wheat) and the remaining is left as fallow. Naturalvegetation in the study area is Pinus spp., Quercus spp., Abiesreligiosa, Agave spp., Prunus serotina, and Crataegus mexicana.The herbaceous stratus is abundant and comprises species mainlyfrom the Asteraceae and Gramineae families.

2.1. Profiles description, soil toposequence and sampling

A toposequence including natural forest and cultivated maizewas selected within La Loma watershed. These are the two mostimportant land uses of the watershed. The toposequence islocated on the slope facing south of the watershed. The forest islocated in the upper part of the toposequence and the maize onein the lower part. This watershed had been instrumented toevaluate the effects in the last 50 years of current agriculturalpractices on water availability and quality.

Two soil profiles were selected for the study, the firstcorresponding to the natural forest and the second to thecultivated maize. The distance between the two profiles isapproximately 150 m, with an altitude variation of 35 m. Themaize profile was excavated within a plot sown with this crop,in a 28% slope. The maize plot has been under intensivecropping during the last 40 years. The forest profile was located

under forest and natural vegetation; the mean slope in this partof the toposequence is 35%. Both sites are instrumental forwater balance and erosion determination.

Soil profiles were described from 0 to 110 cm, and sampleswere taken from each horizon for laboratory analysis. In bothprofiles, intact and oriented cores were also sampled in plasticboxes measuring 10×5×3 cm3 to make thin layers for structureand porosity analysis. To describe and characterise the profiles,the technique recommended by Cuanalo (1990) was used. Partof each sample was kept at field moisture condition and theremainder was air-dried and sieved at 2 mm for characterization.The general features of the two profiles are given in Table 1.

2.2. Soil physical, chemical and mineralogical characteristicsand functioning

The texture was analysed on the soil without organic matter bytwo methods: with a laser granulometer (Mastersizer, MalvernInstruments) and the pipette method. Bulk density was measuredfrom intact cylindrical soil cores of 211 cm3, sampled from eachhorizon and dried at 105 °C until constant weight. To analyse soilstructure and porosity in thin layers, the soil water of the intactsamples was replaced by acetone in a vacuum chamber withoutaltering the original structure. Samples were then impregnatedwith unsaturated polyester resin and left to dry at air temperaturefor about 4 weeks. Once hardened, thin layers were cut andpolished to reach a 30 μm thickness (Murphy, 1986), and the soilsections were observed with a petrography microscope. Soil pHwas measured both in water and in 1 M KCl suspensions using a1:2.5 w/v soil solution ratio. The pH in NaF was measured in asuspension of 1 g soil mixed with a 50 mL 1MNaF solution after2 min stirring. The exchangeable aluminum was extracted with a1 M KCl solution and determined by titration with NaOH 0.1 M.Total soil organic carbon (SOC) was analysed by dry combustion(TOC-5050A Shimadzu). Exchangeable bases and the cationexchange capacity at pH 7 were determined by the ammoniumacetate method. The cations (Ca+2, Mg+2, Na+ and K+) wereremoved by leaching with 1M ammonium acetate (pH 7), using amechanical vacuum extractor, and analysed in the supernatant byatomic adsorption spectroscopy (SpectrAA 220 Varian). Phos-phate retention was determined according to Blakemore et al.(1987). The water retention capacity was measured at 15 bar and0.33 bar by the suction plate and the pressure plate methodsrespectively (NOM-021-RECNAT, 2000).

Table 2Horizons characteristics of both profiles

Depth(cm)

Horizon Sand Silt Clay 15 bar water content ρda Colour

moistureCohesion b Roots c Boundary d

(%) (%) (g cm−3)

Maize profile0–15 Ap 29 62 9 25.5 0.7 10YR 3/2 1 1 d15–20 A1 45 50 5 26.9 0.7 10YR 3/2 1 3 d20–45 A2 23 66 11 30.5 0.5 10YR 3/1 2 3 d45–65 2A1 25 66 9 31.1 0.5 10YR 3/1 3 3 d65–85 2A2 26 63 11 37.9 0.5 10YR 3/1 3 3 d85–110 3A 22 68 10 33.2 0.5 10YR 3/1 3 5

Forest profile0–10 A1 29 63 13 30.1 0.6 5YR2.5/1 2 1 s10–15 A2 19 68 13 27.9 0.8 5YR2.5/2 2 2 s15–37 2Bw1 11 44 45 24.5 1.0 5YR 4/4 3 3 s37–57 2Bw2 10 48 42 24.9 1.1 5YR 3/4 4 4 s57–100 2Bw3 6 57 37 25.4 1.1 5YR 4/6 3 4a ρd = bulk density.b Cohesion 1:very loose; 2:loose; 3:quite compact; 4:compact.c Roots 1: abundant; 2: quite abundant; 3: present; 4: few; 5: very few.d Boundary d: diffuse; s: sharp.


Selective dissolutions of Al, Fe and Si compounds werecarried out using three chemical reactants to get the non-crystalline components. “Active” or short-range-order Al, Feand Si compounds (Alox, Feox, Siox) were extracted aftershaking the soil with an ammonium oxalate solution buffered atpH 3. To extract Fe and Al complexed with organic mattercompounds (Fep, Alp), the soil was shaken with a sodiumpyrophosphate solution (Blakemore et al., 1987). To extract“free” Fe and Al compounds (Fed, Ald), a citrate dithionitesolution was used (Mehra and Jackson, 1960). Concentrationsof Fe, Al and Si in supernatants were determined using anatomic absorption spectrophotometer (SpectrAA 220 Varian).

Table 3Main chemical properties of the soil profiles

Depth(cm)

SOCa

(g kg−1)pH Pret

b

(%)CECc

(cmolc kH2O NaF KCl

Maize profile0–15 54 5.5 11.1 5.1 N90 22.315–20 53 6.1 11.2 5.8 N90 23.020–45 56 6.2 11.2 5.8 N90 20.045–65 53 6.3 11.2 5.9 N90 24.065–85 47 6.3 11.4 5.9 N90 23.185–110 51 6.5 11.4 6.0 N90 23.6

Forest profile0–10 55 6.1 10.9 5.4 N90 21.810–15 39 6.2 10.9 5.5 N90 19.615–37 13 6.2 9.9 5.5 85 11.537–57 8 6.3 9.6 5.4 82 12.157–100 4 6.6 9.6 5.5 82 12.1a SOC = Organic Carbon.b Pret = Phosphate retention.c CEC = Cation exchange capacity.d EB = Exchangeable bases.e EA = Exchangeable acidity (Al3++H+).

The allophane content was estimated according to Parfitt(1990). To calculate ferrihydrite and iron oxide contents,equations from Parfitt et al. (1988) and Gunjigake and Wada(1981) were used. X-ray diffraction (XRD) analyses wereperformed on the soil fraction under 2 mm and under 2 μm sizesto identify primary and secondary minerals, both withoutorganic matter, and the last with and without amorphousmaterials. Infrared (MIR) analyses in the fine fraction withoutorganic matter were performed with a spectrometer (NicoletNexus 470 FT-IR). Non-crystalline and crystalline componentswere observed by using a TEM, the observations were per-formed on a JEOL model 2010 instrument operating at 200 kV.

g−1)EB d (cmolc kg

−1) EA e

(cmolc kg−1)

Ca2+ Mg2+ K+ Na+

7.8 0.5 0.4 0.3 0.0818.2 1.2 0.5 0.3 0.0719.3 1.6 0.1 0.1 0.0520.1 1.4 b10−4 0.1 0.0516.1 0.9 b10−4 b10−4 0.0516.0 1.1 0.2 0.1 0.05

6.6 0.9 0.5 b10−4 0.026.4 0.9 0.7 b10−4 0.024.5 0.7 0.7 b10−4 0.025.0 0.7 1.2 b10−4 0.014.7 0.5 1.3 b10−4 0.01

Table 4Chemical selective dissolution of Al, Fe and Si and mineralogy of the soil samples

Depth of sampling (cm) Alox+1 /2Feox(%)

Allophane a

(%)(Alox-Alp)/Siox(%)

Ferrihydrite b

(%)Feox(%)

Alox(%)

Siox(%)

Fed(%)

Alp Fep Feox/Fed

Maize profile0–15 6.6 18.5 1.7 3.2 1.9 5.7 3.1 2.3 0.5 0.08 0.815–20 7.4 23.1 1.7 2.0 1.2 6.8 3.8 3.1 0.3 0.09 0.420–45 7.0 22.5 1.6 1.8 1.0 6.4 3.7 2.8 0.4 0.08 0.445–65 7.5 25.5 1.5 2.6 1.6 6.7 4.3 2.1 0.4 0.09 0.765–85 6.8 26.0 1.3 2.7 1.6 6.0 4.3 3.1 0.3 0.08 0.585–110 7.9 27.7 1.5 2.1 1.3 7.2 4.6 2.4 0.4 0.09 0.5

Forest profile0–10 6.3 14.8 n.d. c 1.4 0.8 5.9 2.5 2.0 n.d. n.d. 0.410–15 5.6 16.8 n.d. 1.4 0.8 5.1 2.8 2.1 n.d. n.d. 0.415–37 0.9 n.d. 0.5 0.6 0.2 3.6 n.d. n.d. 0.137–57 0.6 n.d. 0.4 0.4 0.2 3.7 n.d. n.d. 0.157–100 0.6 n.d. 0.4 0.4 0.2 3.3 n.d. n.d. 0.1a Allophane=6⁎Siox

(Parfitt, 1990).b Ferrihydrite=1.7⁎Feox

(Childs et al., 1991).c n.d. = not determined.

Table 5Relative abundance of primary's minerals of the soils profiles

Depth(cm)

Horizon Cristobalite(4.07 Å)

Plagioclase(3.2 Å)

Quartz(3.3 Å)

Hornblend(8.5 Å)

Maize profile0–15 Ap ⁎⁎⁎⁎⁎ a ⁎⁎⁎⁎⁎ ⁎⁎ ⁎

15–20 A1 ⁎⁎⁎⁎⁎ ⁎⁎ ⁎⁎⁎⁎ b ⁎

20–45 A2 ⁎⁎⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎

45–65 2A1 ⁎⁎⁎⁎⁎ ⁎⁎ ⁎⁎⁎⁎ ⁎

65–85 2A2 ⁎⁎⁎⁎⁎ ⁎ ⁎⁎⁎ ⁎⁎

85–110 3A ⁎⁎⁎⁎⁎ ⁎ ⁎⁎⁎⁎ ⁎⁎

Forest profile0–10 A1 ⁎⁎⁎⁎⁎ ⁎⁎⁎ c ⁎ ⁎⁎

10–15 A2 ⁎⁎⁎⁎⁎ ⁎⁎⁎ ⁎⁎⁎ ⁎⁎

15–37 2Bw1 ⁎⁎⁎⁎⁎ ⁎⁎⁎⁎ ⁎ e ⁎⁎⁎⁎

37–57 2Bw2 ⁎⁎⁎⁎⁎ ⁎⁎ d n.d. f ⁎

57–100 2Bw3 ⁎⁎⁎⁎⁎ ⁎⁎⁎ n.d. ⁎

a ⁎⁎⁎⁎⁎ = predominant.b ⁎⁎⁎⁎ = very abundant.c ⁎⁎⁎ = abundant.d ⁎⁎ = moderately abundant.e ⁎ = present.f n.d. = not detected.


Mössbauer measurements were conducted at a room tempera-ture with a 57Co in Rh source, using a constant accelerationspectrometer (Wissel) operated in transmission mode. AllMössbauer spectra were fitted using a least-squares minimiza-tion algorithm and Lorentzian lineshapes in order to obtain thevalue of isomer shift (δ), quadrupole splitting (Δ), hyperfinemagnetic field (Hf) and relative area (A). The isomer shift valuesare reported relative to α-Fe at 298 K.

2.3. Sediment and soil organic carbon losses

The soil organic carbon (SOC) concentration was analysedin runoff water and sediment loss at the outlet of the maize plotduring the 2003 rainy season. Sediment loss was estimated bysampling the runoff water and determining the sedimentconcentration in the samples. For each runoff event, 4 to 5runoff samples were used for SOC determination, half of eachsample was filtered so as to obtain the dissolved and particulateSOC in the runoff water. For each runoff event, the averageparticulate SOC concentration was multiplied by the totalsediment loss to obtain the particulate SOC loss by runoff anderosion. During each runoff event, the average SOC concen-tration of two following samples was multiplied by the flowduring the time laps to obtain the dissolved SOC loss by runoff.

2.4. Soil classification and andic properties

Soils were classified according to the FAO-ISRIC and ISSS(1998) and the Soil Survey Staff (1999) systems.


3.1. Soil profile description

The forest and maize soil profiles were developed fromigneous andesitic material. The maize profile is the result of theredistribution of volcanic ash coming from the upper part of the

toposequence. The slope in this site (35%) and heavy summerrains (1300 mm mostly during summer) account for the transportof fine volcanic material from the upper to the lower part of theslope.

The main features of the two profiles, maize and forest, arepresented in Tables 1 and 2, the main chemical properties inTable 3, chemical selective dissolution in Table 4 andmineralogy in Table 5.

The maize profile exhibits six horizons (Fig. 3 and Table 2).The horizons showed little differences in soil texture,consistency and colour. The soil has a granular micro (fluffy)structure with a subangular blocky macrostructure. Consistencyvaries from very friable to friable with depth. The whole profilepresents a dark yellowish colour in humid. On the other hand,

Fig. 3. Main features of the maize and forest soil profiles.


the forest profile is formed by five horizons, with markedchanges in soil texture, consistency and colour between eachhorizon. The two surface horizons have a dark yellow colourwhile the deeper ones have a yellowish red colour related to Feoxides. Below 27 cm depth, the profile displays rocky frag-ments, from frequent to abundant with depth.

Similar physical characteristics: texture, bulk density, waterretention capacity and cohesion in the A horizons of bothprofiles confirm that both derived from similar materials. How-ever, the upper A horizons of the maize profile would berelatively newer than the A horizons of the forest profile.

3.2. Soil physical, chemical and mineralogical characteristicsand functioning

3.2.1. Physical characteristicsThe whole maize profile has a silt loam texture. For clas-

sification purposes (FAO-ISRIC and ISSS, 1998), the texturepresented in Table 2 was determined on the air dry soil by thepipette method, leading to an over and under-estimation of sandand clay fractions respectively. The air drying of Andosols leadsto the irreversible formation of aggregates from its colloidalfraction (Nanzyo, 2002).

The laser diffraction after ultrasonic dispersion has shown tobe more adapted to study particle sizes of allophanic soils(Buurman et al., 1997). The particle size distribution of the soilunder maize estimated with this method (not shown) yielded26% clay, 70% silt and 4% sand. Typically in Andosols, loam isthe dominant fraction. The formation of loamy particle-sizedmicro-aggregates in Mexican Andosols and the difficulty todisperse them have been reported by Etchevers et al. (1986).

The texture of the forest profile surface horizons is silt loam,changing to a texture dominated by clay with depth. In deeperhorizons, the soil water content is higher even during the dry

season, leading to greater weathering and in the end, to theaccumulation of fine neo-formed materials (Porta-Castellanoset al., 2003). The lower part of this profile is stony, the rockfragments appear nearly unaltered, and the altered materialcomes from volcanic ashes deposited on the top and weatheringof the fine material from the breccia.

In the maize profile, the analysis of thin layers shows that thepredominant structure is subangular blocky, with a loosestructure at the surface, becoming more massive and homoge-neous with depth. This shows the effect of soil ploughing whichdamages the structure down to 20–25 cm. Different pore typescould be observed: pores of complex packing, some cavities,holes and cracks along the whole profile. However, soilploughing also destroys pore connection at the soil surface,and macroporosity is no longer well connected. This largelyaffects water and solute movement in the soil, as aggregatearrangement and a better pore connectivity in deep layers favoura preferential flow which can accelerate solute movement to thegroundwater (Prado et al., 2006). On the other hand, in theforest profile, the structure varies from subangular blocky at thesurface, to fluffy at 37 cm deep. Due to the abundance of rocksbelow this depth, it was not possible to sample intact cores toelaborate thin layers.

Table 2 shows the 15 bar water retention capacity of bothprofiles. These values seem rather low for an Andosol but thedescription of the soil profile during the dry season rejected thehypothesis of the soil being in the hydric subgroup. The ratiobetween available water content (0.33 bar water content–15 barwater content) and water content at 0.33 bar in the maize profile,indicates that between 23 to 45% (depending on depth) of0.33 bar water in this soil is available water content. Wada(1985) found a similar relation for Andosols in Alaska, UnitedStates, Chile, Ecuador, and dry parts of Hawaii (20 to 70%). Ahigh water content at 0.33 bar is a feature common to Andosols,


however the values for “La Loma”Andosols are low (35 to 60%depending on depth), compared to values reported by Wada(1985). The reduction of the water holding capacity at 0.33 barmay be due to agricultural practices, especially for the tophorizon.

The water retention capacity was also estimated by automaticmeasurements in the field (with capacitance probes, CS615,Campbell), installed horizontally between 0–30, 30–60 and60–90 cm, on both maize and forest plots. The raw dataobtained from the probes were transformed to volumetric watercontent, using the calibration curve given by the constructor,previously checked using gravimetric samples. Fig. 4 shows thewater content variation over time in three layers of the two plots.The maximum water retention capacity was taken as the watercontent that will persist during the rainy season, i.e. after acouple of days without rain, when the drainage stopped. Undermaize this maximal water retention capacity varies from 0.61,0.68 and 0.67 m3 m−3 for the three layers 0–30, 30–60 and 60–90 cm respectively. The lower water retention capacity in thefirst 30 cm of the soil under maize could be due to the lowercontent of amorphous materials. Furthermore, the cultivatedtopsoil is nearly bare during the 6 months of the dry season and

Fig. 4. Water content variation in the maize and forest plots.

directly exposed to solar radiation. Drying provokes irreversibleprocesses in Andosols, leading to a definitive decrease in thewater retention capacity. Buytaert et al. (2002) studied the effectof soil use on the hydrological properties of volcanic soils fromEcuador, concluding that the loss in water retention capacity isirreversible. They could not find a direct relationship betweenthe number of years the soil was cultivated and the fact that thesoil could not recover its original capacity, even after 50 years ofpasture.

Under forest, for the same depth levels, the maximal waterretention capacity is much lower: 0.51, 0.54 and 0.41 m3 m−3

.

The deepest layer under forest is the one having the lowestwater retention capacity, due to the increase in the number ofstones and the loss of andic properties. It can be noticed at theend of the dry season that the water content is lower under forest(0.15 to 0.19 m3 m−3) than under maize (0.34 to 0.48 m3 m−3).At the beginning of the rainy season, the water content increasesmore slowly on the former than on the latter, to reach a lowermaximum. Afterwards, during the rainy season, the watercontent variation is lower under forest than under maize. Thesedifferent plot behaviours can be explained by their soilproperties but also by higher plant uptake and higher rainfallinterception by canopy under forest (Viramontes et al., 2006).

3.2.2. Chemical characteristicsIn the maize profile, pH H2O increases with depth from 5.5

to 6.5 while SOC decreases slightly; the same trend can beobserved between pH H2O and Siox content (see Tables 3 and4), showing that organic matter carboxyl groups as well as theproton dissociation of non-crystalline aluminosilicates affectsoil acidity. Nanzyo et al. (1993a) pointed out that acidity inallophanic soil poor in organic matter could mostly be attributedto the second factor. Shoji et al. (1996) indicated that ag-ricultural practices strongly affect the soil pH, especially thefertilizer use, which could explain the difference in pH in themaize plot topsoil (pH 5.5) as compared to pH in deeperhorizons (pH 6.1 to 6.5). The pH in the forest profile is similarto the maize one and relatively constant in depth. Below 15 cmdepth of the former profile, the Siox content decreases to nearly0 (as it will be shown in the chemical selective dissolutionsection). Therefore pH is thus mainly controlled by the humifiedorganic matter content.

The ΔpH value (pH KCl–pH H2O) gives an indication aboutthe sign and magnitude of the soil surface charges. The maizeprofile, with a relatively high organic matter content and a pHover 5, had more negative charges than positive ones, with aΔpH slightly negative (−0.35 to −0.45). This result, combinedwith low values of extractable Al, shows that the soil contains amixture of variable and permanent charges, according to thecriteria given by Uehara and Gillman (1981) for tropical soilswith variable charges. In the forest profile, the ΔpH value is alsonegative, but higher (−0.7 to −1.2) than in the maize profile. Inthis profile the surface charges vary with depth from a mixtureof both charge types to variable charges deeper in the profile.

The sum of exchangeable bases is high in both profiles(average values of 18 and 7 cmolc kg

−1 in the maize and forestprofiles respectively (see Table 3). For both profiles, the Ca2+


concentration is higher than other bases concentration, while theNa+ concentration is very low. Ca2+ and Na+ are usually themost abundant cations but also the most mobile ones. Basesaturation is high and variable along the maize profile, and islinked to the pH from 0 to 60 cm depth. On the other hand, in theforest profile base saturation is lower and correlated to the goodpermeability and drainage associated with the very low runoff atthis site, as a result of the rainfall interception by the trees andthe accumulation of litter of pine needles and its products ofdecomposition, which improves water drainage and basesleaching.

In the maize profile, the cation exchange capacity (CEC)does not show any trend with depth, and is quite low comparedto values found on Typic Hapludand from the Meseta Tarasca(Alcalá et al., 2001), but similar to values reported by Cruz-Huerta and Geissert (2000), for the same kind of soil inVeracruz. The content of Siox as well as the CEC level increasewith depth and a better relationship could be found between Sioxand CEC than between SOC and CEC. A possible explanationis that the dissociation of silanol groups contributes more to theCEC than the organic matter carboxyl groups, as noted byNanzyo et al. (1993a) for soil with low organic matter content.A similar trend can be observed in the two first horizons of theforest profile. From the third one, the CEC is reduced by half,and is constant down to 100 m. In the maize profile the effectiveCEC, which is defined as the sum of exchangeable bases plusKCl-exchangeable acidity, is very close to the measured CEC inall A horizons, with the exception of the upper most. Howeverin the forest profile the effective CEC (6–12 cmolc kg

−1) waslower than the measured CEC (12–22 cmolc kg−1). Thedifferences between the CEC and the effective CEC are due tothe action of variable charge characteristics (Nanzyo et al.,1993a). As it will be explained later in this paper in B horizonsthere are certain minerals other than allophane and humicmateriasl that contribute to variable charge. Nanzyo et al.(1993a) suggest that lower values of pH (KCl) (5.4–5.5) are dueto the dissociation of carboxyl groups on soil organic matter.

3.2.3. Chemical selective dissolutionIn A horizons of maize and forest profiles, the major

weathering products are amorphous minerals (such as allo-phane, 15–28%) and in minor ones halloysite and hematite.

Fig. 5. Transmission electron micrographs (TEM) of the fine clay fraction (b2 μm) in shalloysite (horizon 2A1) and (c) tubular particles of halloysite (horizon 3A) in the ma

Fig. 5 shows the TEM of the clay fraction for the maize profile,where allophane, spheroidal and tubular halloysites can beobserved. Halloysite is considered a minor component of theclay complex and the only evidence is the TEM. According toParfitt et al. (1983), halloysite is common in soils with andicproperties located in areas where precipitation is less than1700 mm, resulting from a desilication process (Chartres andvan Reuler, 1985).

In the A horizons of maize profile the ratio Alp/Alox is lowerthan 0.1, meaning that the profile studied is an allophanicAndosol. According to several authors (Parfitt and Saigusa,1985; Quantin, 1986; Shoji et al., 1988; Nezeyimana, 1997), thelow value of the Alp/Alox ratio indicates the presence of anenvironment rich in silica (high content of Siox from 3.8 to4.6%). It also means that the content of humus-bounded Al islow, and the Al content in allophane and imogolite is muchhigher. The small amount of extractable Fep (b0.1%) indicatesthe presence of small quantities of Fe humus complexes in thesehorizons where allophane is predominant. In contrast, thenegative ΔpH values for A and B horizons of forestall profile(−0.7 to −1.1) suggest a relative abundance of humus overallophanic clays (Nanzyo et al., 1993a). Values of 0.2% Sio and0.6–0.4% Alo for B horizons, corroborate smaller contents ofallophanic clays. In A horizons of forestall profile, theweathering products are allophane and Al/Fe humus complexes.

In the maize profile, the Al:Si ratio, determined with therelation (Alox–Alp)/Siox, varies between 1.4 and 1.8. Parfitt (1980)indicates that soils exhibiting Al:Si ratio values between 1 and 2contain both types of allophanes, Si-rich and Al-rich allophanes.

The chemical dissolution shows the presence of bothcrystalline and noncrystalline forms of Fe compounds. Thelow Feox/Fed ratio indicates a high degree of crystallinity of theiron oxides in both profiles (Malucelli et al., 1999). This lattertopic is discussed in the section of minerals.

Ferrihydrite (iron oxyhydroxide), the noncrystalline form ofFe, was present in both profiles. It averages 2.4% in the maizeprofile and 1.4% in the 0–15 cm layer of the forest one.According to Malucelli et al. (1999), the presence of ferrihydritecan be related to the combination of high and constant humidityconditions and factors disturbing Fe crystallization such as highorganic matter content and high Al activity in acid conditions.Both profiles were located in a udic moisture regime.

oil without organic matter: (a) allophane-like mineral (horizon 3A), (b) spheroidalize profile.

