Joint inversion of InSAR and broadband teleseismic waveform data with ABIC: application to the 1997 Manyi, Tibet earthquake

Similar documents
Buried Strike Slip Faults: The 1994 and 2004 Al Hoceima, Morocco Earthquakes.

Hybrid Back-Projection: Theory and Future Direction of Seismic Array Analysis

SOURCE PROCESS OF THE 2003 PUERTO PLATA EARTHQUAKE USING TELESEISMIC DATA AND STRONG GROUND MOTION SIMULATION

Two Contrasting InSAR Studies of Recent Earthquakes in Tibet

Journal of Geophysical Research Letters Supporting Information for

27th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies

Inversion of Earthquake Rupture Process:Theory and Applications

Post-seismic motion following the 1997 Manyi (Tibet) earthquake: InSAR observations and modelling

RELOCATION OF THE MACHAZE AND LACERDA EARTHQUAKES IN MOZAMBIQUE AND THE RUPTURE PROCESS OF THE 2006 Mw7.0 MACHAZE EARTHQUAKE

Moment tensor inversion of near source seismograms

On the Limitation of Receiver Functions Method: Beyond Conventional Assumptions & Advanced Inversion Techniques

27th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies

Widespread Ground Motion Distribution Caused by Rupture Directivity during the 2015 Gorkha, Nepal Earthquake

Empirical Green s Function Analysis of the Wells, Nevada, Earthquake Source

Source rupture process of the 2003 Tokachi-oki earthquake determined by joint inversion of teleseismic body wave and strong ground motion data

SUPPLEMENTARY INFORMATION

Open Access FULL PAPER. Yukitoshi Fukahata * and Manabu Hashimoto

Displacement field and slip distribution of the 2005 Kashmir earthquake from SAR imagery

Seismic Source Mechanism

Complicated repeating earthquakes on the convergent plate boundary: Rupture processes of the 1978 and 2005 Miyagi-ken Oki earthquakes

Dear editors and reviewer(s), thank for your comments and suggestions. Replies as follows:

Geodesy (InSAR, GPS, Gravity) and Big Earthquakes

Received 17 June 2011; revised 20 March 2013; accepted 7 June 2013; published 27 August 2013.

Data Repository: Seismic and Geodetic Evidence For Extensive, Long-Lived Fault Damage Zones

Strong-Motion and Teleseismic Waveform Inversions for the Source Process of the 2003 Bam, Iran, Earthquake

SOURCE MODELING OF RECENT LARGE INLAND CRUSTAL EARTHQUAKES IN JAPAN AND SOURCE CHARACTERIZATION FOR STRONG MOTION PREDICTION

BROADBAND STRONG MOTION SIMULATION OF THE 2004 NIIGATA- KEN CHUETSU EARTHQUAKE: SOURCE AND SITE EFFECTS

14 S. 11/12/96 Mw S. 6/23/01 Mw S 20 S 22 S. Peru. 7/30/95 Mw S. Chile. Argentina. 26 S 10 cm 76 W 74 W 72 W 70 W 68 W

Geodetic data inversion using ABIC to estimate slip history during one earthquake cycle with viscoelastic slip-response functions

Joint inversion of strong motion, teleseismic, geodetic, and tsunami datasets for the rupture process of the 2011 Tohoku earthquake

Dynamic source inversion for physical parameters controlling the 2017 Lesvos earthquake

Journal of Geophysical Research - Solid Earth

Slip distributions of the 1944 Tonankai and 1946 Nankai earthquakes including the horizontal movement effect on tsunami generation

Basics of the modelling of the ground deformations produced by an earthquake. EO Summer School 2014 Frascati August 13 Pierre Briole

Coseismic slip distribution of the 2005 off Miyagi earthquake (M7.2) estimated by inversion of teleseismic and regional seismograms

JCR (2 ), JGR- (1 ) (4 ) 11, EPSL GRL BSSA

EARTHQUAKE SOURCE PARAMETERS OF MODERATELY EARTHQUAKE IN THE SOUTH EASTERN IRAN BASED ON TELESEISMIC AND REGIONAL DISTANCES

