ground motion simulation

Ground Motion Simulation

Case Study: 1906 San Francisco Earthquake

Brad Aagaard

November 2, 2007

Project Personnel

Collaboration among scientists from USGS and 4 other institutions

USGS Menlo Park Brad Aagaard, John Boatwright, Thomas Brocher,Howard Bundock, Russell Graymer, Thomas Holzer, RobertJachens, Dave Keefer, Linda Seekins, Robert Simpson, CarlWentworth, Mary Lou Zoback

USGS Golden Stephen Harmsen, Stephen Hartzell

Stanford University Greg Beroza, Paul Segall, Seok Goo Song

Lawrence Livermore National Laboratory Shawn Larsen, KathleenMcCandless, Stefan Nilsson, Anders Petersson, Arthur Rodgers,Bjorn Sjogreen, Hrvoje Tkalcic

URS Pasadena Robert Graves

UC Berkeley David Dolenc, Doug Dreger1

Outline

• Overview of SF06 simulation project

• Validation with 1989 Loma Prieta earthquake

• Comparison of MMI: synthetics versus instrumental (ShakeMap)• Comparison of velocity waveforms: synthetics versus observed

• Simulations of the 1906 San Francisco earthquake

• Comparison of MMI: synthetics versus Boatwright’s ShakeMap• Ground motions from 1906 and scenario events• Response of 20-story SMRF buildings (brittle vs. ductile welds)

• Data availability

2

The SF06 Simulation Project

Objectives:

• Estimate ground motions for the 1906 earthquake and similarhypothetical events on the San Andreas fault

• Examine the impact if they happened today

3

Project Plan

Wald et al. (1991)

Beroza (1991)

Source Models

ground motionsRecorded

ShakeMap

Constraints

(validation)1989 M6.9 Loma Prieta

GeologicStructure

Ground MotionSimulations

Song et al.

Source Models

Boatwright et al.ShakeMap

Constraints

Earthquake EffectsModeling

1906 M7.9 San Francisco

4

Jachens et al. 3-D Geologic Model

Unified representation of fault surfaces and lithologies

• Fault surfaces and lithologic boundaries

• Active and inactive faults• Depositional surfaces and unconformities• Topography & bathymetry

• Hierarchical structure (how to assemble blocks from surfaces)

• Easy to refine/update model• Easy to extract subsets of features

• Constructed in Earth Vision (Dynamic Graphics)

5

Geologic Model: Geographic Coverage

Detailed model is surrounded by low-resolution, simple model

-126˚ -125˚ -124˚ -123˚ -122˚ -121˚ -120˚ -119˚35˚

36˚

37˚

38˚

39˚

40˚

41˚

0 100 200km

Detailed Velocity Model

Regional Velocity Model

6

Brocher et al. 3-D Seismic Velocity Model

Create seismic velocity model from geologic model

• Assign material properties to lithologies in geologic model

• Develop regressions based on variety of data• Check against tomographic models

• Given longitude/latitude/elevation return material properties

• Vp• Vs• Density• Qp• Qs• Lithology & depth from free surface

7

Ground Motion Modeling Groups

Five groups using variety of domains with different features

-126˚ -125˚ -124˚ -123˚ -122˚ -121˚ -120˚ -119˚35˚

36˚

37˚

38˚

39˚

40˚

41˚

0 100 200km

Harmsen et al.

Aagaard

Larsen et al.G

raves

Petersson et al.

8

Numerical Models

Same general features but minor implementation differences

• Similarities

• Solve dynamic elasticity equation for rupture on finite-fault• Include 3-D variations in physical properties

• Differences

• Simulation domains• Minimum period and minimum shear-wave speed• Spatial & temporal discretization schemes• Attenuation: values and implementation• Accommodating topography

9

Reality Check: 1989 Loma Prieta Eq

Gauge level of accuracy of ground motion modeling

• Simulate earthquake with Wald and Beroza source models

• Compare wave propagation implementation across modelers

• Are source models closer than wave propagation implementations?• What wave propagation implementations match observations?

