Regionally Optimized Background Earthquake Rates from ETAS (ROBERE) for Probabilistic Seismic Hazard Assessment

Date: 4/24/2019

Time: 06:00 PM

Room: Grand Ballroom

U.S. Geological Survey probabilistic seismic hazard assessments (PSHA) often use background earthquake rate estimates determined by spatially smoothing a single declustered earthquake catalog from which aftershocks have been removed using the method of Gardner and Knopoff (hereafter GK74; 1974). Declustering seeks to remove aftershocks and leave behind a suite of independent, or background, events that can be modeled as a Poisson process. We expect declustering to reduce the estimated earthquake rates because it removes earthquakes from the catalog, but this is not always the case. During development of the 2018 1-year PSHA model for the Central and Eastern United States (Petersen et al., 2018), the total number of M≥3 earthquakes in Oklahoma and Kansas decreased from 2016 to 2017 but the number of events in the declustered catalog using GK74 rose. This can partially be explained by a number of spatially scattered events in western Oklahoma, but Petersen et al. also noted that GK74 may not work well in regions of induced seismicity. There are more general problems. GK74 and many declustering methods reduce the Gutenberg-Richter b-value. While declustering lowers the overall number of earthquakes, reducing the Gutenberg-Richter a-value, the estimated rate of large earthquakes can be higher after declustering due to the extrapolation to larger magnitudes using an artificially low b-value. We seek to improve these hazard assessments with a method for determining Regionally Optimized Background Earthquake Rates from ETAS (ROBERE). ROBERE produces a spatially-smoothed background earthquake rate model by applying existing spatial smoothing techniques to a suite of declustered catalogs, each of which was produced with the stochastic Epidemic Type Aftershock Sequence (ETAS) (Zhuang et al., 2002) declustering method. This method optimizes parameters for the region being studied, includes parameter uncertainty, and has the advantage of preserving the original b-value.

Presenting Author: Andrea Llenos

Authors