tsunamigenerationduetosupershear …...7.5 palu earthquake. the epicentre is indicated by a black...

Tsunami generation due to supershearearthquakes: a case study

Faisal AmlaniHarsha S. BhatWim J.F. SimonsAlexandre ShubnelChristophe VignyAres J. RosakisJoni EfendiAhmed ElbannaHasanuddin Z. Abidin

Univ of Southern California, USAÉcole Normale Supérieure, FranceDelft Univ of Technology, NetherlandsÉcole Normale Supérieure, FranceÉcole Normale Supérieure, FranceCalifornia Institute of Technology, USABIG (Geospatial Information Agency), IndonesiaUniv of Illinois Urbana-Champaign, USABIG (Geospatial Information Agency), Indonesia

c© Authors. All rights reserved.

The 2018 Sulawesi earthquake and Palu bay tsunami

Northern S

egment

Saluki Segment

PALP (73.5,5.7) kmGPS Station

(0,0)

1.5°S119.5°E

0°120.25°E

Mw 7.5 PALU

Palu Bay

PANTOLOANTidal Gauge

Palu Segment

PANTALOAN Tidal GaugeWater Height

PALP GPS Station

fault parallelParticle Velocity+1 m/s

-1 m/s0s 10s 20s 30s

fault normal

F Amlani, HS Bhat, WJF Simons, A Schubnel, C Vigny, AJ Rosakis, J Efendi, A Elbanna, HZ Abidin c© Authors. All rights reserved. 1/14

Far field evidence of a supershear rupture at Palu

• Supershear speed inferred from short-duration of moment release, smoothness of faultgeometry and lack of significant aftershocks

Geodetic data Rupture speed between 4.3 and 5.2km/sARTICLESNATURE GEOSCIENCE

of the northern lobe is towards the northwest, which suggests that the rupture turns towards the northwest at latitude 0.1° S to con-tinue offshore and ultimately connect up with the Minahassa trench further north. East of the fault, south of Palu city, the phase gradient indicates a range increase compatible with a left-lateral strike-slip and possible subsidence. North of 0.7° S, the phase gradient shows a range decrease, which suggests that uplift occurred to the east of the fault, leading for local transpression in the area of the Sulawesi neck.

Rupture characteristics and evidence for supershearSubpixel optical image correlation21 from Landsat-8, Sentinel-2 and WorldView satellite imagery (Figs. 1 and 2 and Supplementary Fig. 1) allows us to map the trace of the rupture in great detail and to quan-tify the horizontal component of the coseismic displacement at the surface. Critically, the optical correlation data provide valuable constraints on the near-field displacement pattern (in the region where the InSAR decorrelates), which thus reveals details of how the fault broke the surface and gives resolution to the shallow part of our fault slip model. South of Palu city (0.9° S), the north–south

(N–S) displacement field (Fig. 2) shows that the trace of the 2018 rupture reproduces, to the first order, the shape of the Palu-Koro fault, as documented from tectonic geomorphology and geological investigation5,8,9. South of the Palu coastline (that is, south of 0.9° S), the rupture is linear and strikes ~N172°. Near 1.187° S (~33 km south of the coastline), the rupture bends sharply to the southeast for ~9 km, where it forms a major releasing bend before recover-ing its initial azimuth and continuing for another ~20 km. North of Palu city, the rupture disappears offshore within the Palu bay, and reappears 21 km further north within the Sulawesi neck, where a much smoother displacement gradient can be followed northwards for 60 km (probably due to a buried slip that does not come up to the surface), until it reaches the Balaesang Peninsula releasing bend at latitude 0.1° S. The azimuth of this northern segment (~182°) is rotated ~10° clockwise from that of the Palu segment to the south (~172°); the change occurs somewhere in the bay of Palu.