Fig. 6. Infrared spectrum of the maize horizons.

Fig. 7. Differential X-ray diffractometer patterns of the forest profile (a) in thesoil coarse fraction and (b) in the fine fraction.


The Fed–Fep relationship in the maize profile (2.01–3.01%)shows the presence of iron oxides in this profile (Gunjigake andWada, 1981). The chemical selective dissolution indicates fourdifferent Fe weathering products within the maize profile, (0–15, 15–45, 45–85 and 85–110 cm, see Table 4). The horizonswith lower values for Feox (indirect estimation of ferrihydrite)presented higher content of Fed, and vice versa. In the forestprofile the Feox and Fed define two patterns above and below15 cm depth. According to Dahlgren et al. (1993) the lowerpercentages of Feox as compared to Fed are the result of thetransformation of thermodynamically metastable ferrihydriteinto stable Fe-oxides goethite and hematite. The smalldifferences in the A horizon weathering products of the maizeprofile make it possible to rename the horizons of this profile:Ap (0–15 cm), A1 (15–20 cm), A2 (20–45 cm), 2A1 (45–60 cm), 2A2 (65–85 cm) and 3A (85–110 cm), however, noother differences in morphological characteristics were ob-served in field. These results support the hypothesis that themaize profile is the consequence of a process of accumulation oflayers of volcanic ash transported from the upper part of thetoposequence showing different degrees of alteration of theashes.

3.2.4. Mineralogical characteristicsThe analysis of the soil fine fraction without organic matter

with a polarized light microscopy showed that the volcanicglass content was lower than 10%. The volcanic glass isuncoloured which is characteristic of the andesitic rock(Dahlgren et al., 2004).

Table 5 shows the relative abundance of minerals accordingto the main reflections of XRD diagrams of the maize profile.The most abundant mineral corresponded to cristobalite(4.07 Å) which is frequently found in volcanic rocks. Otherprimary minerals present in the profile were: plagioclases(3.2 Å), quartz (3.3 Å) and hornblend (8.5 Å). Plagioclasesprevail in the surface horizon but decrease in the middle (15–65 cm) and lower parts (65–110 cm) of the profile. The decreaseof sand particles (29–22%) and the increase of silt (66–68%)

and clay ones (9–11%) toward the bottom of the profile confirma greater alteration of minerals at deeper levels. Hornblend, amineral less resistant to weathering, was less abundant thanplagioclases. Quartz is abundant or very abundant in the entireprofile, except in the surface horizon. The fine fraction (b2 μm)of the maize profile horizons is characterised by the abundanceof amorphous materials, as indicated by the X ray diffraction(diagram not presented). The presence of these amorphousmaterials at all depth levels of the profile was evidenced by theselective dissolution treatments and IR spectroscopy. Fig. 6shows broad bands at 3400 cm−1 and near 1000 cm−1 typical ofallophane (Russell et al., 1981). The latter band shows analumina-rich allophane and the previous band an allophane richin silica. The presence of halloysites in the horizons of the maizeprofile was evident only through the TEM technique (seeFig. 5). Halloysite in volcanic ash soils can result from the directweathering of feldspars (Eswaran, 1972; Chartres and vanReuler, 1985) or from the re-silication of allophane (Wada,1989).

Table 6Mössbauer parameters recorded at room temperature of a representative sample(2Bw1)

Assigment Relativearea (%)

Quadrupole splitting(Δ)(mm/s)

Isomer shiftδ/Fe (mm/s)

Magneticsplitting B(T)

Hematiteα-Fe2O3 3.62 −0.284 0.381 51.56Magnetite 22.38 −0.230 0.249 49.2Fe3O4 0.44 −0.535 0.689 44.9High spin (Fe2+) 6.88 2.3 1.04 –High spin (Fe3+) 66.68 0.594 0.53 –

Note: isomer shift (δ) values relatives to α-Fe.


The XRD of the forest profile shows the same primaryminerals than those observed in the maize profile. Cristobalite(4.07 Å) was also the prevailing mineral both in the coarse(Fig. 7a) and fine fraction (Fig. 7b). Quartz (3.3 Å) was onlypresent in the upper horizon. Cristobalite crystallizes at highertemperatures and therefore at earlier stages than quartz. Thus, itmay be considered that the different contents of fine-grainedquartz and cristobalite between horizons (Table 5) actuallyreflect the different stages in the differentiation of magma andthe composition of the resulting ash (Tokashiki and Wada,1975).

Plagioclases were abundant to moderately abundant in theprofile. An increase in the presence of hornblend (8.5 Å) wasobserved in the middle portion of the profile, but it was scarcelypresent in the lower part. As shown in the Bowen's scheme,quartz generally appears at a later stage of the magma fractionalcrystallization, whereas feldspars continually appear as pro-ducts with changes in their composition.

The scarce presence of quartz in B horizons and theabundance of hornblend (XDR) in the A horizons of the forestprofile show that the parent material of these horizons is not thesame. The B horizon is a relatively old volcanic materialcontaining unaltered and partially altered rock fragments, whilethe latter are formed by relatively young volcanic ash. Thedifferences mentioned in the parent material explain thedifferences in the physical and chemical characteristics betweenA and B horizons. This information was the basis for renamingthe horizons.

The IR spectroscopy analysis of the forest profile horizons isshown in Fig. 8. A and B horizons show a different secondarymaterial composition. Gibbsite was the principal mineralcomponent in the B horizons (high frequency bands around3550 cm−1) but small amounts of 1:1 clays were also present.Gibbsite is the result of the relatively short life of allophane in

Fig. 8. Infrared spectrum of the forest horizons.

humid environments (Wada, 1989). Gibbsite formation hasbeen reported under conditions of excessive or moderatedrainage, where the alteration of the minerals in feldspars andamphiboles is favoured in rocks of andesitic nature (Wada,1989; Wilson, 2004). The presence of this mineral in Bwhorizons reveals a process of desilication due to weatheringconditions (Chartres and van Reuler, 1985) and partially explainthe effective CEC (6 to 9 cmolc kg

−1).The Fe chemical selective dissolutions in the A and B

horizons were discussed in a previous section. In the presentsection the results of the Fe-minerals studied by Mössbauerspectroscopic and XRD are reported.

The XRD studies reveal that hematite (α-Fe2O3) prevailed inthe upper and lower part of the maize profile; however, it wasnot found in the middle horizons. Magnetite (Fe3O4) is alsopresent in all horizons in similar amounts of this profile. Lelongand Souchier (1987) quoted in their review Hetier's report in theTMVB Andosols magnetite moves during alteration.

In the forest profile, XRD showed hematite associated withmagnetite (Fe3O4) and maghemite (Fe2O3) in all the horizons(data not shown). The latter mineral was found only in the forestprofile. It is commonly formed from magnetite (Wada, 1987) inthe presence of organic matter coming from forest vegetation(Bigham et al., 2002). Wada (1987) reports associations ofmaghemite or hematite and magnetite, and gibssite in well-weathered volcanic ashes in Hawaii.

In both soil profiles, the Mössbauer spectroscopy confirmedthe presence of hematite and magnetite. This technique alsoshowed Fe2+ and Fe3+ associated with aluminosilicate mineralsand, in spite of a pH around 6, the presence of not wellcrystallized goethite (Table 6, Fig. 9). The goethite could berecrystallized into hematite after the dissolution of the first(Lelong and Souchier, 1987). The reddish colour of the forestprofile B horizons would indicate the presence of hematite.However, the yellow to dark yellowish colour of the A horizonswould be associated with goethite.

The evolution of the different Fe forms observed in theprofiles (ferrihydrite, hematite and goethite) was confirmed bythe chemical selective dissolution. These transformations havebeen reported by Schwertmann (1988). Mizota and vanReeuwijk (1989) have associated the presence of goethite andhematite rather than ferrihydrite with low base saturation as it isthe case in the forest profile. The hematite present in the Bhorizons of the latter profile can be accounted for the FedNFeorelationship typical of the more evolved materials and for the

Fig. 9. Mössbauer spectroscopy in 2Bw1 horizon of the forest profile.


contrasting climate conditions during winter and summer(Lelong and Souchier, 1987).

The XRD shows halloysite in the B horizons of the forestprofile (Fig. 10). This mineral appeared associated with gibbsitein these horizons. A similar association has been reported byWada et al. (1986) in Ustands from Hawaii. Halloysite andkaolinite have been related to low base saturation conditionsand metal–humus complexes where allophane is absent (Wadaet al., 1986; Mizota and van Reeuwijk, 1989). The XRD doesnot show crystalline silica-rich clays in this forest profile. Theabsence of precursor minerals (e.g. chlorite, mica) in thevolcanic ash parent material impedes the formation ofcrystalline clay minerals due to the lack of a template that canserve to catalyze the precipitation of crystalline layer silicatesfrom solution (Dahlgren and Walker, 1993).

3.3. Sediment and soil organic carbon losses

In the maize profile, the average SOC content is approxi-mately 50 g kg−1 and is quite constant down to 110 cm depth;

Fig. 10. The XRD of the fine clay fraction (b2 μm) in 2Bw3 horizon of forestprofile.

however, in the forest profile the SOC content is lower anddecreases below 15 cm depth. The close canopy of the pine treeslimits the development of herbaceous vegetation.

Various mechanisms explain the C accumulation in soils.The main one is the formation of stable complexes betweenhumic substances and Al, Fe, and non-crystalline compounds,which are protected against microbial degradation (Wada, 1989;Nanzyo et al., 1993a; Nezeyimana, 1997). Boudot et al. (1989)and Zech et al. (1997) consider that the protection of organicmatter from microbial decomposition is related to its insolubil-ity and the interaction of the organic compounds with the short-range compounds that form organic-mineral compounds. In the80's, an inverse relationship between SOC accumulation andallophanic clay formation in Andosols under a udic moistureregime was established, since organic matter plays an “anti-allophanic” role (Shoji and Fujiwara, 1984). Al preferablybuilds up Al–humus complexes and thus is not available tocombine itself with Si to form allophanes. Microbial activity inAndosol is quite low, due to the low availability of phosphate,acid to slightly acid pH and high organic matter stability(Nanzyo, 2002).

Batjes and Sombroek (1997) compared the vertical distribu-tion of the SOC content in different tropical soils and pointedout that contrary to other soils, the SOC content in Andosols ishigh in the first 15 cm, then decreases by half at a depth of40 cm, and then does not vary much from 40 to 200 cm depth.The SOC content in the topsoil (0–15 cm) of both profiles of“La Loma” is lower (54 g kg−1) compared to other Andosols:100 g kg−1 in the Andosols from Java Island (Van-Ranst et al.,2002), 140 g kg−1 in the Andosols from South Ecuador(Buytaert et al., 2002), 95 g kg−1 in the Chilean volcanic soils(Escudey et al., 2004).

Fig. 11 shows the dissolved and particulate mean organiccarbon (OC) concentrations for each sampled runoff event.About half of the runoff events were sampled during the 2003rainy season. The sampled runoff events represent 45% of thetotal runoff during the rainy season, and about 68% of theerosion. Particulate OC concentrations were fairly stable, with a7.9% mean concentration (the standard deviation is 0.8). Theparticulate OC loss measured in 2003 was 416 kg C ha−1 whichmeans that the total particulate OC loss by runoff and erosionduring the 2003 rainy season approaches 600 kg C ha−1. It

Fig. 11. Dissolved and particulate mean organic carbon concentrations for eachsampled runoff event of the maize plot.


represents between 15 to 20% of the SOC stock in the first cm ofthe soil profile. Dissolved OC concentrations (in filtered runoffwater) were more variable. The sampled dissolved OC loss inrunoff water is negligible, approaching 3 kg C ha−1 (i.e. a totalof about 6 kg C ha−1 for the 2003 rainy season). Dixon andSchulze (2002) mentioned as well that allophanic soils areresistant to erosion compared to other soils. Allophanes andorganic matter interactions form aggregates resistant to adversemechanical effects of raindrops and runoff. High hydraulicconductivity and water retention capacity in Andosols alsodecrease water runoff.

As local agriculture is practiced in La Loma, most plantresidues are not incorporated at the end of the cropping season,which means that the organic matter pool is not renewed. Theorganic horizon must have been lost due to decreasing fertilityover the cropping years or when clearing the plots for maizecropping. Under forest the superficial runoff and erosionwere measured as negligible, SOC could have been lost bysubsurface flow.

3.4. Soil classification and andic properties

The andic properties analysed in the fine soil fraction (b2mm)were determined following the standards used to classifyAndosols by FAO-ISRIC and ISSS (1998): bulk densityb0.90 g cm−3, phosphate retention≥70%, volcanic glass contentin the fine earth fraction b10% and Alox+1/2 Feox ≥2%.

The whole maize profile and the two first horizons of theforest profile present a soil bulk density lower than 0.9 g cm−3

(see Table 2), which is characteristic of Andosols. In the maizeprofile, bulk density decreases with the increase of the Sioxcontent, the allophanes being among the non-crystallinematerials contributing to the low soil bulk density. Nanzyoet al. (1993b) explain that fresh ashes have a bulk density above1.5 g cm−3, and this value decreases with weathering, devel-oping a soil porous structure thanks to the presence of non-crystalline materials and organic matter.

In the whole maize profile, phosphate retention was higherthan 90%, whereas in the forest profile this retention decreasesfrom over 90% in the two first horizons to 81.7% at 100 cm depth,along with Alox and Feox contents (Table 4). According to Parfittand Clayden (1991), this characteristic corresponds to a soil withwell developed andic properties. Nanzyo et al. (1993a) show thatAndosols have a high adsorption capacity for fluor and phosphate,due to the high content of active Al and Fe compounds (acidoxalate extracted). This capacity can be linked to the pHmeasuredin NaF, which directly determines the adsorption level of fluor.Phosphate retention depends as well on the pH in water, but nodirect relation could be found in this study between these twoparameters. Nevertheless, the high P-retention in the Bw horizonsof the forest profile was attributed to the presence of gibbsite,which adsorbs phosphate through ligand exchange (Parffit, 1978)and Fe oxides (Gunjigake and Wada, 1981).

The Alox+1 / 2 Feox relation is over 2 along the whole maizeprofile and in the two first horizons in the forest profile(Table 4). According to Parfitt and Clayden (1991), Soil SurveyStaff (1999) and FAO-ISRIC and ISSS (1998), this means that

amorphous components such as allophane, imogolite andferrihydrite are dominant in the soil.

According to the FAO-ISRIC and ISSS (1998) system, themaize profile soil was classified as an Andosol (PachicAndosol) while the forest one corresponds to a Cambisol(Dystric Cambisol). According to the Soil Survey Staff (1999),the maize profile was classified as an Andisol (TypicHapludand) and the forest profile as Inceptisol (TypicDystrudept). The Typic Hapludand corresponds to deepersoils located in the middle and lower areas of the volcanicrelief and which are the result of or are affected by volcanicashes (Alvarado and Bornemisza, 1984). These same soils,more developed, where Bw horizons have formed, like theforest profile, have been classified as Inceptisols, with andiccharacteristics, being the Typic one of the most frequent groups.

4. Conclusions

The morphology of the soil profiles was affected by thetopography. Morphological features and chemical and physicalanalyses allowed to determine that the soil under the forest(upper part of the toposequence) was an Inceptisol (TypicDystrudept) and that in the lower portion was an Andisol (TypicHapludand). The first layers of the latter soil profile resultedfrom the redistribution of the soil from the upper part of thetoposequence. The land use change favored this redistribution.Deeper horizons of this profile were developed from volcanicashes deposited in situ. Hematite and ferrihydrite were con-sidered markers of evolution in the redistributed volcanic ashes.Allophane and Fe mineral were the main minerals in TypicHapludand (maize profile) and the Fe–Al complexes andhalloysite were present in a lower proportion. In contrast, in theA horizon of the Typic Dystrudept (forest profile) the mainmineral components were Fe–Al complexes and in the Bhorizon the gibbsite. In both horizons the Fe minerals wereassociated with the presence of gibbsite and halloysite. Theposition in the toposequence and the physical and chemicalcharacteristics of these profiles determined their present func-tioning, like losses from erosion and C dynamics.

Acknowledgements

Blanca Prado is indebted to CONACYT (Mexico) andSFERE (France) for her Ph D. grant. The research was fundedby the “Institut de Recherche pour le Développement” (IRD),France. H. Yee-Madeira is COFAA-Fellow.

References

Aguilera, H.N., 1963. Algunas consideraciones, características, génesis yclasificación de los suelos de Ando. In: Sociedad Mexicana de la Cienciadel Suelo (Ed.), Proc. 1st National Congress of Soil Science, D.F.,Mexico, pp. 233–247.

Alcalá, M., 2003. Los Andosoles de la Meseta Tarasca, Michoacán. Suclasificación taxonómica, propiedades y mineralogía. Ph.D. Thesis, Colegiode Postgraduados, Montecillo, México.

Alcalá, M., Ortiz, C., Gutiérrez, Ma. Del Carmen, 2001. Clasificación de lossuelos de la Meseta Tarasca, Michoacán. Terra 19, 227–239.


Alvarado, A., Bornemisza, E., 1984. Management and classification of andisolsof Costa Rica. In: Beinroth, F.H., Luzio, W., Maldonado, F., Eswaran, H.(Eds.), Proceedings of the Sixth International Soil Classification Workshop,Chile and Ecuador. Paper I: Papers. Sociedad Chilena de la Ciencia delSuelo, Santiago, Chile, pp. 69–96.

Alvarado, H.A., Bertsch, H.F., Bornemisza, S.E., Cabalceta, A.G., Forsythe, H.W.,Henríquez, H.C., Mata, Ch.R., Molina, R.E., Salas, C.R., 2001. Suelosderivados de cenizas volcánicas (Andisoles) de Costa Rica. 1a ed. AsociaciónCostarricense de la Ciencia del Suelo (ACCS). Universidad de Costa Rica,Cetro de Investigaciones Agronómicas (CIA). San José, C. R.

Batjes, N.H., Sombroek, W.G., 1997. Possibilities for carbon sequestration intropical and subtropical soils. Global Change Biology 3 (2), 161–173.

Bigham, J.M., Fitzpatrick, R.W., Schulze, D.G., 2002. Iron oxides. In: Dixon, J.B.,Schulze, D.G. (Eds.), Soil Mineralogy with Environmental Applications.SSSA Book Ser. No, vol. 7. Soil Sci. Soc. Am., Madison, WI, pp. 323–366.

Blakemore, L.C., Searle, P.L., Daly, B.K., 1987. Technical Paper No. 9,Procedures for Soil Analysis. International Soil Reference and InformationCentre, Wageningen, Netherlands.

Boudot, J.P., Bel Hadj, B.A., Steiman, R., Seigle-Murandi, F., 1989.Biodegradation of synthetic organo–metallic complexes of iron andaluminiumwith selected metal to carbon ratios. Soil Biology&Biochemistry21, 961–966.

Buurman, P., Boer, K.D., Pape, T., 1997. Laser diffraction grain-sizecharacteristics of Andosols in perhumid Costa Rica: the aggregate size ofallophane. Geoderma 78, 71–91.

Buytaert, W., Deckers, J., Dercon, G., Bievre, B.D., Govers, J.P., 2002. Impactof land use changes on the hydrological properties of volcanic ash soils inSouth Ecuador. Soil Use and Management 18, 94–100.

Campos-Cascaredo, A., Oleschko, K., Cruz, L., Etchevers, J., Hidalgo, C., 2001.Estimación de alofano y su relación con otros parámetros químicos enAndosoles de montaña del volcán Cofre de Perote. Terra 19 (2), 105–116.

Chartres, C.J., van Reuler, H., 1985. Mineralogical changes with depth in alayered Andosol near Bandung, Java (Indonesia). Journal of Soil Science 36,173–186.

Childs, C.W., Matsue, N., Yoshinaga, N., 1991. Ferrihydrite in volcanic ash soilsof Japan. Soil Science and Plant Nutrition 37, 299–311.

Cortés, L.A., 1966. Estudio pedológico de la zona oriental de Iztaccihuatl ensuelos derivados de rocas y cenizas volcánicas andesitas. M.Sc. Thesis,Colegio de Postgraduados, Chapingo, Mexico.

Cruz, L., Campos, A., 1997. Propiedades químicas y morfológicas de unasecuencia de suelos con material volcánico de diferentes grados dealteración. Terra 15 (3), 267–278.

Cruz-Huerta, L., Geissert, D., 2000. Carga eléctrica de los andosoles del Cofrede Perote, Veracruz México. Terra 18 (2), 115–124.

Cuanalo, C., 1990. Manual para la descripción de perfiles de suelos en el campo,3rd ed. Colegio de Postgraduados, Chapingo, México.

Dahlgren, R.A., Walker, W.J., 1993. Aluminum release rates from selectedSpodsol Bs horizons: effect of pH and solid-phase aluminum pools.Geochimica et Cosmochimica Acta 57, 57–66.

Dahlgren, R.A., Saigusa, M., Ugolini, F.C., 2004. The nature, properties andmanagement of volcanic soils. Advances in Agronomy 82, 113–182.

Dixon, J.B., Schulze, D.G., 2002. Soil Mineralogy with EnvironmentalApplications. Soil Science Society of America, Madison, WI. 866 pp.

Escudey, M., Forster, J.E., Galindo, G., 2004. Relevance of organic matter insome chemical and physical characteristics of volcanic ash-derived Soils.Communications in Soil Science and Plant Analysis 35 (5&6), 781–797.

Eswaran, H., 1972. Morphology of allophane, imogolite and halloysite. Claysand Clay Minerals 9, 281–285.

Etchevers, J., Oleschko, K., Cruz, L., Hidalgo, C., 1986. Complex nature of themechanical elements in Andosols from Mexico. XIII Congress of theInternational Society of Soil Science, Hamburg, Germany.

FAO-ISRIC, ISSS, 1998. World Reference Base for Soil Resources (WRB).World Soil Resources No. 84. FAO, Rome.

Gunjigake, N., Wada, K., 1981. Effects of phosphorus concentration and pH onphosphate retention by active aluminium and iron of Ando soils. SoilScience 132, 347–352.

Hidalgo, C., Etchevers, J., Martinez, B., 1986. Determinación de la densidadaparente en suelos de Ando. Terra 4, 3–9.

Hidalgo, C., Etchevers, J.D., Quantin, P., 1991. Imogolita en un Andisol deMéxico. Turrialba 41 (4), 509–514.

INEGI, 1999. Superficies Continental e Insular del Territorio Nacional. INEGI,D.F., México.

Leamy, M.L., 1984. Andosols of the World. In: Universidad de la Laguna (Ed.),International Congress of Volcanic Soils. Secretariado de Publicaciones, LaLaguna, Spain, pp. 369–387.

Lelong, F., Souchier, B., 1987. Métodos de identificación y cuantificación de losconstituyentes del suelo. In: Bonneau, M., Souchier, B. (Eds.), Constitu-yentes y propiedades del suelo. Masson, S. A, Barcelona, España, pp. 1–16.Versión española.

Malucelli, F., Terribile, F., Colombo, C., 1999. Mineralogy, micromorphologyand chemical analysis of andosols on the island of Sao Miguel (Azores).Geoderma 88, 73–98.

Mehra, O.P., Jackson, M.L., 1960. Iron oxide removal from soil and clays by adithionite–citrate systems buffered with sodium bicarbonate. Clays and ClayMinerals 7, 317–327.

Miehlich, G., 1991. Chronosequences of volcanic ash soils. HamburgerBodenkundliche Arbeiten 15, Hamburg, Germany. 207 pp.

Mizota, C., van Reeuwijk, L.P., 1989. Clay mineralogy and chemistry of soilsformed in volcanic material in diverse climatic regions. Soil Monograph,vol. 2. ISRIC, Wageningen.

Murphy, C.P., 1986. Thin Section Preparation of Soil and Sediments. ABAcademic Publishers, Great Britain.

Nanzyo, M., 2002. Unique properties of volcanic ash soils. Global Environ-mental Research 6 (2), 99–112.

Nanzyo, M., Dahlgren, R., Shoji, S., 1993a. Chemical characteristics of volcanic ashsoils. In: Shoji, S.,Nanzyo,M.,Dahlgren,R.A. (Eds.),Volcanic ash soils, genesis,properties and utilization. Elsevier, Amsterdam, The Netherlands, pp. 145–187.

Nanzyo, M., Shoji, S., Dahlgren, R., 1993b. Physical Characteristic of volcanicash soils. In: Shoji, S., Nanzyo, M., Dahlgren, R.A. (Eds.), Volcanic ashsoils, genesis, properties and utilization, The Netherlands, pp. 189–207.

Nezeyimana, E., 1997. A toposequence of soils derived from volcanic materialsin Rwanda: morphological, chemical and physical properties. Soil Science162, 351–360.

NOM-021-RECNAT, 2000. Especificaciones de fertilidad, salinidad y clasifica-ción de suelos. Estudios, muestreo y análisis, Secretaria del medio ambiente yrecursos naturales, D.F., Mexico.

Parffit, R.L., 1978. Anion adsorption by soils and soil materials. Advance inAgronomy 30, 1–50.

Parfitt, R.L., 1980. Chemical properties of variable charge soils. In: New ZealandSociety of Soil Science (Ed.), Soils with Variable Charge. K. G. Theng, LowerHutt, N. Z.

