HR: 11:20h
AN: U42A-05 INVITED     [Abstracts]
TI: A Bayesian Approach to Correlation of Paleoearthquakes on a Fault
AU: * Biasi, G P
EM: glenn@seismo.unr.edu
AF: University of Nevada Reno, Seismological Laboratory MS-174, Reno, NV 89557 United States
AU: Weldon, R J
EM: ray@uoregon.edu
AF: University of Oregon, Dept. of Geological Sciences 1272, Eugene, OR 97403 United States
AB: We have developed an approach to paleoearthquake correlation for sites on the southern San Andreas fault in California. Paleoseismic sites are spaced kilometers to multiple tens of kilometers apart along the fault. One cannot normally say from paleoseismic dating alone whether a surface rupture spans multiple sites or whether the individual observations are from smaller earthquake ruptures that are merely close in time. Seismic hazard and thus societal consequences depend both on the magnitude and frequency of ground-rupturing earthquakes, so if ruptures correlate, the seismic hazard is weighted toward fewer, larger earthquakes, and if they do not, the hazard is caused by more numerous, somewhat smaller earthquakes. Our approach produces probability of correlation estimates by using additional physical measurements available from the paleoseismic investigation. The most powerful is the measurement of rupture displacement, d_obs, at a site. In outline the idea is that large point displacements generally go with longer ruptures, and longer ruptures have an increased likelihood of spanning from one site to the next. Mapped historical ruptures provide empirical relationships between average slip of a rupture, rupture length, and earthquake magnitude. Assessing the probability of correlation would be straightworward except that rupture displacement can vary along its length by up to a factor of 3 or so from its average, depending on geologic and possibly antecedent slip conditions. Especially for strike-slip faults one cannot say whether d_obs is from a peak for a smaller earthquake, or a low point from a large earthquake. However, the degree of variability of slip for small and large ruptures is usefully similar when normalized by the average slip for the rupture. This provides the basis for a Bayesian inversion for probabilities of rupture length and earthquake magnitude given the point measurement of displacement, p(L|d_obs) and p(M|d_obs), respectively. Geometric considerations are used to form a simple probability of correlation given length, p(correl|L). With p(L|d_obs) this leads to the probability of correlation given a point observation of displacement, p(correl|d_obs). Correlation probabilities decline with increasing site separation and decreasing observed displacement. For the southern San Andreas fault, large site spacings and the relatively small number of slip measurements are leading challenges to a more complete rupture history of the fault.
DE: 1105 Quaternary geochronology
DE: 7221 Paleoseismology (8036)
SC: Union [U]
MN: Fall Meeting 2005