Validation framework for semi-stochastic simulations in Cascadia earthquake early warning

We generate a synthetic earthquake dataset for the Cascadia region and present a framework for validating simulations for use in testing earthquake early warning (EEW) performance. The Cascadia Subduction Zone (CSZ) offshore western North America has hosted ~M9 earthquakes (the most recent being in 1700 C.E.), but few moderate-to-large earthquakes have been recorded here on seismic instruments. Synthetic seismic and geodetic data provide a useful supplement to the paucity of recorded data necessary for testing EEW; however, no published study to date has validated simulations and their output for such applications. We use a set of 1D semi-stochastic forward modeling codes to generate 112 M6.6–9.4 CSZ rupture scenarios and waveforms for 191 sites between Oregon and British Columbia. We validate the waveforms based on features critical to EEW infrastructure, which include event detection success, magnitude estimation, and ground-motion intensities. We also present an example performance test of the ShakeAlert EPIC and Ocean Networks Canada early warning algorithms using six of the simulated events. Through the validation process, we find the simulated data represent real earthquakes scenarios well and to be a valuable component in the training of EEW algorithms for the Cascadia region.