Parfitt, R.L., 1990. Allophane in New Zealand— a review. Australian Journal ofSoil Research 28, 343–360.

Parfitt, R.L., Clayden, B., 1991. Andosols— the development of a new order insoil taxonomy. Geoderma 49, 181–198.

Parfitt, R.L., Saigusa, M., 1985. Allophane and humus–aluminum in Spodosolsand Andepts formed from the same volcanic rock beds in New Zealand. SoilScience 139, 149–155.

Parfitt, R.L., Russell, M., Orbell, G.E., 1983. Weathering sequence of soils fromvolcanic ash involving allophane and halloysite. New Zealand. Geoderma29, 41–57.

Parfitt, R.L., Childs, C.W., Eden, D.N., 1988. Ferrihydrite and allophane in fourAndepts from Hawaii and implications for their classification. Geoderma 41,223–241.

Porta-Castellanos, J., Regueria, M.L.-A., Roquero de, L.C., 2003. Edafologíapara la agricultura y el medio ambiente. Mundiprensa, España.

Prado, B., Duwig, C., Escudey, M., Esteves, M., 2006. Nitrate sorption in aMexican allophanic Andosol using intact and packed columns. Commu-nications in Soil Science and Plant Analysis 37, 2911–2925.

Quantin, P., 1986. Properties and genesis of andosols. Cahier Orstom, ser Pedol,vol XXII no. 1. 6th international soil classification workshop in Chile andEquator, pp. 105–111.

Russell, M., Parfitt, R.L., Claridge, G.G.C., 1981. Estimation of the amounts ofAllophane and other materials in the clay fraction of an Egmont loam profileand other volcanic ash soils. New Zealand. Australian Journal of SoilScience 19, 185–195.


Schwertmann, U., 1988. Occurrence and formation of iron oxides in variouspedoenvironments. In: Stucki, J.W., Goodman, B.A., Schwertmann, U.(Eds.), Iron in Soils and Clay Minerals. D. Reidel Publishing Company,Dordrecht, pp. 267–308.

Shoji, S., Fujiwara, F., 1984. Active aluminum and iron in the humus horizons ofAndosols from Northeastern Japan: their forms, properties, and significancein clay weathering. Soil Science 137, 216–226.

Shoji, S., Takahashi, T., Ito, T., Ping, C.L., 1988. Properties and classification ofselected volcanic ash soils from Kenai Peninsula, Alaska. Soil Science 145,395–413.

Shoji, S., Nanzyo, M., Dahlgren, R.A., Quantin, P., 1996. Evaluation andproposed revisions of criteria for andosols in the world reference base forsoil resources. Soil Science 161 (9), 604–614.

Soil Survey Staff, 1999. Soil Taxonomy. A basic system of soil classification formaking and interpreting soil surveys. Agriculture Handbook 436.Department of Agriculture, Natural Resources Conservation Service.Washington, D. C., United States.

Takahashi, T., Shoji, S., 2002. Distribution and classification of volcanic ashsoils. Global Environmental Research 6 (2), 83–97.

Tokashiki, Y., Wada, K., 1975. Weathering implications of the mineralogy ofclay fractions of two ando soils, Kyushu. Geoderma 14, 47–62.

Uehara, G., Gillman, G., 1981. The Mineralogy, Chemistry, and Physics ofTropical Soils with Variable Charge Clays. Westview Press, Boulder,Colorado.

Valera, M.A., 1993. Fisicoquímica y mineralogía de Andosoles de la region deTeziutlan, Puebla. MSc. Thesis, Universidad Autónoma de México, Mexico.

Van-Ranst, E., Utami, S.R., Schamshuddin, J., 2002. Andosols on volcanic ashfrom Jaa Island, Indonesia: physico-chemical properties and classification.Soil Science 167 (1), 68–79.

Viramontes, D., Rojas, F., Duwig, C., Esteves, M., 2006. Caracterización hidro-erosiva de suelos de tipo andosol e importancia de este recurso en el Estadode Michoacán. Las ciencias del agua en Morelia, aplicaciones frente a losretos del siglo, vol. XXI.

Wada, K., 1985. The distinctive properties of andosols. Advances in SoilScience 2, 174–229.

Wada, K., 1987. Minerals formed and mineral formation from volcanic ash byweathering. Chemical Geology 60, 17–28.

Wada, K., 1989. Allophane and imogolite, In: Dixon, J.B., Weeds, S.B. (Eds.),Minerals in Soil Environment, 2nd ed. Soil Sci. Soc. Am. Book, MadisonWI, pp. 1051–1087.

Wada, K., Kakuto, Y., Ikawa, H., 1986. Clay minerals, humus complexes, andclassification of four “Andepts” of Maui, Hawaii. Soil Science Society ofAmerica Journal 50, 1007–1013.

Wambeke, A.V., Hastings, P., Tolomeo, M., 1986. Newhall Simulation Model.Department of Agronomy, Cornell University, Ithaca, N.Y.

Werner, G., 1985. Volcanic ash soils and erosion. Natural Resources andDevelopment 22, 35–45.

Wilson, M.J., 2004. Weathering of the primary rock-forming minerals:processes, products and rates. Clay Minerals 39, 233–266.

Zech, W., Senesi, N., Guggenberger, G., Kaiser, K., Lehmann, J., Miano, T.M.,Miltner, A., Schroth, G., 1997. Factors controlling humification andmineralization of soil organic matter in the tropics. Geoderma 79, 117–161.

TECHNICAL ARTICLES

TRANSPORT AND SORPTION OF 2,4-DICHLOROPHENOXYACETICACID IN ALLOPHANIC SOILS

Karin Muller1 and Celine Duwig2

Field applications of the herbicide 2,4-dichlorophenoxyacetic acid(2,4-D) are commonly formulated as esters, amine salts, or alkali, butstudies on the environmental impact focus on the 2,4-D acid form. Wehypothesized that the formulation would affect its transport. To betterunderstand 2,4-D transport through allophanic soils, leaching experi-ments with 2,4-D acetic acid, 2,4-D ethylhexyl ester, and tritium wereconducted in 100 mm of repacked soil columns. Three of the four siltloam soils were allophanic. The 2,4-D sorption isotherms and kinetics ofsorption were determined from batch experiments. The symmetricaltritium breakthrough curves (BTCs) showed that water transport was inphysical equilibrium, and these were used to estimate the dispersivity ofthe soils. Tritium was not inert in the allophanic soils. The BTCs of both2,4-D formulations were shifted to the right of tritium and showedextended tailing characteristic typical for sorption nonequilibrium. The2,4-D BTCs were successfully fit with a two-site chemical nonequili-brium model. Batch and leaching experiments showed the sorption of2,4-D ethylhexyl ester to be higher and characterized by strongernonequilibrium than 2,4-D acetic acid sorption in all soils. The sorptioncoefficient values derived from the BTCs were higher than those from24-h batch experiments. For the two allophanic topsoils, the predictedester BTCs were significantly worse when using the independentlyderived Kd compared with the BTCs using the fitted Kd. 2,4-D sorptionwas correlated to pH and organic matter content. Transport of 2,4-Dwas slower than expected in the allophanic soils and was attributed toretention of dissociated 2,4-D by positively charged soil surfaces. (SoilScience 2007;172:333–348)

Key words: 2,4-D, nonequilibrium transport, tritium, allophanic soil.

PESTICIDE transport has been intensivelyinvestigated in permanent charge soils in

temperate climate zones, but for variable chargesoils in subtropical and tropical climate zones,investigations are still scarce. Allophanic soils arerepresentatives of soils with a pH-dependentcharge and are rich in minerals such as kaolinite,

gibbsite, hematite, or allophanes. These crystal-line and noncrystalline components have reac-tive OH groups that develop appreciable anionexchange capacity under acid conditions (Lairdand Sawhney, 2002). Transport of anions isretarded in variable charged soils. Allophanicsoils belong to the most fertile soils and aresubjected to intensive agriculture. With thegrowing world population, these fertile soilswill be exposed to increasing pressure throughfurther intensification, including higher pesti-cide and fertilizer input as well as irrigation. Oneof the potential problems connected withagricultural intensification is the increased risk

333

0038-075X/07/17205-333–348 May 2007

Soil Science Vol. 172, No. 5

Copyright * 2007 by Lippincott Williams & Wilkins, Inc. Printed in U.S.A.

1AgResearch, Ruakura Research Centre, PB 3123, Hamilton, New Zealand.

Dr Muller is corresponding author. E-mail: [email protected]

2Institut de Recherche pour le Developpement, c/o Colegio de Postgraduados,

Laboratorio de Fertilidad de Suelos, CP 56230 Montecillo, Mexico.

Received Sep. 17, 2006; accepted Jan. 10, 2007.

DOI: 10.1097/SS.0b013e31803bbb85

Copyr ight © Lippincott Williams & Wilkins. Unauthorized reproduction of this article is prohibited.

of surface and ground water contamination bypesticides. The interactions between soil char-acteristics and pesticide properties are governingpesticide mobility and sorption. All pesticidesregardless of their characteristics may be sorbedto soil through van der Waals forces or hydro-gen bonds. More specific sorption mechanismsare dependent on the pesticide properties.Molecular nonionic pesticides are partitionedinto hydrophobic soil organic matter (OM).Anionic or dissociated acidic pesticides such as2,4-dichlorophenoxyacetic acid (2,4-D) may besorbed on OM and clay minerals via anionexchange mechanisms and on Al or Fe hydrox-ides through ligand exchange (Sposito, 1989). Inallophanic soils, OM and short-range alumino-silicates are important sorbents for pesticidesbecause of their variable charges. OM hasdifferent functional groups that adsorb H+-andOHj-forming charged surfaces, which bindionic pesticides. Other positively charged sitesin volcanic soils include AlOH2+ of allophanicclays and FeOH2+ of ferrihydrite. Oxide surfa-ces can develop positive charge at low pHvalues. Interactions between OM and highlyreactive minerals are also important for pesticidesorption.

The herbicide 2,4-D is widely used tocontrol broad-leaf weeds in wheat, small grain,corn, rangeland, and pasture in New Zealandand Mexico. 2,4-D acid is the parent herbicidalmoiety in all registered formulated end-useproducts that contain 2,4-D as an active ingre-dient. Commercial formulated products containin general 2,4-D as an inorganic or amine salt oras an ester. Most studies, however, investigatingthe environmental behavior of 2,4-D have used2,4-D acetic acid. Concerns were voiced that

the environmental profile for 2,4-D-basedproducts might not be sufficient because it wasfounded on studies using only a single form.Consequently, for the reregistration of 2,4-D inthe United States, comprehensive field dissipa-tion studies were conducted with an ester andan amine salt formulation of 2,4-D in a range ofdifferent soil types (Wilson et al., 1997). Resultsshowed that ester and amine form had littleeffect on the dissipation rate of 2,4-D. Bothwere converted rapidly (G24 h) to the sameanionic form. Hydrolysis of ester was slowerthan amine dissociation. Studies on the behaviorof different 2,4-D formulations are rare butcould be important for the fate of 2,4-D shortlyafter its application.

The objective of this research project was tostudy sorption and transport of 2,4-D acetic acidand 2,4-D ethylhexyl ester, applied as theformulated product Pasture-Kleen (Dow Elanco,New Plymouth, New Zealand), in two allo-phanic soils from New Zealand and Mexicousing batch and column experiments. The twospecific objectives were to (i) test the hypothesisthat a pesticide’s formulation impacts on itsbehavior in soils and (ii) show that the uniqueproperties of allophanic soils impact on pesti-cides’ fate in soils.

MATERIAL AND METHODS

Soils

Two allophanic soils were collected from0 to 14 cm depth, and two other soils were anallophanic and a nonallophanic subsoil. Theallophanic and nonallophanic subsoils weretaken between 75 and 85 cm and between 24and 34 cm depth, respectively. The first topsoil

TABLE 1

Main physical and chemical characteristics of the soil studied

SoilBulk density

(g cmj3)

Total

carbon.

(%)

Total

nitrogen

(%)

pH

H2O

CEC-

(cmol kgj1)

Clay

content

(%)

Clay content

without OM (%)

Mexican

topsoil

0.75 4.52 0.26 5.95 35.2 26.2 20.7

Ohakune

topsoil

0.85 9.00 0.78 4.97 38.2 11.9 21.3

Ohakune

subsoil

0.80 3.02 0.21 6.08 23.3 34.7 14.4

Te Kowhai

subsoil

1.10 0.32 0.01 5.23 9.7 18.3 n.d.‘

.Because the pH is acidic, there is no carbonate; therefore, total carbon is equivalent to organic carbon.-CEC = cation exchange capacity.‘n.d. = not determined.

334 MULLER AND DUWIG SOIL SCIENCE


was collected under maize, 150 km west ofMexico City in Mexico at 2500 m height. Thesecond topsoil and the first subsoil were sampledunder oat near Ohakune at an altitude of 650 m,in the Central Volcanic Region of the NorthIsland of New Zealand. The two soils developedfrom similar volcanic deposits (andesetic tephra)under different climatic conditions. They werecomparable in clay composition and OM con-tents (Tables 1 and 2). The Mexican soil isclassified as Typic Hapludand, the Ohakune siltloam soil as Typic Hydrudand (Soil Survey Staff,1992). The second subsoil was the Te Kowhaisilt loam soil, classified as a Haplaquept (SoilSurvey Staff, 1992). This soil was collected at analtitude of 54 m under maize at the WaikatoResearch Orchard near Hamilton, North Island,New Zealand. The parent material of the poorlydrained Te Kowhai silt loam is pumice. Soilproperties for the four soils were determinedwith standard methods and are listed in Tables 1and 2. The clay composition of the Te Kowhaisoil is dominated by halloysite, volcanic glass,and kaolinite (no allophane). The clay fractionsof the Mexican and Ohakune soils dominantlyconsist of amorphous phases, mainly allophane,volcanic glass, and amorphous silica.

Solutes

We compared the sorption behavior of2,4-D in its acetic acid form to the ethylhexylester in a formulated product. The two activeingredients differ in their basic physicochemicalproperties: The 2,4-D acetic acid has forexample a water solubility of 27,644 mg Lj1

(http://www.pesticideinfo.org/Detail_Chemical.jsp?Rec_Id=PC35042), which compares to0.0867 mg Lj1 for the ester (http://www.24d.org/Commission.pdf). The vapor pressureof the acid is 1.9 � 10j5 Pa at 25 -C (Wauchopeet al., 1992), and thus much lower than thevapor pressure of the ester (4.8 � 10j4 Pa at25 -C, http://www.24d.org/Commission.pdf).

The acid’s half-life is approximately 10-foldof that of the ester (http://www.24d.org/Commission.pdf). The weakly acidic compoundhas a pKa value of about 2.8. For theexperiments, 14C-labeled [Ring-U-14] 2,4-D(purity 99%, specific activity 100 2Ci mLj1)was obtained from Amersham Biosciences(Auckland, New Zealand). Part of the aceticacid was esterified using conventional EDCcoupling conditions and isolated in 90% yieldafter chromatography on silica gel. The esterwas then formulated as the emulsifiable concen-trate Pasture-Kleen (personal communication,Dr. Fielder, HortResearch, 2001).

Both solutes were dissolved in a 5-mMCaCl2 solution. Batch and column experimentswere carried out with this background solutionmimicking the electrolytic strength of the soilsolution. For the column experiments, tritiatedwater (3H2O, specific activity = 5 mCi mLj1;Amersham Biosciences) was used as an inertwater tracer.

Batch Experiments

We conducted batch experiments to quan-tify herbicide sorption in the four soils. Sorptionwas measured for a wide range of concentrationsof 14C-radiolabeled 2,4-D acid (0.51, 1.04, 2.05,3.12, 4.15, 11.2, 22, 44.8, 67.2, and 89.6 2gmLj1) and a narrower range of 14C-radiolabeled2,4-D ethylhexyl ester (0.0143, 0.0285, 0.057,0.086, and 0.114 2L mLj1; because of thespecific density of Pasture-Kleen of 1.07 mg/2L, these correspond to 15.3, 30.5, 61, 92,and 122 2g mLj1, respectively). The esterconcentrations were aligned to normal fieldapplication rates of Pasture-Kleen (DowElanco). The range of 2,4-D acid concentrationswas chosen considering two aspects: the highconcentrations correspond to the ester concen-trations (2 L haj1 Pasture-Kleen is equivalent to44.8 2g mLj1 2,4-D acid) and the low concen-trations represent more likely soil residue levels.

TABLE 2

Short-range order minerals of the soil studied.

SoilAcid oxalate (%) Pyrophosphate (%)

Allophane. (%) Ferrihydrite- (%)Fe Al Si Fe Al

Mexican topsoil 1.44 5.31 2.54 0.08 0.37 17.8 2.5

Ohakune topsoil 1.40 4.40 1.40 0.06 0.38 16.1 2.4

Ohakune subsoil 2.01 6.01 2.52 0.76 1.09 17.6 3.4

Te Kowhai subsoil 0.16 0.15 0.14 0.05 0.05 0.4 0.3

.Calculated following Parfitt (1990).-Calculated following Parfitt and Childs (1988).

VOL. 172 ~ NO. 5 2,4-D IN ALLOPHANIC SOILS 335


The fresh soil was stored at 4 -C, homo-genized, and sieved (pore size G 2 mm) shortlybefore the experiments. The gravimetric soilwater content of the sieved soil was determined.Suspensions were prepared for each concentra-tion in triplicate with a soil/solution ratio of 1:2using 5 mM CaCl2 as background electrolyte.Fresh soil equivalent to 2 g of dry soil wasequilibrated with 1 mL of the standard solutionand CaCl2 by shaking on an end-over-endshaker for 24 h at 23 T 1 -C. Suspensions werecentrifuged at 1962 relative centrifugal force for5 min. Preliminary experiments showed thatvarying the centrifugation time from 5 to 30min had no significant impact on the separationof solid and liquid phases. An aliquot of 1 mLwas transferred into a scintillation vial, thor-oughly mixed with 9 mL biodegradable scintil-lation cocktail, and left for 24 h at 10 -C in thedark before triplicate 14C measurements witha liquid scintillation counter (Wallace 1409;Wallace, Turku, Finland) were conducted at astandard error below 5%. Sorption kinetics wereanalyzed by altering the contact time rangingfrom 5 min to 24 h for the 44.8-2g gj1 2,4-Dacetic acid and the 61-2g mLj1 2,4-D ethyl-hexyl ester, respectively. The final herbicideconcentration in solution was calculated usingthe following equation (Baskaran et al., 1996):

Finalconcentration

¼ initial concentration�14C final activity

ð1Þ

The amount of herbicide sorbed to the soilwas calculated as the difference between theinitial and the final concentration of herbicide insolution. Blanks were prepared to verify thatsorption to tubes could be excluded.

Mass balance checks were performed onsome samples by analyzing 14C activity in thealiquot as well as the slurry. The determinationof the specific activity remaining in the slurrywas performed via dry combustion of about0.05 g of slurry in a Biological Material OxidizerOX-600 (R. J. Harvey, Hillsdale, New Jersey)and subsequent measurements on the liquidscintillation counter.

Results from the sorption experiments werefit with the linear sorption isotherm [Eq. (2)]and the empirical Freundlich equation [Eq. (3)].The linear sorption isotherm equation is:

Cs ¼ KdCe; ð2Þ

where Kd characterizes the linear sorptionbehavior, Cs is the sorbed concentration(2g gj1), and Ce is the equilibrium concen-tration in solution (2g mLj1). Nonlinear sorp-tion is frequently described by the empiricalFreundlich equation:

Cs¼ Kf Cne ; ð3Þ

with Kf and n being empirical coefficients of theequation. The coefficient n indicates the non-linear dependence of the sorption rate on thesolute concentration.

Column Experiments

The transport of 2,4-D through saturated soilcolumns was analyzed by miscible displacementtechniques. Glass columns (140 � 28.5 mminternal diameter) were packed to a depth, L, of100 mm with fresh soil material, at a bulkdensity, %b, of 0.743 g cmj3 for the allophanicsoils and 1 g cmj3 for the Te Kowhai silt loam.All tubes were sealed at the base with a nylonmesh with a pore diameter of 60 2m to retainthe soil in the columns. The columns weresaturated with three pore volumes of 5-mMCaCl2 solution to establish steady-state hydro-dynamic conditions and equilibrium in the soilsolution composition. The columns were thenspiked with a pulse of 2,4-D acetic acid (C0)using the double field application rate (2.24 mLof the 89.6 2g mLj1 standard with a specificactivity of 6.1 � 105 d.p.m. mLj1), or of theformulated product Pasture-Kleen (2.2 mL ofthe 122-2g mLj1 standard with a specificactivity of 3.3 � 105 d.p.m. mLj1). Tocharacterize the hydrodynamic properties ofthe soil columns, tritiated water (3H2O) wasapplied simultaneously with the 2,4-D spikes(2.5 mL of a 3H2O standard with a specificactivity of 5 � 105 d.p.m. mLj1). The pulse wasleached through the columns by a steady-stateflux of 9 or 12 pore volumes of CaCl2 for 2,4-Dacetic acid and 2,4-D ethylhexyl ester, respec-tively, using a peristaltic pump (Watson-Marlow505U; Watson-Marlow, Cornwall, UK). Col-umn effluents were sampled time-proportionalby taking a leachate sample every 16 min. Analiquot of 1 mL was mixed with 9 mL ofscintillation cocktail, and the specific activitywas measured as in the batch experiments. Forthe leachate samples, a dual-counting protocolfor 3H and 14C specific activity was developed.The conditions for all column experiments aresummarized in Table 3.

14C initial activity



At the end of the experiment, the columnswere sliced into five layers of 2 cm depth each.A subsample from each layer was analyzed forthe 14C specific activity using the oxidizer todetermine the resident concentrations in thesoil. The remainder of the soil was used tomeasure the gravimetric soil water content.

The total recovery of 14C herbicide Mt wascalculated for the column experiments:

Mt ¼ Ml þMs ð4Þ

with Ml being the total 14C herbicide leachedfrom the columns calculated via triangulation:

Ml ¼ ~n

i¼0

ðCli þ Clðiþ1ÞÞ=2Iðtðiþ1ÞjtiÞ ð5Þ

where Cli is the concentration (d.p.m.) of 14Cherbicide in the leachate at the time ti, ti is thetime (min), and n the number of leachatesamples. In Eq. (4), Ms is an estimate of thetotal amount of 14C herbicide remaining in thesoil at the end of the experiments and wascalculated as:

Ms ¼ ~5

m¼1

CdsmMdsm

Mdssubm

100

Oxeffð6Þ

where Mdsm is the wet soil mass of the soil layerm, Mdssubm is the subsample of the soil layer mthat was used for determining the amount of 14Cherbicide in this soil layer (Cdsm), and Oxeff is acorrection factor taking into account the effi-ciency of the oxidizer measurements.

Mathematical Model

Reactive solute transport during steady-stateflow can be mathematically described by theclassical convection-dispersion equation (CDE),which assumes equilibrium transport and sorp-

tion (Kut<lek and Nielsen, 1994). In its dimen-sionless form, the equation can be written as:

Rflc

flT¼ 1

p

fl2c

fl2Z2j

flc

flzð7Þ

where R is the retardation factor (R = 1 +%bKd/E, with E as the volumetric water contentand %b the bulk density), c(Z,T) is the normal-ized solute concentration (C/C0, with C0 as theconcentration of the applied solution), T is thedimensionless time or pore volume (T = vt/L,and v = q/E), P is the column Peclet number(P = vL/D, with L as the column length and Dthe dispersivity), and Z is the normalized length(Z = z/L). All parameters except R and P weremeasured during the column experiments andtransformed into dimensionless parameters(Table 3). Assuming the tritiated water to bean inert tracer (i.e., Kd = 0), the parameter Rwas set to 1 to simulate the 3H breakthroughcurves (BTCs). The remaining parameter P wasobtained by fitting the CDE to the experimental3H BTCs, using the CXTFIT curve-fittingprogram in its inverse mode. CXTFIT usesanalytical solutions for solving the CDE forsteady-state conditions (Leij and Toride, 1998).The inverse procedure was applied to all experi-ments assuming an initially solute-free soilprofile subjected to a pulse application of tritiatedwater under steady-state water flux. For thesoil-surface boundary condition, a first or con-centration type condition was used:

cð0;TÞ ¼ A%ðTÞ ð7bÞ

where %(T) denotes the delta function and A thetotal amount applied.

