Slip-inversion artifacts common to two independent methods J. Zahradník, F. Gallovič Charles University in Prague Czech Republic.

Slides:



Advertisements
Similar presentations
Traditional practice separates seismic data processing and further interpretation. However the most efficient processing methods utilize a-priori information.
Advertisements

Group Velocity Dispersion Curves from Wigner-Ville Distributions Simon Lloyd 1, Goetz Bokelmann 1, Victor Sucic 2 1 University of Vienna 2 University of.
© 2011 Autodesk Freely licensed for use by educational institutions. Reuse and changes require a note indicating that content has been modified from the.
Toward the next generation of earthquake source models by accounting for model prediction error Acknowledgements: Piyush Agram, Mark Simons, Sarah Minson,
Prague, March 18, 2005Antonio Emolo1 Seismic Hazard Assessment for a Characteristic Earthquake Scenario: Integrating Probabilistic and Deterministic Approaches.
1 A New Technique for Deriving Electric Fields from Sequences of Vector Magnetograms George H. Fisher Brian T. Welsch William P. Abbett David J. Bercik.
Single station location Multiple station location
The Calibration Process
UNIVERSITY OF ATHENS Faculty of Geology and Geoenvironment Department of Geophysics and Geothermics A. Agalos (1), P. Papadimitriou (1), K. Makropoulos.
11/02/2007PEER-SCEC Simulation Workshop1 NUMERICAL GROUND MOTION SIMULATIONS: ASSUMPTIONS, VERIFICATION AND VALIDATION Earthquake Source Velocity Structure.
5: EARTHQUAKES WAVEFORM MODELING S&W SOMETIMES FIRST MOTIONS DON’T CONSTRAIN FOCAL MECHANISM Especially likely when - Few nearby stations, as.
Application to Wells Nevada Earthquake
Near-Field Modeling of the 1964 Alaska Tsunami: A Source Function Study Elena Suleimani, Natalia Ruppert, Dmitry Nicolsky, and Roger Hansen Alaska Earthquake.
Surface wave tomography: part3: waveform inversion, adjoint tomography
Teleseismic Location find direction of signals based on Array algorithms backtrace ray paths through the earth simplifications: flat earth, plane waves.
MARsite kickoff meeting December 19-20, 2012, Istanbul WP5 - TASK 2 Near real-time determination of the earthquake finite-fault source parameters and models,
Second degree moments – a tool for the fault plane detection?
Linear(-ized) Inverse Problems
RESOLVING FOCAL DEPTH WITH A NEAR FIELD SINGLE STATION IN SPARSE SEISMIC NETWORK Sidao Ni, State Key Laboratory of Geodesy and Earth’s Dynamics, Institute.
An array analysis of seismic surface waves
1 The Earth’s Shells Quantitative Concepts and Skills Weighted average The nature of a constraint Volume of spherical shells Concept that an integral is.
Focal Mechanisms of Micro-earthquakes in the Little Carpathians - the Influence of the Model and of the Source Time Function Used AIM Second Annual Meeting.
The kinematic representation of seismic source. The double-couple solution double-couple solution in an infinite, homogeneous isotropic medium. Radiation.
MICRO-SEISMICITY AND FOCAL MECHANISMS IN THE MALÉ KARPATY MTS., SLOVAKIA Lucia Fojtíková, Václav Vavryčuk, Andrej Cipciar, Ján Madarás.
IDENTIFICATION OF THE FAULT PLANE AND A SIMPLE 3D VISUALIZATION TOOL Petra Adamová, Jiří Zahradník Charles University in Prague
1 Cythera M6.7 earthquake (January 8, 2006) in southern Aegean: uneasy retrieval of the upward rupture propagation J. Zahradnik, J. Jansky, V. Plicka,
Uncertainty of location and multiple-point source model of M 7.1 Van earthquake, Turkey, 2011 J. Zahradnik 1, E. Sokos 2, J. Jansky 1, V. Plicka 1 1) Charles.
Institute of Geological & Nuclear Sciences Limited, P.O. Box 30368, Lower Hutt, New Zealand Ph: Russell Robinson & Rafael Benites Synthetic.
Complex earthquake directivity during the 2009 L’ Aquila mainshock Tinti E., Scognamiglio L., Cirella A., Cocco M., and A. Piatanesi Istituto Nazionale.
Blue – comp red - ext. blue – comp red - ext blue – comp red - ext.
Quick fault-plane identification by a geometrical method: The M w 6.2 Leonidio earthquake, 6 January 2008, Greece and some other recent applications J.