The high-resolution displacement field (40 m resolution) allows a structural examination of the rupture south of Palu, which is made up of six segments (A–F, Figs. 2 and 3) (detailed descrip-tion in Supplementary Information). Some segments appear to be

b

119° E 119.5° E 120° E 120.5° E

Longitude Longitude

121° E 121.5° E 122° E2.5° S

a

2.0° S

1.5° S

1.0° S

0.5° S

0°

0.5° N

North Sula Block

Makassar BlockMatano Fault

1.0° N

3.5° N

3.0° N

2.5° N

2.0° N

1.5° N

Latitu

de

Latitu

de

LOS

Fligh

t

–750 –500 –250 0 250 500

ALOS-2 InSAR phase (mm)

–5 –4 –3 –2 –1 0 1 2 3 4 5

Seismicity (days to Mw 7.5)

119.5° E 120° E

1.5° S c d

1.0° S

0.5° S

0°

–3 –2 –1 0 1 2 3 4 5

N–S offsets (m)

–0.50 –0.25 0.00 0.25 0.50

Seismicity (days to mainshock)

–2–1012345678Offset (m)

1.5° S

1.4° S

1.3° S

1.2° S

1.1° S

1.0° S

0.9° S

0.8° S

0.7° S

0.6° S

0.5° S

0.4° S

0.3° S

0.2° S

0.1° S

0°

Latitude

–45045Azimuth (°)

Minahassa Trench

Palu-Koro Fault

Fig. 1 | Setting and measured surface displacements associated with Mw 7.5 Palu earthquake. The epicentre is indicated by a black star. The focal mechanism is from the USGS. The coloured dots represent one month of foreshock (blue) and aftershock (yellow and red) seismicity (different temporal colour scales are provided on each panel to show the detailed evolution of the seismic sequence). The topography is from the Shuttle Radar Topography Mission. a, The unwrapped ALOS-2 interferogram (21 August 2018 to 2 October 2018) showing the coseismic displacement in the line of sight (© JAXA). A relative increase between two pixels means a relative displacement towards the satellite. The surface trace of the ruptured fault is shown as a red line. Grey circles represent the background seismicity. The geological traces of the main faults are shown in black. b, Map of the horizontal surface displacement computed by the correlation of Sentinel-2 images. The arrows show the horizontal displacement, whereas the colours correspond to the N–S component of the displacements. c,d, Distribution along the rupture of the surface coseismic offsets extracted at 0.01° of the fault (red, north–south; blue, east–west) (c) and of the azimuth (green, azimuth of the slip vector; black, azimuth of the fault trace) (d). LOS, line-of-sight.

NATURE GEOSCIENCE | VOL 12 | MARCH 2019 | 192–199 | www.nature.com/naturegeoscience 193

ARTICLES NATURE GEOSCIENCE

exceptionally sharp, expressed as a narrow surface rupture, with little coseismic deformation taken up by distributed shear off the main fault trace. Significantly, segments A and B do not follow the long-term fault morphology, lying 2–3 km from the bedrock range front within the basin (Fig. 2d). The coseismic displacement (4–7 m) is remarkably smooth and almost pure left-lateral strike-slip, whereas the fault-normal component is accommodated on sec-ondary structures located off the main fault (Figs. 2 and 3). This is precisely what was observed along the supershear segments of the 1999 Izmit22–24 and 2001 Kunlun25 earthquakes. In both cases, the strike-slip rupture trace was located a few kilometres from the obvious bedrock fault structure, whereas the fault perpendicular component was accommodated on secondary structures through a slip partitioning mechanism26.

Fault segments that have hosted supershear ruptures have been shown to share specific structural characteristics (straight and

without major structural complexity) and very smooth coseismic slip in pure mode II, which reaches the surface (even though some faults may be dipping or accommodating oblique movement over the long term27,28). As pure strike-slip does not produce topography, ancient traces of such ruptures are challenging to identify, in par-ticular in a humid climate setting. However, subtle evidence for lat-erally confined river channels in Palu basin were found9, suggesting that a fault branch crossed straight through the sediments, which in turn led the authors to conclude that the Palu-Koro fault may be capable of generating supershear ruptures.