The two-region transport model focuses onphysical processes and divides the soil water intoa mobile and an immobile water region (van

TABLE 3

Conditions for the column experiments with 2,4-D acetic acid and 2,4-D ethylhexyl ester in the four soils

Soil typeBulk density

(g cmj3)

Soil water content

(m3 mj3)

Flow rate

(mL minj1)

Pulse T03H2O

Pulse T014C

2,4-D acetic acid

Mexican topsoil 0.749 0.644 0.235 0.0609 0.0545

Ohakune topsoil 0.726 0.648 0.235 0.0587 0.0526

Ohakune subsoil 0.710 0.590 0.235 0.0633 0.0567

Te Kowhai subsoil 1.003 0.525 0.145 0.0746 0.0669

2,4-D ethylhexyl ester

Mexican topsoil 0.749 0.651 0.235 0.0602 0.0530

Ohakune topsoil 0.749 0.665 0.236 0.0589 0.0519

Ohakune subsoil 0.735 0.598 0.236 0.0642 0.0565

Te Kowhai subsoil 0.998 0.603 0.233 0.0650 0.0572



Genuchten and Wierenga, 1976). The two-sitesorption model considers the nonequilibriumobserved as a chemical process and suggests thatsorption can be described as a two-site adsorp-tion mechanism. Adsorption on one fraction ofthe sorption sites is regarded an instantaneousprocess, whereas adsorption to the remainingsites is time-dependent. Here, ! is the first-orderrate coefficient for kinetic sorption and f is thefraction of adsorption sites assumed to be atinstantaneous equilibrium. Sorption-relatednonequilibrium may not only result fromchemical nonequilibrium, but also from rate-limited diffusive mass transfer (Brusseau et al.,1991). The dimensionless transport equationsfor physical and chemical nonequilibrium sit-uation are mathematically identical according toNkedi-Kizza et al. (1984) and can be writtenadimensionally as:

"Rflc1

flTþ 1j"ð ÞR flc2

flT ¼ 1

P

fl2c1

fl2Z2j

flc1

flZð8Þ

1j"ð ÞR flc2

flT¼ 5 c1jc2ð Þ ð9Þ

where " is a partition coefficient [Eq. (10)], 5is a mass transfer coefficient [Eq. (11)], c1 is thedimensionless concentration of the mobileregion (two-region transport model) or theentire solution (two-site sorption model), andc2 is the concentration in the immobile region orthe adsorbed concentration for the two-sitesorption model. The total volumetric watercontent E for the two-region model is given asthe sum of the volumetric water content in themobile, Em, and in the immobile, Eim, regions.

" ¼ Eþ f %k

Eþ %kfor the two � site model and

" ¼ Em þ f %k

Eþ %kfor the two � region model ð10Þ

5 ¼ !ð1j"ÞRL8

for the two � site and

5 ¼ !L

E8for the two � region model ð11Þ

The nonlinear sorption parameter Kf fromthe batch experiments was used to obtain alinearized sorption coefficient KL, which wethen used to estimate the retardation factor R inour simulations.

KL ¼2Kf C

nj10

nþ 1ð12Þ

where C0 is the maximum concentration usedfor as input concentration in the batch experi-ments (van Genuchten, 1981).

Initial and boundary conditions from fittingthe tracer 3H BTCs, as well as the determinedPeclet number, were kept for fitting theherbicide BTCs. Therefore, only two parame-ters, namely, " and 5, had to be found throughfitting the curve to the experimental data. Thesetwo parameters are inversely correlated to asystem’s nonequilibrium; the higher their values,the lower the nonequilibrium. To obtain rea-sonable first estimates of the adjustable dimen-sionless parameters, the method of moments wasapplied (Kamra et al., 2001).

The fitted parameter values were then usedto calculate the dimensional transport parametersdispersivity, 1 (mm), the distribution coefficient,Kd (L kgj1), the first-order kinetic rate coef-ficient, ! (sj1), and the fraction of exchangesites, f (j). Finally, the optimized parameter setwas used to predict the depth distribution of theherbicides resident in the soil column at the endof the displacement experiment.

RESULTS AND DISCUSSION

Sorption Isotherms

We did not analyze the chemical nature ofthe radioactivity and assumed that degradationwas negligible during these short-term experi-ments based on information found in theliterature. Estrella et al. (1993) reported that2,4-D mineralization was preceded by a 3-daylag phase in their batch and column experi-ments. In saturated column experiments, theynoticed no buildup of biodegrading populationsdue to lack of time and oxygen. Hornsby et al.(1996) averaged 2,4-D’s half-life to 10 days froma variety of studies with different soil types andconditions and estimated that within 24 h, 93%of the initial 2,4-D should remain.

In the range of the concentrations studied,the shapes of the isotherms for all soils were "L"type (Fig. 1), which is mathematically describedby the Freundlich equation. The isotherm for2,4-D acetic acid and the Te Kowhai subsoilcould also be classified as BC[-type adsorptionisotherm or the linear sorption isotherm model[Eq. (2)], which represents a special linear caseof the BL^ curve characterized by relativelyhomogeneous adsorption sites with constantaffinity. The maximum sorption parameter forboth compounds was determined for the Mex-ican topsoil followed by the Ohakune topsoil



(Tables 4 and 5). The formulated 2,4-D producthad higher sorption coefficients than 2,4-Dacetic acid. In all experiments, the Freundlichsorption isotherm was superior to the linearsorption isotherm. Sorption is stronger at low

concentrations if n G 1 (Chen and Wagenet,1997). This explains that decreasing the maximum2,4-D acetic acid concentration of the concen-tration range used for the isotherms from about90 to below 5 2g mLj1 resulted in an increase

Fig. 1. Sorption isotherms for (A) 2,4-D acetic acid for a high (left) and a low (right) concentration range and(B) 2,4-D ethylhexyl ester for the four soils. MTS = Mexican topsoil; OTS = Ohakune topsoil; OSS = Ohakunesubsoil; TKSS = Te Kowhai subsoil.

TABLE 4

Sorption and regression coefficients for linear and Freundlich isotherms of 2,4-D acetic acid in the four soils.

SoilsFreundlich isotherm Linear isotherm

Error- (%)Kf n R2 Kd (L kgj1) R2 KL

Low range‘

Mexican topsoil 8.79 0.88 0.993 11.32 0.977 7.81 31.0

Ohakune topsoil 6.55 0.78 0.999 10.12 0.966 5.29 47.8

Ohakune subsoil 1.90 0.85 0.989 2.18 0.965 1.64 24.8

Te Kowhai subsoil 0.41 1.00 0.807 0.41 0.651 0.41 0

High rangeP

Mexican topsoil 7.20 0.82 0.998 5.23 0.988 3.52 32.6

Ohakune topsoil 6.05 0.76 0.999 3.92 0.982 2.34 40.4

Ohakune subsoil 2.12 0.95 0.997 1.89 0.990 1.74 8.1


.All measurements were conducted in three repetitions.-Absolute error in Kd relative to KL.‘Equations are based on five concentrations ranging from 0 to 4.15 2g mLj1.PEquations are based on 10 concentrations ranging from 0 to 89.6 2g mL-1.



in Kd values for all four soil types rangingfrom 13% for the Ohakune subsoil to 61% for

both the Ohakune topsoil and the Te Kowhaisubsoil (Table 4). Hu and Brusseau (1998) foundthe same concentration-dependent sorptionbehavior for 2,4-D acetic acid in a sandy loamsoil. Consequently, retardation may increase asconcentrations decrease because of dispersion.For the isotherms of the formulated product andthe two subsoils, n was larger than 1. In this case,sorbed solute mitigates further sorption. Welinearized the nonlinear sorption coefficients byapplying Eq. (12). Using the linear sorptioncoefficients instead of the linearized sorptioncoefficients led to a maximum absolute error of48% for 2,4-D acetic acid sorption in theOhakune topsoil and 24% for 2,4-D ethylhexylester in the Mexican topsoil (Tables 4 and 5).Average absolute errors were 31 and 14% for2,4-D acetic acid and 2,4-D ethylhexyl ester,respectively.

Sorption Kinetics

Pesticide sorption is often described as atwo-phase process (Smith et al., 2003). Fastsurface processes are followed by slower diffu-sion processes into and out of micropores of soilaggregates, OM, and minerals (Pignatello andXing, 1996). In our experiments (using thestandard of 44.8 2g mLj1 2,4-D acetic acid),the same trend was observed for 2,4-D aceticacid: a rapid sorption within the first 5 min of84, 75, 81, and 38% of the total amount sorbedin the Mexican topsoil, the Ohakune topsoil,the Ohakune subsoil, and the Te Kowhaisubsoil, respectively, was followed by a muchslower sorption rate (Fig. 2). The rapid surfacesorption sites were saturated by the weaklysorbed 2,4-D anions within the first 5 min ofcontact time. At all times, sorption rates in theMexican topsoil were highest, and sorptionrates to the Te Kowhai subsoil were the lowest.The sorption rates during the experiments couldbe described by power equations with determi-nation coefficients of more than 0.98 and

TABLE 5

Sorption coefficients and regression coefficients for linear and Freundlich isotherms of 2,4-D ethylhexyl ester in the four soils.

SoilsFreundlich isotherm Linear isotherm

Error- (%)Kf n R2 Kd (L kgj1) R2 KL

Mexican topsoil 9.88 0.84 0.976 7.42 0.909 5.62 24.2

Ohakune topsoil 7.68 0.84 0.987 5.68 0.916 4.37 23.0

Ohakune subsoil 1.87 1.06 0.988 2.14 0.964 2.21 8.1


.All measurements were conducted in three repetitions.-Absolute error in Kd.

Fig. 2. Sorption kinetics for (A) 2,4-D acetic acid and(B) 2,4-D ethylhexyl ester for the four soils. MTS =Mexican topsoil; OTS = Ohakune topsoil; OSS =Ohakune subsoil; TKSS = Te Kowhai subsoil.



exponential coefficients of j0.77, j0.75,j0.75, and j0.69 for the Mexican topsoil, theOhakune topsoil, the Ohakune subsoil, and theTe Kowhai subsoil, respectively. 2,4-D aceticacid continued to be sorbed to all four soilsduring the entire experimental period. Equi-librium was apparently not reached. At the endof the experiment, 37, 35, 25, and 5% of the2,4-D acetic acid were sorbed to the Mexicantopsoil, the Ohakune topsoil, the Ohakunesubsoil, and the Te Kowhai subsoil, respectively.

The 2,4-D ethylhexyl ester sorption kineticswere very different. The experiment wasrepeated to confirm the results. For all soil types,the highest sorption was observed after the first5 min (Fig. 2). During the subsequent 24-hequilibrium time, Bdesorption^ was observedresulting in final amounts of 37, 36, 24, and 7%2,4-D ethylhexyl ester sorbed to the Mexicantopsoil, Ohakune topsoil, Ohakune subsoil, andTe Kowhai subsoil, respectively. We checked thechemical nature of the radioactivity via high-performance liquid chromatography for a fewsamples (Muller et al., 2005) without attemptingquantification of results. The 2,4-D ethylhexylester had a retention time of 12.75 min, and theanalysis of the standard showed that the entireradioactivity was associated with the ester. Forsamples with a shaking time of 5 min, the entireradioactivity was associated with the ester,whereas for samples with incubation time of24 h, the main radioactivity peak was associatedwith a concentration peak at 4.5 min, whichwas identified as 2,4-D acetic acid. Additionally,the percentages of sorbed herbicides at the endof the experiments for the formulated productand 2,4-D acetic acid were very similar. Theseresults confirm previous observations of rapidconversion of the ester to the acetic acid form(Wilson et al., 1997), which has a lower affinity tothe soil surfaces than the hydrophobic compound.However, the Kd values for ester and acetic aciddetermined after 24 h of equilibrium time weresignificantly different, showing that the conver-sion was not completed.

Sorption Mechanisms

The Kd values for 2,4-D acid are inaccordance with the findings on sorptionbehavior of anionic herbicides in allophanic soilsby Baskaran et al. (1996) and Stolpe and Kuzila(2002). The Kd values were positively correlatedwith organic carbon, cation exchange capacity,allophane, and ferrihydrite contents with deter-mination coefficients of 0.89, 0.80, 0.67, and

0.72, respectively. As the pH of the soilsolutions (4 G pH G 6) exceeded the 2,4-D’spKa value of 2.8, the dominating 2,4-D specieswas the 2,4-D anion. Ionic pesticides can form aneutral ion pair with cations from the soilsolution and undergo hydrophobic sorption.Indeed, the observed L-shaped isotherms indi-cate that the compounds were strongly sorbed tothe solid surface either by inner-sphere com-plexation or by significant van der Waalsinteractions in the sorption process (Sposito,1989). However, in the pH range of the soils,ion exchange and hydrogen bonding processeswere also important. Anionic sorption can occuron positively charged soil minerals, hydroxyl-iron, and hydroxyl-aluminum compounds.These have a high sorption capacity for phenoxyherbicides (Spadatto and Hornsby, 2003). Edgesof aluminosilicate minerals, in which ions arenot fully coordinated, and 1:1 clay minerals withedges having higher percentage of total chargecan also adsorb anions. Spadatto and Hornsby(2003) showed that for sorption of ionicpesticides, conformational changes of OM playan important role as well. Higher pH improvesthe accessibility of OM sorption sites. Althoughhaving a relatively high organic carbon contentand high allophane content, the Kd in theOhakune subsoil, however, was relatively small.The particle size analysis of the soil with andwithout OM (Tables 1 and 2) showed that theOM is mainly constituted of small particles(G6 2m), thus Bhumus-like.^ Furthermore, thehigh pyrophosphate-extractable Al content indi-cates aluminum humus complexes, which areknown to decrease the positive charge onallophane (Parfitt, 1990), thus decreasing thesorption sites for anionic 2,4-D.

Sorption coefficients for 2,4-D ethylhexylester in the four soils followed the order of thosefor 2,4-D acetic acid but were all slightly higherthan those for the acid. 2,4-D in its ester formundergoes primarily hydrophobic sorption toOM, which can be described by a linearisotherm (Brusseau and Rao, 1989a). Theobserved nonlinear isotherms demonstrate thatmore specific reactions took place as well. Estersof 2,4-D were shown to hydrolyze rapidly innonsterile soil to the acid form and therespective alcohol (Wilson et al., 1997). Wilsonet al. (1997) reported that more than 72% of2,4-D hydroxylethyl ester hydrolyzed within72 h. Hydrolysis during the equilibrium time of24 h has certainly taken place in the tubes wherethe solution was in excess to the soil, but the



reaction was probably not completed. Specificsorption of the ester might have stabilized thecompound and inhibited a complete hydrolysis.

Simulation of BTCs

The shape of the tritiated water BTCs wasfairly symmetrical and sharp for all columns,signifying ideal hydrodynamic properties (Fig. 3).The BTCs of tritiated water generally appearedslightly after one pore volume. As alreadyexpected from this graphical inspection, treatingtritiated water as an inert tracer in the fittingprocedure by fixing its retardation factor to 1gave poor fitting results. By fitting both R andD, the CDE gave acceptable results (all determi-nation coefficients 90.98) indicating, as expected,physical equilibrium conditions in the saturatedhomogeneously repacked soil columns. All retar-dation factors except those for the Te Kowhaisubsoil were then greater than 1, ranging from1.02 to 1.23. Kamra et al. (2001) pointed out thatsmall errors in the determination of the soil watercontent could have a strong impact on thecalculation of R. Such small errors could explainR values greater than 1. However, retardation

factors greater than unity for tritiated water werereported in the literature (Chemistry et al., 1992;Logsdon et al., 2002), although tritiated water iscommonly treated as an ideal tracer for water.Retardation of tritiated water has been attributedto isotopic exchange to the solid soil phase. Thetritium in the tracer water molecule canexchange with hydrogen atoms naturally presentin the soil. The isotope will mainly enter weakerhydrogen bridges, which are characteristic ofbiopolymers including organic substances. It canbe fixed on clays and other hydrated minerals.This could explain why in our experimentstritiated water was only retarded in the allophanicsoils, which are characterized by amorphouscompounds with very reactive OH groups andhigh contents of OM capable of holding water.

The fitting procedure led to very differentPeclet numbers for the same soil, for example,72 and 110 for the Mexican topsoil. The Pecletnumber describes the ratio between hydrody-namic dispersion and convection and dependsonly on the medium. With identical packing ofsieved soil material and the similar flow rates,the Peclet number should be the same for

Fig. 3. Experimental (symbols) and simulated (lines) BTCs for tritiated water through the four soils. Data fromcolumns treated with 2,4-D ethylhexyl ester and 2,4-D acetic acid are symbolized by solid and hollow circles,respectively.



Brepetitions.^ The difference found could bebecause of differences in the soil pore networkdue to the packing procedure. Low dispersivitiesand high Peclet numbers indicate poor disper-sion and were reported previously for repackedsoils (Beigel and Pietro, 1999).

All BTCs for 2,4-D acid and 2,4-D ethyl-hexyl ester were shifted to the right of the tracerBTCs, indicating reduced mobility and sorptionof the herbicides (Fig. 4). The relative positionof the 2,4-D ethylhexyl ester BTCs comparedwith the 2,4-D acid BTCs was as expected fromthe batch experiments. 2,4-D ethylhexyl esterwas retarded more strongly than 2,4-D acid inall four soils; the 2,4-D ethylhexyl ester BTCsexhibited a stronger tailing front (this means thatthe process of chemical nonequilibrium wasstronger) compared with the 2,4-D acid, espe-cially in the two allophanic topsoils. Tailing ischaracteristic for reactive solutes and is generallya consequence of the kinetic component ofsorption. Sorption-related nonequilibrium trans-port characteristics were reported for variousorganic chemicals in repacked soil columns (Huand Brusseau, 1998; Beigel and Pietro, 1999).

Another observation is that the measured herbi-cide concentrations did not drop back to zeroeven after leaching of 9 to 12 pore volumes ofCaCl2 solution. This was more pronounced inthe two allophanic topsoils and again strongerfor 2,4-D ethylhexyl ester than for 2,4-D acid,reflecting the difference in sorption between thetwo 2,4-D formulations.

Molecular diffusion was neglected in thisstudy because of relatively high Darcy fluxdensities and Peclet numbers much higher than1. It was assumed that the dispersion coefficientfound for 3H2O was a good estimate fordispersion during herbicide transport. Theparameters " and 5 were fitted to the herbicideBTCs (Fig. 4 and Table 6), whereas theremaining parameter R was obtained from thebatch experiments. Sorption was nonlinear;thus, we calculated R based on the linearizedKL values [Eq. (12)]. The use of KL obtainedfrom batch experiments led to a reasonabledescription of the experimental 2,4-D acidBTCs (using KL values from low concentrationrange) and a rather poor description of the 2,4-Dethylhexyl ester BTCs. By treating R as an

Fig. 4. Experimental (symbols) and simulated (lines) BTCs for 2,4-D acetic acid and 2,4-D ethylhexyl ester throughthe four soils. Data from columns treated with 2,4-D ethylhexyl ester and 2,4-D acetic acid are symbolized by solidand hollow diamonds, respectively.



additional free adjustable parameter, the fit tothe experimental data with the two-site modelimproved for all soils, which was expected withone additional free fitting parameter.

The fraction of sorption sites that were atinstantaneous equilibrium f ranged from 0.41 to0.55 for the allophanic soils, indicating thatapproximately half of the sorption sites wereeasily accessible. This can either be explained byfast diffusion in the saturated soil to sorptionsites in the internal voids of OM or by sorptionsites located on sorbent surfaces. In all soils, f wassmaller for the ester than for the acid, indicatingthat rate-limited sorption was more importantfor the ester than for the acid. Consequently, the

kinetic rate constant ! was also higher for theester than for the acid in the two allophanictopsoils. The kinetic rate constant for the esterwas lower for the subsoils that had lower OMcontents than for the topsoils, indicating diffu-sion into sorbent OM of the topsoils. Also, aninverse relationship between the sorption rateconstant, k, calculated as k = (! I v)/[(1 j ") I RL](Nkedi-Kizza et al., 1989), and the sorptioncoefficient Kd was found (log k = j0.91–0.76log Kd, R

2 = 0.92). This indicates intraorganicmatter diffusion (Nkedi-Kizza et al., 1989).According to the data compilation by Brusseauand Rao (1989b), ionic and nonionic compoundscluster into two discrete groups, with smaller k

TABLE 6

Model parameters for simulations of 2,4-D and 2,4-D ethylhexyl ester BTCs using the two-site nonequilibrium model

(Simunek et al., 1999)

Soil KL (batch). (L kgj1) Kd- (L kgj1) 1 (mm) f ! (sj1) R2 (%)

2,4-D acetic acid

Mexican topsoil 3.52 (7.81) 7.57 0.91 0.55 8.81 0.994 (0.996)‘

Ohakune topsoil 2.34 (5.29) 5.86 2.52 0.47 10.98 0.995 (0.893)‘

Ohakune subsoil 1.74 (1.64) 1.73 3.94 0.53 24.35 0.988 (0.988)‘

Te Kowhai subsoil 0.12 (0.41) 0.12 3.04 0.33 43.44 0.996 (0.996)‘


Mexican topsoil 5.62 12.05 0.79 0.50 21.91 0.985 (j0.58)‘

Ohakune topsoil 4.37 10.29 0.78 0.41 17.10 0.973 (j0.64)‘

Ohakune subsoil 2.21 3.73 1.55 0.48 13.05 0.980 (j0.34)‘

Te Kowhai subsoil 0.33 1.85 0.76 0.13 14.52 0.956 (–)‘

.Kd values for the entire concentration range and values in parentheses for the low concentration range (0–4.15 2g mLj1).-Kd values inferred from the experimental BTCs treating R as a fitting parameter.‘R2 when using the independently determined Kd values in batch experiments.

TABLE 7

Mass balance for the column experiments with the four soils calculated as sum of disintegrations per minute measured in total

leachate and disintegrations per minute determined in soil after termination of experiments

Soil 14C in soil 14C in leachate Total 14C Recovery (%)

2,4-D acetic acid

Mexican topsoil 0.0201 0.0345 0.0546 100.1

0.0211 0.0348 0.0559 102.6

Ohakune topsoil 0.0139 0.0388 0.0527 100.1

0.0116 0.0403 0.0519 98.7

Ohakune subsoil 0.0003 0.0572 0.0575 101.4

0.0 0.0567 0.0567 100.0

Te Kowhai subsoil 0.0001 0.0688 0.0689 103.0

0.0 0.0669 0.0669 100.0


Mexican topsoil 0.0424 0.0184 0.608 114.7

0.0336 0.0169 0.505 100.4

Ohakune topsoil 0.0267 0.0238 0.0505 97.0

0.0235 0.0238 0.0473 91.0

Ohakune subsoil 0.0027 0.0552 0.0549 97.0

0.0011 0.0556 0.0567 100.4

Te Kowhai subsoil 0.0098 0.0490 0.0588 102.8

0.0091 0.0507 0.0598 104.5

First row for each soil and herbicide shows adimensional experimental values; second row, adimensional simulated values.



values for ionic compounds. In our experiments,the sorption rate coefficients were in the sameorder of magnitude for the acid and the ester andalso in the same range for ionic compounds foundin the data compilation by Brusseau and Rao(1989b). This is another indication for hydrolysisof the ester during the flow experiments.

Comparison of Batch and Column Experiments

The Kd values determined from the columnexperiments were generally higher than thoseobtained from the batch experiments (Table 6).In the literature, the opposite has often beenfound (Burgisser et al., 1993) and has beenexplained by the impact of shaking and cen-trifugation procedures and the artificially highsolution-sorbent ratios used in batch experi-ments. All those experimental conditions canlead to soil abrasion and disintegration ofaggregates, changing surfaces of sorbents andexposing sorption sites not available undernatural flow conditions. In this study, however,

we found the opposite for 2,4-D ethylhexylester in all four soils and 2,4-D acid in the twoallophanic topsoils. If equilibrium was notreached within 24 h of the batch experiments,then sorption coefficients would have beenunderestimated. Also, local equilibrium in thecolumn experiments might have not beenreached. Truncated BTCs limit the accuracyof parameter estimation by inverse modeling(Altfelder et al., 2001). Another explanation forthe discrepancy for nonlinear sorbing compoundsmight be the choice of concentration ranges forthe batch experiments. 2,4-D acid sorptioncoefficients determined only from the low con-centration range were, indeed, closer to thosedetermined from the column experiments. The2,4-D acetic acid concentrations of the leachatesamples were in the low concentration range ofthe batch experiments.

Using the independently derived Kd valuesfor the simulations generally led to a deterio-ration of the modeling results. The results for

Fig. 5. Depth profiles of resident radioactivity in five depths of the columns treated with 2,4-D acetic acid and2,4-D ethylhexyl ester.



the 2,4-D acid BTCs in the two subsoils were,however, only minimally affected by fixing theretardation factors. The resulting fits for the twoallophanic topsoils and 2,4-D acid were stillreasonable. For the ester, however, using the Kd

values from the batch experiments led tounacceptable results. Sorption equilibrium forthe ester was apparently not reached within 24 h.Conditions for measuring herbicide sorptionduring batch equilibrium experiments are verydifferent from the sorption situation duringdynamic flow experiments (Estrella et al., 1993).Batch equilibrium sorption coefficients measuredat a fixed time do not take the time-dependenceof sorption occurring under flow conditionsinto account, which was more important for theester than for the acid. Qafoku et al. (2000)postulated that it is better to estimate sorptionand transport parameters with the same miscibledisplacement technique. Column experimentsare preferred as they approximate field condi-tions and maintain narrow soil/solution ratios.Experiments with intact cores are desirable.

Column experiments with repacked soil, how-ever, emphasize the effect of soil chemicalproperties on solute movement.

Mass Balance

The measured recovery data range between100 and 103% for the 2,4-D acetic acid and 97to 115% for the 2,4-D ethylhexyl ester (Table 7).The deviations from a perfect mass balance canbe explained by the uncertainty of the 14Cmeasurements with the oxidizer. Only a smallsubsample of wet soil per soil layer (0.05 g) wascombusted in the oxidizer. Changes in themeasured soil water content occurring duringthe time between weighing in a sample and itscombustion have a significant impact in thecalculations of the total 14C activity per soil layer.