DETERMINATION OF SLIP DISTRIBUTION OF THE 28 MARCH 2005 NIAS EARTHQUAKE USING JOINT INVERSION OF TSUNAMI WAVEFORM AND GPS DATA

Fault identification for buried strike-slip earthquakes using InSAR: The 1994 and 2004 Al Hoceima, Morocco earthquakes

Source process of the 2011 off the Pacific coast of Tohoku Earthquake with the combination of teleseismic and strong motion data

The April 6 th 2009, L Aquila (Italy) earthquake: DInSAR analysis and seismic source model inversion

Measurement of differential rupture durations as constraints on the source finiteness of deep-focus earthquakes

Surface displacements and source parameters of the 2003 Bam (Iran) earthquake from Envisat advanced synthetic aperture radar imagery

7 Ground Motion Models

Method to Determine Appropriate Source Models of Large Earthquakes Including Tsunami Earthquakes for Tsunami Early Warning in Central America

3.3. Waveform Cross-Correlation, Earthquake Locations and HYPODD

Location and mechanism of the Little Skull Mountain earthquake as constrained by satellite radar interferometry and seismic waveform modeling

Source rupture process inversion of the 2013 Lushan earthquake, China

Depth (Km) + u ( ξ,t) u = v pl. η= Pa s. Distance from Nankai Trough (Km) u(ξ,τ) dξdτ. w(x,t) = G L (x,t τ;ξ,0) t + u(ξ,t) u(ξ,t) = v pl

JOURNAL OF GEOPHYSICAL RESEARCH, VOL. 116, B01406, doi: /2010jb007849, 2011

Distribution of slip from 11 M w > 6 earthquakes in the northern Chile subduction zone

Introduction to Displacement Modeling

FULL MOMENT TENSOR ANALYSIS USING FIRST MOTION DATA AT THE GEYSERS GEOTHERMAL FIELD

28th Seismic Research Review: Ground-Based Nuclear Explosion Monitoring Technologies

The IISEE earthquake catalog, Catalog of Damaging Earthquakes in the World, IISEE-NET,, and BRI strong motion observation

Vertical to Horizontal (V/H) Ratios for Large Megathrust Subduction Zone Earthquakes

CHARACTERIZING EARTHQUAKE SLIP MODELS FOR THE PREDICTION OF STRONG GROUND MOTION

Spatio-temporal variation in slip rate on the plate boundary off Sanriku, northeastern Japan, estimated from small repeating earthquakes

Geophysical Journal International

Journal of Geophysical Research (Solid Earth) Supporting Information for

Short Note Source Mechanism and Rupture Directivity of the 18 May 2009 M W 4.6 Inglewood, California, Earthquake

Case Study 1: 2014 Chiang Rai Sequence

Source of the July 2006 West Java tsunami estimated from tide gauge records

The combined inversion of seismic and geodetic data for the source process of the 16 October,

Constraining earthquake source inversions with GPS data: 2. A two-step approach to combine seismic and geodetic data sets

Automatic Moment Tensor Analyses, In-Situ Stress Estimation and Temporal Stress Changes at The Geysers EGS Demonstration Project

Songlin Li 1, Xiaoling Lai 1 Zhixiang Yao 2 and Qing Yang 1. 1 Introduction

Tsunami Simulation of 2009 Dusky Sound Earthquake in New Zealand

Spatial and Temporal Distribution of Slip for the 1999 Chi-Chi, Taiwan, Earthquake

An intermediate deep earthquake rupturing on a dip-bending fault: Waveform analysis of the 2003 Miyagi-ken Oki earthquake

The March 11, 2011, Tohoku-oki earthquake (Japan): surface displacement and source modelling

Rupture Process of the Great 2004 Sumatra-Andaman Earthquake

Effect of the Emperor seamounts on trans-oceanic propagation of the 2006 Kuril Island earthquake tsunami

Dynamic Triggering Semi-Volcanic Tremor in Japanese Volcanic Region by The 2016 Mw 7.0 Kumamoto Earthquake

Effects of Surface Geology on Seismic Motion

Building up Seismsic Models for Ground Motion Prediction of Taiwan: Problems and Challenges

Rupture Characteristics of Major and Great (M w 7.0) Megathrust Earthquakes from : 1. Source Parameter Scaling Relationships

