Aa

ESE

taa�B2TiIha

.

©

GEOPHYSICS, VOL. 74, NO. 2 �MARCH-APRIL 2009�; P. O29–O40, 15 FIGS., 1 TABLE.10.1190/1.3078402

passive seismic survey over a gas field: Analysis of low-frequencynomalies

rik H. Saenger1, Stefan M. Schmalholz2, Marc-A. Lambert2, Tung T. Nguyen2, Arnaud Torres3,abrina Metzger3, Robert M. Habiger3, Tamara Müller3, Susanne Rentsch4, andfraín Méndez-Hernández5

hcictscfltTsav

ABSTRACT

Passive seismic low-frequency �from approximately 1–6 Hz�data have been acquired at several locations around the world.Spectra calculated from these data, acquired over fields withknown hydrocarbon accumulations, show common spectralanomalies. Verification of whether these anomalies are commonto only a few, many, or all hydrocarbon reservoirs can be provid-ed only if more and detailed results are reported. An extensivesurvey was carried out above a tight gas reservoir and an adjacentexploration area in Mexico. Data from several hundred stationswith three-component broadband seismometers distributedover approximately 200 km2 were used for the analysis. Several

Bbme�ebswioe

eceivedG, Zuriclz@erdis.com;

co Lonzh@pe

O29

ydrocarbon reservoir-related microtremor attributes were cal-ulated, and mapped attributes were compared with known gasntervals, with good agreement. Wells drilled after the surveyonfirm a predicted high hydrocarbon potential in the explora-ion area. A preliminary model was developed to explain theource mechanism of those microtremors. Poroelastic effectsaused by wave-induced fluid flow and oscillations of differentuid phases are significant processes in the low-frequency range

hat can modify the omnipresent seismic background spectrum.hese processes only occur in partially saturated rocks. We as-ume that hydrocarbon reservoirs are partially saturated, where-s the surrounding rocks are fully saturated. Our real data obser-ations are consistent with this conceptual model.

INTRODUCTION

A growing number of surveys over different oil and gas fieldshroughout the world have established the presence of spectralnomalies in the passive seismic wavefield, i.e., microtremors, withhigh degree of correlation to the location of hydrocarbon reservoirs

Dangel et al., 2003; Holzner et al., 2005; Akrawi and Bloch, 2006;irialtsev et al., 2006; Rached, 2006; Suntsov et al., 2006; Graf et al.,007; Lambert et al., 2008; van Mastrigt and Al-Dulaijan, 2008�.hese microtremors can be used as a reservoir indicator for optimiz-

ng well placement during exploration, appraisal, and development.n contrast to conventional seismic technologies, the investigation ofydrocarbon reservoir-related microtremors is generally passivend does not require artificial seismic excitation sources.

Manuscript received by the Editor 7 November 2007; revised manuscript r1ETH Zurich, Geological Institute, Zurich, Switzerland, and SpectraseisA2ETH Zurich, Geological Institute, Zurich, Switzerland. E-mail: schmalho3Spectraseis AG, Zurich, Switzerland. E-mail: arnaud.torres@spectrase

mueller@spectraseis.com.4Formerly Freie Universität Berlin, Berlin, Germany; presently WesternGe5Pemex Exploratión y Producción, Villahermosa, Mexico. E-mail: emende2009 Society of Exploration Geophysicists.All rights reserved.

The ever-present seismic background noise of the earth �e.g.,erger et al., 2004� most likely acts as the driving force for hydrocar-on-indicating signals. Some possible underlying rock-physicsechanisms that generate spectral anomalies are discussed in Graf

t al. �2007� and are considered in this paper. The main observatione.g., Dangel et al., 2003; van Mastrigt and Al-Dulaijan, 2008� is annergy anomaly in the low-frequency band of passive seismic dataetween approximately 1 and 6 Hz. When measured at the surface,pectral energy is elevated above a hydrocarbon reservoir, comparedith spectral energy measured at positions away from a reservoir. It

s important, theoretically, that the generating mechanism and thebserved anomaly may be present in a wider-frequency range. How-ver, between 1 and 6 Hz there is a typical noise trough in the back-

20 October 2008; published online 3 March 2009.h, Switzerland. E-mail: erik.saenger@erdw.ethz.ch.

w.ethz.ch; marc.lambert@erdw.ethz.ch; tung.nguyen@erdw.ethz.ch.sabrina.metzger@spectraseis.com; rob.habiger@spectraseis.com; tamara

don Technology Centre, Gatwick, U. K. E-mail: SRentsch@slb.com.p.pemex.com.

gba

atleptuccvFm

ceaaarcnwrar

oWlrakm

tstftn

L

TMCcmqkhst

••

•

•

snribtev

tsletsvLtt

A

b

Fr4eitp

O30 Saenger et al.

round spectrum �Peterson, 1993; Berger et al., 2004�, which mighte the only frequency window where hydrocarbon-related effectsre visible.

In addition to the energy anomaly, Lambert et al. �2007� describenother independent spectral attribute. They find that the spectral ra-io between horizontal and vertical components can show an anoma-y in the presence of hydrocarbons. Because of the well-known pres-nce of surface waves, locating the source of these anomalies is ofrimary importance. Steiner et al. �2008� show results that indicatehe reservoir zone is the origin of low-frequency microtremors. Bysing a time-reverse wave-propagation method, they suggest to lo-ate the corresponding source of the anomaly in depth. All of theseharacteristics make it possible to distinguish the hydrocarbon reser-oir-related indicators from well-known site effects �Bard, 1999;äh et al., 2001� and volcanic tremors �Schick, 1988� in passive seis-ic data sets.The analysis of low-frequency microtremors for detecting hydro-

arbon reservoirs is an emerging technology with ongoing researchspecially focusing on data-analysis techniques. Ambiguous resultsnd possible pitfalls also are reported �Ali et al., 2007; Berteussen etl., 2008; Hanssen and Bussat, 2008�. Analyzing microtremorsround reservoirs with considerable noise �e.g., production noise�equires careful data analysis because �a� anomalies caused by noisean be misinterpreted as caused by the reservoir or �b� such a high-oise environment can overwhelm the signal. The question ofhether spectral anomalies in the passive surface wavefield are cor-

elated to all, many, or only a few hydrocarbon reservoirs can only benswered if more and detailed results of passive seismic surveys areeported.

We describe results of a passive low-frequency survey carried outver a tight gas reservoir and an adjacent exploration area in Mexico.e briefly describe the acquisition and main processing steps fol-

owed to map the spectral energy anomaly. Next, we analyze the dataecorded at two locations above and away from the reservoir as ex-mples of the anomaly behavior in this data set. We extract severaley attributes to characterize the modifications of the passive seis-ic wavefield caused by the presence of hydrocarbons. The distribu-

70575

Station (seismometer)

Drainage area

Zone 4 (horst)

Zone 3 (graben)

Zone 2(horst)

Zone 1(graben)

70139

km

igure 1. Main geologic structures at reservoir depth and drainageadii around producing wells. Zones 1 and 3 are grabens, zones 2 andare horsts. Zone 2 is well developed and contains most of the prov-n recoverable reserves of dry gas. Production started 25 years agon the top right of zone 2. It continues today and generally has movedoward station 70139. The small stars indicate the measurementoints for the survey. Stations 70139 and 70575 are circled in red.

ion of these attributes with spatial location is compared with the ob-erved energy anomaly for the whole survey. The goal of this charac-erization is twofold: to improve the processing and interpretationor future surveys and to verify or refute theoretical explanations ofhe origin of hydrocarbon reservoir-related microtremors.Aprelimi-ary rock-physics model is discussed in a separate section.

