scec broadband platform (bbp) simulation methods validation for

SCEC Broadband Platform (BBP) Simulation Methods Validation for NGA-East

BBP Validation Team: N. Abrahamson, P. Somerville, F. Silva, P. Maechling, R. Archuleta, J. Anderson, K. Assatourians, G. Atkinson, J. Bayless, J. Crempien, C. Di Alessandro, R. Graves, T. Hyun, R. Kamai, K. Olsen, R. Takedatsu, F. Wang, K. Wooddell,, D. Dreger, G. Beroza, S. Day, T. Jordan, P. Spudich, J. Stewart and their collaborators…

BBP_v14.3 Validation Exercise

Menu du jour n  Introduction n  Validation framework and schemes

n  Part A. Validation against recorded ground motions n  Event and record selection n  Correcting for site conditions

n  Part B. Validation against GMPEs n  Evaluation


Large collaborative validation of simulations using the SCEC BroadBand Platform

Driven by need of seismic hazard projects to supplement recorded datasets n  South-Western U.S. utilities (SWUS) n  PEER NGA-East project (new CENA hazard model) n  PEER NGA-West projects

n  Southern California Earthquake Center (SCEC) BroadBand Platform (BBP) n  Set of computational tools for ground motion

simulations, including post-processing

Collaboration of SWUS-SCEC-PEER critical to success.

3


Past validations:

Source: Graves and Pitarka (2010)

4

This exercise: Quantitative validation for forward simulations in engineering problems


Objective relevant to NGA-East n  Quantitative validation for forward

simulations n  To supplement recorded data for development

of GMPEs and hazard analyses n  Key focus: 5% damped elastic “average” PSA

(f=0.1-100 Hz/ T=0.01-10 s) n  Validation for mean ground motions only –

not for standard deviation!

5


Key lessons learned – past validations n  Need more transparency... n  Need to validate against many events n  Need clear documentation of fixed and optimized

parameters from modelers for each region n  Need source description that is consistent

between methods n  Use unique crustal structure (V, Q) for all models n  Consider multiple source realizations n  Run simulations for reference site conditions –

correct data with empirical site factors n  Make all validation metrics computation and plots

in uniform units/format – implement post-processing pipeline on BBP

n  Need to tie-in to specific code/BBP version 6


Validation schemes n  A. Validation against recorded earthquake

ground motions

n  B. Validation against GMPE for generic scenarios

Validation allows for development of region-specific rules (source scaling, path)

7


Obs, TS, Surf. Sims, TS, Ref.

Site factors

Obs, IM, Ref.

GOF, Ref.

GMPE, IM, Ref.

Sims, IM, Ref.

IM Processors

Obs, IM, Surf.

IM Processors

Event and stations

Obs: observed/recorded Sims: simulated TS: time series IM: Intensity measures (e.g. PSA) Ref.: reference rock site Surf.: surface, soil site GOF: goodness-of-fit

This validation exercise

Part A.

Part B.

8

BBP_v14.3 Validation Exercise 9

Method Name(s) Method type – Finite fault models Contact(s) and Institution

Composite Source Model (CSM) Broadband deterministic

J. Anderson (UNR)

UCSB R. Archuleta, J. Crempien

(UCSB)

EXSIM Stochastic Brune spectrum K. Assatourians, G. Atkinson (UWO)

Graves and Pitarka Hybrid: deterministic LF and stochastic HF

R. Graves (USGS)

SDSU (BB Toolbox) K. Olsen (SDSU)

Simulation Methods and Modelers

Point source methods validated outside BBP: SMSIM Silva’s stochastic models (also FF)

GMPEs used for comparison: NGA-West1: As, BA, CB, CY (2008) CENA: Silva SCVS 2003, Atkinson 2008-2001, Pezeshk et al. (2011)


Selection of events and stations Part A (comparison with recordings)

§  Large dataset (~25 EQs)

§  Many regions & tectonic environments

§  Span wide magnitude range

(Mw 4.64 to 7.62)

§  Variety of mechanisms

§  Well-recorded (17 EQs with> 40 records)

§  Select a large subset of stations (~40) that are consistent with mean and standard deviation PSa of the full dataset.

