California Regional Adjustments to Ground Motion Models Used in Earthquake Early Warning Algorithms
Date: 4/25/2019
Time: 06:00 PM
Room: Grand Ballroom
Source-based earthquake early warning systems (EEWs), such as ShakeAlert EEW for the West Coast of the US, rely on ground-motion models (GMMs) to predict the expected ground motion at a user’s location. We explore how (or whether) regional adjustments to GMMs improve ground-motion estimates, and hence the accuracy of EEW alerts, in California. Our approach is to first analyze the residuals (difference) between ground-motion data and GMMs, and then decompose those residuals into contributions of source, site, and path. Specifically, we propose a novel way to define the event term, using only records close to the event, as farther data is more representative of the path attenuation rather than the source.
We use the Abrahamson et al. (2014; ASK14) NGA-West2 compared to NGA-West2 data only in California, consisting of 374 earthquakes of M3 to M7.3, a total of 15,178 records. We consider only residuals in close proximity to determine event terms, inherently different from what is currently practiced. We look for events that were recorded on at least three stations within a 20 km radius, and increasing the radius until five station recordings are available, out to a maximum of 40 km. The goal is to use enough data to distinguish between event and site residuals but with close proximity. Both the Bay Area and Southern California exhibit regional trends; therefore, to improve the GMM, we implement regional corrections.
To test this improved GMM in an EEW application, we apply the Aagaard et al. (2018) cost-benefit matrix – which tells us how accurate the GMM is for early warning purposes – on several recent large California earthquakes not included in the NGA-West2 data set. We compare the predicted ground motion to ShakeMaps (considered ground truth) to determine the improvement with the adjusted GMM. We find the corrected GMM more accurately reflects the observed ShakeMap, showing that regional corrections can improve earthquake early warning alerts.
Presenting Author: Lauren Abrahams
Authors
Lauren Abrahams labraha2@stanford.edu Stanford University, Stanford, California, United States Presenting Author
Corresponding Author
|
Annemarie S Baltay abaltay@usgs.gov U.S. Geological Survey, Menlo Park, California, United States |
Brad Aagaard baagaard@usgs.gov U.S. Geological Survey, Menlo Park, California, United States |
California Regional Adjustments to Ground Motion Models Used in Earthquake Early Warning Algorithms
Category
Next Generation Earthquake Early Warning Systems: Advances, Innovations and Applications