THE SURVEY

ocation and geology

The survey area lies in the Burgos basin in northeastern Mexico.he origin of this basin is associated with the opening of the Gulf ofexico during the Jurassic period, beginning sedimentation in theallovian with evaporitic deposits. The sedimentation conditionshanged in the Cenozoic when a great regression occurred. A sedi-entary sequence of at least 8000 m was accumulated. This se-

uence has strong gas potential, represented by rocks with type IIIerogen. The complex fault system in the survey area comprisesorst and graben structures and is part of a larger-scale half-grabenystem. The main structural features are the four major listric faultshat divide the survey area into four zones �see Figure 1�:

Zone 1: Graben and production area.Zone 2: Horst — the main producing structure that has been ex-ploited for more than 25 years. The average reservoir is at about2000 m in depth. These known gas accumulations provide a testlocation to corroborate the correlation of low-frequency anoma-lies with the presence of hydrocarbons.Zone 3: Graben — where the deepest block of the area lies. How-ever, a small-scale horst structure is present in the middle of thegraben with producing wells.Zone 4: Horst — an interesting prospect for further exploration.Some wells were drilled on the uppermost parts of the horst struc-ture and some 2D seismic information was acquired, but much ofthe zone remains unexplored.

A secondary fault trend compartmentalizes the whole system intomaller blocks of varying size, volume, depth, and reservoir thick-ess. This is particularly the case in zones 1 and 2 �see Figure 1�. Theeservoir system �Paleocene Wilcox� consists of four main produc-ng intervals; the top deltaic sequence is the best producer, followedy three other sandstone layers. The net sand thickness varies be-ween 120 and 30 m, caused by block-nose erosion on the shallow-st compartments, faulting of some blocks, and lateral thicknessariations of the sediments.

Above the Play-Lobo unconformity that seals the reservoir zone,he Eocene sequence is regular and homogeneous �rather constant ineismic velocity�, with synsedimentary fan-shaped thickening. Theate Eocene layers outcrop almost everywhere. The surface soil lay-rs show parallel trends along the main fault direction correspondingo variations in the composition, forming alternating regosol/xerosoltripes of approximately 1 km wide, especially in zone 4 of the sur-ey. They consist mostly of reddish lutites and shales from the Mid-ate Cenozoic. No detailed information is available on the soil

hickness variations, but it appears homogeneous from the surface inhis semiarid environment.

cquisition and grid layout

Using 20 ultrasensitive portable three-component �3-C� broad-and seismometers �frequency range, 0.03–50 Hz; sampling rate,

1sattfig

oo9av

whpsd

N

adtt

scAvsawv

tt

fd1d

P

tcvocssc

Fwfb�iTcth

FtltowPr

Spectral anomalies above reservoirs O31

00 Hz; sensitivity, 1500 V/m/s�, we acquired more than 700 mea-urements of the omnipresent seismic wavefield at the surface overpproximately 200 km2. Two grid layouts were acquired consecu-ively over a 3-month period. Both used 1000-m node spacing, buthe second grid was staggered offset with respect to the middle of therst, reducing the average spacing between nodes to 700 m. Someaps in the grid are caused by access issues with landowners.

Four permanent reference stations were installed for the durationf the survey. However, because of health and safety restrictions, allther measurements were recorded during the daytime �i.e., from:00 to 19:00�. After each measurement, with a minimum durationround 3 hours, the raw 3-C sensor data �measuring surface particleelocities� were stored with an individual station number.

An important step during the survey was quality control, whichas performed directly in the field. Bad measurements caused byardware problems or an artificially high noise level were identifiedromptly. In this case, the location was remeasured.Additionally, weystematically remeasured some data points as a repeatability test toetermine the variability of the passive seismic wavefield with time.

oise identification

The raw data may include strong perturbations �noise, artifacts�nd discontinuities �data gaps�. To obtain a clean signal in the timeomain, we cut out all time intervals with obvious strong artificial in-erferences. This is an important step in the workflow. However, it ishe first interpretive, nonautomatic routine in the data processing.

Figure 2 shows a typical example.All well-known, transient noiseignatures �e.g., vehicle noise� are removed from the data. From theleaned data �green�, we calculate the power spectral density �PSD�.standard procedure is applied to determine the PSD for 40-s inter-

als and then to calculate the arithmetic average for the whole mea-urement. For this survey, records of at least 2 hours led to a stablend reproducible result in the frequency domain. Measurementsith less than 90 minutes of clean data �i.e., with less than 135 inter-als of 40 s� were not used for further analysis. Figure 3 shows spec-

15:50 16:19 16:48 17:16 17:45 18:14 18:43−0.25

−0.2

−0.15

−0.1

−0.05

0

Time (hh:mm)

Par

ticle

velo

city

(arb

.uni

ts)

igure 2. Time series of vertical particle velocity �raw sensor dataith DC shift� of seismic motion measured at 100-Hz sampling rate

rom station 70139. Time intervals that do not contain large distur-ances are selected �green� from the original recorded time seriesblue� and used for further processing. The decision on which cleanntervals are chosen is based on several quality control procedures.he high amplitudes at the very beginning and the very end areaused by the crew switching the sensor on and off. The long-timerend ��0.15 to �0.07� is the technical drift of the instrument andas no influence on our applied analysis.

ra of the vertical component, and the corresponding standard devia-ion, for two different measured stations, marked in Figure 1.

RESERVOIR-RELATED SPECTRAL ATTRIBUTES

A primary goal of the data processing is to identify and map low-requency energy anomalies in the expected total bandwidth of hy-rocarbon reservoir-related microtremors �i.e., approximately–6 Hz�. Dangel et al. �2003� suggest this could be accomplished byetermining an attribute related to the strength of spectral peaks.

SD-IZ

In contrast with Dangel et al. �2003�, we recommend an integra-ion technique along a linear frequency scale that considers a singleomponent of the signal. Figure 3 illustrates the calculation using theertical �z-� component, but the anomaly itself can be observed onther components as well. Noise-floor variations are taken into ac-ount by determining the individual minimum amplitude of eachpectrum between 1 and 1.7 Hz. For this survey, we typically ob-erve a minimum in this range; for other surveys, one may slightlyhange these values.

Frequency (Hz)Frequency (Hz)

Frequency (Hz)

70575

70139

Frequency (Hz)

1 7654320

0.5

1

1.5

2

2.5

3

3.5

4

1 7654320

0.5

1

1.5

2

2.5

3

3.5

4

PS

D(a

rbitr

ary

units

)P

SD

(arb

itrar

yun

its)

a)

b)

igure 3. Spectrum �solid line� of the passive seismic wavefield �ver-ical surface velocities� from 0.5 to 7.4 Hz frequency. The dashedine displays the standard deviation of the mean spectrum. �a� Sta-ion 70139 was recorded over a known gas field; �b� station 70575 isver an area with no hydrocarbon potential. The shaded surface is al-ays calculated using a linear frequency scale and illustrates theSD-IZ value. The amplitudes for both stations can be compared di-ectly; no scaling factor is applied.

Ptb

nbtF

oqmTsb

aPpHlzisa

V

lesfprviHsg

cssp�l2ctlTpoe

opstTo

FtuRopt

F0fiT

O32 Saenger et al.

The integral above this minimum amplitude level defines theSD-IZ value, where IZ stands for the integral of the z-component. It

akes into account the whole energy anomaly above a well-definedackground level �i.e., the minimum between 1 and 1.7 Hz� and is

kmPSD−IZ (interpolated)

0−150

150−500

500−1240

1240−5000

igure 4. Survey map of the PSD-IZ attribute. Green circles displayhe PSD-IZ value of each station. Four levels of signal strength aresed for the interpolated map using a standard kriging algorithm.ed indicates the location of the highest values. There are, amongthers, two areas with relatively high values: One is in the producingart of the survey �solid ellipse�, and the other is in a chosen explora-ion zone �dashed circle�.