10

Region Event Name Year Mw # Records < 200 km (* <1000 km)

Note on Selection # Selected Records

Actual (final)

WUS Loma Prieta 1989 59 40

WUS Northridge 1994 6.73 124 All stations within 10 km selected 39

WUS Landers 1992 7.22 69 All stations within 100 km selected 40

WUS Whittier Narrows 1987 5.89 95 Truncate stations

at 40 km 39

WUS North Palm Springs 1986 6.12 32 31

JAPAN Tottori 2000 6.59 171 40

JAPAN Niigata 2004 6.65 246 40

WUS Alum Rock 5.45 40 40

WUS Chino Hills 5.39 40 40

CENA Saguenay 1988 5.81 14* All records

selected, use only within 200 km

11

CENA Riviere-du-Loup 2005 4.6 98*

All records selected, use only

within 200 km 21

CENA Mineral, VA 2011 5.68 94* All records

selected, use only within 300 km

10


Selection of stations GOAL: Reduce the number of records (stations) per event while maintaining the mean and standard deviation of the full dataset. CRITERIA: §  Site Vs30 >= 300 m/s to reduce strong non-linear site

effects §  Well-Recorded Earthquakes (40 stations per. event is ideal) §  Close-in records (<200 km; <1000 km for CENA)

METHODOLOGY: 1.  Develop simple regression model for all stations for each

earthquake (PGA, T=0.2 sec, T=1.0 s) 2.  Randomly sample 10 stations from 4 equal distance bins

(log space) to generate ~1000 to 10,000 random samples. 3.  Perform regressions on samples from population. 4.  Develop a penalty function to select station sample with

small mean bias and standard deviation over all spectral periods (PGA, T=0.2 sec, T=1.0 sec)

Part A (comparison with recordings)


Selection of stations Northridge Example


−120 −119.5 −119 −118.5 −118 −117.5 −117 −116.533.6

33.8

34

34.2

34.4

34.6

34.8

35

35.2

35.4

Longitude

Latit

ude

Northridge Station Locations

101 102102

103

Distance (km)

Vs30

Northridge

Vs30 Full SetSampled Stations (Vs30 > 300 m/s)

300 m/s

103

102


Selection of stations Northridge Example


100 101 10210−4

10−3

10−2

10−1

100

Distance (km)

PGA

(g)

Northridge

Full Model PredictionSample PredictionPGA ObservationsRandom Samples

σfull = 0.442 σselected= 0.444

100 101 10210−4

10−3

10−2

10−1

100

Distance (km)

Spec

tral A

ccel

erat

ion

at T

=0.2

sec

(g)

Northridge

Full Model PredictionSample PredictionT=0.2 sec ObservationsRandom Samples

100 101 10210−4

10−3

10−2

10−1

100

Distance (km)

Spec

tral A

ccel

erat

ion

at T

=1.0

sec

(g)

Northridge

Full Model PredictionSample PredictionT=1.0 sec ObservationsRandom Samples

100

10-4

10-2




Correcting data to reference condition

n  WUS: Recorded data corrected to Vs30=863 m/s n  Empirical amplification factors from Stewart

and Seyhan (Boore et al. NGA-West2 model) n  Flin=Linear scaling (Vs30 effect) n  Fnl=Nonlinear scaling relative to Flin

n  Fsite=1/exp(Flin+Fnl) n  Basin correction through Z1.0=f(Vs30) based on

Chiou and Youngs 2008 (NGA-West1 model) n  Fbasin=1/exp(f(Φ5,Φ6,Φ7,Φ8)

n  Corrected PSA=PSARec*Fsite*Fbasin



Correcting data to reference condition

n  CENA: Recorded data corrected to Vs30,ref=1000 m/s

n  Empirical amplification based on GWG memo (Stewart et al.) n  Flin=Linear scaling (Vs30,rec to 760 m/s) n  F760-Vs30,ref= Linear scaling using updated AB06

factors (Boore, personal communication)

n  Fsite=1/exp(Flin+F760-Vs30,ref)

n  Corrected PSA=PSARec*Fsite



Nomenclature For each scenario, specification of: n  Source: Mw, geometry, location, hypocenter n  Path: consistent with 1D velocity model n  Site (as-recorded to reference): empirical site

correction factors from Boore et al. 2013 NGA-West2

For each scenario, seismograms generated for: n  50 source realizations n  ~ 40 stations n  2 horizontal dir.


