default

USGS Bay Area Seismic Velocity ModelConstruction and Earthquake Simulations

Brad AagaardCollaborators: Thomas Brocher, Robert Jachens, Robert Simpson

July 15, 2013

default

Outline

Overview of Earth Structure ModelsUSGS Bay Area Geologic and Seismic Velocity Models

Methodology of constructionQuerying the seismic velocity modelExamples from validationApplication to scenario earthquakesIdeas for improving the models

Introduction Overview

default

Earth Structure Models

Objective: Describe 3-D geologic structure

Geometry of faultsGeometry of major lithologic boundariesPhysical properties

Elastic propertiesAttenuation

Introduction Overview

default

Earth Structure ModelsUSR approach provides framework for integrating multiple geophysical datasets

Tomographic approachRelate seismic velocity to point in spacePermits arbitrarily complex variations in wave speedsDensity not independently constrained

Unified structural representation (USR) approachGeologic block model describes geometry (faults, layers, etc)Seismic velocity model relates elastic properties to geologyPermits sharp lithologic boundariesConstraints from surface traces, gravity, tomographyOften relies on rules to convert rock type to elastic properties

Introduction Overview

default

3-D Bay Area Earth Structure ModelsRegion of Coverage

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

36˚

37˚

38˚

39˚

40˚

41˚

42˚

0 100 200 300

km Bakersfield

Visalia

Fresno

Reno

South Lake

Coalinga

Sonora

Merced

Alturas

Susanville

Paso Robles

Modesto

Grass Valley

King City

Hollister

Sacramento

Chico

Monterey

San Jose

Redding

San Francisco

Yreka

Santa Rosa

Ukiah

Fort Bragg

Eureka

Crescent City

Detailed Model

Regional Model

Introduction Overview

default

3-D Bay Area Earth Structure Models: History

2005 Models developed for the 1906 earthquakeground-motion modeling

Detailed model for the Bay AreaCoarse resolution model for surrounding areaThurber tomographic model (much coarser)

2006 Initial analysis by Rodgers et al. and Kim and DregerShear wave speed about 5% too low in East BayDifficult to isolate regions needing improvement

2008 Minor updates to correct significant discrepanciesIncrease shear wave speed in East BayCorrect significant discrepancies with Thurbertomographic model

Introduction Overview

default

Bay Area Geologic Block ModelConstructed from geologic mapping, gravity, seismicity, etc using Earth Vision

23 faults29 lithologic units130 blocks

Introduction Block Model

default

Bay Area Geologic Block ModelVertical slice through Santa Clara Valley shows basin structure

Introduction Block Model

default

Geologic Block Model→ Seismic Velocity Model

GeologicBlock Model

SurfaceExtraction

Block ModelSurfaces

Lithologyto ElasticProperties

Depth

SeismicVelocityModel

RasterizeSeismicVelocity

Model File

Lon, Lat,Elev

Vp, Vs,Density,Qp, Qs

C/C++/Fortran

Introduction Seismic Velocity Model

default

Geologic Block Model→ Seismic Velocity ModelFranciscan (Foothills) elastic properties as a function of depth

Vp(km/s) =

a + 2.5 + 2.0d 0 ≤ d ≤ 1.0kma + 4.5 + 0.45(d − 1) 1.0km ≤ d ≤ 3.0kma + 5.4 + 0.0.0588(d − 3) 3.0km ≤ d

a = 0.13

density = 1.74Vp0.25

Vs(km/s) = 0.7858− 1.2344Vp + 0.7949Vp2

−0.1238Vp3 + 0.00064Vp4

Qs =

{−16 + 104.13Vs − 25.225Vs2 + 8.2184Vs3 Vs > 0.3km/s13 Vs ≤ 0.3km/s

Qp = 2Qs

Introduction Seismic Velocity Model

default

Seismic Velocity Model: Deep SedimentsDepth to Vs 2.5 km/s isosurface

-123˚ -122˚ -121˚

37˚

38˚

0 50km

0

1

2

3

4

Isos

urfa

ce D

epth

(km

)

