Over- or Under-detection: How do Models of Paleoseismic Rupture Detectability Affect Estimates of Earthquake Probabilities?

Abstract:

Paleoseismic records are often considered incomplete due to the difficulty of detecting ruptures in trenches. Events become increasingly difficult to detect as slip decreases, large earthquakes may be missed if the rupture did not pass through the trench, and it is difficult to differentiate between events that occur very close in space and time. One might assume that these difficulties would result in missing or fewer events. Alternatively, misinterpretation of paleoseismic features could result in over-counting, which may be a more significant problem. We test the effects of different models of detectability in paleoseismic studies on the probabilities of large earthquakes in California. We employ the 3D boundary element code RSQSim with a new California fault model, based on the UCERF3 report, to generate synthetic catalogs with millions of events. The simulations incorporate rate-state fault constitutive properties, in complex, fully interacting fault systems. Our catalogs are tuned to match the recurrence intervals at the paleoseismic sites in the UCERF3 report by making adjustments to the normal stress in the model. We compare earthquake probabilities at paleoseismic sites using catalogs that were thinned, prior to tuning, based on different models of event detectability. The first catalog was tuned assuming 100% detectability. The second catalog was thinned using the UCERF3 [Appendix I] probability model of detection, which is based on amount of observed slip at each site. The third catalog was thinned based on a model of detectability that assumes fewer detectable events than given by the UCERF3 model. Finally, the fourth catalog was thinned based on the UCERF3 probability model, but tuned to 25% longer mean recurrence intervals. Comparisons of the different catalogs suggest that the mean recurrence intervals from paleoseismic trenches may be too short, resulting in over-estimation of probabilities.