For all four soils, a higher percentage of thetotal radioactivity was left in the soils treatedwith the 2,4-D ester than for the 2,4-D aceticacid, as expected from the herbicide propertiesand the batch experiments (Table 7). In the twosubsoils, the 14C residue levels were, in general,

Fig. 6. Depth profiles of the experimental and simulated resident radioactivity in the two allophanic topsoilstreated with 2,4-D ethylhexyl ester (top) and 2,4-D acetic acid (bottom).



very low at the end of the experiment (Fig. 5),reflecting the fact that the breakthrough wascompleted. For the Mexican and the Ohakunetopsoil, the soil resident radioactivity was 80 and52% of the ester input and 36 and 26% of theacid input, respectively. For both soils andherbicides, the resident radioactivity was highestat the bottom of the column, increasing fromthe 2- to the 10-cm depths, indicating theincomplete BTCs at the end of the experiments.The depth distribution of the total radioactivityresident within the soil columns was fairly wellreproduced with the two-site model that wascalibrated to the BTCs (Fig. 6).

CONCLUSIONS

The important finding of this study is that2,4-D ethylhexyl ester leaching was slower than2,4-D acetic acid leaching in all four soils.Although there was evidence for fast hydrolysisof the ester, it was apparent that this reactionwas not completed during the experiments.Formulation of the herbicide had a significantimpact on its transport behavior directly afterthe surface application. Both batch and displace-ment experiments also showed that sorption ofthe ester was characterized by stronger non-equilibrium than 2,4-D acid sorption in theslightly acidic allophanic soils. This was explainedby the wider range of 2,4-D forms available dueto hydrolysis of the ester. Leaching was describedbest by the two-site nonequilibrium model. Theextended slow and continuous release of 2,4-Dacid from the two allophanic topsoils and of2,4-D ester from all four soils and the high resi-due fractions remaining in the topsoil columnssuggest that sorption was rate limited.

In contrast to general statements, sorptioncoefficients were found to be higher in thecolumn than in the batch experiments using awide concentration range. These results under-line the nonlinearity of 2,4-D sorption but alsoshow that sorption equilibrium was probablynot reached within the 24 h of the batchexperiments. Sorption was mainly correlated toorganic carbon content and pH. The presenceof allophanes, however, increased sorptioncompared with the nonallophanic soil. Theimpact of organomineral interactions in variablecharge soils requires further investigation.

The study also demonstrated that usingtritiated water as an inert water tracer forsaturated flow through allophanic soils is inap-propriate. Considerable retardation of tritiated

water was measured for transport through allallophanic soils used in this study, whereastransport in the nonallophanic soil was perfectlywell described by the CDE and R fixed to 1.

REFERENCES

Altfelder, S., T. Streck, M. A. Maraqa, and T. C.Voice. 2001. Nonequilibrium sorption of dime-thylphthalate—Compatibility of batch and columntechniques. Soil Sci. Soc. Am. J. 65:102–111.

Baskaran, S., N. S. Bolan, A. Rahman, and R. W.Tillman. 1996. Pesticide sorption by allophanicand non-allophanic soils of New Zealand. N.Z. J.Agric. Res. 39:297–310.

Beigel, C., and L. D. Pietro. 1999. Transport oftriticonazole in homogeneous soil columns: Influ-ence of nonequilibrium sorption. Soil Sci. Soc.Am. J. 63:1077–1086.

Brusseau, M. L., R. E. Jessup, and P. S. C. Rao.1991. Nonequilibrium sorption of organic chemi-cals: Elucidation of rate-limiting processes. Envi-ron. Sci. Technol. 25:134–142.

Brusseau, M. L., and P. S. C. Rao. 1989a. Sorptionnonideality during organic contaminant transportin porous media. CRC Crit. Rev. Environ.Control 19:33–99.

Brusseau, M. L., and P. S. C. Rao. 1989b. Theinfluence of sorbate-organic matter interactionson sorption nonequilibrium. Chemosphere 18:1691–1706.

Burgisser, C. S., M. Cernik, M. Borkovec, and H.Sticher. 1993. Determination of nonlinear adsorp-tion isotherms from column experiments: Analternative to batch studies. Environ. Sci. Technol.27:943–948.

Chemistry, E. T. A., O. H. Jacobsen, F. J. Leij, andM. T. van Genuchten. 1992. Parameter determi-nation for chloride and tritium transport inundisturbed lysimeters during steady flow. NordicHydrol. 23:89–104.

Chen, W., and R. J. Wagenet. 1997. Description ofatrazine transport in soil with heterogeneous non-equilibrium sorption. Soil Sci. Soc. Am. J. 61:360–371.

Estrella, M. R., M. L. Brusseau, R. S. Maier, I. L.Pepper, P. J. Wierenga, and R. M. Miller. 1993.Biodegradation, sorption, and transport of 2,4-dichlorophenoxyacetic acid in saturated andunsaturated soils. Appl. Environ. Microbiol. 59:4266–4273.

Hornsby, A. G., R. D. Wauchope, and A. E. Herner.1996. Pesticide Properties in the Environment.Springer Verlag, New York.

Hu, M. Q., and M. L. Brusseau. 1998. Coupledeffects of nonlinear, rate-limited sorption andbiodegradation on transport of 2,4-dichlorophe-noxyacetic acid in soil. Environ. Toxicol. Chem.17:1673–1680.



Kamra, S. K., B. Lennartz, M. T. van Genuchten, andP. Widmoser. 2001. Evaluating non-equilibriumsolute transport in small soil columns. J. Contam.Hydrol. 48:189–212.

Kut<lek, M., and D. R. Nielsen. 1994. Soil Hydrology.Catena Verlag, Cremlingen-Destedt, Germany.

Laird, D. A., and B. L. Sawhney. 2002. Reactions ofpesticides with soil minerals. In Soil MineralogyWith Environmental Applications. J. B. Dixonand D. G. Shulze (eds.). Soil Science Society ofAmerica, Madison, WI, pp. 765–793.

Leij, F. J., and N. Toride. 1998. Analytical solutionsfor nonequilibrium models. In Physical Non-equilibrium in Soils: Modeling and Application.H. M. Selim and L. Ma (eds.). Ann Arbor Press,Chelsea, MI, pp. 117–156.

Logsdon, S. D., K. E. Keller, and T. B. Moorman.2002. Measured and predicted solute leachingfrom multiple undisturbed soil columns. Soil Sci.Soc. Am. J. 66:686–695.

Muller, K., R. Stenger, and A. Rahman. 2005.Herbicide runoff on a pastoral hill slope—A casestudy. In Bringing Groundwater Quality Researchto the Watershed Scale. N. R. Thomson (ed.).IAHS Publ. 297, Wallingford, Oxfordshire, UK,pp. 81–88.

Nkedi-Kizza, P., J. W. Biggar, H. M. Selim, M. T.van Genuchten, P. J. Wierenga, J. M. Davidson,and D. R. Nielsen. 1984. On the equivalence oftwo conceptual models for describing ionexchange during transport through an aggregatedoxisol. Water Resour. Res. 20:1123–1130.

Nkedi-Kizza, P., M. L. Brusseau, P. S. C. Rao, andA. G. Hornsby. 1989. Nonequilibrium sorptionduring displacement of hydrophobic organicchemicals and 45Ca through soil columns withaqueous and mixed soil. Environ. Sci. Technol.23:814–820.

Parfitt, R. L. 1990. Allophane in New Zealand—Areview. Aust. J. Soil Res. 28:343–360.

Parfitt, R. L., and C. W. Childs. 1988. Estimation offorms of Fe and Al: A review, and analysis ofcontrasting soils by dissolution and Moessbauermethods. Soil Chem. Mineral. 26:121–144.

Pignatello, J. J., and B. Xing. 1996. Mechanisms ofslow sorption of organic chemicals to naturalparticles. Environ. Sci. Technol. 30:1–11.

Qafoku, N. P., M. E. Sumner, and D. E. Radcliffe.2000. Anion transport in columns of variablecharge subsoils: Nitrate and chloride. J. Environ.Qual. 29:484–493.

Simunek, J., M. T. van Genuchten, N. Toride, andF. J. Leij. 1999. The STANMOD computersoftware for evaluating solute transport in porousmedia using analytical solutions of the CDE.Manual. US Salinity Laboratory AgriculturalResearch Service, US Department of Agriculture,Riverside, California.

Smith, M. C., D. R. Shaw, J. H. Massey, M. Boyette,and W. Kingery. 2003. Using nonequilibriumthin-disc and batch equilibrium techniques toevaluate herbicide sorption. J. Environ. Qual. 32:1393–1404.

Soil Survey Staff. 1992. Keys to Soil Taxonomy.SMSS Technical Monograph No. 19. PocohontasPress, Blacksberg, VA.

Spadatto, C. A., and A. G. Hornsby. 2003. Soilsorption and acidic pesticides: Modeling pHeffects. J. Environ. Qual. 32:949–956.

Sposito, G. 1989. The Chemistry of Soils. OxfordUniversity Press, New York.

Stolpe, N. B., and M. S. Kuzila. 2002. Relativemobility of atrazine, 2,4-D and dicamba involcanic soils of south-central Chile. Soil Sci.167:338–345.

van Genuchten, M. T. 1981. Non-EquilibriumTransport Parameters From Miscible Displace-ment Experiments. Research Report No. 119.U.S. Department of Agriculture Science andEducation Administration, U.S. Salinity Labora-tory, Riverside, CA.

van Genuchten, M. T., and P. J. Wierenga. 1976.Mass transfer studies in sorbing porous media I.Analytical solutions. Soil Sci. Soc. Am. J. 40:473–480.

Wauchope, R. D., T. M. Buttler, A. G. Hornsby,P. W. M. Augustijn-Beckers, and J. P. Burt. 1992.The SCS/ARS/CES Pesticide Properties Databasefor Environmental Decision-Making. Rev. Envi-ron. Contam. Toxicol. 123:1–164.

Wilson, R. D., J. Geronimo, and J. A. Armbruster.1997. 2,4-D dissipation in field soils after applicationsof 2,4-D dimethylamine salt and 2,4-D 2-ethylhexylester. Environ. Toxicol. Chem. 16:1239–1246.



Quantifying fluorescent tracer distribution in allophanicsoils to image solute transport

C. DUWIGa,b , P. DELMAS

c , K. MULLERd , B. PRADO

b , K. RENc , H. MORIN

a,b & A. WOODWARDc

aUMR 5564 LTHE/IRD, BP S3, 38041 Grenoble Cedex 9, France, bLaboratorio de Fertilidad de Suelo, Colegio de Postgraduados, CP

56230, Montecillo, Mexico, cDepartment of Computer Science, The University of Auckland, Private Bag 92019, Auckland, New Zealand,

and dAgResearch, Private Bag 3123, Hamilton, New Zealand

Summary

The accurate prediction of solute transport through soils is a necessity to counter the worldwide degradation

of aquifers. Dye tracers are widely used to visualize active flow paths in cross-sections of soil, but methods

previously proposed to map concentrations have been very costly, demanding, or of coarse resolution and

not always applicable in dark allophanic soils. We have developed a cheap and fairly easy experimental

procedure and usedmultiple regression to map dye concentrations in two dimensions.We tested the method

using the fluorescent dye, pyranine, in intact cores of an allophanic soil. The method requires a calibration

step, which we made using eight dye concentrations. The main difficulty was to mix the soil homogeneously

with the dye and to pack it evenly before acquisition of the images. The pyranine was infiltrated in soil cores

under unsaturated conditions: its distribution on the vertical core faces was highly heterogeneous with

fingered penetration. The maps of dye concentration obtained from each core section achieved fine spatial

resolution (e.g. 0.25 mm2 per pixel) and satisfactory dye concentration localization and estimation. We

could achieve better spatial resolution by sectioning the soil cores at finer intervals, and estimate the dye

concentration more accurately by improving the correction for illumination variations.

Quantification d’un traceur fluorescent dans des sols allophaniques

Resume

Pouvoir predire avec precision le transport des solutes a travers les sols est essentiel pour contrecarrer la

degradation mondiale des aquiferes. Les colorants sont communement utilises comme traceur pour visual-

iser les chemins de flux preferentiels dans des profils de sol, mais les methodes proposees anterieurement pour

obtenir la carte des concentrations etaient soit onereuses, complexes ou de faible resolution, et pas toujours

applicables a des sols allophaniques de couleur sombre. Nous developpons ici une experience peu couteuse et

facile a mettre en œuvre, et nous utilisons la regression multiple pour obtenir les cartes de concentrations du

colorant. Nous testons la methode proposee a l’aide du colorant fluorescent, la pyranine, dans des mono-

lithes intacts d’un sol allophanique. Nous avons premierement procede a une calibration, en utilisant huit

concentrations. La principale difficulte a ete de melanger le sol avec le colorant de facxon homogene puis de le

compacter avant l’acquisition des images. La pyranine a ete infiltree dans les monolithes de sol sous con-

ditions non saturees: la distribution du colorant sur les faces verticales dumonolithe est tres heterogene, avec

un front d’infiltration « en doigts ». Les cartes de concentrations de colorant obtenues pour chaque section

du monolithe presentent une bonne resolution spatiale (0.25 mm2 par pixel) ainsi qu’une localisation et

quantification des concentrations satisfaisantes. Une meilleure resolution spatiale pourrait etre obtenue en

sectionnant les monolithes plus finement, et l’estimation des concentrations pourrait etre plus precise en

ameliorant la correction des inhomogeneites d’eclairage.

Introduction

Solute transport through soil is topical and of public concern

because aquifers have been contaminated by agrochemicals in

many parts of the world. The reliable prediction of the transportCorrespondence: C. Duwig. E-mail: [email protected]

Received 1 February 2007; revised version accepted 4 September 2007

94# 2007 The Authors

Journal compilation # 2007 British Society of Soil Science

European Journal of Soil Science, February 2008, 59, 94–102 doi: 10.1111/j.1365-2389.2007.00970.x

of agrochemicals through soils is necessary for managing land

and protecting aquifers, yet it is still a scientific challenge (Jarvis,

2007). Classical transport equations are for homogeneous soil

structure and uniform flow that is controlled by convection

and dispersion. In reality, solute transport is confounded by

the non-uniformity of soils, with the result that non-equilibrium

or heterogeneous transport of solute is more the rule than the

exception (Fluhler et al., 1996). The main impediment for

developing good predictive models is our limited understand-

ing of the mechanisms that govern transport. We must be able

to take into account the complexity of soil structure and

porous networks. Applying dye tracers is a popular method

that provides insight at a fine spatial resolution (around 10 mmper pixel), and offers the prospect of quantifying the static and

dynamic aspects of flow and sorption.

We think that the transport of reactive solute through allo-

phanic soils formed from volcanic ash involves unique processes

that require modifications of simulation models. Such soils con-

tain non-crystalline colloidal materials that confer on them

unique physical and chemical properties, such as low bulk den-

sity, large anion exchange capacity, and large proportions of

organic matter that are rare in other soil kinds. Allophanes

and other amorphous materials have large specific surface area

andare highly reactive. This combination leads to abehaviour of

reactive solutes that is different from that in non-allophanic soils

(Baskaran et al., 1996; Muller & Duwig, 2007). Ash soils are

distributed over about 0.8% of the earth’s land surface and are

highly regarded for agriculture; many of the most productive

densely populated regions are near volcanoes. They are subject

to increasingly intensive use, which affects their filtering capac-

ity. We therefore studied solute transport in such soils to

understand the process with a view to protecting them and the

associated aquifers.

Soil scientists have been applying dye tracers to reveal active

flow paths for 40 years (Reynolds, 1966; Bouma et al., 1977).

Various dyes, both fluorescent and non fluorescent, have

been used, including Brilliant Blue FCF (Ewing & Horton,

1999; Forrer et al., 2000; Kulli et al., 2003; Morris & Mooney,

2004; Javaux et al., 2006), Brilliant Sulfaflavine (BF) and

Sulforhodamine B (SB) (Aeby et al., 2001; Vanderborght

et al., 2002). Fluorescent tracers have an advantage over

these in that several can be quantified simultaneously by fluo-

rescence spectroscopy, with sophisticated filters and light

(Aeby et al., 2001). They also reveal flow paths in allophanic

soils despite the soils’ dark colour. Due to large exchange

capacities of allophanic soils, any dye is expected to be

strongly retarded in them. Visualizing dye patterns linked

with image analysis allows one to differentiate between

stained and unstained soil (Ghodrati & Jury, 1990; Flury et al.,

1994; Petersen et al., 1997).

Exploiting this semi-quantitative classification, researchers

have classified transport mechanisms and related them to soil

properties and structure (e.g. Kulli et al., 2003). Further quan-

titative analysis of concentration patterns is necessary to inves-

tigate the effect of soil and chemical heterogeneity and to

understand sorption and flow processes in field soil. The con-

centration maps so obtained help to evaluate transport mod-

els, which is usually done indirectly by analysis of a solute’s

breakthrough curve obtained from leachate samples. Process-

ing of dyed images was first applied in laboratory column

experiments with homogeneous media, such as glass beads and

sand, to determine the concentration distribution of dye trac-

ers (Schincariol et al., 1993; Abey et al., 2001). Subsequently,

various research teams including Forrer et al. (2000), Vander-

borght et al. (2002) and Javaux et al. (2006) successfully quan-

tified dye infiltration patterns in heterogeneous soils. However,

the step from qualitative evidence of heterogeneous flow phe-

nomena to quantifying dye tracer distributions at a fine spatial

resolution remains difficult despite recent promising progress.

Furthermore, to include the impact of chemical heterogeneity

of soils on solute sorption and transport pattern in the inter-

pretation of dye patterns is still a challenge.

The successful quantification of dye distribution by image

processing always involves a calibration step that links the cor-

rected image parameters to the dye concentrations. Calibration

can be done with samples taken from homogeneously stained

areas of soil profiles that have been photographed and analysed

chemically (Forrer et al., 2000), or in an independent calibra-

tion experiment in which soil samples with known dye concen-

trations are prepared and photographed together with the

profile (Vanderborght et al., 2002; Persson, 2005). Both meth-

ods require sufficient samples for calibration (Persson, 2005).

The main drawback of the first technique is the difficulty of

obtaining homogeneous stained samples even on small vol-

umes. Furthermore, each sample provides an average concen-

tration for the whole volume, whereas only a surface

concentration is estimated by image analysis. The difficulty

with the second procedure is the difficulty of mixing soil

homogeneously with dye. For both methods it is crucial that

the extractant chosen to obtain the total dye concentration

enables the dye sorbed to the soil to be recovered. Currently

the most popular method of analysing the data is to fit regres-

sions to them (Ewing & Horton, 1999). Persson (2005) margin-

ally improved estimates of the concentrations using neural

networks. Furthermore, experiments with fluorescent dyes

usually involve photographic and optical equipment that is

too expensive for most researchers.

We propose an experimental procedure and method of anal-

ysis to quantify dye distribution in allophanic soils using intact

soil cores. We focus on:

1 demonstrating that reactive solute transport in allophanic

soils is strongly heterogeneous;

2 developing a dye tracer method applicable in allophanic

soils, which will eventually help to improve our understanding

of transport processes in these unique soils; and

3 improving the robustness and precision of quantifying

dye concentrations by introducing an independent calibration

method.

Dye tracer quantification in allophanic soils 95

# 2007 The Authors

Journal compilation # 2007 British Society of Soil Science, European Journal of Soil Science, 59, 94–102

For the latter, the colour characteristics of images from cali-

bration boxes with known concentrations were linked to the

concentrations via multiple regression analysis. We used stan-

dard cameras and lighting equipment to photograph the distri-

bution of the fluorescent tracer. We describe the methodology,

including the experimental set-up, data acquisition, and data

analysis. We then describe a specific application of our method

to an allophanic soil and present concentration maps of vertical

core faces.

Methodology

Experimental set-up

Leaching experiments were made on intact soil monoliths of

0.25 m diameter and 0.35 m depth, taken between 0.35 and

0.70 m from an agricultural field. The field bulk density between

0.35 and 0.70 m is 0.51 g cm�3 (SD 0.03, obtained from samples

taken every 5 cm on five dates). The monoliths were excavated

and sculpted in the field with a spirit level, so as to have a per-

fectly horizontal surface on top, and then wrapped in thin

plastic film. A metal cylinder of 0.3 m further protected each

monolith’s structure from disturbance as it was transported to

the laboratory. The space between the cylinder and the mono-

lith was filled with polyurethane foam as described by Morris

& Mooney (2004). The foam when hard stabilized the soil

cores during transport and prevented flow down the walls of

the cylinders during the leaching experiments. The monoliths

were kept in a cool room (at 4°C) to prevent their drying

before the experiments. Initial soil water content varied

between 0.80 and 0.96 g g�1. Using a disc infiltrometer of 0.2

m diameter with a tension set at –10 mm, we passed 3 litres of

3 g l�1 dye (pyranine) solution through the intact soil cores

under unsaturated conditions during about 2 hours. A thin

sand layer on top of the soil core ensured good contact

between the soil and the base of the infiltrometer. Final water

content varied between 0.88 and 1.16 g g�1 (average 0.98

g g�1: SD 0.08, 11 samples). The images were taken immedi-

ately at the end of the infiltration experiment in a dark room.

Cutting and photographing the core faces took about 6 hours.

An independent calibration procedure was developed to re-

late the image parameters taken of vertical soil sections to the

dye concentrations. Calibration boxes were prepared at similar

water content and bulk density as those of the soil core at the end

of the leaching experiment. Forty-six grams of fresh soil (28.6 g

dry weight) was mixed with 11 ml of pyranine solution at eight

concentrations (5, 10, 15, 20, 25, 30, 35 and 40 g l�1). In this

mixture the soil was at field capacity (0.99 g g�1). For pre-

paring the mixtures, fresh soil was used to keep the natural soil

water retention and ion exchanges capacities, which tend

to decrease irreversibly when allophanic soils dry (Uehara &

Gillman, 1981). Field capacity was chosen as final water content

of the mixtures as it allowed the direct comparison between

images taken from the calibration boxes and images from

faces of the soil core. Furthermore, only at this water content

could we homogeneously mix the soil with the dye and pack it

evenly into transparent plastic boxes (19.6 � 7.1 � 1 cm). The

soil was packed into the boxes at a depth of 0.4 cm and a bulk

density of 0.51 g cm�3, matching the bulk density of the soil in

the field. The surface of the soil in these boxes was photo-

graphed under identical conditions (i.e. the same distance,

height, illumination) as the core sections. New calibration

samples had to be prepared for each infiltration experiment as

pyranine degrades quickly in soil (Omoti & Wild, 1979).

To acquire images the experimental set-up consisted of a tri-

pod, a 15-W UV-A fluorescent black light neon tube

(F15T8BLB, Ushio, Tokyo) with spectral output ranges

between 300 and 400 nm peaking at 365 nm and a black cloth

to cover any reflective material within the camera’s field of view.

We used a Canon PowerShot A80 (Canon Inc., Tokyo) con-

trolled by a computer laptop to record the images. The Canon

driver and control software enable remote acquisition of images

from any computer. The digital camera was fitted with a CDD

sensor (width 7.18 mm, height 5.32 mm) with approximately 4.1

million effective pixels. The lens, in 35 mm film equivalent, was

38 mm (Wide angle) to 114 mm (Telephoto). The camera pro-

vides a ‘manual mode’, which allows one to set the focus, shutter

speed and aperturemanually.Under black light illumination the

digital camera’s automatic focusing usually fails. So we focused

the camera manually at the beginning of the calibration exper-

iment (the distance from camera to object was fixed). To allow

for the largest range of dye concentration to be detected and

tested, we had to find a balance between decreasing the shutter

speed to detect the faintest dyed soil, while limiting saturation

effects (i.e. making sure the histogram of dye pixels did not peak

at grey level 255). An optimal set-up spreads the range of R, G

and B values from small to large dye concentrations as much

as possible. We found the optimal camera settings by checking

the histograms of the calibration box images for the whole

range of dye concentrations tested, and we kept them constant

for the whole experiment. They were an exposure time of 4 s,

ISO 100 mode, aperture value of F-stop 2.8 and focal length of

7.81 mm. White balancing was manually set with the grey card

as a ‘white reference’ under experimental illumination con-

ditions and fixed throughout the experiment. The image size

was 2272 � 1704 pixels, with a horizontal and vertical resolu-

tion equivalent to 0.25 mm2 per pixel. The locations of every

element of the set-up, namely camera, lamp and soil core, in

the dark-room were accurately recorded. To maximize pixel

resolution the camera height and to-core-distance parameters

were tuned so that the soil core or calibration box occupied

almost the whole image. The distance between the camera and

object was kept at 0.4 m for all images.

Data acquisition

We acquired two sets of images: one for the calibration method

and one for each core face. For the two sets, we maintained

96 C. Duwig et al.

# 2007 The Authors


identical conditions, taking both sets of images directly after

preparation of the calibration boxes or after infiltration to avoid

degradation of the dye through time (Omoti & Wild, 1979).