Our observations of the Mw7.5 surface rupture south of Palu city are remarkably consistent, and together suggest that a rup-ture at a supershear velocity on some segments is highly probable. Segments A, B and E are all characterized by minor normal slip, a very straight azimuth, no differential slope, nearly uniform slip (especially when considering the primary and secondary slip) and

Palu

e

Mudslide

d

A

0 5

119° 50′ ELongitude

5

3

1

0

–2

–5

–3

0

–90

–30

30

90

150

210

27 0

500

1,000

1,500

2,000

2,500

–2

–1

0

1

2

3

0° 50′ S

1° 10′ SLatitu

de

1° 20′ S

1° S

Distance (km)10

B

C

D

E

F

f

f

a b

c

Mudslide

Fig. 2 | Detailed features of the Palu rupture. a, Slip vectors determined from the near-field (primary rupture, blue arrows) and medium field (primary!+ !secondary, red arrows) ruptures. b,c, N–S (b) and E–W (c) displacement (m) fields from the correlation of pre- and post-earthquake Sentinel-2 images. The fault rupture is shown in black (bold is the primary rupture and dashed is the secondary one). d, Topography with the different segment lengths (m) labelled as in Fig. 3. e, Azimuth of the ground displacement (°). f, High-resolution E–W displacement map for the northern Palu section of the rupture, from correlation of WorldView satellite images (©2018 DigitalGlobe, a Maxar company; pre-image, 20 February 2018; post-image, 2 October 2018). The white decorrelated patch corresponds to a large mudslide, visible in the post-event WorldView imagery.

NATURE GEOSCIENCE | VOL 12 | MARCH 2019 | 192–199 | www.nature.com/naturegeoscience194

A Socquet, J Hollingsworth, E Pathier, M Bouchon, Nature Geoscience (2019)


Far field evidence of a supershear rupture at Palu

• Supershear speed inferred from back-projection of teleseismic data and from far-fieldRayleigh mach waves

Teleseismic data Rupture speed of 4.1km/s

ARTICLES NATURE GEOSCIENCE

0°

a b

0.4° S

0.8° S

1.2° S

1.6° S

0°

0.4° S

0.8° S

1.2° S

1.6° S

119.4° E 119.7° E 120° E 120.3° E 120.6° E

–3,000 0Elevation (m)

3,000

120° E119.6° E 120.4° E

Epicentre

BP location

Mainshock BPs

Aftershock relocation

Fig. 2 | Calibration of back-projection based on aftershock data. a,b, Back-projection (BP)-inferred (green circles) and relocated (red stars) locations of nine M 5.0+ aftershocks that span the rupture region, and back-projection radiators (grey circles) before (a) and after (b) the slowness calibration. The results shown are for the Australia array.

100° E 120° E 140° E 160° E

30° S

20° S

10° S

0°

JAGI

PSAA

MULG

SAUI

Epicentre (NEIC)

0.5 1 .00.9Correlation coefficient

0.80.70.6

BBJI

Fig. 3 | Evidence of a far-field Rayleigh-wave Mach cone. The coloured area within the green lines is the predicted area scanned by the Mach cone with the maximum possible Mach angle, based on the observed rupture velocity (4.1!km!s–1) and considering the uncertainty of Rayleigh wave phase velocity. The locations of the broadband stations are indicated by triangles. Their colour indicates the correlation coefficients between 15 to 25!s Rayleigh wave displacement seismograms of the Palu earthquake and its M 6.1 foreshock. Rayleigh waves recorded by the five stations with labelled names are shown in Fig. 4.

NATURE GEOSCIENCE | VOL 12 | MARCH 2019 | 200–205 | www.nature.com/naturegeoscience202 H Bao, JP Ampuero, L Meng, EJ Fielding, C Liang, CWD Milliner, T Feng, H Huang, Nature Geoscience(2019)


Searching for unmistakable signatures in the near field

• Trailing Rayleigh wave • Fault parallel velocity > fault normal • Dilatationalyfield

2002 Denali accelerometer data (PS10, 250Hz) 2002 Denali lab experiments

Fault normal

Fault parallel

Vertical

WL Ellsworth, M Celebi, JR Evans, EG Jensen, R Kayen,MC Metz, DJ Nyman, JW Roddick, P Spudich, CDStephens, Earthquake Spectra (2004)

M Mello, HS Bhat, AJ Rosakis, H Kanamori, Earth andPlanetary Science Letters (2014)E Dunham & R Archuleta, Bulletin of the SiesmologicalSociety of America (2004)