V/H

(dim

ensi

onle

ss)

V/H

(dim

ensi

onle

ss)

1 7654320

0.5

1

1.5

2

2.5

3

3.5

4

1 7654320

0.5

1

1.5

2

2.5

3

3.5

4

Frequency (Hz)

Frequency (Hz)

70575

70139

a)

b)

igure 5. The V/H ratio of the passive seismic wavefield from.5 to 7.4 Hz. �a� Station 70139 was recorded over a known gaseld; �b� station 70575 is over an area with no hydrocarbon potential.he horizontal thin line indicates a value of V/H � 1.

ot restricted to peak strength at specific frequencies. Therefore, am-iguous high-amplitude peaks, which may be caused by human ac-ivity at the surface �e.g., the narrow-band peaks at 2.5 or 3 Hz inigure 3�, do not contribute significantly to the PSD-IZ value.For this survey, we calculate the integral up to 3.7 Hz because of

bvious artificial monochromatic noise identified above this fre-uency for stations not shown here. The attribute defined in thisanner for the PSD of station 70575 �Figure 3� is effectively zero.he procedure is applied to the passive seismic data of each mea-urement point. As an output, we generate an energy anomaly mapased on this specific integral attribute.

In Figure 4, we show such a map and classify four different anom-ly levels to characterize their spatial distribution. The strongestSD-IZ values are well aligned with the zone where most of the gas-roducing wells are concentrated �zone 2, marked in Figure 1�.owever, the main trend of high PSD-IZ values crosses perpendicu-

ar to the horst-and-graben trend and picks up the proven reserves inones 1 and 3 as well. Moreover, there is no acquisition footprint vis-ble in the map, although the measurements are performed on twotaggered grids with a time difference of six weeks �see Acquisitionnd Grid Layout�.

ertical/horizontal signal

The PSD-IZ value is one attribute to characterize spectral anoma-ies above hydrocarbon reservoirs.Another signature can be extract-d by analyzing spectral ratios. They are more stable in time than ab-olute spectra �Bard, 1999�. This spectral ratio attribute is developedrom an investigation that found there can be a trough rather than aeak in the horizontal/vertical �H/V� ratio within the frequencyange that Dangel et al. �2003� considers for the hydrocarbon reser-oir-related spectral anomaly. Therefore, we develop an attribute us-ng the vertical/horizontal �V/H� ratio in contrast to the well-known/V ratio method used by others to identify soil layers with passive

eismic data sets �e.g., Fäh et al., 2001�. Lambert et al. �2007� sug-est that V/H anomalies might correlate with the reservoir.

As shown in Figure 5, we observe a V/H peak �i.e., values signifi-antly above one� in the frequency band between 1 and 6 Hz for atation placed above hydrocarbons. To map this attribute for all mea-urement locations, we calculate the surface area under the ratio ofower spectral densities that are above one and between 1 and 6 Hzsee Figure 5�. As for the energy anomaly, we observe relativelyarge anomaly patches throughout the known reservoir area in zone. As the V/H spectral ratio normalizes the vertical to the horizontalomponents, this attribute is independent of the PSD-IZ attribute. Ifhe general noise level is low, weak seismic signals associated with aow energy anomaly can induce a significant V/H signal value.herefore, the V/H signal values shown in Figure 6 must be inter-reted in conjunction with the PSD-IZ map �Figure 4�. More detailsn how we calculate the spectral ratio are summarized from Lambertt al. �2006� inAppendix A.

Surface layers can induce a peak and a trough in the H/V spectrumf the passive seismic wavefield �Konno and Ohmachi, 1998�. Aeak in the V/H spectrum caused by surface layers can mask our V/Hignal attribute. The trough, which appears as a peak in the V/H spec-rum, can induce misleading high values of the V/H signal attribute.he specific frequency in which this effect occurs depends on the ge-metry and the seismic properties of the surface layers. The trough

fHq

wl

F

ebia�ttaafesTc

ttTzaaw

naotf

P

faarfitt

wi

ib

f

wt

Fpsittz

FagPtql

Spectral anomalies above reservoirs O33

requency is approximately double the peak frequency fH/V in the/V spectrum �Konno and Ohmachi, 1998�, which is given by theuarter-wavelength rule �e.g., Pujol, 2003�:

fH/V �VS

4H, �1�

here VS is the S-wave velocity and H is the thickness of the surfaceayer.

requency shift of maximum spectral peaks

Dangel et al. �2003� observe that spectra above hydrocarbon res-rvoirs usually contain spectral peaks within a narrow-frequencyand of 1.5–4 Hz, characteristic of the oil- and gas-bearing locationsn the field. In this survey, spectral peaks are indeed present on sever-l spectra of the acquired data, as shown for station 70139 in Figure 3e.g., at 2.5, 3, 4, and 5 Hz�. However, the number of the peaks andheir relative/absolute amplitudes fluctuate a lot with time and loca-ion, making it difficult to generate a consistent map based on theirverage amplitudes, as suggested by Dangel et al. �2003�. Instead ofnalyzing the amplitudes, this method focuses on determining therequency value corresponding to the maximum spectral peak forach measurement point within a certain frequency window on thepectrum. The approach provides information about the anomalies.his is the frequency of the most significant peak within the frequen-y band of interest.

Figure 7 shows the distribution of the frequencies correspondingo maximum spectral peaks between 1.5 and 3.7 Hz. Red indicateshe areas with maximum spectral peaks appearing above 2.4 Hz.hese areas match well with known hydrocarbon-bearing areas inone 2. A possible explanation of this frequency-shift observationnd its relation to hydrocarbon reservoirs is discussed by Lambert etl. �2008�. It is based on the assumption that reservoirs emit seismicaves at low frequencies, induced by a directed force.The technique also can be used to identify locations of artificial

oise sources. For example, in this survey, a peak at 5 Hz frequentlyppears on the spectra. This feature is monochromatic and stablever time, which leads us to conclude it is a man-made component ofhe wavefield with no relation to the subsurface. As such, this strongeature must be avoided.

olarization

Principal-components analysis �PCA� of particle motion as aunction of time is described by Jurkevics �1988� as polarizationnalysis. Figure 8 illustrates the four polarization attributes of thispproach. The first step in our procedure is to band-pass filter the al-eady cleaned 3-C data in the time domain. We apply a zero-phaselter that passes the frequencies from 1 to 3.7 Hz. Considering any

ime interval of 3-C data ux, uy, and uz containing N time samples, au-o- and crossvariances can be obtained with

Cij � � 1

N�s�1

N

ui�s�uj�s�� , �2�

here i and j represent the component index x, y, z and where s is thendex variable for a time sample. The 3�3 covariance matrix

C � �Cxx Cxy Cxz

Cxy Cyy Cyz

Cxz Cyz Czz� �3�

s real and symmetric and represents a polarization ellipsoid withest fit to the data.

The principal axis of this ellipsoid can be obtained by solving Cor its eigenvalues �1, �2, �3 and eigenvectors p1, p2, p3:

�C � �I�p � 0, �4�

here I is the identity matrix. The solid red vector in Figure 8 illus-rates the largest eigenvector.

kmV/H−Signal

0−20

20−80

80−160

160−500

igure 6. Survey map of the V/H attribute. The V/H attribute is inde-endent of the PSD-IZ attribute mapped in Figure 4. Four levels ofignal strength are displayed using a standard kriging algorithm. Redndicates the location of the highest values. There are, among others,wo areas with relatively high values: One is in the producing part ofhe survey �solid ellipse�, and the other is in a chosen explorationone �dashed circle�.

kmPeak frequency

< 1.8

1.8−2.0

2.0−2.4

Hz

Hz

Hz

> 2.4 Hz

igure 7. Survey map of the frequency of the maximum peak withinwindow from 1.5 Hz �blue� to 3.7 Hz �red�. A standard kriging al-orithm is applied. The displayed signature is independent of theSD-IZ and V/H attributes mapped in Figure 4 and Figure 6, respec-

ively. There are, among others, two areas with a relatively high-fre-uency signature: One is in the producing part of the survey �solid el-ipse�, and the other is in a chosen exploration zone �dashed circle�.

tl

IirtT

Ip

att

rvasthis

pvr�

arscgt

ocwa

Fthdrs

FphhtaDF

Fp

O34 Saenger et al.