Before cutting a monolith, we took the soil core out of its

metallic envelope, and we removed the top sand layer to avoid

contaminating the soil with stained sand. The core was sliced

vertically roughly every centimetre. Tobeginwith, we used a saw

to remove the bulk of soil material. To finish the cut, a spatula

was more convenient, enabling us to prepare the surface as

smoothly as possible. At every new profile we moved the core

to the cutting front to ensure that the distance between the cam-

era and the photographed profile remained constant.

Before taking photographs of the core, we had to take several

other images for later corrections. These included:

1 an image with the core replaced by a grey card under the

same illumination conditions as for the core, providing a white

balance and illumination anisotropy benchmark image; and

2 an image with a grid (0.01-m squares) seated in the same

position as the core under the same illumination conditions to

estimate optical distortion by the lens.

Grey scales were placed above and on the side of the column’s

cutting front to verify the stability of the illumination conditions

in the dark room. Our apparatus was to be monitored for the 6

hours an experiment lasted. There was no evident variation in

illumination over time, and so the grey scales were not used to

correct images acquired.

The set of images from the calibration procedure and for each

profile of the soil core were then pre-processed before being

integrated into the data base set as described below.

A protocol for pyranine extraction from soil had to be tested

to enable us to verify the total dye concentration in the calibra-

tion boxes. Pyranine was extracted from soil samples by a pro-

cedure based on the method of Forrer et al. (2000). Fresh soil

with a water content of 0.6 g g�1 was mixed with the volume

of dye solution necessary to reach the water content at field

capacity (0.99 g g�1). We studied nine concentrations ranging

between 0 and 45 g l�1. A sub-sample of each mixture was

taken to check the soil’s water content. Forrer et al. (2000)

extracted the dye after drying the mixture in the oven. To sim-

plify the method, we tested the efficiency of extracting the dye

directly from a wet soil mixture. Therefore, we extracted the

pyranine from a sub-sample of either 1 g of the wet mixture or

0.5 g of the mixture previously dried at 80°C for 24 hours by

shaking them with 10 ml of an acetone-water mixture (volume

ratio 1:4) on an orbital shaker for 16 hours. After centrifuga-

tion at 1050 g for 25 minutes, filtering and buffering of the

supernatants to pH 7 with 0.1 M NaOH, the tracer concentra-

tion was determined by spectrophotometric analysis (on a HP

845 X UV visible system, Hewlett Packard, Palo Alto, CA,

USA) at a wavelength of 402 nm. In the range of concen-

trations studied (solutions buffered at pH 7), the optimal pyr-

anine absorbance range was between 370 and 410 nm, with

a maximum at 402 nm. The concentration of pyranine extrac-

ted depended linearly on the concentration added in the range

studied: Cextracted ¼ 0.96 Cadded (R2 ¼ 0.94) for the wet mixture

and Cextracted ¼ 0.92 Cadded (R2 ¼ 0.99) for the dry mixture.

Data analysis

The images acquired were stored as a set of R, G, B (red, green,

blue) colour characteristics. The following operations were

necessary to create our data base.

First we passed the images through a standard low-pass filter

(a Gaussian filter) to remove spurious noise. Then we applied

Tsai calibration (Tsai, 1987) principles to compute the first-

order radial distortion parameter responsible for most of the

optical aberrations in modern cameras. The distortion was less

than 1 pixel within 500 pixels of the optical centre (e.g. image

centre) and less than 5 pixels within 800 pixels. Variation in

illumination proved to be themain source of error as recognized

previously by Persson (2005). Illumination anisotropy was cor-

rected as described in Aeby et al. (2001). A GIN-ICHI Silk

Grey Card (Neutral Grey 18% reflection) was placed in front

of the cutting profile and filled the image.

We used the direct background subtraction in the RGB space.

This was preferred to a background subtraction involving

conversion from and to the HSV space (Forrer et al., 2000) due to

the presence of very small R values, prone to noise. Under dark

fluorescence, illumination of un-dyed soil appeared with colour

values that were darker and well separated from the dyed pixels

in the colour space. We removed the un-dyed soil pixels using

Otsu’s technique (Otsu, 1979). The image histogram was mod-

elled as the sum of two independent sources with mono-dimen-

sional signal, each modelling the opposite side of the histogram

spectrum (e.g. background versus foreground). A threshold

(separating darker pixels from lighter pixels) was obtained by

minimization of the inter-class variance of the histogram sour-

ces. Un-dyed pixels were displayed in black in the concentration

maps (Figures 3b, 4). Saturated (in a colour sense, e.g. at least

one colour channel (R, G or B) value at 255) pixels carry biased

information and therefore could not be included in our statisti-

cal analysis. They were removed and displayed in white in the

concentration maps (Figures 3b, 4).

In our second step we modelled the known concentrations of

the calibration boxes by fitting a second-order polynomial

model to their corrected R, G and B values. We prepared one

calibration box for each dye concentration: this gave us eight

specific dye concentrations. For each concentration, R, G and

B pixel values were computed by averaging 25 5 � 5 pixel sub-

regions generating the calibration data set. The data were

standardized to zero mean and unit variance to R;G; and B.

We fitted the two second-order polynomials as given in Pers-

son (2005) to the standardized data to predict either the con-

centration C or its logarithm by least squares techniques. Non-

significant predictor variables (P values greater than 0.05) were

removed from the full second-polynomial by stepwise back-

ward elimination. We selected five relationships to predict the

concentration: the full polynomial models predicting C and ln


# 2007 The Authors


C and two polynomial models with a reduced number of sig-

nificant parameters. We also tested a linear regression on

R;G; and B. We validated these models with a set of 1000 R, G,

B values per calibration box taken from different locations

than the sub-regions for the calibration. Following Forrer

et al. (2000), we assessed the precision of the predictions of

each model by calculating the bias and the mean square error

of prediction (MSEP). We selected the model giving the largest

coefficient of determination and the lowest bias and MSEP.

We created a dye concentration map for each profile. Outside

the range of R, G, B values obtained from the calibration

boxes, models in ln C gave predictions of C obviously wrong.

Models in C allowed a better extrapolation for values of C

exceeding 7.84 g cm�3. We displayed C between 7.84 and 8.5 g

cm�3 in pink and larger than 8.5 g cm�3 in red (Figures 3b, 4).

Method application

Soil and dye characteristics

The soil we studied is an Andisol developed from andesitic

tephra. It is a Pachic Andosol (FAO-ISRIC & ISSS, 1998) and

was characterized in detail by Prado et al. (2007). It was under

maize at La Loma, in the Valle de Bravo catchment 150 km

west of Mexico City (19°16¢48.6²N and 99°58¢13.7²W) at

a height of 2500 m. The Valle de Bravo catchment is one of the

main sources of drinking water for Greater Mexico City. The

land is used mainly for agriculture and most of the water bod-

ies there are enriched in nutrients. A collaborative programme

between France and Mexico started in 2001 to study sediments

and contaminant transport in the catchment, at column and

plot scales. The soil’s main properties are listed in Tables 1, 2.

The large amounts of amorphous minerals and organic matter

are responsible for the low bulk density. The only significant

change in this otherwise homogeneous soil is the decrease in

organic matter content with depth. Another important charac-

teristic of the soil is its dark colour, which forced us to work

with a fluorescent tracer. No visible variation could be

observed in the colour below 20 cm (Table 2).

For our experiments we chose the fluorescent tracer pyranine

120% (trisodium 8-hydroxypyrene-1,3,6-trisulfonate, chemical

formula C16H7Na3O10S3) with a molecular weight of 524 g

mol�1. It is readily visible with yellow-green fluorescence with

black light and has a low toxicity (oral toxicity 5 g kg�1). It is

sensitive to pH between 6 and 10 (Zhu et al., 2005), and so we

had to buffer all soil solutions.

We determined the pyranine adsorption isotherm by mixing

30ml of dye solutionswith concentrations between 0 and20 g l�1

with 2-mm sieved fresh soil, equivalent to 15 g dry soil. Three

replicates were studied for each concentration. The suspen-

sions were shaken for 24 hours and subsequently centrifuged

at 1050 g for 25 minutes. The tracer concentrations were deter-

mined in the supernatant by UV-spectrophotometric analysis

(HP 845 X UV visible system) at 402 nm after the soil sol-

utions had been buffered at pH 7 with 0.1 M NaOH. The iso-

therms were described by the Langmuir adsorption model:

Csorb ¼KLSmaxCeq

1 þ KLCeq; ð1Þ

where Csorb is the adsorbed tracer mass in g per kg of dry soil,

Smax is the maximum adsorbed tracer mass per g of dry soil at

maximum sorption, KL is a constant and Ceq is the equilibrium

Table 1 Chemical characteristics of the soil profile

Depth

Horizon

OCa

pH in H2O

CECb Pretc Allophaned Ferrihydritee

/cm /g kg�1 /cmolc kg�1 /%

0–15 Ap 54 5.5 22.3 >90 18.5 3.2

15–20 A1 53 6.1 23.0 >90 23.1 2.0

20–45 A2 56 6.2 20.0 >90 22.5 1.8

45–65 2A1 53 6.3 24.0 >90 25.5 2.6

65–85 2A2 47 6.3 23.1 >90 26.0 2.7

85–110 3A 51 6.5 23.6 >90 27.7 2.1

aOC ¼ Organic carbon.bCEC ¼ Cation exchange capacity.cPret ¼ Phosphate retention.dAllophane ¼ 6 � Siox (Parfitt, 1990).eFerrihydrite ¼ 1.7 � Feox (Childs et al., 1991).

Table 2 Physical characteristics of the soil profile

Depth

Horizon

Sand Silt Clay qda Colour

Cohesionb Rootsc/cm /% /g cm�3 Moist

0–15 Ap 29 62 9 0.7 10YR 3/2 1 1

15–20 A1 45 50 5 0.7 10YR 3/2 1 3

20–45 A2 23 66 11 0.5 10YR 3/1 2 3

45–65 2A1 25 66 9 0.5 10YR 3/1 3 3

65–85 2A2 26 63 11 0.5 10YR 3/1 3 3

85–110 3A 22 68 10 0.5 10YR 3/1 3 5

aqd ¼ bulk density.bCohesion ¼ 1, very loose; 2, loose; 3, quite compact; 4, compact.cRoots¼ 1, abundant; 2, quite abundant; 3, present; 4, few; 5, very few.

98 C. Duwig et al.

# 2007 The Authors


concentration in the solution (g l�1). The Langmuir equation

was fitted to the experimental isotherm by non-linear least

squares.

Figure 1 shows the experimental data and the fittedLangmuir

isotherm model. Pyranine was sorbed strongly, clearly nonli-

nearly, and was well described by the Langmuir sorption iso-

therm (R2 ¼ 0.97) with a sorbed concentration of 5.66 g kg�1

at the maximum sorption and KL ¼ 7.11.

Verification of calibration procedure

Figure 2 shows a selected area of the calibration box for each

concentration.Largedye concentrationswere difficult to discern

as their colour characteristics were very close one to another and

as well as to saturated (in a grey-level sense) level. The mean and

standard deviation of the R, G, B values for each calibration

box are given in Table 3. G and B increase with increasing con-

centration while R decreases. The standard deviations of R and

B are fairly constant whereas the standard deviation of G

increases with increasing values of G.

We applied the fivemodels that fitted best the calibration data

to the validation data and predicted C. As shown in Table 4,

the bias is negative for all models. Looking at the bias per cali-

bration box (i.e. on a 1000 R;G; and B sub-sample, result not

shown), the bias is either negative or positive, depending on

the box and the model. However, the calibration box with the

largest dye concentration gives by far the most negative bias

for all models, confirming that large dye concentrations were

difficult to analyse. Although the linear regression gives the

smallest coefficient of determination, it also has the least bias

and one of the smallest MSEPs. We selected the two models

with the smallest MSEPs and bias, the linear regression and

the second-order polynomial, to predict C on the soil core

images. Both models gave comparable and satisfactory con-

centration maps. For the sake of simplicity, we decided to

work with the linear relationship. Table 5 gives the estimates

of the parameters fitted and the associated standard errors.

Results from intact cores

Five intact cores were excavated at the same depth (0.35 m) and

leached with pyranine. Similar dye infiltration patterns could be

observed. We explain here the results of the two-dimensional

quantification of dye distribution for one of the cores: Figure 3a

shows a selection of serial images taken of various profile faces

with the same conditions as for the calibration images. The dye

provided a stainedpattern thatwas easily visible against the dark

soil with a black light (Figure 3a). As expected from the batch

Ceq /g l-10 1 2 3 4 5 6 7

Cso

rb /g

kg-1

0

1

2

3

4

5

6

7

Figure 1 Langmuir isotherm for pyranine in the fresh allophanic sub-

soil (0.35–0.70 m), Equation (1) with KL ¼ 7.11 and Smax ¼ 5.66 g kg�1.

Figure 2 Selected area of the calibration images for each concentra-

tion studied (in mg cm�3).

Table 3 Means and standard deviations of the R, G and B values for

the 25 sub-samples of each calibration box

CR G B

/mg cm�3 Mean SDa Mean SDa Mean SDa

0.99 26.61 2.99 19.79 2.31 64.10 11.18

1.96 27.41 2.73 34.58 5.34 67.52 12.96

2.94 23.63 1.76 53.43 10.08 71.19 12.31

3.92 16.42 1.25 75.01 12.31 73.52 11.63

4.90 12.09 2.06 96.13 13.30 87.04 12.06

5.88 7.55 2.40 115.28 16.49 92.31 14.11

6.86 4.62 1.46 142.46 18.51 106.95 15.46

7.84 2.28 0.62 149.45 19.37 106.21 16.55

aSD ¼ standard deviation.

Table 4 The coefficients of determination, R2, bias and mean square

errors of prediction (MSEP) for the five models tested on the valida-

tion data

Model 1a 2b 3c 4d 5e

R2 0.992 0.971 0.992 0.99 0.99

Bias �0.016 �0.005 �0.018 �0.081 �0.078MSEP 0.134 0.212 0.135 0.309 0.310

aModel 1: second-order polynomial in R, G, B predicting C.bModel 2: linear model in R, G, B predicting C.cModel 3: second-order polynomial in R, G, B with reduced number of

parameters predicting C.dModel 4: second-order polynomial in R, G, B predicting ln C.eModel 5: second-order polynomial in R, G, B with reduced number of

parameters predicting ln C.


# 2007 The Authors


experiments, pyranine was strongly sorbed into the soil. The

water front, which reached the bottom of the soil core, was well

below the pyranine front. The dye distribution was very hetero-

geneous in this soil core showing ‘finger flow’. In some slices dye

movement through worm or root channels was clearly visible.

From Figure 3a, border flows are clearly visible. Their pertur-

bation on the profile’s inner pixels is not visible. Note that the

faces at 19.5 and 18 cm from the back of the soil core exhibited

many saturated pixels (in the colour sense, e.g. one or more

colour value at 255). In these two profiles only the un-sliced

external face of the cores, where border flow occurs, is visible.

The dye concentration profiles were then obtained (Figure 3b)

by the method explained in the ‘Data analysis’ section. A close-

up from a face is shown in Figure 4 to exhibit gradations of dye

concentration values.

Discussions, conclusions and future recommendations

The movement of pyranine was very heterogeneous, with

marked fingering. Even in this fine textured soil with homoge-

neous structure, the movement of a reactive solute follows pref-

erential flow pathways. Our method to map fluorescent dye

concentrations in soil cores in allophanic soils should eventually

improve our understanding of sorption and transport processes

in these soils. The method is to be deployed at cores with dyes

that can be used as tracers for specific contaminants.

Figure 3 (a) Original dyed profiles – thickness

measured from the back of the core, (b) Esti-

mated concentration profiles – thickness mea-

sured from the back of the core, black

represents un-dyed pixels, white corresponds

to colour-saturated pixels where the multiple

regression was not applicable (N/A), dye con-

centration in mg cm�3.

Table 5 Estimates of the parameters and their approximate standard

errors (SE) for the linear model (model 2)

Response variables

C

Estimates SE

Intercept 4.411 0.028

R 0.347 0.147

G 3.702 0.225

B �1.367 0.106

100 C. Duwig et al.

# 2007 The Authors


The resolution in the horizontal direction could be further

improved by sectioning the soil profiles at finer intervals. A newer

cutting device is currently being designed to allow thinner cuts in

the horizontal direction, as well as repetitive and semi-automatic

acquisition of soil core profiles, thus limiting errors from slicing

and photographing. The images acquired could then be used to

reconstruct the flow pathway in three dimensions.

We achieved good gradation for eight calibration levels,

although our estimates were inaccurate at the largest concentra-

tion. Potential solutions are to use a 12-bit CCD sensor camera,

polarizing filters to avoid reflection effects, as well as a better

compensation for variations in illumination for each image, but

at an increased experimental cost.

The difficulty of homogeneously mixing the soil with the dye

solution resulted in non-homogeneous images and a larger than

expected colour distribution for a given concentration. It

reduced the range of tested concentrations and increased the

error rate of the statistical analysis. Also, we still have to verify

the concentrations found by image analysis with our proposed

dye extractionmethod, but first we have to find amethod to take

homogeneously dyed samples in the soil profile or core sections.

The sampling poses another methodological problem, namely

volume concentration by chemical analysis versus surface con-

centration by image analysis. One other problem we may face is

the instability of the fluorescence with changing pH, light, water

content (Banninger et al., 2006), and time. However, using cal-

ibration boxes proved to be easier and more accurate than tak-

ing samples on the core face to relate colour parameters to

concentration. Samples taken from soil profiles or core faces,

even of a small volume (less than 1 cm3) have heterogeneous

surface concentrations over the sampling area leading to large

uncertainties in the calibration (Aeby et al., 2001).

Overall, our method offers a cheap alternative to costly light-

ing, camera andfilter equipment. Furthermore, it is applicable to

allophanic soils, the darkness of which does not allow the use of

non-fluorescent dyes.

Acknowledgements

We thank Rosia Maria Lopez from the Laboratorio de Fertil-

idad de Suelo for help in analysing dye in solution and soil

samples. This research was done as a preliminary study to the

CONACYT Project no 47293-F.

References

Aeby, P., Schultze, U., Braichotte, D., Bundt, M., Moser-

Boroumand, F., Wydler, H. et al. 2001. Fluorescence imaging of

tracer distributions in soil profiles. Environmental Science and Tech-

nology, 35, 753–760.

Banninger, D., Lehmann, P., Fluhler, H. & Guglielmetti, M. 2006.

Modelling the effect of soil water content and sorption on dye-

tracer fluorescence. European Journal of Soil Science, 57, 808–815.

Figure 4 Close-up of the 13.3 cm profile

showing local variations of dye concen-

trations along paths of preferential flow.


# 2007 The Authors


Baskaran, S., Bolan, N.S., Rahman, A. & Tillman, R.W. 1996. Pesti-

cide sorption by allophanic and non-allophanic soils of New

Zealand. New Zealand Journal of Agricultural Research, 39, 297–310.

Bouma, J., Jongerius, A., Boersma, O., Jager, A. & Schoonderbeek,

D. 1977. The function of different types of macropores during sat-

urated flow through four swelling soil horizons. Soil Science Society

of America Journal, 41, 945–950.

Childs, C.W., Matsue, N. & Yoshinaga, N. 1991. Ferrihydrite in

volcanic ash soils of Japan. Soil Science and Plant Nutrition, 37,

299–311.

Ewing, R.P. & Horton, R. 1999. Discriminating dyes in soil with

color image analysis. Soil Science Society of America Journal, 63,

18–24.

FAO-ISRIC & ISSS 1998. World Reference Base for Soil Resources

(WRB). World Soil Resources No 84, FAO, Rome.

Fluhler, H., Durner, W. & Flury, M. 1996. Lateral solute mixing pro-

cesses – a key for understanding field-scale transport of water and

solutes. Geoderma, 70, 165–183.

Flury, M., Leuenberger, J., Studer, B., Fluhler, H., Jury, W.A. &

Roth, K. 1994. Pesticide Transport through Unsaturated Field Soils:

Preferential Flow. Ciba Ltd, Basel.

Forrer, I., Papritz, A., Kasteel, R., Fluhler, H. & Luca, D. 2000.

Quantifying dye tracers in soil profiles by image processing. Euro-

pean Journal of Soil Science, 51, 313–322.

Ghodrati, M. & Jury, W.A. 1990. A field study using dyes to charac-

terize preferential flow of water. Soil Science Society of America

Journal, 54, 1558–1563.

Jarvis, N.J. 2007. A review of non-equilibrium water flow and solute

transport in soil macropores: principles, controlling factors and

consequences for water quality. European Journal of Soil Science,

58, 523–546.

Javaux, M., Kasteel, R., Vanderborght, J. & Vanclooster, M. 2006.

Interpretation of dye transport in a macroscopically heteroge-

neous, unsaturated subsoils with a one-dimensional model. Vadose

Zone Journal, 5, 529–538.

Kulli, B., Stamm, C., Papritz, A. & Fluhler, H. 2003. Discrimination

of flow regions on the basis of stained infiltration patterns in soil

profiles. Vadose Zone Journal, 2, 338–348.

Morris, C. & Mooney, S.J. 2004. A high-resolution system for the

quantification of preferential flow in undisturbed soil using obser-

vations of tracers. Geoderma, 118, 133–143.

Muller, K. & Duwig, C. 2007. Transport and sorption of 2,4-dichloro-

phenoxyacetic acid in allophanic soils. Soil Science, 72, 333–348.

Omoti, U. & Wild, A. 1979. Use of fluorescent dyes to mark the

pathway of solute movement through soils under leaching con-

ditions: 1. Laboratory experiments. Soil Science, 128, 28–33.

Otsu, N. 1979. A threshold selection method from gray-level histo-

grams. IEEE Transactions on Systems, Man, and Cybernetics, 9,

62–66.

Parfitt, R.L. 1990. Allophane in New Zealand – a review. Australian

Journal of Soil Research, 28, 343–360.

Persson, M. 2005. Accurate dye tracer concentration estimations

using image analysis. Soil Science Society of America Journal, 69,

967–975.

Petersen, C.T., Hansen, S. & Jensen, H.E. 1997. Tillage-induced hori-

zontal periodicity of preferential flow in the root zone. Soil Science

Society of America Journal, 61, 586–594.

Prado, B., Duwig, C., Hidalgo, C., Gomez, D., Prat, C., Yee, H.

et al. 2007. Characterization, classification and functioning of two

profiles under different land uses in a volcanic sequence in Central

Mexico. Geoderma, 139, 300–313.

Reynolds, E.R.C. 1966. The percolation of rainwater through soil

demonstrated by fluorescent dyes. Journal of Soil Science, 17,

127–132.

Schincariol, R.A., Herderick, E.E. & Schwartz, F.W. 1993. On the

application of image analysis to determine concentrations dis-

tributions in laboratory experiments. Journal of Contaminant

Hydrology, 12, 197–215.

Tsai, R.Y. 1987. A versatile camera calibration technique for high-

accuracy 3D machine vision metrology using off-the-shelf TV

cameras and lenses. IEEE Journal of Robotics and Automation, 3,

323–344.

Uehara, G. & Gillman, G. 1981. The Mineralogy, Chemistry, and Phys-

ics of Tropical Soils with Variable Charge Clays. Westview Press,

Boulder, CO.

Vanderborght, J., Gahwiller, P. & Fluhler, H. 2002. Identification of

transport processes in soil cores using fluorescent tracers. Soil Sci-

ence Society of America Journal, 66, 774–787.

Zhu, H., Derksen, R.C., Krause, C.R., Fox, R.D., Brazee, R.D. &

Ozkan, H.E. 2005. Effect of solution pH conditions on fluores-

cence of spray deposition tracers. Applied Engineering in Agricul-

ture, 21, 325–329.

102 C. Duwig et al.

# 2007 The Authors


Agricultural Water Management 96 (2009) 1377–1386

Image processing-based study of soil porosity and its effect on water movementthrough Andosol intact columns

B. Prado a,f,*, C. Duwig b, J. Marquez c, P. Delmas d, P. Morales e, J. James d, J. Etchevers f

a Universidad Nacional Autonoma de Mexico, Instituto de Geologıa, Ciudad Universitaria, 04510, Mexico, D.F., Mexicob UMR 5564 LTHE/IRD, BP 53, 38041 Grenoble Cedex 9, Francec CCADET, UNAM, A.P 70-186, 04510 D.F., Mexicod Department of Computer Science, The University of Auckland, Private Bag 92019, New Zealande Laboratorio de Geoquımica Isotopica, Instituto de Geologıa, UNAM, D.F., Mexicof Colegio de Postgraduados, Laboratorio de Fertilidad de Suelo, CP 56230 Montecillo, Mexico

A R T I C L E I N F O

Article history:

Received 15 September 2008

Accepted 12 April 2009

Available online 30 May 2009

Keywords:

Macroporosity

Image analysis

Displacement experiment

A B S T R A C T

The soil pore network and marcoporosity are important factors affecting water and solute transport. The

transfer of contaminants to water resources is of particular importance in the Valle de Bravo watershed

as it provides 10% of the drinking water for the 20 million inhabitants of Mexico City. This watershed is

composed mainly of Andosols with unique mineralogical and physical characteristics. Soil porosity is

usually examined on thin sections, using various image analysis techniques. We propose a novel

methodology combining image analysis and a displacement experiment to study relationships between

soil structure and water tracer transport parameters. H218O displacement experiments were conducted

through intact soil columns sampled at three depths from a representative cultivated Andosol profile.