• Test geologic and seismic velocity models

• What areas seem to be well-characterized by models?• What areas seem to be poorly-characterized by models?

10

MMI: Synthetics (Beroza Source) versus Instrumental

Synthetics capture large length-scale variations in shaking

Graves (f < 1.0 Hz) ShakeMap

-123˚ -122˚ -121˚

37˚

38˚

0 50km

IIIIIIIVVVIVIIVIIIIXX

MMI

San Francisco

San Jose

Santa Cruz

Livermore

Concord

Monterey

Hollister

-123˚ -122˚ -121˚

37˚

38˚

0 50km

IIIIIIIVVVIVIIVIIIIXX

MMI

San Francisco

San Jose

Santa Cruz

Livermore

Concord

Monterey

Hollister

11

MMI Residuals: Graves Synthetic-Observed

Long-period synthetics underpredict MMI due to limited bandwidth

-123˚ -122˚ -121˚

37˚

38˚

0 50km

-2

-1

0

1

2

MM

I(sy

n)-M

MI(

obs)

San Francisco

San Jose

Santa Cruz

Livermore

Concord

Monterey

Hollister

0

100

200

300

400

500

Cou

nt

-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0MMI(syn)-MMI(ref)

mean = -1.27std dev. = 0.66

12

MMI Residuals: Graves Broadband Synthetic-Observed

Broadband synthetics give reasonable fit to MMI (PGV/PGA)

-123˚ -122˚ -121˚

37˚

38˚

0 50km

-2

-1

0

1

2

MM

I(sy

n)-M

MI(

obs)

San Francisco

San Jose

Santa Cruz

Livermore

Concord

Monterey

Hollister

0

100

200

300

400

500

Cou

nt

-3.0 -2.5 -2.0 -1.5 -1.0 -0.5 0.0 0.5 1.0 1.5 2.0 2.5 3.0MMI(syn)-MMI(ref)

mean = -0.34std dev. = 0.56

13

Waveforms at AGNW: Modelers vs. Observed

Reasonable match in amplitude & duration for both source models

Time (s)

Vel

ocity

(m

/s)

AGNW

East Component

5 10 15 20 25 30

−0.2

0.0

0.2

Time (s)

North Component

5 10 15 20 25 30

−0.2

0.0

0.2

Time (s)

Up Component

5 10 15 20 25 30

−0.2

0.0

0.2

ObservedGraves [Beroza]Harmsen [Beroza]Dolenc [Beroza]Aagaard [Beroza]Graves [Wald]Harmsen [Wald]Dolenc [Wald]Aagaard [Wald]

0 20 40

km

AGNW

14

Waveforms at WATS: Modelers vs. Observed

Beroza source fits amplitude & duration better but not polarity

Time (s)

Vel

ocity

(m

/s)

WATS

East Component

5 10 15 20 25 30

−0.2

0.0

0.2

Time (s)

North Component

5 10 15 20 25 30

−0.2

0.0

0.2

Time (s)

Up Component

5 10 15 20 25 30

−0.2

0.0

0.2

ObservedGraves [Beroza]Harmsen [Beroza]Dolenc [Beroza]Aagaard [Beroza]Graves [Wald]Harmsen [Wald]Dolenc [Wald]Aagaard [Wald]

0 20 40

km WATS

15

Summary of Loma Prieta Modeling

3-D geologic model successfully captures important 3-D effects

• We reproduce the duration and amplitude of shaking within the SanFrancisco Bay area

• Good agreement among wave propagation implementations (modelers)

• Better agreement among modelers than among source models• Finite-element implementation does better for vertical component

• We can tell the difference b/t the Beroza and Wald source models

• Wald source model radiates energy primarily to the north• Beroza source model radiates energy bilaterally