The rectilinearity parameter L, sometimes called linearity, relateshe magnitudes of the intermediate and smallest eigenvalue to theargest eigenvalue:

L � 1 � ��2 � �3

2�1 . �5�

t measures the degree of how the linear incoming wavefield is polar-zed. It yields values between zero and one �random ball and needle,espectively; see Figure 8�. Two angles, dip and azimuth, describehe orientation of the largest eigenvector, p1 � �p1�x�,p1�y�,p1�z��.he dip is calculated with

� � arctan� p1�z�p1�x�2 � p1�y�2 . �6�

t is zero for a horizontal polarization and is defined positive for theositive z-direction. The azimuth is specified as

x

y

z

x

y

z

a) b)

igure 8. Polarization attribute sketch demonstrating variability inhe rectilinearity. In this schematic, the solid blue line illustrates a 3Dologram of particle velocity. �a� High rectilinearity and mediumip. �b� Low rectilinearity and relatively high dip. The length of theed arrow is given by the largest eigenvalue �1, referred to as thetrength of the signal.

igure 9. Time variations of the four polarization attributes for zero-hase band-pass-filtered data �1–3.7 Hz� from station 70139 �highydrocarbon potential�. Time intervals of 40 s are analyzed. Theorizontal solid line represents the value using data of the wholeime period. A dip of � � 90° indicates vertical particle oscillationnd an azimuth of � � 0° indicates north-south particle oscillation.ip, azimuth, rectilinearity, and largest eigenvalue are illustrated inigure 8.

� � arctan� p1�y�p1�x�

�7�

nd is measured counterclockwise from the positive x-axis. In addi-ion, we consider the strength variations of the signal by analyzinghe largest eigenvalue �1.

Figures 9 and 10 plot dip �, azimuth � , largest eigenvalue �1, andectilinearity L as a function of time �by analyzing 40-s time inter-als� for microtremor data of stations 70139 and 70575. The stationsre located in an area with high and low hydrocarbon potential, re-pectively. It is clear that we have, on the surface, a complicated mix-ure of different wave types with different origins. However, weave identified stable trends in the survey area regarding the polar-zation attributes in the frequency range between 1 and 3.7 Hz. Weummarize our observations in Table 1.

This polarization analysis is useful for a detailed analysis of theassive seismic wavefield. It provides information about the timeariability of the microtremor phenomena related to hydrocarboneservoirs. For example, the time variation of the largest eigenvalue

1 and the azimuth � seen in Figure 9 seems typical for stationsbove a reservoir, whereas a relatively low largest eigenvalue and aelatively stable azimuth are more typical for an anthropogenic noiseource. Furthermore, the polarization analysis may be applied to lo-ate the source area of the microtremors in depth by applying a mi-ration approach. Such a technique is used by Rentsch et al. �2007�

o locate microseismic events by analyzing P-wave arrivals.

CORRELATION OF ATTRIBUTES WITHRESERVOIR LOCATION

Theoretically, it is possible to compare measurements directlynly where all affecting conditions are known precisely. This, ofourse, is not the case for a real passive seismic survey in an areaith complex reservoir geology. Therefore, the interpretation of all

ttribute maps �Figures 4, 6, and 7� must be undertaken with caution.

igure 10. Same as Figure 9 but for station 70575 �low hydrocarbonotential�.

A�hbog

S

awvwmlPti

tsnlalsubaeo

I

tscerHTtaagvv

sascotmis

est

C

mztfau1whips�

a

b

c

Fb�wcof

Spectral anomalies above reservoirs O35

s for all exploration methods, additional subsurface informatione.g., reflection seismic, well logs, or electromagnetic methods� iselpful. A complete and quantitative interpretation for the hydrocar-on-potential estimate must consider local and surveywide behaviorf the detected spectral anomaly. Therefore, we first discuss someuidelines of interest for this and other surveys.

imilar noise and source strength conditions

An ideal survey would consist of synchronous measurements forll grid locations over a very long time interval �longer than oneeek�. For practical reasons, this is not always possible. In our sur-ey, only daytime measurements with a maximum of 16 sensorsere performed �except for reference stations�. Several remeasure-ents for different days at exactly the same location with very simi-

ar results justify this procedure �e.g., variability was �20% for theSD-IZ value�. This is in line with the observation that the energy in

he frequency band between 0.07 and 0.5 Hz �oceanic microseism�s relatively constant �Figure 11�.

However, we expect better results for hydrocarbon reservoir de-ection if the data set also contains nighttime measurements for eachtation �Lambert et al., 2008�. The anthropogenicoise level, which may influence our analysis, isower during the night �Bard, 1999�. Typical vari-tions of the seismic background wavefield are il-ustrated for this survey in Figure 11. The ob-erved variations indicate that, in general, it is notseful to mix day and night data in one hydrocar-on reservoir-related attribute map. However, ancquisition footprint caused by significant chang-s in the seismic background wavefield cannot bebserved for this survey.

nfluence of subsurface heterogeneities

As shown in Figure 1, the reservoir geology ofhis survey area is a complex horst-and-grabenystem. An additional perpendicular fault systemompartmentalizes the reservoir formation inven smaller blocks. In contrast, the layers above this formation areelatively homogeneous.An analysis not shown here indicates weak/V peaks above 10 Hz in some areas of the considered survey.hese may be induced by soft-soil layers. However, we conclude

hat this effect is not able to significantly influence our V/H-signalnalysis below 6 Hz. Moreover, the topography in the investigatedrea is relatively flat. Therefore, the known effect that seismic ener-y can accumulate in topographic heights �Hestholm et al., 2006;an Mastrigt andAl-Dulaijan, 2008� cannot be observed for this sur-ey.

In general, complex underground features must influence the pas-ive seismic wavefield measured at the surface �Bard, 1999�. For ex-mple, body waves can be converted to surface waves at steps andlots �Saffari and Bond, 1987�. The identification of independent mi-rotremor attributes in this paper increases the confidence level inur ability to detect hydrocarbon reservoirs. The quality of interpre-ation can be increased further by applying other passive seismic

ethods on similar acquired data sets. For example, passive seismicmaging �Artman, 2006� or velocity inversion from 3-C array mea-urements �Fäh et al., 2008� can be used to determine subsurface het-

Table 1. ComPSD-IZ valu

Station

Figure

Dip �

Azimuth �

Largesteigenvalue �1

Rectilinearity

rogeneities. With the help of forward-modeling studies, one cantudy their particular influence on hydrocarbon reservoir-related at-ributes.