NEW VERSION OF ISOLA SOFTWARE TO INVERT FULL WAVEFORMS INTO SEISMIC SOURCE MODELS Efthimios Sokos 1) and Jiri Zahradnik 2) 1) University of Patras, Greece.
Disputable non-DC components of several strong earthquakes Petra Adamová Jan Šílený.
University of Colorado Boulder ASEN 5070: Statistical Orbit Determination I Fall 2014 Professor Brandon A. Jones Lecture 26: Singular Value Decomposition.
March 2006 WGCEP Workshop Ruth A. Harris U.S. Geological Survey.
Colorado Center for Astrodynamics Research The University of Colorado 1 STATISTICAL ORBIT DETERMINATION ASEN 5070 LECTURE 11 9/16,18/09.
Large Earthquake Rapid Finite Rupture Model Products Thorne Lay (UCSC) USGS/IRIS/NSF International Workshop on the Utilization of Seismographic Networks.
Jayne Bormann and Bill Hammond sent two velocity fields on a uniform grid constructed from their test exercise using CMM4. Hammond ’ s code.
Earthquake source parameters inferred from teleseismic source time functions Orfeus Workshop “Waveform Inversion” June, 19th, 2008 Martin Vallée and Jean.
The January 2010 Efpalio earthquake sequence in Western Corinth Gulf: epicenter relocations, focal mechanisms, slip models The January 2010 Efpalio earthquake.
Sarah Minson Mark Simons James Beck. TeleseismicStrong motionJoint km Delouis et al. (2009) Loveless et al. (2010) Seismic + Static.
Time reversal imaging in long-period Seismology
SPICE Research and Training Workshop III, July 22-28, Kinsale, Ireland Seismic wave Propagation and Imaging in Complex media: a European.
Antonella Cirella, Alessio Piatanesi, Elisa Tinti, Massimo Cocco Ground Motion and Source Process of the 6 th April 2009 L’Aquila, central Italy, Earthquake.
Václav Vavryčuk Rosalia Daví Institute of Geophysics, Academy of Sciences, Praha Seismic network calibration for retrieving accurate.
Ground motion simulations in the Pollino region (Southern Italy) for Mw 6.4 scenario events.
1 Cythera M6.7 earthquake (January 8, 2006) in southern Aegean: uneasy retrieval of the upward rupture propagation J. Zahradnik, J. Jansky, V. Plicka,
The seismogram U = Source * Propagation * Site.
Classification Ensemble Methods 1
Synthetic tests of slip inversion (two methods: old = ISOLA, new = conj. gradients) J. Zahradník, F. Gallovič MFF UK.
Near Fault Ground Motions and Fault Rupture Directivity Pulse Norm Abrahamson Pacific Gas & Electric Company.
1 Wavefield Calibration Using Regional Network Data R. B. Herrmann Saint Louis University.
Focal mechanisms and moment tensors of micro-earthquakes in the Malé Karpaty (Little Carpathians) Mts., Slovakia Lucia Fojtíková 1, Václav Vavryčuk 2,
Earthquake source modelling by second degree moment tensors Petra Adamová Jan Šílený Geophysical Institute, Academy of Sciences, Prague, Czech Republic.
Alexandra Moshou, Panayotis Papadimitriou and Kostas Makropoulos MOMENT TENSOR DETERMINATION USING A NEW WAVEFORM INVERSION TECHNIQUE Department of Geophysics.
A new prior distribution of a Bayesian forecast model for small repeating earthquakes in the subduction zone along the Japan Trench Masami Okada (MRI,
Seismic phases and earthquake location
MOMENT TENSOR INVERSION OF POSSIBLY MULTIPLE EVENTS AT REGIONAL DISTANCES Petra Adamová 1, Jiří Zahradník 1, George Stavrakakis 2 1 Charles University.
On constraining dynamic parameters from finite-source rupture models of past earthquakes Mathieu Causse (ISTerre) Luis Dalguer (ETHZ) and Martin Mai (KAUST)
Teleseismic Location find direction of signals based on Array algorithms backtrace ray paths through the earth simplifications: flat earth, plane waves.
Kinematic Modeling of the Denali Earthquake
The Calibration Process
1-D Mississippi embayment sediment velocity structure and anisotropy: constraint from ambient noise analysis on a dense array Chunyu,Liu1; Charles A. Langston1.
High-Performance Computing (HPC) IS Transforming Seismology
Larry Braile, Purdue University
Slip pulse and resonance of Kathmandu basin during the 2015 Mw 7
Cultural Confusions Show that Facial Expressions Are Not Universal
Two M5 earthquakes in Corinth Gulf, January 2010
by J. Galetzka, D. Melgar, J. F. Genrich, J. Geng, S. Owen, E. O
by Satoshi Ide, Annemarie Baltay, and Gregory C. Beroza
Presentation transcript:

Slip-inversion artifacts common to two independent methods J. Zahradník, F. Gallovič Charles University in Prague Czech Republic

Guess the correct answer: If two methods yield a stable slip feature, that feature is likely true. If two methods yield a stable slip feature, that feature might still be wrong.

Guess the correct answer: If two methods yield a stable slip feature, that feature is likely true. If two methods yield a stable slip feature, that feature might still be wrong.

Two methods: iterative deconvolution of the point-source contributions (Kikuchi and Kanamori, 1991) and ISOLA code (Sokos & Zahradnik, 2008); modified to allow a less concentrated distribution of the slip (Zahradnik et al., JGR, in press) a new technique (Gallovic et al., GRL 36, L21310, 2009), iterative back-propagation of the waveform residuals by the conjugate gradients technique; gradient of the waveform misfit with respect to the model parameters being expressed analytically; the positivity and fixed scalar moment constraint applied The two methods do not need prior knowledge of the nucleation point and rupture velocity.

Part 1 An incorrect rupture velocity and spurious (false) patches from error-free synthetic data. Synthetic data mimic Mw 6.3 earthquake, Greece 2008 discussed at the end.

HYPO and DD relocation: A. Serpetsidaki, Patras PSLNET BB and SM (SER, MAM, LTK, PYL co-operated by Charles Univ.) ITSAK SMNOA BB Synthetic data mimic the Mw 6.3 Movri Mountain 2008 (Andravida) earthquake, Greece

Low-frequency inversion (f<0.2 Hz) of synthetic near-regional data: a line source The station distribution fixed (as in real data case). Three scenarios of the rupture propagation direction. Two asperities symmetric with respect to the fault center.

Low-frequency inversion (f<0.2 Hz) of synthetic near-regional data: a line source The station distribution fixed (as in real data case). Three scenarios of the rupture propagation direction. Two asperities symmetric with respect to the fault center.