Source studies of the ongoing ( ) sequence of recent large earthquakes in Canterbury

Simulation of Strong Ground Motions for a Shallow Crustal Earthquake in Japan Based on the Pseudo Point-Source Model

EXAMINATION ON CONSECUTIVE RUPTURING OF TWO CLOSE FAULTS BY DYNAMIC SIMULATION

to: Interseismic strain accumulation and the earthquake potential on the southern San

Measuring seismicity in the Groningen Field. Bernard Dost, Elmer Ruigrok, Jesper Spetzler, Gert-Jan van den Hazel, Jordi Domingo

Simulation of earthquake rupture process and strong ground motion

by A. Ozgun Konca, Sebastien Leprince, Jean-Philippe Avouac, and Don V. Helmberger

Earthquake patterns in the Flinders Ranges - Temporary network , preliminary results

The source process of the 2001 July 26 Skyros Island (Greece) earthquake

RAPID SOURCE PARAMETER DETERMINATION AND EARTHQUAKE SOURCE PROCESS IN INDONESIA REGION

Extending the magnitude range of seismic reservoir monitoring by Utilizing Hybrid Surface Downhole Seismic Networks

The Combined Inversion of Seismic and Geodetic Data for the Source Process of the 16 October 1999 M w 7.1 Hector Mine, California, Earthquake

Numerical Modeling for Earthquake Source Imaging: Implications for Array Design in Determining the Rupture Process

Co-seismic slip from the July 30, 1995, M w 8.1 Antofagasta, Chile, earthquake as constrained by InSAR and GPS observations

Was the February 2008 Bukavu seismic sequence associated with magma intrusion?

A non-linear geodetic data inversion using ABIC for slip distribution on a fault with an unknown dip angle

Interseismic strain accumulation across the Manyi fault (Tibet) prior to the 1997 M w 7.6 earthquake

Source model of the 2005 Miyagi-Oki, Japan, earthquake estimated from broadband strong motions

Sendai Earthquake NE Japan March 11, Some explanatory slides Bob Stern, Dave Scholl, others updated March

A systematic measurement of shear wave anisotropy in southern California Report for SCEC Award #15081 Submitted March 15, 2016

Source Characteristics of Large Outer Rise Earthquakes in the Pacific Plate

Rapid magnitude determination from peak amplitudes at local stations

2008 Monitoring Research Review: Ground-Based Nuclear Explosion Monitoring Technologies

Transcription:

Joint inversion of InSAR and broadband teleseismic waveform data with ABIC: application to the 1997 Manyi, Tibet earthquake Gareth Funning 1, Yukitoshi Fukahata 2, Yuji Yagi 3 & Barry Parsons 4 1 University of California, Berkeley, USA 2 University of Tokyo, Japan 3 University of Tsukuba, Japan 4 COMET, University of Oxford, UK

Overview Introduction Rationale for joint inversions The 1997 Manyi, Tibet earthquake Properties of the earthquake source model Inverting for source parameters with ABIC Brief description of the method InSAR-only, seismology-only and joint inversion results

Why attempt a joint inversion? InSAR has excellent spatial resolution (< 100 m) can map and model deformation in fine detail

Example: the Bam earthquake Deformation pattern consistent with slip on two parallel faults. Bam, Iran: track 385 (ascending), frame 0575 (detail) Funning et al., 2005, JGR

Why attempt a joint inversion? InSAR has excellent spatial resolution (< 100 m) can map deformation in fine detail but temporal resolution is poor ( 35 days) no information on timing.

Why attempt a joint inversion? InSAR has excellent spatial resolution (< 100 m) can map deformation in fine detail but temporal resolution is poor ( 35 days) no information on timing. Body wave seismology has good temporal resolution, and is also sensitive to location.

Deformation pattern consistent with slip on two parallel faults. Jackson et al., 2005, submitted to GJI

Deformation pattern consistent with slip on two parallel faults. Jackson et al., 2005, submitted to GJI

Why attempt a joint inversion? InSAR has excellent spatial resolution (< 100 m) can map deformation in fine detail but temporal resolution is poor ( 35 days) no information on timing or sequence of events. Body wave seismology has good temporal resolution, and is also sensitive to location. Aim is to generate a solution that is well constrained in space and time.