orrelation of attributes with producing reservoir area

The hydrocarbon distribution in zones 1 and 2 is not easy to deter-ine precisely for several reasons. The reservoir has multistack pay

ones and major unconformities that contribute to relatively largehickness variations. The low formation permeability also requiresracture stimulation to exploit the field. The radius of the drainedrea around a producing well can therefore be roughly estimated andsed to map the approximate location of the reservoir �Figures 1 and2�. In addition, the production history is quite different, with someells drilled recently and others producing for the past 25 years. Theydrocarbon-related PSD-IZ energy anomaly shown in Figure 4 isncreasing toward station 70139 in zone 2. This coincides with theroduction history. For nonstacked reservoirs, it is reported that thetrength of the energy anomaly is proportional to the pay thicknessDangel et al., 2003; Holzner et al., 2005�. However, this behavior

on of polarization attributes for stations with high and low

70139�above hydrocarbons�

70575�not above hydrocarbons�

10

able, high value ��80°� Stable, low value ��10°�

nstable, as expected for suchgh dip values

Relatively stable; may point to asurface noise source

rying, but relatively highring measurement period

Relatively low with some spikes

lative high, relatively stable,eakly correlated with �1

Lower in comparison to valuesobserved above the hydrocarbons

Noise Range (10−20Hz)

Signal Range (1−3Hz)

OWP Range (0.07−0.5Hz)

−150

−130

−130

−150

Mea

nva

lue

(dB

)

−110

−130

Mea

nva

lue

(dB

)M

ean

valu

e(d

B)

Mon Tue Wed Thu Fri Sat Sun

)

)

)

igure 11. Typical strength variations for one week of the seismicackground field of this survey for three different frequency bands:a� 10–20 Hz, �b� 1–3 Hz, and �c� 0.07–0.5 Hz. The processed dataere recorded with a permanent station in an area with low hydro-

arbon potential. Transients were not removed from these data. Thecean-wave peak �OWP� band is relatively constant; the noise rangerom 10–20 Hz shows typical day-night variations.

parise.

9

St

Uhi

Vadu

L Rew

cm

�sstkmcB

antbme

Hs

euo

dscdtdsictP

gorutlft

tlviTt

anwwtta

S

swgptqhf

1c

F�Pis

O36 Saenger et al.

annot be confirmed for the survey discussed in this paper becauseultistack pay zones are present.Given these aspects, we note that all microtremor attribute maps

PSD-IZ in Figure 4, V/H signal in Figure 6, and peak frequencyhift in Figure 7� show consistent patterns above zone 2. One of thetrongest microtremor anomalies �represented by measurement sta-ion 70139 and considering all attributes� is observed above the areanown, from production, to be hydrocarbon bearing. Therefore, ourultiattribute analysis shows that hydrocarbon reservoir-related mi-

rotremors are correlated spatially with this tight gas reservoir in theurgos basin in Mexico.In Figure 12, we show a direct comparison of the PSD-IZ values

nd the estimated drainage radii of the producing wells. Productionoise sources �e.g., tube waves or pumps� are an unlikely explana-ion for the observed pattern at the surface, which is marked with alack ellipsoid in Figures 4, 6, and 7. However, some ambiguity re-ains because a producing reservoir is not strictly a seismically qui-

t area �e.g., production facilities�.

ydrocarbon attributes versus wells drilled after theurvey

In the exploration zone of this survey, there is one big PSD-IZ en-rgy anomaly pattern in zone 4, marked with a dashed circle in Fig-re 4. We interpret the V/H signal and the relatively high frequencyf the maximum peak in Figures 6 and 7, respectively, as positive in-

igure 12. PSD-IZ attribute versus reservoir location information.a� Drainage radii of producing wells, reservoir fault system andSD-IZ values for each station. �b� Interpolated PSD-IZ values us-

ng a standard kriging algorithm and status of wells drilled after oururvey.

icators for the presence of hydrocarbons. It was a success for theurvey that two wells drilled in this area after data acquisition indi-ated gas-bearing sediments �see Figure 12�. Note that the low pro-uction rate of these wells may be caused by the low permeability ofhe reservoir rock. The observed anomaly cannot be induced by pro-uction activity because the mentioned wells were drilled after theurvey and there was no production or other significant human activ-ty in the vicinity �at least within the area marked with a dashed cir-le�.Also, two producing wells drilled after the survey in the produc-ion zone 2 are located where we have observed a relatively highSD-IZ value �Figure 12�.

PRELIMINARY MODEL OF HYDROCARBONRESERVOIR-RELATED MICROTREMORS

Graf et al. �2007� discuss possible mechanisms of microtremorsenerated or modified by hydrocarbon reservoirs. Because this is anngoing research field, we expect a continuous refinement of ourock-physics model. The model proposed here is based on well-doc-mented observations and well-known rock-physics wave-propaga-ion theories. Although other mechanisms could contribute to theow-frequency observations, we assume that the rock-physical ef-ects discussed below contribute to the observed signal characteris-ics.

Steiner et al. �2008� report an important observation regarding theheoretical understanding of this phenomenon. They show that theow-frequency anomaly may originate from the hydrocarbon reser-oir itself. In contrast to interferometry, this time-reverse algorithmmages the locus of an energy source rather than imaging reflectors.hat result connects the frequency-domain anomaly at the surface to

he reservoir at depth �Lambert et al., 2008�.We split our consideration into three parts: sources, mechanism,

nd observations. First, we look at possible sources. Because we doot use any active source, we must consider the seismic backgroundavefield. Second, we review possible rock-physics mechanismsithin a hydrocarbon deposit that are able to modify the spectra in

he low-frequency range above it. Third, we compare spectral at-ributes identified above to the theoretical description of the sourcend mechanism questions.

eismic background spectrum

Peterson �1993� and Berger et al. �2004� consider in detail thetrength of ambient earth noise. They develop a low-noise model,hich predicts the worldwide minimum energy for seismic back-round noise for a large-frequency band. This spectrum has two im-ortant features with respect to microtremors. First, there is a rela-ively quiet interval from 1 to 6 Hz �i.e., a minimum�. This is the fre-uency window where hydrocarbon reservoir-related microtremorsave been observed. Related physical effects may be present in otherrequency bands but may be much more difficult to discriminate.

Second, there is a dominant peak around 0.14 Hz �Friedrich et al.,998�. The origin of this peak is ocean waves interacting with theoast structure. This produces the so-called ocean-wave peak, which

csa2wtufafBa

R

fttprrfli

fiawwebtaiwb

R

b2asfdrapa

csVcc

R

r

botntf

rsst

1

2

3

P

mtotp

Fpls

Spectral anomalies above reservoirs O37

an be observed at all locations around the world. The correspondingurface waves propagate through entire continents and can, for ex-mple, be used to determine seismic velocities down to a depth of0 km �e.g., Shapiro et al., 2005�. Interestingly, Rayleigh wavesith frequencies around 0.14 Hz oscillate at reservoir depth �deeper

han 500 m� mainly in a vertical direction. This is illustrated in Fig-re 13. This preferred particle oscillation direction is also observedor the microtremors above a reservoir, which show V/H valuesbove 1 and a strong vertical polarization. Some more detailed low-requency borehole measurements of ambient noise are reviewed inonnefoy-Claudet et al. �2006�. Note that hydrocarbon reservoirsre always perturbed by the seismic background waves.

ock-physics low-frequency mechanism

Theoretically, it is very hard to explain specific low-frequency ef-ects of a hydrocarbon reservoir with elastic properties only. Weherefore consider poroelastic effects, which can cause high attenua-ion within the reservoir only and consequently increase the com-lex impedance contrast between the reservoir and the surroundingocks. In that case, the reservoir acts as a scatterer, and we refer to theelated effects as resonant scattering. We also consider microscaleuid oscillations caused by the surface tension between two pore flu-

ds and refer to such oscillation effects as resonant amplification.The mechanisms causing resonant scattering and resonant ampli-

cation can occur only in multiphase or partially saturated rocks. Wessume that the hydrocarbon reservoir is partially saturated �e.g.,ith gas and water� and the surrounding rocks are fully saturatedith water. The low-frequency resonant scattering and amplification

ffects therefore occur only within the reservoir and may modify theackground seismic wavefield in a characteristic way. These charac-eristic modifications can be observed in the spectral attributesbove hydrocarbon reservoirs.Another possibility would be a higherntensity of low-frequency fracture and/or fluid-migration processesithin the reservoir compared to outside the reservoir. Further possi-le nonlinear mechanisms are discussed in Zhukov et al. �2007�.

esonant scattering

Seismic low-frequency effects of hydrocarbon reservoirs haveeen known for many years �Castagna et al. 2003; Chapman et al.006; Goloshubin et al., 2006; and references therein�. Chapman etl. �2006� state, “Abnormally high reservoir attenuation is the ob-erved ground truth.” A high seismic attenuation of reservoirs in therequency range between 1 and 6 Hz may be caused by wave-in-uced flow in partially saturated rocks �see Mavko et al. �1998� andeferences therein�. Following this argument, the reservoir itself actss a strong scatterer of seismic waves because of high complex im-edance in contrast to the surrounding rocks, which have small or nottenuation �Quintal et al., 2009�.

