Source Inversion Validation (SIV)Current Progress and Future Directions
Morgan Page, Martin Mai, Mathieu Causse, Juerg Hauser, Danijel Schorlemmer, and SIV Modelers
Different inversion methods, model parameterizations,
constraint choices, and data sets lead to different results.
Mai, P. M., G. Festa, C. Francois-Holden, J. Burjanek, S. Di Carli, A. Emolo, A. Zollo, B. Delouis, and R. Madariaga, 2006.
Modelers were given:• seismic moment• strike, dip, and rake• hypocentral depth• velocity-density structure• noise-free synthetics
Unknown to modelers, input model had a constant rise time, constant rupture velocity, and a triangular SVF
Testing kinematic source inversionsMartin Mai’s SPICE Blind Test
Blind Test Results
!"#$%&'()*&$+,")'-./,0.1")'2!(-3'Martin Mai, Morgan Page, and Danijel Schorlemmer
Previous Source Inversion Validation Exercise (SPICE Test) showed large variability between models
First Steps:
! Problem 1a: Point-Source Green’s function test) ! Problem 1b: Extended Source Green’s function test) ! Problem 2: Simple Inversion
SIV Goals:
• Verify that the forward problem is being done correctly by all modelers
• Assess reliability, resolution, and robustness of different inversion strategies
• Develop testing center to host inversion “experiments” of varying complexity
! !"#$%&'()*'+,,-'./0/'12#34)256'$('174'8"*492%*:'$;'7<<'"54'57#9=$579%&'$%'()*'><$%:'(*4('=7<="<7(*:'()*'?#**%@4';"%=92%4'=2##*=(<A'
! B)*'.CDE5#2F*=('()"4'4(7#(4'1$()'7'(*4('(2'G*#$;A'?HE=2I5"(792%4J'! !"#$%&'(#)*+,-./&.01.23.4,"&56.."/*/3,&,*$7,4./(./.0.8.0.239.(:$"."/&+5.;$&5.
5#3#<,%,#)(.(:$"./%4.=#8+/*.(:$"'>)%+?#%.#>.4)*/?#%..&*[email protected].(,+.
! B5,.(5,/*'3#4):)(./&.&5,.<$C,%.4,"&5.*,():&.$%D.E;[email protected]..
!*5() J23K
DK J23L(K
D. J23L(K
!*%4$(A J<L+3HK
1A1 FAM 9AI 9AH '9A1 FAM 9AI 9AH '9A1 NAN HA0 9AN 'FAM NAN HA0 9AN 'FAM IA9 HAI 9AO '0MA1 IA9 HAI 9AO '0MA1 IAM HAM 9AM '9FA1 IAM HAM 9AM '9FA1 MA1 FAI9 HA9 'FNA1 MA1 FAI9 HA9
Initial Results Reveal Large DiscrepanciesDip-Slip Case
Current “Point Source” Test ResultsDip-Slip Case
Current “Point Source” Test ResultsStrike-Slip Case
Initial Extended Source Results 0-5 Hz
• Significant differences appear in the whole frequency range
• Discrepancies mainly arise from the various techniques used to interpolate the rupture parameters on the fault plane
Slide: Mathieu Causse
Initial Extended Source Results 0-1 Hz
Slide: Mathieu Causse
Slide: Mathieu Causse
SIV website: http://siv.usc.edu
• Download problem descriptions, submitted results, talks from workshops
• Discussion pages for codes to work out bugs and share documentation
Benchmark Database with Upload & Plotting ToolsIn development
Website being developed by Juerg Hauser
Benchmark Database with Upload & Plotting ToolsIn development
Website being developed by Juerg Hauser
! "#$%&#%'()&%*))())%$*+(,#-.%./)0&1%23/)45)/./6*-/&7%/8%-9:&9-(%.#;(6)1%
!"#$%&'()*+&%+*,-+.//0+
12345+1#6&78"&9:&;<=+&;)&,(>&+6#$?%+
12@45+A6&78"&9:&;<=+>B*$&+6#$?%+
18
Types of Inversion Tests
• Kinematic Models (like original SPICE-test)
• Real Earthquakes• construct earthquake from empirical
Green’s functions, or• use subset of stations to produce
synthetics at remaining stations
• Dynamic Models • best of both worlds: can be realistic;
true slip model is known
120.5 120.4 120.2
35.7
35.8
35.9
36
GH3E GH2E
GFU SC3E SC2E SC1E
CH3E FZ3
TEMB
CH2E
KFU
CH1E FZ1
C2W
C3W
C4W
RFU C4AW
GH3W GH5W
PHOB
FFU
FZ9
FZ6
FZ15
FZ4
FZ12
FZ7
JFU DFU
GH1W
FZ11
FZ8
CAND
CARH
HOGS
HUNT
LAND
MASW
MIDA
MNMC
POMM
RNCH
TBLP
10km
Latitu
de (˚
N)Longitude (˚W)
120.3
VC5W
COAL MFU
VC2E
LOWS
VC4W
VC3W VFU
VC2W VC1W EFU
Model 3 ! Input
GPS model Inverted Kinematic Models
Joint:Symmetric Rise time Joint: Asymmetric rise time
Seismic
Inversion Exercise #1Spontaneous Dynamic Crack-like Rupture
“Racetrack” Station Distribution:Modelers will invert some synthetics, predict others
Available at SIV website: siv.usc.edu
! !"#$%&"'#()"#%$*%*+,&'--.%/('0,1.%$2"%&3#4$%30%'&+-3$(5"%'05%+2'#"%
! 6*7(&"0$%$2"%(#"%*1%$2"%1*)8')59&*5"-30:%7*5"#%;*0%$2"%<=>9#3$"?%
! @A+'05%B%3&+)*C"%$2"%<=>%8"D#3$"%! "#$%&'()&#&*+$+,-.(
! /%%$.(0%1(+22-'+&/-(%3$+3-(1-."$/()%2#&1+.%3.(
! E"A$%:"0")',*0%*1%&*5"-30:%F%30C")#3*0%"A")73#"#%! 4+3-2&,)(1"#/"1-(2%'-$.5(6+/7(+3)1-&.+38(.%"1)-()%2#$-9+/:5(6+/7(&3'(6;%(1&3'%2(3%+.-(
! ':3&2+)(1"#/"1-(2%'-$.5(6+/7(+3)1-&.+38(.%"1)-()%2#$-9+/:5(6+/7(&3'(6;%(1&3'%2(3%+.-(
! "343%63(#-1/"1*&,%3.(+3(<&1/7(./1")/"1-(
! G)*C35"%'553,*0'-%5'$'%#"$#%! =/&,)('+.#$&)-2-3/.>(/-$-.-+.2+)(6&?-0%12.(
! H*8%$*%'##"##%"+3#$"&37%(07")$'30,"#%30%&*5"-#%I%
! J*,C'$"%&*)"%$"'&#%$*%+'),73+'$"%%
SIV: The Next Steps