The 1997 Manyi, Tibet earthquake 11th November 1997, M w 7.5 largest event of that year Occurred in a remote area of N Tibet; no casualties reported, no field analysis. Ideal remote earthquake to study in this way both InSAR and seismology require no field visits.

The 1997 Manyi, Tibet earthquake 36 8th November 1997, M w 7.5 Aftershock pattern spans 3 satellite tracks 35 SRTM topography in 34 shaded relief 86 87 88 89

The 1997 Manyi, Tibet earthquake Peak LOS offset is 2.4 m ( 7 m of left-lateral offset)

Earthquake source model properties Define a fault geometry using the InSAR data.

The fault geometry Assumed fault geometry: Strike 258 Dip 90 Rake -5 Length 180 km Width 18 km 180 km 18 km (schematic)

The fault model Assumed fault geometry: Strike 258 Dip 90 Rake -5 Length 180 km Width 18 km 180 km Fault is divided into 6 6 km patches. 18 km (schematic)

Earthquake source model properties Define a fault geometry using the InSAR data. Define a source model with multiple time steps.

The source model The amount of slip is solved for at each of five time steps. t 1 t 2 Commencement of slip on each patch occurs after a rupture front has reached the patch. t 3 t 4 t 5 Each time step has a duration of 2 seconds. increasing time

Earthquake source model properties Define a fault geometry using the InSAR data. Define a source model with multiple time steps. Model must be smooth in space oscillations imply unphysical strains on the fault.

Spatial smoothing The total slip on a given patch a 1 a 4 a 7 a M a 2 a 5 a 8 a 3 a 6 a 9 (schematic)

Spatial smoothing The total slip on a given patch is related to that of patches adjacent to it, by a finitedifference Laplacian approximation. (a 2 a 5 ) (a 5 a 8 )+(a 4 a 5 ) (a 5 a 6 ) = 0 a 2 + a 4 4a 5 + a 6 + a 8 = 0 a 1 a 4 a 7 a 2 a 5 a 8 a M a 3 a 6 a 9 (schematic)

Earthquake source model properties Define a fault geometry using the InSAR data. Define a source model with multiple time steps. Model must be smooth in space oscillations imply unphysical strains on the fault Model must be smooth in time oscillations imply changes in acceleration.

Temporal smoothing The slip on a given patch, at a given time step t 2 a 95 (schematic)

Temporal smoothing The slip on a given patch, at a given time step t 1 t 2 a 5 is related to that for the same patch at the previous and subsequent time steps. a 95 a 185 t 3 increasing time (schematic)

Temporal smoothing The slip on a given patch, at a given time step is related to that for the same patch at the previous and subsequent time steps. a 5 a 95 a 185 t 1 t 2 t 3 Laplacian smoothing in 1D is used here (a 5 a 95 ) (a 95 a 185 ) = 0 a 5 2a 95 + a 185 = 0

Earthquake source model properties Define a fault geometry using the InSAR data. Define a source model with multiple time steps. Model must be smooth in space oscillations imply unphysical strains on the fault Model must be smooth in time oscillations imply changes in acceleration Model must also be consistent with the available data, of course

Inverting data using ABIC Akaike s Bayesian Information Criterion (Akaike, 1980) What level of resolution is appropriate, given the input data and fault geometry? Calculation of ABIC allows the optimal estimation of the relative weighting of different constraints on the inversion.

Inversions using ABIC Model, a, has two types of constraints: Observations: d = Ha + e o d = observed data H = data kernels a = fault slip e o = errors in observations Smoothing: Sa + e s = 0 Ta + e t = 0 S = 2D Laplacian operator e s = errors in smoothing T = 1D Laplacian operator e t = errors in smoothing

Inversions using ABIC Observation equation: Smoothing equations: d = Ha + e o Sa + e s = 0 Ta + e t = 0

Inversions using ABIC Data distribution: Prior distribution: p(d a;σ 2,γ 2 ) = (2πσ 2 ) -N/2 E -1/2 exp[-(1/2σ 2 )(d-ha) T E -1 (d-ha)] p(a;ρ s2, ρ t2 ) = (2π) -M/2 (1/ρ s2 )G s + (1/ρ t2 )G t 1/2 exp{-a T [(1/2ρ s2 )G s + (1/2ρ t2 )G t ]a}