Therefore, a reservoir may become visible at the surface by typi-al scattering phenomena such as single scattered body waves ortanding waves. However, standing shear waves would not generate/H values above one, and the dominant frequency of the mi-

rotremor will be depth dependent and relatively low. These effectsannot be observed at the gas field analyzed in this paper.

esonant amplification

Oil bubbles can oscillate in pore spaces �Hilpert et al., 2000; Be-esnev, 2006; Holzner et al., 2009�. The main restoring force of the

ubbles driving these oscillations is the surface tension between theil and water. Theoretically, all systems with a wetting and a nonwet-ing fluid exhibit a typical resonance frequency. Therefore, this reso-ant amplification effect also can be present for reservoirs with par-ial gas saturation. The resonance frequencies can be in the 1–6-Hzrequency band �Holzner et al., 2005; Holzner et al., 2009�.

Seismic background waves reaching the reservoir can induce aesonant amplification of those frequencies. Frehner et al. �2009�

how that those oscillations at the pore scale can be visible in theeismic spectra, measured at the surface above a reservoir. This hashree important consequences:

� These systems will emit energy after excitation �i.e., there is noperfect time correlation with the triggering source�. This is con-sistent with considerations using an active seismic vibratorsource �Kouznetsov et al., 2005; Turuntaev et al., 2006� and ob-servations indicating that hydrocarbon-related spectral anoma-lies can be stimulated by an earthquake �Nguyen et al., 2008�.

� Those systems will act as secondary sources; as such, it shouldbe possible to locate them. A successful localization approachhas been performed by Steiner et al. �2008�.

� Preferred direction of the triggering waves will be inherited inthe radiation pattern of the emitted wavefield �i.e., V/H valuesabove one�.

reliminary rock-physics model

We summarize our theoretical review in a preliminary interpretiveodel about the origin of hydrocarbon reservoir-related tremors.Al-

hough it may be necessary to modify this model, it is consistent withur theoretical investigations and experimental observations �i.e.,he identified seismic attributes� of the survey discussed in this pa-er.

−0.4 −0.2 0 0.2 0.4 0.6 0.8 1−5000

−4500

−4000

−3500

−3000

−2500

−2000

−1500

−1000

−500

0

Amplitude of ground motion (normalized)

Dep

th(m

)

VerticalHorizontal

igure 13. Amplitude versus depth for a Rayleigh wave of 0.14 Hzropagating through a homogeneous half-space. P- and S-wave ve-ocities are set to 3000 and 1730 m/s, respectively. The analyticalolution is adapted from Pujol �2003�.

wa1aepberPatfiws

ccsPptia

opittqgtdcsptr

ptfrtaprfe

CahtssSigtcA

isadflc��1a

FoiIpor�s

O38 Saenger et al.

Figure 14 illustrates and summarizes the main points. Oceanaves generate low-frequency, high-amplitude Rayleigh waves

round 0.14 Hz, which are observable worldwide �Friedrich et al.,998�. The strength of those waves varies in time; therefore, theylso contain energy around 3 Hz. As discussed, they oscillate at res-rvoir depth mainly in the vertical direction. So we also expect thisreferred direction for a resonant amplification effect of hydrocar-ons in the pore space. Whether and the degree to which nonlinearffects are important in this process is part of ongoing research. Theesulting radiation pattern of this secondary source will emit mainly-waves in vertical and S-waves in horizontal directions. Addition-lly, any kind of body waves hitting the reservoir also contribute tohe excitation of resonance effects. This is consistent with the identi-ed microtremor attributes above the gas reservoir of this survey andith the observation that the reservoir itself may act as a secondary

ource �Steiner et al., 2008�.We observe a strong energy anomaly between 1 and 6 Hz in all

omponents above hydrocarbons. This is illustrated for the verticalomponent of station 70139 in Figure 3. A peak above one in thepectral V/H ratio �Figure 5� is also an expected characteristic for-waves originating from the reservoir. The seismic attributes of theolarization analysis above hydrocarbons — a constant high dip ofhe particle velocity, a relatively high rectilinearity, a strongly vary-ng azimuth, and a nonvanishing largest eigenvalue �Figure 9� —lso agree with the model in Figure 14.

CONCLUSIONS

We have described a passive low-frequency microtremor surveyver a gas field in Mexico with a complex subsurface geology. In theroven hydrocarbon-bearing area, we observed an energy anomalyn the frequency range from approximately 1–6 Hz. To characterizehis anomaly in more detail, we extract several reservoir-related at-ributes: PSD-IZ, V/H signal, polarization, and maximum peak fre-uency. These attributes exhibit anomalous values above the knownas accumulations. We use this knowledge in an unexploited area ofhe survey to estimate the hydrocarbon potential. In the center of aetected anomaly in this region, at least two wells indicating hydro-arbons were drilled after the data were recorded and analyzed. Thisupports the theory that low-frequency anomalies are related to theresence of hydrocarbon reservoirs and can be used as complemen-ary information to structural imaging methods to reduce drillingisk and assist well positioning.

background spectin the seismic

present everywhsurface waves a

Surface

Reservodepth

Ocean-wave-gen

igure 14. A preliminary model that explains therigin of hydrocarbon-indicating tremors and thats consistent with the spectral attributes �i.e., PSD-Z and V/H signal� described in this paper. One im-ortant observation is that the vertical polarizationf the ocean-wave-generated Rayleigh waves ateservoir depth is also present in the low-frequencyLF� hydrocarbon reservoir-related microtremorignal.

Additionally, we propose a preliminary rock-physics model to ex-lain the origin of hydrocarbon reservoir-related tremors. Poroelas-ic effects caused by wave-induced fluid flow and oscillations of dif-erent fluid phases are considerable effects in the low-frequencyange, which can modify the omnipresent seismic background spec-rum. Both can contribute independently to the specific signal char-cteristic, and both are based on the assumption that the reservoir is aartially saturated multiphase system. The surrounding rocks of theeservoirs are only saturated with one single fluid where those ef-ects are not present. Our observed microtremor attributes above res-rvoirs are consistent with the preliminary model.

ACKNOWLEDGMENTS

We thank assistant editor J. Carcione, associate editor D. Gao, M.hapman, four anonymous reviewers, B.Artman, P. West, M. Kelly,nd B. Hardley for very useful comments and suggestions whichelped to improve the paper. We also thank N. Riahi, who enhancedhe quality of figures. E. H. Saenger thanks the DFG �Deutsche For-chungsgemeinschaft� for supporting him through a Heisenbergcholarship �SA 996/1-1�. M.-A. Lambert, T. T. Nguyen, and S. M.chmalholz were supported by the CTI �the Swiss Confederations

nnovation promotion agency� and ETH Zurich. The authors arerateful to Petroleos Mexicanos �PEMEX� for the permission to useheir dataset and their collaboration. In particular we thank J. A. Es-alera Alcocer, A. Oviedo Pérez, J. F. Gonzales Pineda, H. Córdovaguayo,A. Salas Zapata, and E. Castro Medellin.

