P E E R Ground Motions Basin Modeling

P E E R Ground Motions: Basin Modeling and Rapid Response Douglas Dreger Dept. of Earth and Planetary Science University of California, Berkeley 2002 PEER Annual Meeting

Outline • Validation of numerical codes for basin response modeling • Rapid near-fault strong ground motion estimation and reporting

Forecasting Ground Motions Assuming a kinematic source model and 3 D velocity structure we can simulate strong ground motion time histories. Simulated PGV • What is the variation in possible rupture kinematics? • How well known is the 3 D velocity structure? • How well calibrated are the numerical routines?

Numerical Code Validation • PEER/Lifelines funded research project to verify performance of finite-difference and finite-element algorithms for the simulation of time histories for 3 D velocity structure • Multi-Institutional Working Group: – Steve Day (SDSU) - Coordinating PI – Robert Graves & Arben Pitarka (URS) – Douglas Dreger & Shawn Larsen (UCB/LLNL) – Kim Olsen (UCSB) – Jacobo Bielak (CMU)

Summary of Project • Validation with simple geometries and sources – halfspace & layered velocity models – point-source and propagating-source models – anelastic loss • Validation with SCEC 3 D reference model – minimum Vs of 500 and 200 m/s tested • Northridge simulation and comparison with observations – Work is underway

Methods • 4 Finite Difference (structured regular grid) – UCSB, LLNL/UCB, URS – Main Variations: Parallelism Memory management Gridding strategies Surface Boundary conditions Absorbing boundaries Material interface treatment Source formulation • Finite Element (unstructured irregular grid) – CMU – Serial grid generation – Parallel execution via automated domain decomposition

• Low-velocity elastic layer • Propagating dislocation Problem 4

Problems 5 & 6 • SCEC So. Calif. Seismic Velocity Model • Problem 5: Min. S wavespeed 200 m/s • Problem 6: Min. S wavespeed 500 m/s

SCEC Model--Min. S velocity 500 m/s

Discretization of Shallow Velocity Profile Actual Profiles • very slow nearsurface velocity • strong nearsurface velocity gradients • Considerable site variability The various numerical codes differ in gridding strategy Resulting in quite different cellcentered sampling

Future Plans • Apply methodologies to Northridge to: – Validate source models – Validate wave propagation models – Characterize uncertainties • Document the results – Detailed problem descriptions are available by ftp – down-loadable waveforms will also be made available – Journal article • Proposed PEER/SCEC project to – application of the codes to both northern and southern California ground motion problems – further coordinate testing of modeling procedures

Rapid Response Ground Motions • PEER/Lifelines funded research project to develop a nearrealtime system for automated finite-source inversion and near -fault ground motion estimation • Berkeley Seismological Laboratory Team – Douglas Dreger, PI – Asya Kaverina, Pete Lombard, Lind Gee, and Douglas Neuhauser

Rapid Characterization of Strong Ground Shaking: Shake. Map

Station Coverage is a Concern • Shakemaps depend strongly on station coverage. – Robust in well instrumented regions – Dependent upon estimated ground motions in poorly instrumented regions • Regional variations in station coverage, and telemetry outages can impact the ability to rapidly characterize strong shaking levels

Northridge PGA Shake. Map in Deteriorating Coverage Tri. Net Stations 40 km spacing Very sparse Hypothetical case for Northridge but a real problem for other regions. Can we simulate near-fault strong ground motions in areas of sparse coverage?

Automated Processing Stream at the Berkeley Seismological Laboratory Hypocenter & magnitude Finite-fault 30 s-4 min 11 -20 min Ground-motion map Seismic Moment Tensor 6 -9 min 12 -35 min

Example of kinematic finite-source model for the Northridge earthquake

Shake. Map improvement with finitesource information Tri. Net Stations Raw Data Finite-source augmented 40 km spacing Very sparse

Hector Mine Earthquake: Tri. Net Shake. Maps Automatic Map Revised Map

Hector Mine Earthquake - Preliminary Result Comparison of PGV observations and predictions Combining observations & predictions to produce a Shake. Map

Comparison of Tri. Net and Our Shake. Maps

Hector Mine Earthquake - Revision Surface fault offset, GPS and SAR data also used Kaverina et al. (2002)

Hector Mine Kinematics

Hector Mine: Finite-source PGV Shake. Maps Automated Result Utilized seismic data only at realtime reporting stations Dreger and Kaverina (2000) Revised Result Utilized improved seismic coverage, surface fault offsets, GPS, and SAR data Kaverina et al. (2002)

Summary of Results • Our approach is designed to rapidly characterize near-fault strong ground motions in both well and poorly instrumented regions. • The 1999 Hector Mine earthquake provided a successful test of the concept. • The approach has been implemented as a processing stage of the Rapid Earthquake Data Integration (REDI) realtime processing system at the Berkeley Seismological Laboratory (BSL) – Currently being evaluated in a test mode • As implemented there are three Shake. Map stages: – Tri. Net Shake. Map (4 -6 minutes) – Directivity Corrected Shake. Map (~15 minutes) – Finite-source Augmented Shake. Map (~30 minutes) • We are working with the southern and northern California Shake. Map working groups towards integration of our method into the authoritative Shake. Map system for California