MMI-Based Site Amplification Factors for USGS Earthquake Early Warning Algorithms

Abstract:

We develop empirical, intensity-based site corrections using recorded earthquake ground motions from Southern California for application to earthquake early warning (EEW) and consider their relationship to V_S30. Propagation of Locally Undamped Ground Motion (PLUM) is an EEW system currently operational in Japan and under development in the U.S. by the USGS for use on the West Coast (Kodera et al. 2018; Cochran et al. 2019). Rather than using earthquake source information to estimate an alerting area, PLUM uses observed ground motion to directly estimate shaking in the surrounding area. To improve the timeliness and accuracy of alerts, site corrections should be applied to the stations used in the algorithm to account for local amplification of ground motion. Site corrections are often approximated using V_S30 (the time-averaged shear wave velocity in the top 30m of the crust) but this method suffers from non-uniqueness, as local ground motion can be affected by a variety of seismic phenomena occurring between the surface and seismic bedrock, and does not uniquely correlate with average shear-wave velocity. Instead, we compute site-specific correction factors in Modified Mercalli Intensity (MMI) units for compatibility with USGS EEW algorithms. We use peak ground velocity (PGV), peak ground acceleration (PGA), rupture distance, and magnitude to compute MMI (Worden et al. 2012) from a database of 46,657 instrumental ground-motion records of M ≥ 3 earthquakes in Southern California. We use an intensity prediction equation (Atkinson et. al 2014) to predict the MMI for each record using only magnitude and rupture distance, and then compute residuals. We then partition these residuals into site, event and epsilon (error) terms using a linear mixed-effects analysis. The site terms represent empirical amplification factors for each site, and we find that site response can modify ground motion by up to 2 MMI units at some stations. We implement the site terms in the PLUM algorithm to quantify any changes to alerting performance for earthquakes in Southern California. Lastly, we use our MMI site terms to develop a generalized relationship with V_S30 for forward predicting the site response at stations not considered in our study and also for use in the ShakeAlert grid product, which is not station-based like PLUM.