4,000 time series

16


Ground motion products Part A (comparison with recordings)

17

Northridge PSA, station SCE Vs. 1000034

PSA (

g)

Qualitative evaluation of velocity time series and Husid plot based on Arias intensity


Ground motion products

n  Average GOF with T for all stations within an event


18

n  Goodness-of-fit measures for PSA

Period (s)

n  Average GOF with T for all realizations (all stations)

Period (s)



n  Average GOF with distance (all realizations)


19




n  Map of average GOF (all realizations)


20



n  GOF plots also developed for n  NGA-West1

(2008) GMPEs n  SMSIM

Allows to see trends/event terms



21


Part B: Scenario selection n  Scenarios from NGA-West1&2 well

constrained by data at 20 and 50 km Rrup n  M6.2 SS, Vertical, Ztor= 4 km n  M6.6 SS, Vertical, Ztor= 0 km n  M6.6 REV, Dip=45 deg., Ztor= 3 km n  M5.5 REV, Dip=45 deg., Ztor= 6 km

n  50 realizations of the source, WITH randomized hypocenter location for each

n  Simulations for two velocity models: NorCal and SoCal

Part B (comparison with GMPEs)

22


Summary of Simulated Events

Tottori

Niigata

Chino Hills

Landers

Loma Prieta

Northridge

Alum Rock

* Part B: 4 scenarios

Summary - Parts A and B

23

Saguenay Mineral

Riviere-du-Loup

Whittier

North Palm Springs


Evaluation n  Review panel (June 2013)

n  Douglas Dreger (Chair), UC Berkeley n  Gregory Beroza, Stanford n  Steven Day, SDSU n  Christine Goulet, UC Berkeley n  Thomas Jordan, USC n  Paul Spudich, USGS n  Jonathan Stewart, UCLA

n  Input for review n  Modeler’s documentation and self-assessment n  BBP results (parts A and B)

n  Part A: criteria based on binned GOF according to M (event), R, T

n  Part B: simple pass-fail

Evaluation

24


Evaluation – Part A 1. Comparison of PSA GOF for each event

Mean bias Mean absolute bias

n  Failure threshold is ln(2)=0.69 n  Thresholds of 0.5 and 0.35 were considered as passing

criteria

Evaluation Part A

25

70-2

00 k

m

27

0.01 to 0.1 s 0.1 to 1 s 1 to 3 s More than3 s

0-5

km

5-20

km

20

-70

km

70-2

00 k

m


Evaluation – Part A 1. Comparison of PSA GOF (mean and mean absolute bias) 2. Combined metric: mean and mean absolute bias

n  Used alone n  Used with GMPEs

3. Evaluation of attenuation bias

n  Distance dependence slope of zero within 95% confidence interval

Evaluation Part A

28


Combined Metric & Comparison with GMPEs

CGOF = 12ln data

model( ) +12ln data

model( )

Evaluation Part A

29

CGOF

30

0-5

km

5-20

km

20

-70

km

70-2

00 k

m


Combined Metric & Comparison with GMPEs

CGOF = 12ln data

model( ) +12ln data

model( )

Evaluation Part A

31

CGOFNormalized =CGOFsims CGOFGMPE

CGOFNormalized

32

0-5

km

5-20

km

20

-70

km

70-2

00 k

m


Evaluation – Part A 1. Comparison of PSA GOF (mean and mean absolute bias) 2. Combined metric: mean and mean absolute bias

n  Used alone n  Used with GMPEs

3. Evaluation of attenuation bias n  Distance dependence slope of zero within 95%

confidence interval n  Also check systematic bias (qualitative)

Evaluation Part A

33

ln SaobsSasyn

!

"#

$

%&= a+ b ⋅ ln R( )

Fit a line through distance binned GOF values

Determine whether slope b=0 lies within 95% confidence interval

Attenuation Bias Evaluation Part A

34


Bias with distance

Red shows a ratio of abs(b)/b95% greater than 1.0, the zero slope does not lie within the 95% CI from simulations. Green shows cases where b=0 lies within the 95% CI from simulations. Smaller numbers are generally controlled by small estimates of slope.


Part B – Design and Evaluation criteria Part B (comparison with GMPEs)

36

n  Scenarios from NGA-West1&2 well constrained by data at 20 and 50 km Rrup n  M5.5 REV n  M6.2 SS n  M6.6 SS & REV

n  50 realizations of the source, WITH randomized hypocenter location for each

n  Simulations for two velocity models: NorCal and SoCal


v14.3 CG Update of Review Panel Findings n  Details in SRL Focus issue (8 papers) under review n  The BBP objective of a version-controlled numerical test bed with

common post-processing tools was successful in producing results enabling straightforward analysis and review.

n  All of the currently implemented methods should continue to be refined and improved to provide a variety of options for users and to capture epistemic uncertainty.

n  Four methods, EXSIM, G&P, SDSU and UCSB were found to be suitable for simulation of spectral acceleration from T=0.01 to 3 s over R= 0 to 200 km within the validation magnitude range (Mw 5.9-7.2) for WUS events.

n  The methods are deemed suitable up to Mw 8 for purposes of assessing relative effects of changes in source geometry, rupture direction, presence of secondary slip on splays, hanging wall effects, etc. Additional work is needed for absolute amplitudes.

n  At T>1 s there is increased bias, and for T> 3 s there are significant deviations from GMPEs, based on WUS events.