Inversions using ABIC Data distribution: Prior distribution: p(d a;σ 2,γ 2 ) = (2πσ 2 ) -N/2 E -1/2 exp[-(1/2σ 2 )(d-ha) T E -1 (d-ha)] p(a;ρ s2,ρ t2 ) = (2π) -M/2 (1/ρ s2 )G s + (1/ρ t2 )G t 1/2 exp{-a T [(1/2ρ s2 )G s + (1/2ρ t2 )G t ]a} Combined (posterior) distribution: p(a;σ 2,α 2,β 2,γ 2 d) = cp(d a;σ 2,γ 2 ) p(a;ρ s2,ρ t2 )

Inversions using ABIC Data distribution: Prior distribution: p(d a;σ 2,γ 2 ) = (2πσ 2 ) -N/2 E -1/2 exp[-(1/2σ 2 )(d-ha) T E -1 (d-ha)] p(a;ρ s2,ρ t2 ) = (2π) -M/2 (1/ρ s2 )G s + (1/ρ t2 )G t 1/2 exp{-a T [(1/2ρ s2 )G s + (1/2ρ t2 )G t ]a} Combined (posterior) distribution: p(a;σ 2,α 2,β 2,γ 2 d) = cp(d a;σ 2,γ 2 ) p(a;ρ s2,ρ t2 ) ABIC: ABIC(α 2,β 2,γ 2 )= -2 log p(a;σ 2,α 2,β 2,γ 2 d) da

Inverting data using ABIC ABIC is a function of (in this case) up to three hyperparameters, α 2, β 2 and γ 2. Evaluate ABIC numerically for a range of values. The hyperparameter values that give a minimum are the optimum values. Recover best-fitting set of model parameters, a*, for each model, as part of this process.

Modelling the Manyi earthquake I calculate three models for the earthquake, using the ABIC technique: 1. InSAR data only 2. Seismic data only 3. Both datasets jointly In all cases I use the same simplified fault geometry

InSAR data inversion Simplest case 1 time step. No temporal smoothing. Only 1 hyperparameter α 2, the relative weighting of data and smoothing.

ABIC for the InSAR model Evaluate ABIC for a range of values of α 2 until a minimum is obtained

InSAR model results Majority of significant slip occurs in the upper 12 km of the fault

InSAR model fit to data Fit to data is reasonable for this very crude geometry.

Seismic data inversion 5 time steps, temporal smoothing is required. 2 hyperparameters α 2 and β 2, controlling spatial and temporal smoothing, respectively. Seismic kernels calculated by Yuji Yagi using the method of Kikuchi and Kanamori (1991).

Broadband teleseismic data Broadband vertical component data used from 10 GSN stations

Broadband teleseismic data 80 s time window, Band-pass filtered between 0.01 and 0.8 Hz Sample interval 0.25 s.

ABIC for the seismic model Minimum at α 2 = 0.008, β 2 =0.8, again obtained numerically

Seismic model results Greater slip in the lower 6 km of the fault, and less at the upper 6 km than for the InSAR model.

Fit to data of the seismic model The fit to the waveforms is generally good.

Fit to data of the seismic model Total slip gives a poor fit to the InSAR data

Joint inversion 5 time steps; temporal smoothing required. 3 hyperparameters α 2, β 2 and γ 2 controlling spatial smoothing, temporal smoothing and relative weighting of the datasets.

ABIC for the joint model Minimum at α 2 = 0.0015, β 2 =0.0006, γ 2 =117

Joint model results Less slip in the lower 6 km than for the seismic-only inversion it is relocated to the surface.

Fit to data of the joint model Total slip gives a fit comparable to that for the InSAR-only inversion

Fit to data of the joint model The fit to the waveforms is very similar to that of the seismic-only inversion. (degradation of 2% in misfit)

Slip history of the Manyi earthquake

Comparison of model results InSAR Seis. Joint

Conclusions It is possible to find a solution with ABIC which satisfies both InSAR and seismic datasets. InSAR is the main control on the spatial location of slip, which occurred mainly at shallow depths. This is a viable technique to apply to other large continental earthquakes where near-source information is not available.