HR: 1330h
AN: S42D-0198    [PDF]
TI: Estimating the Variations in the Finite-Fault Source Models Determined from Inversion of Ground Motion Data
AU: * Liu, p
EM: pcliu@crustal.ucsb.edu
AF: Institute for Crustal Studies, University of California, Santa Barbara, CA 93106
AU: Archuleta, R
EM: ralph@crustal.ucsb.edu
AF: Institute for Crustal Studies, University of California, Santa Barbara, CA 93106
AB: Although significant uncertainty is embodied in the source model deduced from the inversion of seismograms, it is unclear how much the uncertainty exists in practice. In general the uncertainty in the final solution can arises from every major element of the inversion procedure. This study primarily focuses on estimating the variation of finite fault solution induced by incomplete data set and by uncertainty in inversion process when the velocity structure is known. We apply a bootstrap statistical method to quantify this variation. The bootstrap method is a re-sampling procedure through which an arbitrarily large number of bootstrap data sets are generated from the original data. Having a bootstrap data set, we can determine a fault model. By statistically analyzing all the resulting solutions we can derive the variance and covariance of the best solution. To mitigate the large computational effort required by a global method to invert the data, a quasi-Newton method, based on the trust region algorithm, is used to invert the bootstrap data sets after the global inversion of the original data finds a starting model. As an application of this method we re-analyze the ground motion data from the 1989 Loma Prieta M 6.9.
DE: 7200 SEISMOLOGY
DE: 7203 Body wave propagation
DE: 7209 Earthquake dynamics and mechanics
DE: 7260 Theory and modeling
SC: Seismology [S]
MN: 2003 Fall Meeting