Evaluation - Results

37


References n  Atkinson, G. M., D. M. Boore, K. Assatourians, K. Campbell and D. Motazedian (2009). A guide to differences between stochastic

point-source and stochastic finite-fault simulations, Bull. Seism. Soc. Am. 99, 3192-3201. n  Beresnev, I., and G. Atkinson (1998a).FINSIM: a FORTRAN program for simulating stochastic acceleration time histories from finite

faults, Seism. Res. Lett. 69, 27–32. n  Beresnev, I., and G. Atkinson (1998a).FINSIM: a FORTRAN program for simulating stochastic acceleration time histories from finite

faults, Seism. Res. Lett. 69, 27–32. n  Boore, D. M. (2005). SMSIM--Fortran Programs for Simulating Ground Motions from Earthquakes: Version 2.3--A Revision of OFR

96-80-A , U.S. Geological Survey Open-File Report n  Boore, D. M. (2009). Comparing stochastic point-source and finite-source ground-motion simulations: SMSIM and EXSIM, Bull.

Seism. Soc. Am. 99, 3202-3216. n  Brune, J. N. (1970). Tectonic stress and the spectra of seismic shear waves from earthquakes, J. Geophys. Res., 76, 5002. n  Graves, R. A. Pitarka (2010). Broadband ground motion simulation using hybrid approach, Bulletin of Seismological Society of

America. Bull. Seism. Soc. Am., 100, 5A, 2095-2123. n  Irikura, K. and H. Miyake (2010). Recipe for Predicting Strong Ground Motion from Crustal Earthquake Scenarios, Pure and Applied

Geophysics, DOI 10.1007/s00024-010-0150-9. n  Liu, P., R. J. Archuleta and S. H. Hartzell (2006). Prediction of broadband ground-motion time histories: Hybrid low/high-frequency

method with correlated random source parameters, Bull. Seismol. Soc. Am., 96, 2118-2130, doi: 10.1785/0120060036. n  Mai, P.M., Imperatori, W., Olsen, K.B. (2010), Hybrid Broadband Ground-Motion Simulations: Combining Long-Period Deterministic

Synthetics with High-Frequency Multiple S-to-S Backscattering, BSSA, 100(5A), pp. 2124-2142. n  Mena, B. and Mai, P. M., Olsen, K. B., Purvance, M. D. and Brune, J. N. (2010). Hybrid Broadband Ground-Motion Simulation Using

Scattering Green's Functions: Application to Large-Magnitude Events, Bull. Seism. Soc. Am. 100, 2143-2162. n  Motazedian, D., and G. M. Atkinson (2005). Stochastic finite-fault modeling based on a dynamic corner frequency, Bull. Seismol.

Soc. Am. 95, 995 – 1010. n  point source in multilayered media, Geophysical Journal International, Volume 148, Issue 3, pp. 619-627. n  Schmedes, J., R. J. Archuleta, and D. Lavallée (2010). Correlation of earthquake source parameters inferred from dynamic rupture

simulations, J. Geophys. Res., 115, B03304, doi:10.1029/2009JB006689. n  Schmedes, J., R. J. Archuleta, and D. Lavallée (2010). Dependency of supershear transition and ground motion on the

autocorrelation of initial stress, Tectonophysics, 493, 222-235, doi: 10.1016/j.tecto.2010.05.013 n  Schmedes, J., R. J. Archuleta, and D. Lavallée, (2012). A kinematic rupture model generator incorporating spatial interdependency of

earthquake source parameters, Geophys. J. Int., doi: 10.1093/gji/ggs021 n  Somerville, P. G., Callaghan, S., Maechling, P., Graves, R. W., Collins, N., Olsen, K. B., Imperatori, W., Jones, M., Archuleta, R.,

Schmedes, J., And Jordan, T.H. (2011). The SCEC Broadband Ground Motion Simulation Platform, SRL, 82(2), p. 275, 10.1785/gssrl.82.2.273.

n  Zeng, Y., J. G. Anderson and G. Yu (1994). A composite source model for computing realistic synthetic strong ground motions, Geophysical Research Letters 21, 725-728.

n  Zhu, L. & Rivera, L. A. (2002) A note on the dynamic and static displacements from a

scec broadband platform (bbp) simulation methods validation for

Documents