S33C-4545:
Implementation of GNSS-Constrained Real-Time Finite Fault Modeling for Improved Earthquake Early Warning: Current Activities at USGS, Menlo Park

Wednesday, 17 December 2014
Deborah E. Smith1, Sarah E Minson2, John O Langbein1, Jessica R Murray1 and Christian Guillemot1, (1)US Geological Survey, Earthquake Science Center, Menlo Park, CA, United States, (2)California Institute of Technology, Seismological Laboratory, Division of Geological and Planetary Sciences, Pasadena, CA, United States
Abstract:
Currently implemented Earthquake Early Warning (EEW) algorithms based on seismic data alone should provide the most robust warnings for most M<6 earthquakes, since real-time GPS positions are too noisy to aid in EEW. However, for larger events, which generate larger fault offsets, GPS data can provide a direct on-scale displacement measurements and has sufficient precision. In such situations, the GPS observations may enable more accurate estimation of magnitude and rupture extent than seismic data.

The USGS Earthquake Science Center in Menlo Park currently obtains real-time data from approximately 100 GNSS stations in northern California. These stations, which span the San Andreas fault system from the Mendocino Triple Junction to San Juan Bautista, are operated by USGS-Menlo Park, UC Berkeley, and UNAVCO. We have developed software tools for monitoring and troubleshooting data acquisition and quality. We have evaluated the latency and precision of position estimates obtained through real-time processing and we have found these results satisfactory for EEW.

We are now implementing the BEFORES algorithm (Minson et al., 2014) that uses Bayesian analysis to determine the best-fitting coseismic fault orientation and finite fault slip distribution (from which moment and rupture extent are obtained) in real-time. BEFORES has been tested extensively on both simulated and real data (retrospectively) for a variety of earthquakes. We are now focusing on three aspects of its implementation: 1) receiving real-time earthquake locations from independent seismic EEW algorithms, that are obtained through multiple TCP/IP connections, and 3) optimizing the computation of elastic Green’s functions. Completion of these tasks plus additional tests using simulated waveforms of earthquakes displacements superimposed on actual data will prepare the algorithm for implementation in the West Coast EEW system.