UC Berkeley

Timely alerts sent through earthquake early warning (EEW) programs allow those alerted to take protective actions to mitigate their risk from potentially damaging shaking. Over the past few years, ShakeAlert, the EEW program focused on the west coast of the con-tiguous United States, has grown, alerting communities within California, Oregon, and Washington about earthquakes where damaging shaking is expected. ShakeAlert uses a set of algorithms including the point-source algorithm, earthquake point-source integrated code (EPIC), to determine the location, magnitude, and origin time of potential earthquakes. Although EPIC produces low-latency and low error solutions for many events originating within the seismic network on land, numerous recent small earthquakes rupturing offshore of northern California have EPIC location solutions with high error ( > 50 km compared to USGS locations). Because most events are occurring offshore, there is a lim-ited number of stations that can trigger and contribute information in a timely manner for use in EEW. To better constrain location solutions in this region, we propose to include information about contemporary past seismicity into EPIC’s grid-search algorithm through a Bayesian framework. This prior information layer downweights high error locations where EPIC’s proposed event location coincides with an area of low prior seismicity in pref-erence for locations with a similar level of data fit that also have higher past seismicity. This addition to EPIC lowers the mean location error offshore northern California from 58 to 14 km.