• Identified locations where velocity model may need adjustment

• Vs in La Honda basin too slow• Vs in Great Valley sequence too slow

16

1906 and Scenario Earthquakes

Simulations of the 1906 earthquake and hypothetical variations

• 1906 earthquake

• USGS 3-D geologic and seismic velocity models• Song et al. source model

• 1906-like scenarios

• Other hypocenters• Rockport: north to south rupture• Bodega Bay: bilateral rupture w/central hypocenter• San Juan Bautista: south to north rupture

• Random slip• Random realization with different distribution than 1906

17

Song et al. 1906 Source Model

Construct updated source model of the 1906 earthquake

• Find unique source model that satisfies all datasets

• Geodetic (triangulation data)• Seismic (teleseismic waveforms)

• Constrain distribution of slip and rupture duration

18

Song et al. Source Model: Inversion Results

Inversion constrains slip and rupture duration

19

Source Implementation: Slip and Rupture Time

Roughen up slip distribution and rupture propagation

20

Comparison of MMI with Boatwright & Bundock

Simulations and intensity map agree on major features

Aagaard (f < 0.5 Hz) Boatwright & Bundock

-123˚ -122˚ -121˚

37˚

38˚

39˚

0 50 100km

IIIIIIIVVVIVIIVIIIIXX

MMI

1906San Francisco

San Jose

Santa Rosa

Sacramento

-123˚ -122˚ -121˚

37˚

38˚

39˚

0 50 100km

IIIIIIIVVVIVIIVIIIIXX

MMI

1906San Francisco

San Jose

Santa Rosa

Sacramento

21

Misfit in Boatwright & Bundock Intensities

Simulations have larger intensities near rupture

-123˚ -122˚ -121˚

37˚

38˚

39˚

0 50 100km

-3

-2

-1

0

1

2

3M

MI(

syn)

-MM

I(re

f)

1906San Francisco

San Jose

Santa Rosa

Sacramento

22

Shaking in San Jose

23

Shaking in Santa Rosa

24

Waveforms (T > 2 s) in San FranciscoV

eloc

ity (

m/s

) East Component1906

0 20 40 60 80 100

−0.3 0.0 0.3

North Component

0 20 40 60 80 100

Up Component

0 20 40 60 80 100

Larsen et al.Petersson et al.GravesAagaardHarmsen et al.

Vel

ocity

(m

/s) RandomHypo06

0 20 40 60 80 100

−0.3 0.0 0.3

0 20 40 60 80 100 0 20 40 60 80 100

Vel

ocity

(m

/s) 1906HypoC

20 40 60 80 100 120

−0.3 0.0 0.3

20 40 60 80 100 120 20 40 60 80 100 120

Time (s)

Vel

ocity

(m

/s) RandomHypoC

20 40 60 80 100 120

−0.3 0.0 0.3

Time (s)20 40 60 80 100 120

Time (s)20 40 60 80 100 120

25

Waveforms (T > 2 s) in San JoseV

eloc

ity (

m/s

) East Component1906

20 40 60 80 100 120

−0.3 0.0 0.3

North Component

20 40 60 80 100 120

Up Component

20 40 60 80 100 120

Larsen et al.Petersson et al.GravesAagaardHarmsen et al.

Vel

ocity

(m

/s) RandomHypo06

20 40 60 80 100 120

−0.3 0.0 0.3

20 40 60 80 100 120 20 40 60 80 100 120

Vel

ocity

(m

/s) 1906HypoC

40 60 80 100 120 140

−0.3 0.0 0.3

40 60 80 100 120 140 40 60 80 100 120 140

Time (s)

Vel

ocity

(m

/s) RandomHypoC

40 60 80 100 120 140

−0.3 0.0 0.3

Time (s)40 60 80 100 120 140

Time (s)40 60 80 100 120 140

26

Waveforms (T > 2 s) in LivermoreV

eloc

ity (

m/s

) East Component1906

20 40 60 80 100 120

−0.4 0.0 0.4

North Component

20 40 60 80 100 120

Up Component

20 40 60 80 100 120

Larsen et al.Petersson et al.GravesAagaardHarmsen et al.