APPENDIX A

CALCULATING V/H SPECTRA

Calculating the V/H spectra consists of several steps, includ-ng time-signal windowing, Fourier transformation, and spectramoothing. Figure A-1 summarizes the data-processing workflownd illustrates the variety of commonly used combinations of the in-ividual processing algorithms. The gray boxes indicate the work-ow used in this work. The numbers correspond to the specific publi-ations and contain more information. �1�: Al Yuncha et al., 2004;2�: Almendros et al., 2004; �3�: Bard, 1999; �4�: Bour et al., 1998;5�: Frischknecht et al., 2005; �6�: Ibs-von Seht and Wohlenberg,999; �7�: Maresca et al., 2003; �8�: Moya et al., 2000; �9�: Parolai etl., 2004; �10�: SESAME, 2004; �11�: Teves-Costa et al., 1996.

S−waves

P−waves

saturated rocksin hydrocarbon

origin of thehydrocarbon LFsignal

polarizedLF signal

ce: ocean waves peak Hydrocarbon reservoir LF signal

Vertically

Reservoir is the

bubbles within pores)(e.g., by oscillating oil

Body waves

(0.05−0.2 Hz) (1−6 Hz)

frequencies (1−6 Hz)

Stimulation ormodification of low

Rayleigh waves

s illustrate particle oscillation

Sour

rum

erere

ir

Red arrow

erated

A

A

A

A

A

B

B

B

B

FbB2 et al., 1

Spectral anomalies above reservoirs O39

REFERENCES

krawi, K., and G. Bloch, 2006, Application of passive seismic �IPDS� sur-veys inArabian Peninsula: EAGE Workshop on Passive Seismic: Explora-tion and MonitoringApplications, Dubai, UnitedArab Emirates,A28.

li, M. Y., K. A. Berteussen, J. Small, and B. Barkat, 2007, A low frequency,passive seismic experiment over a carbonate reservoir in Abu Dhabi: FirstBreak, 25, 71–73.

lmendros, J., F. Luzon, and A. Posadas, 2004, Microtremor analyses atTeide volcano �Canary Island, Spain�: Assessment of natural frequenciesof vibration using time dependent horizontal-to-vertical ratios: Pure andApplied Geophysics, 161, 1579–1596.

l Yuncha, Z., F. Luzon, A. Posadas, J. Martin, G. Alguacil, J. Almendros,and S. Sanchez, 2004, The use of ambient seismic noise measurements for

Rem

DC offset removal(4), (10)

W

No smoothing Low-pa(6

No averaging(2), (3), (4)

No averaging(11)

Smooth spectra

Remove transients

Time signal filter

Windowing

Fourier transform

Merge horizontalcomponents

Average spectra

Calculate spectral ratio

Average spectral ratio

Arithmetic mean(6)

Geometri(10

igure A-1. Data processing flow chart for the V/H and H/V spectraloxes denote the steps applied in our processing flow. Numbers correard, 1999; �4� Bour et al., 1998; �5� Frischknecht et al., 2005; �6� Ib000; �9� Parolai et al., 2004; �10� SESAME, 2004; �11� Teves-Costa

the estimation of surface soil effects: The Motril City case �southernSpain�: Pure andApplied Geophysics, 161, 1549–1559.

rtman, B., 2006, Imaging passive seismic data: Geophysics, 71, no. 4,SI177–SI187.

ard, P.-Y., 1999, Microtremor measurements: A tool for site effect estima-tion?: Balkema, 1251–1279.

eresnev, I. A., 2006, Theory of vibratory mobilization on nonwetting fluidsentrapped in pore constrictions: Geophysics, 71, no. 6, N47–N56.

erger, J., P. Davis, and G. Ekström, 2004, Ambient earth noise: A survey ofthe global seismographic network: Journal of Geophysical Research, 109,B11307.

erteussen, K. A., M. Y. A. Ali, J. S. Small, and B. Barkat, 2008, A low fre-quency, passive seismic experiment over a carbonate reservoir in Abu

Arithmetic mean(9)

ansients), (7)

Don’t remove transients(2), (8)

Detrend signal(5), (9)

Band-pass filtering(11)

ing(11)

No windowing(8)

FouriertransformFouriertransform

Moving window(fixed bandwidth)

(2), (5)

Moving window(variable bandwidth)

(3), (4)

Arithmetic mean Geometric mean

Build spectral ratio( H/V or V/H )

Build spectral ratio( H/V or V/H )

Geometric mean(3)

“Pythagoras”(1), (11)

Complex Fourier transform(3), (7)

rms average(4), (8)

ethod. Different authors use slightly different approaches. The grayreferences. �1� AlYuncha et al., 2004; �2� Almendros et al., 2004; �3�eht and Wohlenberg, 1999; �7� Maresca et al., 2003; �8� Moya et al.,996.

ove tr(3), (6

indow(7), (9),

ss filter)

c mean)

ratio mlate tos-von S

B

B

B

C

C

D

F

F

F

F

F

G

G

H

H

H

H

H

I

J

K

K

L

L

L

M

M

M

N

P

P

P

Q

R

R

S

S

S

S

S

S

T

T

v

Z

O40 Saenger et al.

Dhabi: Wavefront and particle motion study: 70th Annual InternationalMeeting, EAGE, ExtendedAbstracts, B046.

irialtsev, E. V., I. N. Plotnikova, I. R. Khabibulin, and N. Y. Shabalin, 2006,The analysis of microseisms spectrum at prospecting of oil reservoir onRepublic Tatarstan: 68th International Conference and Exhibition, EAGE,ExtendedAbstracts.

onnefoy-Claudet, S., F. Cotton, and P.-Y. Bard, 2006, The nature of noisewavefield and its applications for site effects studies: A literature review:Earth-Science Reviews, 79, 205–227.

our, M., D. Fouissac, P. Dominique, and C. Martin, 1998, On the use of mi-crotremor recordings in seismic microzonation: Soil Dynamics and Earth-quake Engineering, 17, 465–474.

astagna, J. P., S. Sun, and R. W. Siegfried, 2003, Instantaneous spectralanalysis: Detection of low-frequency shadows associated with hydrocar-bons: The Leading Edge, 22, 120–127.

hapman, M., E. Liu, and X.-Y. Li, 2006, The influence of fluid-sensitive dis-persion and attenuation on AVO analysis: Geophysical Journal Interna-tional, 167, 89–105.

angel, S., M. E. Schaepman, E. P. Stoll, R. Carniel, O. Barzandji, E.-D.Rode, and J. M. Singer, 2003, Phenomenology of tremor-like signals ob-served over hydrocarbon reservoirs: Journal of Volcanology and Geother-mal Research, 128, 135–158.

äh, D., F. Kind, and D. Giardini, 2001, Atheoretical investigation of averageH/V ratios: Geophysical Journal International, 145, 535–549.