The soil structure and pore characteristics were obtained by image analysis on thin sections obtained

from each column at the end of the displacement experiment. The total 2D porosity (for pores larger than

50 mm) varied from 80% of the total section area in the topsoil to around 60% in the subsoil. Tubular pores

were the most abundant in the soil profile, but ploughing of the topsoil had destroyed sections of these

pores and replaced them with packing pores. Water transport in the intact subsoil columns was always

in physical non-equilibrium, showing the existence of preferential flow pathways. In the topsoil, one

column out of three showed no preferential flow, demonstrating that soil ploughing also homogenised

pore connections. Pore connectivity was larger in the ploughed topsoil than in their deeper soil horizon

counterparts. Our methodology offers a 2D quantitative characterisation of the macroporous network at

50 mm resolution and the determination of water transport parameters on the same intact soil samples.

3D characterisation of soil porosity using X-ray computed tomography (CT) gives a better picture of pore

connection but usually has lower spatial resolution and a larger cost.

� 2009 Elsevier B.V. All rights reserved.

Contents lists available at ScienceDirect

Agricultural Water Management

journa l homepage: www.e lsev ier .com/ locate /agwat

1. Introduction

Solute transport through soil has become a key research topicsince contamination of groundwater sources has been observedworldwide. The Valle de Bravo watershed in Mexico needs to beprotected from further contamination as it provides 10% of thedrinking water for the 20-million inhabitants of Mexico City.

Water and solute transport through soil is a complex processthat can be directly related to the pore network. Both soil porosity

* Corresponding author at: Universidad Nacional Autonoma de Mexico, Instituto

de Geologıa, Ciudad Universitaria, 04510, Mexico, D.F., Mexico.

Tel.: +52 56 22 42 86x159.

E-mail address: [email protected] (B. Prado).

0378-3774/$ – see front matter � 2009 Elsevier B.V. All rights reserved.

doi:10.1016/j.agwat.2009.04.012

and other soil characteristics such as structure and texture affecttransport processes (Strock et al., 2001). Literature on soil structurecharacteristics and its relationship with solute transport isambiguous. Seyfriend and Rao (1987) reported that solutedispersion is related to the soil structure and its water content.Bejat et al. (2000) observed that in an unsaturated soil, there is alinear relationship between the soil water content and thedispersion of a non-reactive solute, as well as between thedispersion and the pore water velocity. They did not find a directrelationship between the soil’s structural properties and hydro-dynamic dispersion. The pore network, which depends on the soilstructure, plays a decisive role in water and solute movementthrough soil. Walker and Trudgill (1983) found significantcorrelations between geometric variables describing soil porosityand solute transport parameters. For example, the dispersivity

mailto:[email protected]

http://www.sciencedirect.com/science/journal/03783774

http://dx.doi.org/10.1016/j.agwat.2009.04.012

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–13861378

coefficient is exclusively determined by pore geometry (Yule andGardner, 1978). Poletika and Jury (1994) demonstrated that thepresence of macropores is one of the factors responsible forheterogeneity in water and solute movement through soil. Someauthors have related saturated hydraulic conductivity and residualwater content to pore size distribution in soil (Holtham et al.,2007). Macropore networks differ depending on the soil type,morphology, agricultural practices, and faunal activity. Active(i.e. functional) pores change with water content and water porevelocity (Andreini and Steenhuis, 1990; Shipitalo et al., 1990;Edwards et al., 1992; Quinsenberry et al., 1994).

Soil thin section analysis (e.g. Walker and Trudgill, 1983) anddye application in intact soils (Seyfriend and Rao, 1987; Hatanoet al., 1992; Vanderborght et al., 2002; Tarquis et al., 2006) arethe most common techniques for studying soil structure,porosity, and transport parameters. The application of imageanalysis techniques to determine pore characteristics and soilstructure in thin sections has become an indispensable tool forresearch in soil science (Protz et al., 1987). The characterisationof the porous network from the image analysis of soil thinsections allows an independent and direct evaluation of thewater dynamic in a structured soil (Hallaire et al., 1997, 1998;Cervantes et al., 2003; Holtham et al., 2007). It has been used tomeasure the pore size distribution in the various soil horizons(Ismail, 1975), to characterise the orientation, shape, and size ofdifferent pores (Murphy et al., 1977), and to quantify dyetransport in preferential flow pathways (Forrer et al., 2000;Duwig et al., 2008). Image analysis has also been used todetermine solute transport parameters (Persson, 2005) and toexplain the shape of breakthrough curves obtained from solutetransport experiments in soil columns (Walker and Trudgill,1983; Sugita et al., 1995).

One technique for evaluating solute transport and sorptionprocesses in soil is the displacement of the solute through a soilcolumn. The soil is immobile, and the solute moves through thecolumn only once. Transport and sorption processes are affected bythe soil’s structure and other properties. Leachates at the bottom ofthe column are collected and analysed; the breakthrough curveshape is determined by the different processes occurring duringthe displacement of the solute through the soil matrix (e.g.adsorption and preferential flow). Experimental studies where thebreakthrough curve and soil structure characteristics are deter-mined in the same intact soil sample are scarce, and have only beenconducted using expensive non-destructive technologies such assoft X-ray radiography (Mori et al., 1999), and a combination ofCAT and SPECT scanning (Perret et al., 2000). In the case ofAndosols they are nonexistent. This type of study allows thedetermination of a direct relationship between solute transportparameters (by applying a water tracer as the solute of interest)and the soil pore network (by analysing soil thin sections) (Sugitaand Gillham, 1995; Sugita et al., 1995; Bejat et al., 2000).

Table 1Selected soil properties.

Depth (cm) SOCa (g/kg) Texture

Sand Silt (%) Clay

0–15 54 29 62 9

15–20 53 45 50 5

20–45 56 23 66 11

45–65 53 25 66 9

65–85 47 26 63 11

85–110 51 22 68 10

a SOC, Soil Organic Carbon.b CEC, cation exchange capacity (cmolc kg�1).c WC, 15 bar water content.d Allophane = 6 � Si extracted by oxalate (Parfitt, 1990).

Thanks to their unique physical characteristics (e.g. low bulkdensity, high water retention capacity, and usually a high contentof organic matter), which derive from the presence of amorphousmaterials, Andosols usually offer good conditions for agricultureand can support high population densities. However, the presenceof these amorphous materials renders their study quite complexand requires the use of adapted methodologies. Andosol structureand porosity has to be studied at field moisture to avoid theirreversible formation of aggregates, when drying. Its dark colourmakes it impossible to use dyes other than fluorescent ones.

In the current study, displacement experiments were con-ducted with the water tracer H2

18O through intact columnssampled at different soil depths of an Andosol profile. The soilstructure and pore network were obtained by image analysis onthin sections obtained from each column once the displacementexperiment concluded. The image analysis technique consistedof three steps: image segmentation, identification of pores(i.e. labelling), and the calculation of geometric and morphologicparameters, namely their perimeters, areas, shapes (through theirshape factor), bi-dimensional connectivity and tortuosity.

The specific objectives of this work were to analyse the soil porenetwork and its variation with depth within the Andosol profile,and to evaluate the relation between these soil properties andwater transport processes.


2.1. The soil studied

The soil studied is located in the elementary catchment la Loma,part of the Valle de Bravo basin in Mexico State, 150 km westof Mexico City, at a height of 2500 m (19816048.600N and99858013.700W). La Loma has been the location for various studieson agrochemical transport through the soil vadoze zone (e.g. Pradoet al., 2006; Duwig et al., 2006, 2008; Muller and Duwig, 2007). TheValle de Bravo basin is the most important reservoir of theCutzamala system, which provides a significant amount ofdrinking water (19 m3 s�1 or 21% of the daily supply, Tortajadaand Castelan, 2003) to Mexico City.

A plot under maize in the la Loma catchment was selected, andthe water balance, runoff, erosion, and nutrient losses monitored.The soil was characterised and classified as a Pachic Andosol (WRB,2001). The physical, chemical, and mineralogical characteristicsare described in detail in Prado et al. (2007). The whole soil profilepresents andic properties: bulk density < 0.9 g cm�3, phosphateretention � 70%, Alox + (1/2)Feox � 2%, volcanic glass content inthe fine earth fraction < 10%. Table 1 shows some selectedproperties of this soil. The unsaturated soil hydraulic conductivity(at h = �100 mm) and saturated conductivity vary with depth:they decrease from 0.05 and 0.11 cm min�1 at the soil surface to0.012 and 0.08 cm min�1 at 55 cm depth (Prado, 2006).

CECb WCc % pH H2O Allophaned (%) Horizon

22.3 25.5 5.5 18.5 Ap

23 26.9 6.1 23.1 A1

20 30.5 6.2 22.5 A2

24 31.1 6.3 25.5 2A1

23.1 37.9 6.3 26.0 2A2

23.6 33.2 6.5 27.7 3A

Table 2Experimental column parameters.

Column Depth (cm) Length (cm) rsa (Mg m�3) rb (g cm�3) Porosityc (%) ud (cm3 cm�3) Darcy flow (cm cm�1) Voe (cm3)

C1 5–30 25 2.44 0.79 0.68 0.66 0.032 392

C2 5–20 15 2.44 0.74 0.70 0.83 0.008 296

C3 5–20 15 2.44 0.63 0.74 0.70 0.005 250

C4 30–45 15 2.60 0.61 0.77 0.82 0.004 291

C5 30–55 25 2.60 0.79 0.70 0.77 0.016 459

C6 30–53 23 2.60 0.82 0.68 0.79 0.013 424

C7 80–95 15 2.39 0.70 0.71 0.88 0.012 314

C8 80–105 25 2.39 0.69 0.71 0.77 0.016 458

C9 80–95 15 2.39 0.60 0.75 0.84 0.005 298

a rs = soild density estimated by Mercury Intrusion Porosimetry.b r = bulk density.c Porosity = 1 � (r/rs).d u = water content.e Vo = pore volume.

B. Prado et al. / Agricultural Water Management 96 (2009) 1377–1386 1379

2.2. Water displacement experiments

Displacements experiments using the water tracer H218O

(normalized water with 95.1% H218O, Euriso-top, Gif-Sur-Yvette

France) were conducted on intact soil columns. Nine soil columnswere excavated in the field at three different depths: between 5and 30 cm, between 30 and 55 cm, and between 80 and 105 cm.The columns were excavated by carefully pushing a cylindrical PVCtube with a length of 25 cm and an internal diameter of 5.5 cm. Thegood contact and the absence of any artificial pore space in theinterface between the PVC tube and the soil was verified using aCAT scan (Delmas, personal communication). The bottoms of fiveout of nine soil columns were cut to obtain columns 15 cm long inorder to reduce the flow velocity without increasing theexperiment time. The different column lengths are specified inTable 2. The columns were stored at 4 8C until the experiments.Columns were sampled across several horizons depending on thelength and depth of the columns: column C1 across Ap, A1 and A2horizons, columns C2 and C3 across Ap and A1, column C4 acrossA2, columns C5 and C6 across A2 and 2A1, columns C7, C8 and C9across 2A2 and 3A.

During the displacement experiments, a rainfall simulatormade of nine hypodermic needles and a reservoir was used at thetop of the soil columns. The columns were oriented vertically andtwo peristaltic pumps (one at the top feeding the solution into thesimulator and one at the bottom sucking the leachates) created aconstant flow through the columns. At the base of the column therewas a grid to maintain the soil. The space between the pump andthe soil core was hermetically closed to maintain the suction. Inorder to analyse the effect of the flow on the hydro-dispersiveparameters, different Darcy flows were applied, varying from 0.004to 0.03 cm min�1. The same Darcy flow was applied to twocolumns out of three per depth. However, due to the heterogeneityof the structure and porous network of the intact soil, it was notpossible to inject exactly the same flux in each column of the samedepth and thus have exact repetitions. Flux was checked at theentry and exit of the columns by regularly weighing the appliedsolution and the collected samples. Table 2 presents the length, thebulk density, the soil porosity, the final water content, the Darcyvelocity, and the pore volume of each column studied. For sevenout of the nine columns, the volumetric water content was largerthan the porosity calculated from the bulk and solid density. Thisresult means that the soil in the columns was saturated and insome cases, water was probably entrapped due to soil stratification(Table 1).

At the beginning of the experiment, a solution containing thesame major cations (2.4 mM of CaCl2, 0.2 mM of MgCl2 and 0.1 mMof KCl) as the field soil solution was injected during two to threepore volumes so as not to alter the natural ionic strength of the soil.

Electrical conductivity of the leachates was monitored at thebottom of the column. Once a constant value of this parameter wasreached and the flux became stable, about half a pore volume of thewater tracer H2

18O diluted in the same cationic solution wasinjected at the top of the column. This was followed by several porevolumes of the cationic solution, until the electrical conductivityreached its initial value. Samples were automatically collected forH2

18O analysis by mass spectroscopy (Thermo Finnigan MAT 253according to the method of Epstein and Mayeda, 1953). The codeCXTFIT 2.1 (Toride et al., 1999) was used in inverse mode to obtainthe hydrodynamic parameters of the convection dispersionequation (CDE) coupled with the mobile–immobile model. Ananalytical solution of the CDE was fitted to the experimental databy the least-squares optimisation method. By setting the retarda-tion factor R to 1, with H2

18O being an inert tracer, values for thedispersion D (cm2 min�1), the mobile water fraction f(um/u), andthe dimensionless mobile–immobile region exchange coefficient(v) were obtained.

The first-order mass transfer coefficient (a, s�1) that representsthe rate of solute exchange between the mobile and immobileliquid regions is estimated by the equation:

a ¼ vunL

; (1)

where n = average pore water velocity (cm min�1), L = columnlength (cm).

The characteristic exchange time ta (min) between the mobileand immobile regions is calculated as the ratio between immobilewater content (uim) and a.

The resident water time tw (min) is defined as the time for onepore volume to travel through the soil column.

2.3. Thin sections preparation and image acquisition

Soil structure and porosity were analysed on thin sectionsextracted from the same intact columns used in the displacementexperiments. Once the experiments finished, the soil column wascut into 5 cm long segments. The intact samples were then placedin a closed chamber on a 7 cm high grid placed above a plasticcontainer filled with a 20% acetone solution. Every week theacetone solution was replaced by a new acetone solution 20% moreconcentrated and the acetone concentration in the old solutionwas analysed. The water contained in the soil pores was graduallyreplaced by acetone, and the whole pore network was consideredto be filled with acetone when the acetone concentration in theplastic container did not vary for one whole week. The entireprocess took about 6 weeks. The intact samples were thenimpregnated with a polyester resin by applying a vacuum toensure the resin penetrated every pore. The impregnated samples


were dried at room temperature for 4 weeks before being cut intovertical sections. These were polished down to a thickness of30 mm. A set of three thin sections were obtained and analysed forthe intact columns C1, C5, and C8 (see Table 2). The thin sectionswere digitized with an optical scanner (Hewlet Packard Scanjet7400c) to obtain a digital image with an 11 mm-per-pixelresolution. This pixel length was experimentally confirmed usinga measuring scale, placed on the thin section during scanning. Theminimal size of the pores visible on the image was about 11 mm.Due to image processing segmentation (low-pass 3 � 3 filter) andmathematic morphology filters (opening/closing procedures) mostparameters could only be computed for pores which spanned morethan five pixels across. Therefore only pores larger than 50 mm (thewidely accepted lower threshold characterising macroporosity)could be analysed.

2.4. Digital image processing

Digital image processing of the thin slices was completed inseveral stages (Fig. 1).

In the first stage the scanned images of soil sections (Fig. 1a)were segmented (Fig. 1b). They were separated into binarycomponents, according to a criterion based on intensity andcolour thresholds. Automatic segmentation algorithms were tried,but failed as the grey levels of pixels representing pores in similarmaterials changed considerably among images. Classical techni-ques of image segmentation did not provide useful results. For thisreason we employed a semi-automatic procedure known as‘‘Colour Prediction’’ (Barton and Delmas, 2002). First, pixel samplesfrom pores and non-pores were selected manually (using themouse to choose specific pixels), in order to define a 2D matrix ofcolour prediction for the Hue and Saturation colour channels(Duwig et al., 2008). Empirical tests showed that the HSL colourspace (Barton and Delmas, 2002) offered better results than theoriginal RGB colour space. A large Gaussian window was applied asa low-pass filter incrementing the neighbourhood of each pixelclassified as pores, while a narrow, negative-weighted windowdecremented the neighbourhood of pixels classified as non-pores.The Gaussian smoothing, as well as the inclusion of pixels not

Fig. 1. Stages of digital image processing of the soil core thin sections: (a) scanned image o

to soil (no-pores); (c) distance function regarding the bi-dimensional connectivity; (d)

belonging to the porous network (by means of negative values)considerably enhanced the segmentation results. The colourpredictor in the Hue-Saturation space generated a table of thepositive and negative values of Hue-Saturation. Separation of thepore and non-pore regions was achieved by thresholding the Hue-Saturation colour predicate plane (as positive–negative) to create abinary mask image. The colour white was assigned to pores andblack to no-pores (soil) for further processing (Fig. 1b). Themathematical morphology operators (erosion, dilation, opening,closing, etc., see Serra, 1982) were applied to these binary imagesto filter out misclassified pixels and compute the parameterswhich characterise the pore network.

Next, the two-pass sequential labelling algorithm (Haralick andShapiro, 1985) was applied to classify all pore pixels into labelledconnected components (represented by different grey-level valuesin Fig. 1d). The iterative algorithm propagated labels from top tobottom, starting at the left superior corner and moving down to theright inferior corner of the image. A new label was assigned when,for a given pixel, its top, top-left, and left neighbours (on a 3 � 3neighbourhood window) were not labelled (i.e. did not belong tothe pore pixels). Otherwise the smallest label of the threeneighbourhood pixels was assigned to the pixel studied. Followingthis, a second propagation from bottom to top replaced redundantlabels. The algorithm converged after two iterations. Once all poreswere labelled, their individual perimeter and area were calculated.Other parameters such as the shape factor and the bi-dimensionalconnectivity were also calculated.

Discrete border tracking using the so-called criterion of‘‘8-connectivity’’ was used to calculate the perimeter of eachpore. This considers that boundary pixels are connected either byedge or by vertex. The tracking took into account eight possibleorientations, summing up two different distances betweenneighbours: the vertical–horizontal distance (equal to 1) and thediagonal distance (equal to

ffiffiffi2p

).

2.5. Analysis at the micro-morphological scale

The parameters selected to describe the soil porosity are thosereported in the bibliography: the superficial porosity in 2D, the

f soil thin slices; (b) segmented image: white colour was assigned to pores and black

individual pore labeling; (e) labelling last eroded (Nu)

Fig. 2. Tortuosity. Geodesic length (plain line), direct path (dotted line).

Fig. 3. Injected H218O pulse, and observed and simulated dimensional H2

18O

breakthrough curves for three soil columns (a) 5–20 cm depth (C1), (b) 30–45 cm

depth (C5), and (c) 80–95 cm depth (C8).


pores size and shape, their connectivity, and tortuosity (Hallaireet al., 1997, 1998; Cervantes et al., 2003; Holtham et al., 2007).

2.5.1. The 2D porosity

Two-dimensional porosity corresponds to the ratio between thetotal area occupied by the pores and the total area of the cross-section in the image. It is the count of pore pixels divided by theoverall count of pixels in the image.

2.5.2. The pore size

The pore size was estimated from the area of its cross-section inthe image (first in pixels, then converted to mm2). The pores wereclassified into four size classes: class 1, area > 1 mm2; class 2,0.09 mm2 < area < 1 mm2; class 3, 0.01 mm2 < area < 0.09 mm2;class 4, area < 0.01 mm2.

Using the pore size computed from the image, the equivalentdiameter (assuming all pores to be circular) was derived from theircomputed area. All pores larger than 50 mm in diameter or with anequivalent area of 0.002 mm2 were considered to belong to themacropore class.

2.5.3. The pore shape

The shape of each pore was estimated with the longitudinal

index Ia (Coster and Chermant, 1985), estimated from the area A

and the perimeter P of each pore in the image with the equation:

Ia ¼P2

4pA(3)

This index also known as shape factor takes the minimum valueof 1 for a perfect round pore and increases as the pore is longer orhas an irregular profile. It is suited for the distinction of differentcategories of pores (Hallaire and Cointepas, 1993). Considering theclassification proposed by Ringrose-Voase (1996), three categoriesof pore shapes were defined: tubular pores, Ia < 5; fissure pores,5 < Ia < 10; and packing pores, Ia > 10.

2.5.4. The pore bi-dimensional connectivity

This parameter, introduced by Hallaire et al. (1997), wasestimated with an original index, exploiting the notion of the‘‘ultimate eroded’’, proposed in mathematical morphology by Serra(1982).

On a binary image (Fig. 1b), pores are treated as particles(in white) and the material as background (in black). Theconnectivity number Nc corresponds to the number of ‘‘connectedparticles’’. The number of the last eroded, Nu, corresponds to thenumber of convex components in porosity (Fig. 1c and e). Thesetwo numbers allow the definition of the index of bi-dimensionalconnectivity Ic by

Ic ¼1� Nc

Nu(4)

Ic varies between 0 and 1; it is small if the pores are isolated andincreases its value when the pores are interconnected or forminglinks.

2.5.5. Tortuosity

Tortuosity is defined as the length of the true path followed(geodesic length) between two points (Fig. 2, plain line) divided bythe apparent or direct path (Fig. 2, dotted line) between those twopoints. It is the shortest path entirely contained inside the pore,divided by the shortest Euclidean path (that is, the straight line),between the ends of the first path. The mean tortuosity gives anidea of the sinuosity of the pore network. This parameter is equal to1 for a circular or rectilinear pore (when Euclidean and Euclidean

paths coincide) and takes larger values, for pores that producecomplex paths.


3.1. Water displacement experiments: breakthrough curves analysis

Fig. 3 shows the adimensional experimental breakthroughcurves (BTCs), the simulated ones, and the injected H2

18O pulse forthree columns: (a) at 5–30 cm (C1), (b) at 30–55 cm (C5) and (c) at80–105 cm (C8) depth (see Table 2). The H2

18O pulse duration

Table 3Model fitted column parameters.

Column Depth

(cm)

Pea lb

(cm)

f = um/uc tad

(min)

twe

(min)

Modelf

C1 5–30 25 0.98 1.00 – – 1

C2 5–20 8 1.9 0.80 1657 1026 2

C3 5–20 5 3.1 0.70 2272 1254 2

C4 30–45 5 2.9 0.66 2379 1084 2

C5 30–55 15 1.6 0.73 795 534 2

C6 30–53 11 2.4 0.78 685 487 2

C7 80–95 16 1.9 0.75 667 461 2

C8 80–105 20 1.2 0.76 577 544 2

C9 80–95 5 2.8 0.72 2352 1073 2

a Pe: Peclet number.b l: Dispersivity.c f = um/u: mobile water fraction.d ta: exchange characteristic time.e tw: water resident time.f Model 1: in physical equilibrium, 2: in physical non-equilibrium.


varied from one column to the other, because of the difficulty inestablishing the same flux in every column. Before injecting thepulse, the Darcy velocity was calculated from the flux at thecolumn entry. This flux was later corrected as the mean fluxbetween the column entry and exit. The BTCs were asymmetricalin all the extracted columns except for column C1. Column C1’s BTCwas symmetrical (Fig. 3a), with a gravity centre located at one porevolume after half of the H2

18O pulse was injected. This signifiesthat the water tracer movement in C1 column was not retardedand was in physical equilibrium. Asymmetrical BTCs for an inertwater tracer (retardation factor equals 1) means the existence ofphysical non-equilibrium, or preferential fluxes.

The two types of water transport, physical equilibrium and non-equilibrium were only found at the soil layers surface (columnsC1–C3). This was due to the structural differences observed at thisdepth. In column C1, water movement in physical equilibriummeans that the soil was homogeneous. In columns C2 and C3,water movement in physical non-equilibrium signifies theexistence of preferential pathways caused by well connectedmacropores (fissures, pores due to soil fauna and decaying roots).One explanation for these two types of water transport is theeffects of soil ploughing. Columns C1–C3 were excavated in a plotsown with maize where the ploughing affected the first 20 cm ofthe soil profile. Ploughing destroys the aggregates and mixes andhomogenises the soil; process that can be mimicked by sieving thesoil at a uniform particle size. Prado et al. (2006) demonstrated thatwater movement in packed columns with the same soil occurred inphysical equilibrium, due to sieving and homogenising the soilbefore packing it. The excavation of the intact columns in thepresent study was done 5 months after the soil was left bare andnearly a year after the last ploughing. Roots from the previousmaize crop and from weeds growing afterwards, plus thereorganisation of soil fauna during these 5 months are factorsthat can explain the formation of preferential pathways in someplaces in the first 20 cm of the soil profile.

Another explanation of the contrasting results betweencolumns C1 and C2/C3 can be related to soil stratification. Soilstratification was more important in the 5–30 cm length C1column, which included three soil horizons (Ap, A1, A2) than in the5–20 cm C2 and C3 columns that contained only two soil horizons(Ap and A1). These two horizons are characterised by similarphysical characteristics (Table 1). However, the last 10 cm of the C1column came from the 20–45 cm A2 soil horizon which retainsmuch more water than its Ap/A1 topsoil counterpart, because of itslarger clay content (Table 1) (Bartoli et al., 2007): this bottomhorizon of the C1 column therefore regulated water transport inthis column.