San Francisco

San Jose

Santa Rosa

Livermore

Fremont

Napa

Concord

Gilroy

Sacramento

Stockton

Introduction Seismic Velocity Model

default

Seismic Velocity Model: Shallow SedimentsDepth to Vs 1.0 km/s isosurface

-123˚ -122˚ -121˚

37˚

38˚

0 50km

0.0

0.2

0.4

0.6

0.8

Isos

urfa

ce D

epth

(km

)

San Francisco

San Jose

Santa Rosa

Livermore

Fremont

Napa

Concord

Gilroy

Sacramento

Stockton

Introduction Seismic Velocity Model

default

Seismic Velocity Model: Spatial ResolutionFiner resolution near the surface

Detailed Model (213 million points, 8.2 GB)

Depth Horiz. Resolution Vert. Resolution0–0.4km 100m 25m0.4km–3.2km 200m 50m3.2km–6.4km 400m 100m6.4km–45km 800m 200m

Regional Model (155 million points, 6.0 GB)

Depth Horiz. Resolution Vert. Resolution0–6.4km 400m 100m6.4km–45km 800m 200m

Introduction Seismic Velocity Model

default

Seismic Velocity Model Stored as Etree DatabaseEfficient access to multi-resolution binary file via simple API

Etree database developed by Euclid Project at CMUStore data points as octree grid and order points in file usingtree structureSimple API to access Etree file

Very efficient access to variable resolution dataSet cache size for amount of model stored in memory

Wrap seismic velocity model API around Etree APIGeoreferencing and conversion among geographic projectionsRemove topography via flattening/bulldozingAnti-aliasing

Introduction Seismic Velocity Model

default

Validation via Ground-Motion Modeling

Waveform modeling of moderate and large earthquakes1989 M6.9 Loma Prieta (SF06 project)2007 M5.4 Alum Rock (Hayward08 project)2007 M4.2 Oakland (Hayward087 project)2008 M4.1 Alamo (Frankel)

PGV for 10 M4.1–5.4 earthquakes (Kim, Larsen, and Dreger)Travel-time for 12 M4.0–5.1 earthquakes (Rodgers et al.)

Validation Waveform Modeling

default

Validation: 2007 Mw 5.4 Alum Rock, T > 2.0sNeed to refine elastic properties near edge of Evergreen basin

0 10 20 30 40 50 60201510

505

101520

0 10 20 30 40 50 60201510

505

101520

0 10 20 30 40 50 60201510

505

101520

ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation Q32 (East San Jose Array Q32)

Validation Waveform Modeling

default

Validation: 2007 Mw 5.4 Alum Rock, T > 2.0sVelocity model nicely captures characteristics of Cupertino basin

0 10 20 30 40 50 60201510

505

101520

0 10 20 30 40 50 60201510

505

101520

0 10 20 30 40 50 60201510

505

101520

ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation H30 (South Saratoga Array H3)

Validation Waveform Modeling

default

Validation: 2007 Mw 5.4 Alum Rock, T > 2.0sVelocity model captures structure in southern end of Santa Clara Valley

0 10 20 30 40 50 6010

5

0

5

10

0 10 20 30 40 50 6010

5

0

5

10

0 10 20 30 40 50 6010

5

0

5

10ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation 2034 (Morgan Hill, El Toro Fire Station)

Validation Waveform Modeling

default

Validation: 2007 Mw 5.4 Alum Rock, T > 2.0sLivermore basin needs significant improvement

0 10 20 30 40 50 6054321012345

0 10 20 30 40 50 6054321012345

0 10 20 30 40 50 6054321012345

ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation CDOB (Doolan Road)

Validation Waveform Modeling

default

Validation: 2007 Mw 4.2 Oakland, T > 2.0sNo problem for short travel path along west side of Hayward fault

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation BRK (Berkeley, Haviland Hall)

Validation Waveform Modeling

default

Validation: 2007 Mw 4.2 Oakland, T > 2.0sVelocity model is missing shallow sediment in Oakland

0 10 20 30 40 50 602

1

0

1

2

0 10 20 30 40 50 602

1

0

1

2

0 10 20 30 40 50 602

1

0

1

2ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation 2190 (Oakland, Fire Station 29)

Validation Waveform Modeling

default

Validation: 2007 Mw 4.2 Oakland, T > 2.0sVelocity model captures main features of San Pablo basin

