Assimilation of Paleoseismic Data for Earthquake Simulation

Simulation of the complete earthquake generation process requires assimilation of observational data on long-term behavior of faults over multiple rupture cycles. Paleoseismology, the study of pre-instrumental earthquakes from geological field investigations, provides the only observations of long-term fault behavior. Paleoseismic data present challenges for assimilation into models because data sets are small, sparse, analog, and contain qualitative uncertainties. Observations can be categorized as primary measurements from field data, or non-primary extrapolations or interpretations of primary data. Assimilation of non-primary data could yield biased simulation results. We present a primary data set for the San Andreas and Imperial faults for comparison with non-primary data in existing databases, and propose standard formats and structure for paleoseismic data assimilation into numerical simulation models .