Event Date
Testing and Refining Earthquake Forecasts on Fault Networks
Morgan Page, U.S. Geological Survey, Pasadena CA
Modern seismic hazard models such as the 2023 National Seismic Hazard Model (NSHM23) represent fault-based earthquake sources on complex, interconnected fault networks. Some of these faults, such as the San Andreas fault in California, have low rates of recent seismicity compared to the geologically inferred rate of large earthquakes. This discrepancy led to the characteristic magnitude distribution hypothesis, which predicts a higher rate of large earthquakes than a Gutenberg-Richter extrapolation of the small-earthquake rate would predict (Schwartz and Coppersmith, 1984). We test this hypothesis by comparing the fault-based magnitude-frequency distributions from the NSHM23 model in the Western U.S. with observed seismicity over the past 93 years. We find that observations fall outside of the confidence limits predicted by the model in “characteristic” regions. These results challenge the widespread use of characteristic magnitude distributions in fault-based seismic hazard models and suggest, alternatively, large temporal changes in seismicity rates near faults.
We also present a Bayesian framework for inferring dates of last event (DOLEs) throughout a fault network from paleoseismic data. Given the long-term, time-independent rupture rates of an underlying model and paleoseismic DOLE probability density functions at trench sites, we calculate the posterior probability that a given fault section last co-ruptured with any subset of fault sections with paleoseismic data. The resulting posterior co-rupture probabilities are used to construct DOLE probability distributions throughout the fault system, including on faults that lack paleoseismic trenches. Furthermore, this method can be used to update DOLE distributions at the paleoseismic sites themselves, since nearby paleoseismic sites provide information about a site’s most recent event if there is a possibility of co-rupture. This work can be used to apply elastic rebound calculations in a consistent way throughout the fault system.