0 10 20 30 40 50 60

0.4

0.2

0.0

0.2

0.4

0.6

0 10 20 30 40 50 60

0.4

0.2

0.0

0.2

0.4

0.6

0 10 20 30 40 50 60

0.4

0.2

0.0

0.2

0.4

0.6ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation MAR (Mare Island)

Validation Waveform Modeling

default

Validation: 2007 Mw 4.2 Oakland, T > 2.0sVelocity model is too slow, but waveform amplitudes are close

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2

0 10 20 30 40 50 60

0.8

0.4

0.0

0.4

0.8

1.2ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation BRIB (Orinda, Briones Reserve)

Validation Waveform Modeling

default

Validation: 2007 Mw 4.2 Oakland, T > 2.0sVelocity model is slightly slow but under-predicts waveform amplitudes

0 10 20 30 40 50 602

1

0

1

2

0 10 20 30 40 50 602

1

0

1

2

0 10 20 30 40 50 602

1

0

1

2ObservedAagaard

Time (s)

Velo

city

(mm

/s)

East-West North-South VerticalStation 3992 (Pleasant Hill, Fire Station 2)

Validation Waveform Modeling

default

Validation: Peak Ground VelocityError in PGV increases significantly for f > 0.25 Hz (T < 4 s)

Kim, Larsen, and Dreger, BSSA, 2010Validation Peak Velocity

default

Seismic Hazard and 3-D Simulations3-D simulations allow more detail but require greater understanding

USGS National Seismic Hazard MapsEarthquake description: magnitude and fault boundaryGround motions from empirical regressions

Fault orientation, slip direction, dist. from faultPath and site corrections

Ground motion metrics: PGV, PGA, SA3-D ground-motion simulations

Earthquake description: earthquake rupture time historyComplex fault geometrySpatial and temporal evolution of slip

Ground motions from wave propagation3-D physical properties (basin effects)Rupture directivity

Displacement, velocity, and acceleration time histories

Validation Peak Velocity

default

Ground-Motion Simulations3-D ground-motion simulations can include rupture physics

Prescribed slip rupture modelsDriven by source inversions and spontaneous rupturesimulationsDeterministic + stochastic slip fieldsComplex nonplanar fault geometryComplex rupture pathsNot necessarily consistent with underlying physics

Spontaneous rupture modelsSlip evolves based on stress conditions and fault constitutivemodelDeterministic + stochastic stress fieldsComplex nonplanar fault geometryInvolves more parameters and knowledge of conditions in thelithosphere

Validation Peak Velocity

default

UCERF Bay Area Probabilities30 yr probability for Hayward / Rodgers Creek is now 31%

Hayward Fault Scenarios Introduction

default

Hayward FaultFault runs along the edge of the densely populated East Bay

Hayward Fault Scenarios Introduction

default

Hayward FaultFault runs underneath UC Berkeley’s Memorial Stadium

Hayward Fault Scenarios Introduction

default

Jim Lienkaemper’s Tyson’s Lagoon TrenchEvidence for 12 ruptures over the past 1900 years

Lienkaemper et al. USGS Open File Report 03-488Hayward Fault Scenarios Introduction

default

Paleoseismic Record at Tyson’s LagoonCurrently in middle of time window for next expected event

Hayward Fault Scenarios Introduction

default

Hayward Scenario Earthquakes Project

Compute ground motions for a suite of 39 scenarioearthquakes involving the Hayward fault

Rupture lengthHypocentersDistribution of slipRise timeRupture speed

Develop rupture models based on geophysical dataconsistent with NGA ground-motion prediction models

Spatial variation in slipSpatial variation in rise timeSlower rupture speed in areas with little slip

Account for aseismic creep in prescribed slip rupture models

Hayward Fault Scenarios Project Overview

default

Project PersonnelCollaborative effort to develop realistic ruptures and ground motions

USGS Menlo Park Brad Aagaard, John Boatwright, ThomasBrocher, Russell Graymer, Ruth Harris, ThomasHolzer, Dave Keefer, Jim Lienkaemper, David Ponce,David Schwartz, Robert Simpson, Paul Spudich,Janet Watt