First near field evidence of a supershear rupture at Palu

• Trailing Rayleigh wave • Fault parallel velocity > fault normal • Dilatationalyfield

2018 Palu GPS fault parallel data (PALP, 1Hz) 2018 Palu GPS fault normal data (PALP, 1Hz)

PALP GPS StationFault Parallel Velocity

Particle Velocity+1 m/s

-1 m/s0s 10s 20s 30s

1 m/s

PALP GPS StationFault Normal Velocity


-1 m/s0s 10s 20s 30s

-0.7 m/s


Further verification via dynamic rupture models

Supershear rupture (1.6cs)

PALP GPS StationFault Parallel VelocitySupershear Model

Fault Parallel Velocity


-1 m/s0s 10s 20s 30s

1 m/s


Supershear ModelFault Normal Velocity


-1 m/s0s 10s 20s 30s

-0.7 m/s

Subshear rupture (0.8cs)

PALP GPS StationFault Parallel Velocity

Subshear ModelFault Parallel Velocity


-1 m/s0s 10s 20s 30s

1 m/s


Subshear ModelFault Normal Velocity


-1 m/s0s 10s 20s 30s

-0.7 m/s


A few words on supershear theory

shear Mach cone from one pointalong rupture front

shearRayleigh wave Mach frontson free surface

shear Mach cone from source at locked-slipping interface

freesurface

slipping

locked

part ofrupture front moving

at speed vr > cs fault

Mach fronts

EM Dunham & HS Bhat, Journal of GeophysicalResearch (2008)

Vertical velocity

Vertical displacement


Classical displacement-based tsunami models

Vertical bathymetry displacement No Mach signatures

Tsunami

Subduction Zone Fault

CLASSICAL tsunami source

STATIC VERTICALDISPLACEMENT

from Supershear Earthquake


Classical displacement-based models applied to Palu

1D bathymetry (static displacement)

2D bathymetry (static displacement)

A Jamelot, A Gailler, Ph Heinrich, A Vallage, JChampenois, Pure and Applied Geophysics (2019)

2D bathymetry (time-dependent displacement)

T Ulrich, S Vater, EH Madden, J Behrens, Y VanDinther, I Van Zelst, EJ Fielding, C Liang, A-AGabriel, Pure and Applied Geophysics (2019)


Incorporating dynamic ground motion

x

z y

Free surface

Topography (bathymetry)

H(x, y, t) = η(x, y, t) + h(x, y, t)

z = η(x, y, t)

z = − h(x, y, t)

u := velocity

H := total height

η := height from free surface

h := bathymetry height (rest + source)

g := gravitational constant (9.8m/s2)

∂H∂t

+∇·(Hu) = 0

∂(Hu)∂t

+∇·(Hu⊗u+ 1

2gH2)

= gH∇h⇐=

H(x, y, t) = η(x, y, t)+ h(x, y, t)

∂H(x, y, t)∂t

=∂η(x, y, t)

∂t+ ∂h(x,y,t)

∂t


Sourcing with supershear dynamics

DYNAMIC tsunami source

Shear & Rayleighshock fronts

DYNAMIC VERTICALVELOCITY

from Supershear EarthquakeCLASSICAL tsunami source

STATIC VERTICALDISPLACEMENT



Modeling the Palu bay configuration in 1D

Northern S

egment

Palu Segment

PALP GPS Station

Palu Bay

PANTOLOANTidal Gauge

DYNAMIC tsunami source

Shear & Rayleighshock fronts

DYNAMIC VERTICALVELOCITY


Non-Linear Shallow Water Wave Model

, Palu Bay

Robust numerical resolution via explicit Fourier continuation:

F Amlani & NM Pahlevan, Journal of Computational Physics (2020)F Amlani & OP Bruno, Journal of Computational Physics (2016)


Capturing the first motions and arrivals


• The Palu earthquake went supershear (Socquet et al, Bao et al, this work)

• The supershear shock fronts caused the tsunami motion (this work)

• Palu’s shallow bay amplified the tsunami displacement (Ulrich et al, Jamelot et al, etc.)

Preprint available on arxiv:1910.14547


tsunamigenerationduetosupershear …...7.5 palu earthquake. the epicentre is indicated by a black...

Documents