Iterative method (color; slip velocity) ISOLA free and modified (green circles; proportional to moment) x t ‘free’ ‘modified’ ‘Free’ = very concentrated ‘Modified’ = better distributed

Vr = 3.28 km/s (instead of 3 km/s) Unilateral propagation (from the left) 3 km/s

Vr = 3.68 km/s (instead of 3 km/s) Unilateral propagation (from the right) 3 km/s

Vr = 5.68 and 5.26 km/s (instead of 3 km/s), i.e. a larger temporal delay closer to the fault center and a FALSE ASPERITY at the center ! Common to both methods. Bilateral propagation (from the center), no slip at the fault center in the input model 3 km/s

Vr = 5.68 and 5.26 km/s (instead of 3 km/s), i.e. a larger temporal delay closer to the fault center and a FALSE ASPERITY at the center ! Common to both methods. Bilateral propagation (from the center), no slip at the fault center in the input model 3 km/s the worst case

Where the problems may arise from? Explanation in terms of concepts of the ‘source tomography’ (80’s), e.g., Ruff (1984), Menke (1985), Frankel & Wennerberg (1989) kinematic approach

Forward simulation of two asperities (2 x 5 point sources) and two stations directive station anti-directive station Slip Rupture propagation along fault (x) time displacement ZAK SER

Forward simulation of two asperities (2 x 5 point sources) and two stations directive station anti-directive station Slip Rupture propagation along fault (x) time x t ‘locating’ the 2x5 sources back to source ZAK SER

Forward simulation of two asperities (2 x 5 point sources) and two stations directive station anti-directive station Slip Rupture propagation along fault (x) time x t ZAK SER Kinematic Projection Lines (trade-off between source position and time)

x t

True asperity x t

False asperity The unilateral case: False asperity biases the rupture velocity. x t

The bilateral case: False asperity appears as a separate ‘event’ on the intersection of two directive strips. true false x t SER station is directive for one asperity ZAK station is directive for the other asperity

Partial results The projection lines of the individual stations explain the spurious patches. We need a generalization of the Kinematic Projection Lines, or Strips (KPS) for complete wavefields in heterogeneous media. New: thus we introduce the Dynamic Projection Strips (DPS).

Dynamic Projection Strips (still on synthetic data) Key concept: Mapping the correlation between a complete observed waveform at a station and a synthetic waveform due to a single x-t point source. It works like a ‘multiple-signal detector’. The waveform is mapped into equivalent x-t points, similar to kinematic location, hence analogy with the projection lines.

Part 2 The directive station SER is strongly affected by both patches, but ‘sees’ them as a single one. The anti-directive (backward) station ZAK ‘sees’ both patches. x t Unilateral rupture toward x > 0 Dynamic Projection Strips derived from synthetic waveforms

Part 2 The DPS (at right), derived form waveforms, are analogical to kinematic projection lines (dashed). Unilateral rupture toward x > 0

Part 2 Intersecting DPS’ of the individual stations (so-called ‘dark spots’) delimit the source region. Unilateral rupture toward x > 0

Part 2 Final inversion result, already understandable in terms of the station contributions. Unilateral rupture toward x > 0

Part 2 Final inversion result, already understandable in terms of the station contributions. Unilateral rupture toward x > 0

Part 2 x t Unilateral rupture toward x > 0 Dynamic Projection Strips derived from synthetic waveforms

Part 2 This scenario gives similar result, but not exactly ‘mirror-like’. It is because the station network is not symmetric with respect to fault. Unilateral rupture toward x < 0

Part 2 SER is a directive station for one asperity. ZAK is directive for the other asperity. Bilateral rupture from x=0

Part 2 Intersection of the two directive strips attracts the solution to the fault center (false). Bilateral rupture from x=0

Part 2 FALSE ! Thus the false asperity is explained by separately analyzing waveforms of individual stations in terms of DPS. Bilateral rupture from x=0

Part 2 Thus the false asperity is explained by separately analyzing waveforms of individual stations in terms of DPS. FALSE ! Bilateral rupture from x=0

Partial results (still synth. data) The dynamic projection strips (DPS) can be constructed from complete waveforms. The strips illuminate the role played by each station in the slip inversion. The strips enable quick identification of the major slip features: the predominant rupture direction, multiple asperities, etc.

Possible constraints to reduce artifacts Position of the nucleation point Position of a partial patch Caution: Constraining with wrong parameter values may bias the solution! (if known …)

Part 2 Real earthquake data (Mw 6.3 strike-slip) Can the Dynamic Projection Strips be extracted from real waveforms ?