äh, D., G. Stamm, and H.-B. Havenith, 2008, Analysis of three-componentambient vibration array measurements: Geophysical Journal Internation-al, 172, 199–213.

rehner, M., S. M. Schmalholz, and Y. Y. Podladchikov, 2009, Spectral mod-ification of seismic waves propagating through solids exhibiting a reso-nance frequency:A1D coupled wave propagation-oscillation model: Geo-physical Journal International, 176, 589–600.

riedrich, A., F. Krüger, and K. Klinge, 1998, Ocean-generated microseis-mic noise located with the Gräfenberg array: Journal of Seismology, 2,47–64.

rischknecht, C., P. Rosset, and J.-J. Wagner, 2005, Towards seismic micro-zonation 2-D modeling and ambient seismic noise measurements: Thecase of an embanked, deep alpine valley: Earthquake Spectra, 21,635–651.

oloshubin, G., C. Van Schuyver, V. Korneev, D. Silin, and V. Vingalov,2006, Reservoir imaging using low frequencies of seismic reflections: TheLeading Edge, 25, 527–531.

raf, R., S. M. Schmalholz, Y. Podladchikov, and E. H. Saenger, 2007, Pas-sive low frequency spectral analysis: Exploring a new field in geophysics:World Oil, 228, 47–52.

anssen, P., and S. Bussat, 2008, Pitfalls in the analysis of low frequency pas-sive seismic data: First Break, 26, 111–119.

estholm, S., M. Moran, S. Ketcham, T. Anderson, M. Dillen, and G. Mc-Mechan, 2006, Effects of free-surface topography on moving-seismic-source modeling: Geophysics, 71, no. 6, T159–T166.

ilpert, M., G. H. Jirka, and E. J. Plate, 2000, Capillarity-induced resonanceof oil blobs in capillary tubes and porous media: Geophysics, 65, 874–883.

olzner, R., P. Eschle, S. Dangel, M. Frehner, C. Narayanan, and D. Lakehal,2009, Hydrocarbon microtremors interpreted as nonlinear oscillationsdriven by oceanic background waves: Communications in Nonlinear Sci-ence and Numerical Simulation, 14, 160–173.

olzner, R., P. Eschle, H. Zürcher, M. Lambert, R. Graf, S. Dangel, and P. F.Meier, 2005, Applying microtremor analysis to identify hydrocarbon res-ervoirs: First Break, 23, 41–46.

bs-von Seht, M., and J. Wohlenberg, 1999, Microtremor measurements usedto map thickness of soft sediments: Bulletin of the Seismological SocietyofAmerica, 89, 250–259.

urkevics, A., 1988, Polarization analysis of three component array data:Bulletin of the Seismological Society ofAmerica, 78, 1725–1734.

onno, K., and T. Ohmachi, 1998, Ground-motion characteristics estima-ted from spectral ratio between horizontal and vertical components ofmicrotremor: Bulletin of the Seismological Society of America, 88,228–241.

ouznetsov, O. L., I. A. Chirkin, V. V. Dryagin, S. L. Aroutunov, and B. Y.Meltchouk, 2005, Induced seismoacoustic emission: Basis for new tech-nologies of fluid identification: 67th Annual International Meeting,EAGE, ExtendedAbstracts, G048.

ambert, M., S. M. Schmalholz, Y. Y. Podladchikov, and E. H. Saenger,

2007, Low frequency anomalies in spectral ratios of single station mi-crotremor measurements: Observations across an oil and gas field in Aus-tria: 77thAnnual International Meeting, SEG, ExpandedAbstracts, 1352–1356.

ambert, M., B. Steiner, S. M. Schmalholz, R. Holzner, and E. H. Saenger,2006, Soft soil amplification of ambient seismic noise — Field measure-ments and numerical modeling of H/V ratios: EAGE Workshop on PassiveSeismic, ExtendedAbstracts,A10.

ambert, M.-A., S. M. Schmalholz, E. H. Saenger, and B. Steiner, 2008,Low-frequency microtremor anomalies at an oil and gas field in Voits-dorf, Austria: Geophysical Prospecting, doi:10.1111/j.13652478.2008.00734.x.aresca, R., M. Castellano, R. D. Matteis, G. Saccorotti, and P. Vaccariello,2003, Local site effects in the town of Benevento �Italy� from noise mea-surements: Pure andApplied Geophysics, 160, 1745–1764.avko, G., T. Mukerji, and J. Dvorkin, 1998, The rock physics handbook:Cambridge University Press.oya, A., V. Schmidt, C. Segura, I. Boschini, and K. Atakan, 2000, Empiri-cal evaluation of site effects in the metropolitan area of San Jose, CostaRica: Soil Dynamics and Earthquake Engineering, 20, 177–185.

guyen, T. T., E. H. Saenger, S. M. Schmalholz, and B. Artman, 2008, Earth-quake triggered modifications of microtremor signals above and nearby ahydrocarbon reservoir in Voitsdorf, Austria: 70th Annual InternationalMeeting, EAGE, ExtendedAbstracts, P025.

arolai, S., S. M. Richwalski, C. Milkereit, and P. Bormann, 2004, Assess-ment of the stability of H/V spectral ratios from ambient noise and compar-ison with earthquake data in the Cologne area �Germany�: Tectonophys-ics, 390, 57–73.

eterson, J. 1993, Observations and modeling of seismic background noise:U.S. Geological Survey Open-File Report 93-322.

ujol, J., 2003, Elastic wave propagation and generation in seismology:Cambridge University Press.

uintal, B., S. M. Schmalholz, and Y. Y. Podladchikov, 2009, Low-frequen-cy reflections from a thin layer with high attenuation caused by interlayerflow: Geophysics, 74, no. 1, N15–N23.

ached, G. R., 2006, Surface passive seismic in Kuwait: EAGE Workshopon Passive Seismic: Exploration and MonitoringApplications,A27.

entsch, S., S. Buske, S. Lüth, and S. A. Shapiro, 2007, Fast location of seis-micity: A migration-type approach with application to hydraulic-fractur-ing data: Geophysics, 72, no. 1, S33–S40.

affari, N., and L. J. Bond, 1987, Body to Rayleigh wave mode-conversion atsteps and slots: Journal of Nondestructive Evaluation, 6, 1–22.

chick, R., 1988, Volcanic tremor-source mechanisms and correlation witheruptive activity: Natural Hazard, 1, 125–144.

ESAME, 2004, Guidelines for the implementation of the H/V spectral ratiotechnique on ambient vibration: Measurements, processing and interpre-tation: SESAME European research project, http://sesame-fp5.obs.ujf-grenoble.fr/SES_Reports.htm, accessed 14 January 2009.

hapiro, N. M., M. Campillo, L. Stehly, and M. Ritzwoller, 2005, High-reso-lution surface-wave tomography from ambient seismic noise: Science,307, 1615–1618.

teiner, B., E. H. Saenger, and S. M. Schmalholz, 2008, T ime reverse model-ing of low-frequency microtremors: A potential method for hydrocarbonreservoir localization: Geophysical Research Letters, 35, L03307.

untsov, A. E., S. L. Aroutunov, A. M. Mekhnin, and B. Y. Meltchouk, 2006,Passive infra-frequency microseismic technology: Experience and prob-lems of practical use: EAGE Workshop on Passive Seismic: Explorationand MonitoringApplications,A25.

eves-Costa, P., L. Matias, and P. Y. Bard, 1996, Seismic behaviour estima-tion of thin alluvium layers using microtremor recordings: Soil Dynamicsand Earthquake Engineering, 15, 201–209.

uruntaev, S. B., V. N. Burchik, and D. S. Turuntaev, 2006, Microseismicbackground study for gas field exploration: 68th International Conferenceand Exhibition, EAGE, ExtendedAbstracts, P245.

an Mastrigt P., and A. Al-Dulaijan, 2008, Seismic spectroscopy using am-plified 3-C geophones: 70th International Conference and Exhibition,EAGE, ExtendedAbstracts, B047.

hukov, A. P., K. I. Loginov, M. B. Shneerson, V. E. Shulakova, R. G. Khari-sov, and V. A. Ekimenko, 2007, Nonlinear properties of vibrator-generat-ed wavefields and their application to hydrocarbon detection: The LeadingEdge, 26, 1395–1402.