Vel

ocity

(m

/s) RandomHypo06

20 40 60 80 100 120

−0.4 0.0 0.4

20 40 60 80 100 120 20 40 60 80 100 120

Vel

ocity

(m

/s) 1906HypoC

40 60 80 100 120 140

−0.4 0.0 0.4

40 60 80 100 120 140 40 60 80 100 120 140

Time (s)

Vel

ocity

(m

/s) RandomHypoC

40 60 80 100 120 140

−0.4 0.0 0.4

Time (s)40 60 80 100 120 140

Time (s)40 60 80 100 120 140

27

Comparison with NGA Models, Soft Soil Sites

Super-shear rupture creates discrepancies at distances of 10–100 km

T = 3.0 s T = 5.0 s

0.01

0.1

1

Spe

ctra

l acc

eler

atio

n (g

)

1 10 100Distance to rupture (km)

SyntheticsCampbell and Bozorgnia (2007)Chiou and Youngs (2006)

0.01

0.1

1

Spe

ctra

l acc

eler

atio

n (g

)

1 10 100Distance to rupture (km)

28

Comparison with NGA Models, Hard Rock Sites

Hard rock sites more sensitive to directivity and super-shear

T = 3.0 s T = 5.0 s

0.01

0.1

1

Spe

ctra

l acc

eler

atio

n (g

)

1 10 100Distance to rupture (km)

SyntheticsCampbell and Bozorgnia (2007)Chiou and Youngs (2006)

0.01

0.1

1

Spe

ctra

l acc

eler

atio

n (g

)

1 10 100Distance to rupture (km)

29

Implications for Tall Buildings

Anna Olsen et al. study with 20-story SMRF buildings

• 20-story, steel, special moment-resisting frame buildings

• UBC 1994• JBC 1992 (satisfies 1997 UBC static lateral force requirements)• Brittle welds (distribution of yield strain for welds)• Ductile welds (welds do not fracture)

• Nonlinear, 2-D finite-element model

• Fiber elements use nonlinear, hysteretic, steel model• Panel zones use nonlinear, hysteretic, moment-shear strain model• Only deterioration mechanism is weld fracture

• Evaluate performance using peak inter-story drift ratio

30

Peak Interstory Drift: 1906 Earthquake

Tall buildings with brittle welds in SF & SCV would be vulnerable

31

Peak Interstory Drift: Bodega Bay epicenter

Even some tall buildings without brittle welds would be vulnerable

32

Peak Interstory Drift: San Juan Bautista epicenter

Tall buildings with brittle welds in SF & SCV would be vulnerable

33

Implications for Tall Buildings: Summary

Buildings with brittle welds could collapse in large SAF earthquakes

% Urban Area with Simulated CollapsesScenario JBC UBC

Brittle Ductile Brittle DuctileLoma Prieta 0 0 0 01906 0.31 0 0.83 0Bodega Bay 1.1 0.061 6.7 0.21San Juan Bautista 0.031 0 0.64 0

34

Ground Motions: Data Availability

Data will be accessible so anyone can use it

• PGV/PGA: 1/60 degree uniform lat/lon grid

• Velocity waveforms (resolution varies with population density)

• Boatwright intensity sites• Census track centroids (minimum dist b/t points is 2 km)

• Will be distributed as USGS Data Series

• Data has been collected• Time/workload is limiting factor in publishing Data Series• Have capable computer scientist contractor, but no funding

35

Future Work

Ground motions from scenario events involving Hayward fault

• Similar collaborative effort to 1989/1906 simulations

• Expect to produce around 30 scenarios

• 5 combinations of Rodgers Creek, Hayward, Calaveras• 1-3 hypocenters for each rupture length• 3 slip distributions for each hypocenter• Hayward: vary average rupture speed and rise time

• Use latest tools for generating kinematic rupture models

• Target completion date: June 2008

36