Application Movri Mountain (Andravida) Mw 6.3 earthquake, June 8, 2008 NW Peloponnese, Greece Gallovic et al., GRL 36, L21310, 2009

ITSAK, Greece 2 victims hundreds of injuries More details of the practical application in the presentation by Sokos et al. (T/SD1/MO/06)

HYPO and DD relocation: A. Serpetsidaki, Patras PSLNET BB and SM (SER, MAM, LTK, PYL co-operated by Charles Univ.) ITSAK SMNOA BB Near-regional slip inversion

Dynamic projection strips: real data Near-regional stations (< 200 km) f < 0.2 Hz

Aggregated strips of all 8 stations and the slip inversion: real data Data indicate a predominant unilateral rupture propagation, with an almost 5-sec delay of the rupture at the hypocenter. Zahradnik and Gallovic, JGR 2010, in press

Non-unique results: examples of slip models (green) equally well matching real waveforms (var. red. 0.7): Black circles: an (assumed) patch used to initialize the inversion. Zahradnik and Gallovic, JGR 2010, in press

The intention was to improve insight in the slip-inversion ‘black box’. Conclusions:

The intention was to improve insight in the slip-inversion ‘black box’. We suggest complementing the inversions by analyses of the Dynamic Projection Strips (constructed from complete waveforms). DPS illuminate the individual station roles and indicate the major slip features. They also explain the origin of possible artifacts, e.g. biased rupture velocities, or false asperities. Conclusions:

The intention was to improve insight in the slip-inversion ‘black box’. We suggest complementing the inversions by analyses of the Dynamic Projection Strips (constructed from complete waveforms). DPS illuminate the individual station roles and indicate the major slip features. They also explain the origin of possible artifacts, e.g. biased rupture velocities, or false asperities. Spurious asperities may be very stable and common to independent methods; thus easily misinterpreted as ‘real’ features in standard resolution checks. The same station distribution may create artifacts, or not, dependent on the true slip model. Conclusions:

The intention was to improve insight in the slip-inversion ‘black box’. We suggest complementing the inversions by analyses of the Dynamic Projection Strips (constructed from complete waveforms). DPS illuminate the individual station roles and indicate the major slip features. They also explain the origin of possible artifacts, e.g. biased rupture velocities, or false asperities. Spurious asperities may be very stable and common to independent methods; thus easily misinterpreted as ‘real’ features in standard resolution checks. The same station distribution may create artifacts, or not, dependent on the true slip model. For a mathematical counterpart of DPS in terms of Singular Value Decomposition, see Gallovic & Zahradnik (JGR submitted) and poster ES5/P9/ID112 in this session. Conclusions:

Examples of slip models (A to E) equally well matching real waveforms Black circles: an (assumed) slip patch used to initialize the inversion

Part 2 x t Unilateral rupture toward x > 0 Dynamic Projection Strips derived from synthetic waveforms

Part 2 This scenario gives similar result, but not exactly ‘mirror-like’. It is because the station network is not symmetric with respect to fault. Unilateral rupture toward x < 0

Part 2 Thus the false asperity is explained by separately analyzing waveforms of individual stations in terms of DPS. FALSE ! Bilateral rupture from x=0

Possible constraints to reduce artifacts Position of the nucleation point Position of a partial patch Caution: Constraining with wrong parameter values may bias the solution! (if known …)

Part 2 … or increasing frequency (if the structural model is known)

Trade-off between source position and time Station Y True position of a point asperity Xa Trial position of a point asperity X Tr (Xa) + T(Xa,Y) = const = Tr (X) + T(X,Y) Knowing the asperity position and time, Xa and Tr(Xa), we can calculate all equivalent positions X and times Tr (X) characterized by the same arrival time (=const): a hyperbola. For a station along the source line, the Tr = Tr(X) degenerates to a straight line.