Lawrence Livermore National Laboratory Shawn Larsen,Arthur Rodgers

URS Pasadena Robert Graves (now at USGS Pasadena)UC Berkeley Doug DregerStanford University Shuo Ma (now at SDSU)

Hayward Fault Scenarios Project Overview

default

Animation of Shaking Intensity

Hayward Fault Scenarios Ground Motions

default

Comparison with Boore-Atkinson NGA GMPEBias with NGA ground-motion prediction models increases with period

SA T=0.3 s SA T=1.0 s SA T=3.0 s

−123˚ −122˚ −121˚

37˚

38˚

39˚

0 50 100

km

−2

−1

0

1

2

log 2

(SA

T=

0.3s

syn

/ref

)

San Francisco

San Jose

Santa Rosa

Livermore

Fremont

Napa

Concord

Gilroy

Sacramento

Stockton

−123˚ −122˚ −121˚

37˚

38˚

39˚

0 50 100

km

−2

−1

0

1

2

log 2

(SA

T=

1.0s

syn

/ref

)

San Francisco

San Jose

Santa Rosa

Livermore

Fremont

Napa

Concord

Gilroy

Sacramento

Stockton

−123˚ −122˚ −121˚

37˚

38˚

39˚

0 50 100

km

−2

−1

0

1

2

log 2

(SA

T=

3.0s

syn

/ref

)

San Francisco

San Jose

Santa Rosa

Livermore

Fremont

Napa

Concord

Gilroy

Sacramento

Stockton

0

1000

2000

Cou

nt

−2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 2.0 2.5

log2(SA T=0.3s syn/ref)

mean = 0.10std dev. = 0.34

0

1000

2000

Cou

nt

−2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 2.0 2.5

log2(SA T=1.0s syn/ref)

mean = −0.03std dev. = 0.65

0

1000

2000

Cou

nt

−2.5 −2.0 −1.5 −1.0 −0.5 0.0 0.5 1.0 1.5 2.0 2.5

log2(SA T=3.0s syn/ref)

mean = 0.30std dev. = 0.78

Hayward Fault Scenarios Comparison w/NGA

default

Urban Seismic Hazard MapsHigh resolution alternative to National Seismic Hazard Maps

Reduce uncertainty in ground-motions by includingBasin amplificationRupture directivityComplex interaction between rupture directivity and basins

Gaining momentum in USGS and SCECRequires propagating uncertainties

Median values of most parameters are well-constrainedMore work needed to constrain probability distributions andincorporate them into models and simulations

Requires at least hundreds to thousands of simulationsStoring waveform output with proper metadata is challengingRequires better models of Earth structure andearthquake rupture

Moving Forward Future Directions

default

3-D Bay Area Earth Structure: Moving ForwardNeed to improve both the block model and the seismic velocity model

Refine geologic block modelFiner subdivision of lithologic units

Quaternary sediments (e.g., Bay Mud)Tertiary-Cenozoic sediments (e.g., east of Hayward fault)

Regional model is too simpleSan Andreas fault surfaceUpper crust, lower crust, mantle

Refine physical propertiesSystematic application of constraints from seismic data

Iterate on model with full waveform tomographyLocal analysis using dense arrays

Consistency with Vs30 modelsSmall scale (stochastic) variability

Moving Forward Improving Models

default

Improving the Seismic Velocity ModelExcellent coverage in urban area with current instrumentation

-124˚ -123˚ -122˚ -121˚ -120˚36˚

37˚

38˚

39˚

0 100

km

Coalinga

Sonora

Merced

Modesto

Grass Valley

King City

Hollister

Sacramento

Monterey

San Jose

San Francisco

Santa Rosa

Ukiah

Fort Bragg

Detailed ModelCGS

NCBKNSMP

-122˚37˚

38˚

0 50

km

San Jose

San Francisco

Santa Rosa

Moving Forward Improving Models

default

Improving the Seismic Velocity ModelM4+ seismicity (2000–present) is limited and concentrated in space

-124˚ -123˚ -122˚ -121˚ -120˚36˚

37˚

38˚

39˚

0 100

km

Coalinga

Sonora

Merced

Modesto

Grass Valley

King City

Hollister

Sacramento

Monterey

San Jose

San Francisco

Santa Rosa

Ukiah

Fort Bragg

Detailed Model

Moving Forward Improving Models