Adapting ElarmS Earthquake Early Warnings for Cascadia: Development and Testing of ShakeAlerts in the Pacific Northwest

As a first step in establishing an earthquake early warning system in Cascadia, we have installed the ElarmS component of the ShakeAlert system at the Pacific Northwest Seismic Network. In Cascadia our initial focus is primarily on the development of a seismo-geodetic-based real-time finite fault rupture algorithm to detect and characterize a large plate-boundary rupture in progress (see Crowell et. al., this session). In this regard the goal of the purely seismic-data-based ElarmS implementation is to 'trigger' the finite fault rupture algorithm. At the same time, however, the Cascadian ElarmS will also produce warnings for smaller onshore crustal earthquakes. While warnings from these smaller and closer earthquakes will provide shorter warning times for communities, and for less dramatic earthquakes, we intend to use them for educational purposes, and to coordinate with our regional and collaborating partners. They will also help to guide us to shorten data latencies and learn where additional instrumentation is most needed to increase performance. The accuracy of ElarmS in Cascadia is another major concern, because the current ElarmS model presumes an initial focal depth for earthquakes of 8 km based on California experience, while in Cascadia earthquakes of major concern may be as deep as 50 km, and/or occur beyond the western fringe of the seismic network. To this purpose our testing protocol is aimed at determining what changes are required to ensure top performance of an ElarmS-based warning system in Cascadia. Because of Cascadia's relatively low seismicity rate, and the paucity of data from plate boundary earthquakes there of any size, we have prioritized the development of a test system. The test system permits us to: 1) replay segments of actual seismic waveform data recorded from the PNSN and contributing seismic network stations to represent both earthquakes and noise conditions, and 2) broadcast synthetic data into the system to simulate signals we anticipate from earthquakes for which we have no actual ground motion recordings. The test system lets us also simulate various error conditions (latent and/or out-of-sequence data, telemetry drop-outs, etc.) to explore how to protect the system from them. We have also been testing the ElarmS system on real-time seismic network data for about 6 months as of the time of writing of this abstract. Using 268 channels of streaming strong motion and broad-band data, the system has produced very few false alarms and generally performed well for earthquakes between about magnitudes 2.5 and 4.5. Warning times are shorter (and the 'blind zone' smaller) in parts of the network where station density is higher and/or telemetry more fleet. One significant problem we find is that the discriminant used in northern California to differentiate local earthquake signals from teleseisms often fails in Cascadia. We are working to produce a valid teleseism detector.