3.2. Modelling parameters

Table 3 shows the parameters obtained from inverse modellingof the experimental data.

3.2.1. Dispersivity

We calculated the Peclet number (Pe) from the equation nL/D(where n is the pore water velocity in cm min�1 and L is the columnlength in cm) to compare the relative importance of moleculardiffusion and convective dispersion in the columns. Kutılek andNielsen (1994) presented four zones for the Pe values, defining therelative importance of both processes: zone 1: Pe < 0.3, zone 2:0.3 < Pe < 5, zone 3: 5 < Pe < 20 and zone 4: Pe > 20. In zones 1and 2, molecular diffusion is the most important process, while inzones 3 and 4, it is weak. The Pe values obtained in this study weresuperior to 5 for all columns (Table 3). We concluded that in therange of pore water velocities obtained in our experiments, themolecular diffusion was weak compared to the convective

dispersion and that the dispersivity coefficient (l) could becalculated from the equation: D/n.

When comparing adimensional breakthrough curves, widercurves with lower peaks indicate larger soil dispersivity. Fig. 3compares the C5 curve (30–55 cm depth) with the C8 curve(80–105 cm depth). The dispersivity values are shown in Table 3.

The structural differences found in the surface columns are alsoevidenced by the differences in dispersivity values. In column C1where water movement was in physical equilibrium, the soildispersivity was 0.98 cm (Table 3), a value which stands at thelower limit for an intact soil, according to Magesan et al. (1995). Allthe other columns had dispersivity values between 2.3 � 0.66 cm(at 30–55 cm depth) and 1.96 � 0.8 cm (at 80–105 cm depth). Theseresults are similar to the dispersivity found by Rao et al. (1980). Theseauthors found a dispersivity value of 2.14 cm for a well structured andaggregated soil. The soil from la Loma catchment had a homogeneousloamy texture through the whole profile. However soil ploughing ofthe first 20 cm affected the structure as well as some physical andchemical characteristics. Prado et al. (2007) observed that thepredominant structure was subangular blocky, with a loose structure(fluffy) at the surface.

3.2.2. Mobile water fraction

Table 3 shows the different f values in all studied columns. Incolumn C1 physical equilibrium water transport meant the wholesoil water content was mobile (f = 1), while in the other columnsthe mobile fraction varied from 0.66 to 0.80. This parameterexplains the exit of most of the water tracer H2

18O before one porevolume in columns C5 and C8 (see Fig. 3). Columns C5 and C8 had f

values of 0.73 and 0.76, respectively. This means the gravity centreof the H2

18O BTC was located before one pore volume. Thedifferences in l and f values also modified the breakthroughcurves’ shapes. The f value depends on the soil structure andtexture. Vanderborght et al. (2000) found that the volume fractionof the preferential flow region is roughly in the order of 1–3% in aloam soil profile. For a silt loam soil with subangular blockystructure, Lee et al. (2001) reported values of f ranging between0.42 and 0.82; Kamra et al. (2001) found an f value of 0.75 in a siltloam soil.

3.2.3. Mobile–immobile region exchange coefficient

The mobile–immobile region exchange coefficient (a) isaffected by the pore water velocity (n), the aggregates’ size, andthe length of the pathway the water molecules have to travel(Pallud, 2000). We found the exchange coefficient increasedproportionally with the pore water velocity. The correlationcoefficient between a and n was 0.93. Despite being affected by

Fig. 4. Pore size distribution expressed as area in mm2, for each depth and each

class.


the pathway length, the column length did not affect the a value.For example, at 80 cm depth, a varies from 0.00033 min�1 for a15 cm long column to 0.00032 min�1 for a 25 cm long column. Thisresult signifies that the length of porous pathways was smallcompared to the length of the columns studied, confirming thevolume representativeness of the columns.

Table 3 shows that for the columns in physical non-equilibrium,the exchange characteristic time (ta) was greater than the waterresidence time (tw). The water did not stay inside the soil columnlong enough to reach physico-chemical equilibrium with the soil.Preferential flows reduce the contact time between the water andsoil matrix.

3.3. Analysis at the micro-morphological scale: porous network

characteristics from thin section analysis

Pore size and shape factors (averages and standard deviations)for each pore class (as defined in Section 2) are presented in Table 4and discussed hereafter. These values were obtained from threethin sections per column: only columns C1, C5 and C8 werestudied. We calculated the overall number of pores (with adiameter larger than 11 mm) found in the thin sections for eachdepth, and the percentage of macropores (pores with an equivalentdiameter larger than 50 mm or with an area larger than0.002 mm2). The maximum pore count occurred at 5–30 cm(80% of the thin section), decreased at 30–55 cm (40%), andincreased slightly at 80–105 (57%). The macroporosity variationsfollowed the same pattern.

The soil parameters estimated through 2D image analysis arerepresentative only if the soil is an isotropic medium. Indeed, theporosity values obtained from the 2D image analysis of the thinlayers and those determined in a three-dimensional space(estimated from the bulk and solid density) are of the same orderof magnitude. This result validates the hypothesis of the studiedsoil being an isotropic medium.

Table 4 shows average pore size and shape factors for each depthstudied. Across all depths the percentage of pores in class 1 (largestpores) was the lowest and the percentage in class 4 (smallest pores)was the highest. The soil macroporosity was characterised by amajority of relatively small pores (area < 0.01 mm2) and 96% of all

Table 4Average and standard deviation (in bracket) of the pore size and shape for each

depth studied and for each pore class.

Depth (cm)

5–30 30–55 80–105

Class1a (%) 0.84 (0.23)c 0.78 (0.56) 0.41 (0.15)

tpb – – –

fpb – – –

ppb 100 100 100

Class2a (%) 3.2 (0.48) 4.4 (1.6) 4.1 (0.97)

tpb 7 (1.2) 18 (2.6) 7 (4.5)

fpb 25 (5.3) 36 (4.3) 40 (3.4)

ppb 68 (6.5) 46 (16) 53 (7.9)

Class3a (%) 34.8 (2.6) 38.5 (3.9) 39.8 (8.9)

tpb 77 (2.3) 81 (4.5) 76 (4.5)

fpb 21 (2.6) 17(3.9) 21 (3.1)

ppb 2 (0.8) 2 (0.7) 2 (0.1)

Class4a (%) 61.2 (2.69) 56.3 (4.45) 56.1 (13.5)

tpb 100 100 100

fpb – – –

ppb – – –

a Class 1, pores area > 1 mm2; class 2, 0.09 mm2 < pores area < 1 mm2; class 3,

0.01 mm2 < pores area < 0.09 mm2; class 4, pores area < 0.01 mm2.b tp: tubular pores; fp: fissure pores; pp: packing pores.c Standard deviation.

pores were smaller than 0.09 mm2. These results are similar toresults found by Oleschko and Chapa Guerrero (1989) and CabreraCarvajal and Okeschko (1995), which are the only studies on poresizes of Andosols we are aware of. They studied the topsoil of a mollicAndosol either on thin sections or using the water retention curveand found that the majority of macropores (detection limit: 0.5 mmin pore radius) had equivalent pore radii between 15 and 150 mm(i.e. our pore size classes 3 and 4). In general, pore size did not changesignificantly with depth. Fig. 4 shows the pore size distributionexpressed as area in mm2, for each depth and class.

Fig. 5 shows the shape distribution expressed in number ofpores according to the classification of Ringrose-Voase (1996)presented previously. Tubular pores are by far the most abundantacross the three depths. At 80–105 cm, the fissure pores are moreabundant than in the other depths. By determining the pore shapedistribution in each pore size class we made the followingobservations (Table 4). In class 1 (the largest pores) there are onlypacking pores. In class 2 packing pores are the most abundant,along with some fissure pores and a minimal amount of tubularpores. In class 3, we observe the contrary; tubular pores are themost abundant, with some fissure pores and a lower proportion ofpacking pores. In class 4 (the most abundant class), there are onlytubular pores.

Fig. 5. Pore shape distribution according to the classification of Ringrose-Voase

(1996), for each depth and each class.

Table 5Summary of thin section and column parameters for columns C1, C5 and C8.

Column Depth (cm) la (cm) fb (um/u) ac (min�1) nd (cm min�1) Ice Tortuosity

C1 5–30 0.98 1 – 0.05 0.23 (0.01) 1.27 (0.26)g

C5 30–55 1.6 0.73 0.00026 0.02 0.18 (0.09) 1.17 (0.19)

C8 80–105 1.2 0.76 0.00032 0.02 0.12 (0.03) 1.2 (0.24)

a l: dispersivity.b f: um/u: mobile water fraction.c a: first-order mass transfer coefficient.d n = pore velocity.e Ic = bi-dimensional connectivity.g Standard deviation.


The tortuosity did not vary between pore size classes (resultsnot shown). The average value was 1.2 (Table 5) for all depths. Thebi-dimensional connectivity Ic decreased significantly between thesurface and the deepest layers.

3.4. Comparison between soil properties, porous network

characteristics and hydro-dispersive parameters

Tubular pores were the most abundant pores found on thinsections from all depths. The existence of amorphous mineralsand the high organic matter content in Andosols make theirstructure granular and very porous (Nanzyo et al., 1993), with animportant proportion of intra-aggregate pores belonging to sizeclasses 3 and 4. In these classes, we found mainly tubular pores.Tubular pores have a lower mean area (the majority are in class 4,thus the area is lower than 0.01 mm2) than fissure pores andpacking pores. During water transport, these tubular pores ofsmall size act like a water reservoir (Walker and Trudgill, 1983),contributing to the large plant available water holding capacity(Shoji et al., 1993).

The results obtained from the displacement experiments at5–30 cm depth showed that soil ploughing increased thepore connection, as we found that in one column (C1) all thewater was mobile. A result that coincides with a larger value ofthe bi-dimensional connectivity at 5–30 cm than for the deeperlayers (Ic, Table 5). In the deeper layers, there is a smalleramount of macropores and we found preferential flow in allcolumns at 30–55 and 80–105 cm depth. Preferential pathwaysare created by well connected pores that by-pass areas ofimmobile water where pores are not well connected. This canbe related to the greater amount of fissure pores found at80–105 cm. It did not affect the Ic value. Hallaire et al. (1997)observed that the method for calculating Ic was not as welladapted to fissure pores as to pores with other shapes. Themethod can potentially consider several connected fissure poresas a single pore, which would artificially decrease the Ic value fora fissured soil.

Fig. 6. Relationship between the mobile water fraction coefficient and the Darcy

flow.

In spite of the small variation in the number of packing poresbetween the three depths studied (Table 4, classes 1 and 2), theseare still more important in the 5–30 cm layer. Packing pores are theresult of the destruction of well defined pores such as fissures ortubular pores by agricultural practices (Gutierrez, 2004). Hubertet al. (2007) evaluated the relative influence of biological andmechanical processes on the structure of cultivated soils and foundthat mechanical soil ploughing produces continuous packingpores.

We mentioned above that the mobile–immobile regionmass exchange coefficient (a) is affected by the pore watervelocity, the soil aggregates size, and the transport pathway (whichcan be described by the tortuosity). At the macroscopic levelwe have shown that the column length did not affect the a value.The similar tortuosity values found in columns at 30–55 and80–105 cm depth with similar velocity and different columnlengths resulted in similar a values. Similar tortuosity values werealso related to similar soil structure.

A positive linear relationship (R2 = 0.8) was found between themobile water fraction coefficient and the Darcy flow: the f valueincreases when Darcy flow increases (Fig. 6). The immobile waterin the soil column is retained by capillarity, absorbed and trappedin the air bubbles or in the pores. The effect of the partitionbetween mobile and immobile water fraction is accentuated bythe desaturation process, as the soil water content decreases, theimmobile water fraction increases because the mobile water is themoving water.

4. Conclusions

The objective of this study was to investigate potentialrelationships between hydro-dispersive parameters (obtainedfrom the displacement experiments in intact columns), and themorpho-geometrical parameters describing the soil core porousnetwork (as obtained from thin section image analysis).

The soil’s hydro-dispersive parameters were determined bydisplacement experiments in intact columns conducted withthe water tracer H2

18O. Using image analysis, the morphologicaland geometrical parameters (surface porosity, pore size dis-tribution, pore shape, tortuosity, and connectivity) of the soilporous network were computed on the soil columns thinsections.

The total 2D porosity (for pores larger than 11 mm) variedfrom 80% of the total section area at 5–30 cm depth to around60% at 80 cm depth. Tubular pores were the most abundantin the whole soil profile. This can be linked to the importantwater retention capacity of Andosols. Soil ploughing increasedthe porosity and pore connection in the first 30 cm comparedto the deeper layers. Soil ploughing was associated withthe disappearance of preferential flow in one surface column,and a greater bi-dimensional connectivity in the first depthstudied. It also destroyed the pores of well defined shapes


(tubular and fissures), replacing them with packing pores.Preferential flow, found in all but one column, was associatedwith a greater occurrence of fissure pores in the deepest layers.

The combined method proposed in this study allows quanti-tative characterisation of the macroporous network and theevaluation of its role in water and solute transport. This articleis the first to present a methodology coupling image analysis anddisplacement experiments on an Andosol. This method isparticularly suited to evaluating the impact of agriculturalpractices on contaminant leaching, and could be of great interestin Andean countries where volcanic soils are common andagricultural practices are being intensified.

Acknowledgements

The authors thank the personnel of ‘‘Laboratorio de Fertilidadde Suelo’’ of the ‘‘Colegio de Postgraduados, Montecillo’’ for thehelp with soil analyses. The research was funded by the ‘‘Institut deRecherche pour le Developpement’’ (IRD), France and TheUniversity of Auckland, New Zealand (SRF 3608705).

References

Andreini, M.S., Steenhuis, T.S., 1990. Preferential paths of flow under conventionaland conservation tillage. Geoderma 46, 85–102.

Bartoli, F., Regalado, C.M., Basile, A., Buurman, P., Coppola, A., 2007. Physicalproperties in European volcanic soils: a synthesis and recent developments.In: Arnalds, O., Bartoli, F., Buurman, P., Oscarsson, H., Stoops, G., Garcıa-Rodeja,E. (Eds.), Soils of Volcanic Regions in Europe. Springer-Verlag, Berlin,Heidelberg, pp. 515–537.

Barton, G., Delmas, P., 2002. A semi-automated colour predicate for robust skindetection. In: Proceedings of IVCNZ 2002, Palmerston North, New Zealand, pp.121–125.

Bejat, L., Perfect, E., Quisenberry, V.L., Coyne, M.S., Haszler, G.R., 2000. Solutetransport as related to soil structure in unsaturated intact soil blocks. SoilSci. Soc. Am. J. 64, 818–826.

Cabrera Carvajal, F., Okeschko, K., 1995. Efecto de la labranza sobre la estructurainterna de dos tipos de suelo. Agric. Tec. Mex. 21, 139–158.

Cervantes, G.G., Sanchez-Cohen, I., Rossignol, J.P., 2003. Soil water dynamics studiesusing image analysis. In: Proceedings of the First Interagency Conference onResearch in the Watersheds (ICRW), Benson, Arizona, USA, 27–30 October 2003,pp. 263–271.

Coster, M., Chermant, J.L., 1985. Precis d’analyse d’images. CNRS, Paris, France.Duwig, C., Muller, K., Vogeler, I., 2006. 2,4-D movement in allophanic soils from two

contrasted climatic regions. Commun. Soil Sci. Plant Anal. 37, 2841–2855.Duwig, C., Delmas, P., Muller, K., Prado, B., Morin, H., Ren, K., 2008. Quantifying

fluorescent tracer distribution in allophanic soils to image solute transport. Eur.J. Soil Sci. 59, 94–102.

Edwards, W.M., Shipitalo, M.J., Dick, W.A., 1992. Rainfall intensity affects transportof water and chemicals through macropores in no-till soil. Soil Sci. Soc. Am. J. 56,52–58.

Epstein, S., Mayeda, T.K., 1953. Variations of the 18O/16O ratio in natural waters.Geochem. Cosmochim. Acta 4, 213–224.

Forrer, I., Papritz, A., Kasteel, R., Fluhler, H., Luca, F., 2000. Quantifying dye tracers insoil profiles by image processing. Eur. J. Soil Sci. 51, 313–322.

Gutierrez, C., 2004. Micromorfologıa de suelos. Colegio de Postgraduados, Mon-tecillo (Mexico), Soil Science lecture notes (unpubl.) 40 p.

Hallaire, V., Curmi, P., Widiatmaka, 1997. Morphologie de la porosite et circulationspreferentielles en sature. Cas des horizons d’un systeme pedologique armor-icain. Etude et Gestion des Sols 4, 115–126.

Hallaire, V., Hachicha, M., Cheverry, C., 1998. Evolution structurale d’un horizon desurface argileux sous irrigation. Caracterisation de la macroporosite par analysed’images. Etude et Gestion des Sols 5, 107–116.

Hallaire, V., Cointepas, J.P., 1993. Caracterisation de la macroporosite d’un sol deverger par analyse d’image. Agronomie 13, 155–164.

Haralick, R., Shapiro, L., 1985. A survey: image segmentation techniques. J. Comput.Vis. Graph. Image Process. 29, 100–132.

Hatano, R., Kawamura, N., Ikeda, J., Sakuma, T., 1992. Evaluation of the effect ofmorphological features of flow paths on solute transport by using fractaldimensions of methylene blue staining pattern. Geoderma 53, 31–44.

Holtham, D.A.L., Matthews, P.G., Scholefield, D.S., 2007. Measurement and simula-tion of void structure and hydraulic changes caused by root-induced soilstructuring under white clover compared to ryegrass. Geoderma 142, 142–151.

Hubert, F., Vincent, H., Sardini, P., Caner, L., Heddadj, D., 2007. Pore morphologychanges under tillage and no-tillage practices. Geoderma 142, 226–236.

Ismail, S.N.A., 1975. Micromorphometric Soil Porosity Characterization by Meansof Electro-optical Image Analysis (Quantimet 720). Soil Survey Institute,Wageningen.

Kamra, S.K., Lennartz, B., van Genuchten, M.Th., Widmoser, P., 2001. Evaluatingnon-equilibrium solute transport in small soil columns. J. Contam. Hydrol. 48,189–212.

Kutılek, M., Nielsen, D.R., 1994. Soil Hydrology. GeoEcology. Catena Verlag, Crem-lingen-Destedt, Germany.

Lee, J., Horton, R., Noborio, K., Jaynes, D.B., 2001. Characterization of preferentialflow in undisturbed, structured soil columns using a vertical TDR probe. J.Contam. Hydrol. 51, 131–144.

Magesan, G.N., Vogeler, I., Scotter, D.R., Clothier, B.E., Tillman, R.W., 1995.Solute movement through two unsaturated soils. Aust. J. Soil Res. 33,585–596.

Mori, Y., Iwama, K., Maruyama, T., Mitsuno, T., 1999. Discriminating the influence ofsoil texture and management-induced changes in macropore flow using soft X-rays. Soil Sci. 164, 467–482.

Muller, K., Duwig, C., 2007. The transport and sorption of 2,4-D in allophanic soils.Soil Sci. 72, 333–348.

Murphy, C.P., Bullock, P., Turner, R.H., 1977. The measurement and characterizationof voids in soil thin sections by image analysis. I. Principles and techniques. J.Soil Sci. 28, 498–508.

Nanzyo, M., Shoji, S., Dahlgren, R., 1993. Physical characteristic of volcanic ash soils.In: Shoji, S., Nanzyo, M., Dahlgren, R. (Eds.), Volcanic Ash Soils, Genesis, Proper-ties and Utilization. Elsevier, The Netherlands, pp. 189–207.

Oleschko, K., Chapa Guerrero, J.R., 1989. Structural dynamics of a mollic Andosol ofMexico under tillage. Soil Till. Res. 15, 25–40.

Pallud, C., 2000. Etude multi-echelles du fonctionnement hydrodynamique etmicrobiologique d un sol soumis a un apport de solutes: application a l’ammo-nium et au 2,4-D. Ph.D. diss. Joseph Fourier Univ., Grenoble, 259 p.

Parfitt, R.L., 1990. Allophane in New Zealand—a review. Aust. J. Soil Res. 28, 343–360.

Perret, J., Prasher, S.O., Kantzas, A., Hamilton, K., Langford, C., 2000. Preferentialsolute flow in intact soil columns measured by SPECT scanning. Soil Sci. Soc. Am.J. 64, 469–477.

Persson, M., 2005. Accurate dye tracer concentration estimations using imageanalysis. Soil Sci. Soc. Am. J. 69, 967–975.

Poletika, N.N., Jury, W.A., 1994. Effects of soil management on water flow distribu-tion and solute dispersion. Soil Sci. Soc. Am. J. 58, 999–1006.

Prado, B., 2006. Etude du mouvement de l’eau et du transfert reactif du nitrate dansles sols volcaniques du bassin versant elementaire de la Loma, Mexique.Doctoral dissertation, University Joseph Fourier, Grenoble, France.

Prado, B., Duwig, C., Escudey, M., Esteves, M., 2006. Nitrate sorption in a Mexicanallophanic Andisol using intact and packed columns. Commun. Soil Sci. PlantAnal. 37, 2911–2925.

Prado, B., Duwig, C., Hidalgo, C., Gomez, D., Prat, C., Etchevers, J.D., Esteves, M., 2007.Characterization, classification and functioning of two profiles under differentland uses in a volcanic soil sequence in Central Mexico. Geoderma 139, 300–313.

Protz, R., Shipitalo, M.J., Mermut, A.R., Fox, C.A., 1987. Image analysis of soils—present and future. Geoderma 40, 115–125.

Quinsenberry, V.L., Philliphs, R.E., Zeleznik, J.M., 1994. Spatial distribution of waterand chloride macropore flow in a well structured soil. Soil Sci. Soc. Am. J. 58,1294–1300.

Rao, P.S.C., Rolston, D.E., Jessup, R.E., Davidson, J.M., 1980. Solute transport inaggregated porous media: theoretical and experimental evaluation. Soil Sci.Soc. Am. J. 44, 1139–1146.

Ringrose-Voase, A.J., 1996. Measurement of soil macropore geometry by imageanalysis of sections through impregnated soil. Plant Soil 183, 27–47.

Seyfriend, M.S., Rao, P.S.C., 1987. Solute transport in undisturbed columns of anaggregated tropical soil: preferential flow effects. Soil Sci. Soc. Am. J. 51, 1434–1444.

Shipitalo, M.J., Edwards, W.M., Dick, W.A., Owens, L.B., 1990. Initial storms effects ofmacropore transport of surface applied chemicals in no-till soil. Soil Sci. Soc.Am. J. 54, 1530–1536.

Shoji, S., Nanzyo, M., Dahlgren, R.A., 1993. Volcanic Ash Soils. Genesis, Propertiesand Utilization. Elsevier Science Publishers, The Netherlands.

Serra, J., 1982. Image Analysis and Mathematical Morphology. Academic Press,London.

Strock, J.S., Cassel, D.K., Gumpertz, M.L., 2001. Spatial variability of water andbromide transport through variably saturated soil blocks. Soil Sci. Soc. Am. J.65, 1607–1617.

Sugita, F., Gillham, R.W., Mase, C., 1995. Pore scale variation in retardation factor asa cause of nonideal reactive breakthrough curves. 2. Pore network analysis.Water Resour. Res. 31, 113–119.

Sugita, F., Gillham, R.W., 1995. Pore scale variation in retardation factor as a cause ofnonideal reactive breakthrough curves. 1. Conceptual model and its evaluation.Water Resour. Res. 31, 103–112.

Tarquis, A.M., McInnes, K.J., Key, J.R., Saa, A., Garcıa, M.R., Dıaz, M.C., 2006. Multi-scaling analysis in a structured clay soil using 2D images. J. Hydrol. 322, 236–246.

Toride, N., Leij, F.J., van Genuchten, M.Th., 1999. The CXTFIT code for estimatingtransport parameters from laboratory or field tracer experiments. Version 2.1.U.S. Salinity Laboratory Research Report 137, Riverside, California.

Tortajada, C., Castelan, E., 2003. Water management for a megacity: Mexico Citymetropolitan area. Ambio 32, 124–129.

Vanderborght, J., Timmerman, A., Feyen, J., 2000. Solute transport for steady-stateand transient flow in soils with and without macropores. Soil Sci. Soc. Am. J. 64,1305–1317.


Vanderborght, J., Gahwiller, P., Wydler, H., Schultze, U., Fluhler, H., 2002. Imagingfluorescent dye concentrations on soil surfaces: uncertainty of concentrationestimates. Soil Sci. Soc. Am. J. 66, 760–776.

Walker, P.J.C., Trudgill, T.S., 1983. Quantimet image analysis of soil pore geometry:comparison with travel break breakthrough curves. Earth Surf. Process. Land-forms 8, 465–472.

WRB (World Reference Base for Soil Resources), 2001. Lecture Notes onthe Major Soils of the World. In: Driessen, P., Deckers, J., Spaargaren,O., Nachtergaele, F. (Eds.), FAO World Soil Resources Reports 94,Rome.

Yule, D.F., Gardner, W.R., 1978. Longitudinal and transverse dispersion coefficientsin unsaturated plain field sand. Water Resour. Res. 14, 582–588.