Upload
jocelyn-augusta-kelly
View
215
Download
1
Tags:
Embed Size (px)
Citation preview
1
Validation of physics-based ground motion earthquake simulations using a velocity model improved by tomographic inversion results
Ricardo Taborda,1 En-Jui Lee,2 David Gill,3 Po Chen,4 Philip Maechling,3 Thomas H. Jordan2,3
1 Center for Earthquake Research and Information, and Department of Civil Engineering, University of Memphis2 Department of Earth Sciences, University of Southern California3 Southern California Earthquake Center, University of Southern California4 Department of Geology and Geophysics, University of Wyoming
2
southernCaliforniamodels
CVM-H+GTLCVM-S4.26
CVM-HCVM-S
Alternative velocity modelsfor southern California
Magistrale et al. (1996, 200)Kohler et al. (2003)
(…) +Chen et al. (2011)
Lee et al. (2011, 2013)Gil et al. (2013, 2014)
Süss and Shaw (2003)Süss et al. (2005)Plesch et al. (2007, 2009)
(…) +Ely et al. (2010)
5
Taborda and Bielak (2014)BSSA, 104(4): in press
Recent work using different velocity modelscase study: 2008 Chino Hills earthquake
Taborda and Bielak (2013)BSSA, 103(1): 131–156
6
southernCaliforniamodels
CVM-H+GTLCVM-S4.26
CVM-HCVM-S
Alternative velocity modelsfor southern California
Magistrale et al. (1996, 200)Kohler et al. (2003)
(…) +Chen et al. (2011)
Lee et al. (2011, 2013)Gil et al. (2013, 2014)
7
The latest CVM-S4.26 velocity modeltomographic inversion results and merge into CVM-S
CVM-S4.26(Final Model)
CVM-S(Base Model)
Magistrale et al. (1996, 200)Kohler et al. (2003)
Built as a model with“arbitrary” resolution
(…) +Chen et al. (2011)
Lee et al. (2011, 2013)
Starting model
discrete version of themodel with fixed resolution:regular grid every 500 mand minimum Vs = 1000 m/s
3D tomographicinversion yieldsperturbations tostarting model
inversion processincluded 26 iterationsPo Chen and En-Jui Lee
Recovering and merging process
various scheme(s) devised to recover model featurestruncated by the startingmodel and merge the pertur-bations back into the modelfor “arbitrary” queryingresolution
Distributedvia SCEC’s UCVM
(…) +Gil et al. (2013, 2014)
8
The CVM-S4.26 velocity modelmerging alternatives
Option 1
Base Vs < 1 km/s
UseStarting Props.+ Perturbation
NegativePerturbation
Final Model
NO YES
UseBase Props.
+ Perturbation
UseBase Props.
NO YES
checks whether base model is softer than starting model—inside a basin?—
if the base model is softerchecks whether the perturbationwill make it even softer
prevents the perturbation from making softer than the base model—preserves floor base props.—
otherwise ithardens thebase model
otherwise ithardens thestarting model
9
The CVM-S4.26 velocity modelmerging alternatives
Option 2
Base Props.< Starting
UseBase Props.
NegativePerturbation
Final Model
NO YES
Starting Values+
Perturbation
UseBase Props.
NO YESPositivePerturbation
YES
checks whether base model is softer than starting model—inside a basin?—
if the base model is softer, checks whetherthe perturbation will make it even softer
prevents the perturbation from making softer than the base model
prevents the perturbationfrom making stiffer than
the base model
if the base model is stiffer, checks whether
the perturbation will make it even stiffer
10
The CVM-S4.26 velocity modelmerging alternatives
Option 3
Base Props. < Starting
UseStarting Props.+ Perturbation
Final Model
NO YES
UseBase Props.
checks whether base model is softer than starting model—inside a basin?—
the base model ispreserved everywhereit is softer than thestarting model
the inversion results are used everywhere else
11
Comparison between base and merged models
CVM-S CVM-S4.26 Option 1
CVM-S4.26 Option 2 CVM-S4.26 Option 3
Surface shear wave velocity (Vs) in m/s
12
• Largest earthquake in the L.A. region since the 1994 Northridge earthquake.
• Combination of thrust and strike-slip faulting between the Whittier and Chino faults.
• No significant damages, no fatalities.
• Excellent opportunity for testing assumptions and methodologies.
• Recorded in over 450 strong motion station from different seismic networks. 336 surface stations within simulation domain.
The 2008 chino hills earthquakeand region of interest
14
“High”-frequency (0–4 Hz) ground motionselected locations
DataBase model
Merge Option 1Merge Option 2
Merge Option 3
15
6 – 8Good
4 – 6Fair
0 – 4Poor
8 – 10Excellent
» Anderson (2004)13th World Conf. Earthq. Eng.
» as modified inTaborda and Bielak (2013)Bull. Seismol. Soc. Am. 103(1):131–156
AriasIntensity
EnergyIntegral
Duration
PGA PGV PGD
FourierSpectrum
ResponseSpectrum
CrossCorrelation
Validation criteriagoodness of fit
17
CVM-S4.26 Option 1
CVM-S4.26 Option 2 CVM-S4.26 Option 3
GOF scores improvementwith respect to the base model(0–0.25 Hz)
Scale corresponds to change in the GOFscore with respect to the values obtained forthe simulation using the base CVM-S model
20
Improvements beyond inversion fmax
with respect to the base model (0–4 Hz)
CVM-S CVM-S4.26 (Option 1)
21
Improvements beyond inversion fmax
with respect to the base model (0–4 Hz)
GOF score change w.r.t. base model validation
22
» general improvements in the synthetics are obtained but some areas will need further attention
» changes in GOF scores are of the order of 1 to 4 points maximum
» additional improvements are unlike to come from marginal changes to the velocity models at this point, therefore other aspects (like frequency dependent attenuation and coherency in the source model) will need to be considered
Closing remarks and future work
23
Validation of physics-based ground motion earthquake simulations using a velocity model improved by tomographic inversion results
Ricardo Taborda,1 En-Jui Lee,2 David Gill,3 Po Chen,4 Philip Maechling,3 Thomas H. Jordan2,3
1 Center for Earthquake Research and Information, and Department of Civil Engineering, University of Memphis2 Department of Earth Sciences, University of Southern California3 Southern California Earthquake Center, University of Southern California4 Department of Geology and Geophysics, University of Wyoming