Real-time modeling of GPS and accelerometer data for earthquake early warning and rapid hazard assessment

Real-time GPS networks provide the perfect complement to seismic networks, which operate with lower noise and higher sampling rates than GPS networks, but only measure accelerations or velocities, putting them at a disadvantage for ascertaining the full extent of slip during a large earthquake in real-time. Real-time GPS networks also have the advantage of capturing motions throughout the entire earthquake cycle (interseismic, seismic, coseismic, postseismic),.Here we report on three examples of rapid modeling of recent large earthquakes near large regional real-time GPS networks and combined GPS/seismic networks. The first utilizes 416 stations in Japan's GEONET during the 2003 Mw 8.3 Tokachi-Oki earthquake about 100 km offshore Hokkaido Island, the second investigates the 2010 Mw 7.2 El Mayor-Cucapah earthquake recorded by 95 stations in the California Real Time Network and the final one examines the 2011 Mw 9.0 Tohoku-Oki earthquake recorded by over 800 stations in GEONET. We utilize two inverse approaches and one forward approach to ascertain the extent of fault-slip in a simulated real-time environment. The first inverse approach uses predefined fault planes from a catalogue of generalized faults while the second one computes fault planes from a separate centroid moment tensor (CMT) inversion that operates once per second. The forward approach runs a grid search over a suite of possible 2-D Gaussian slip distributions. In all three approaches we are able to roughly characterize all three earthquakes using about a minute or two of data, greatly enhancing the time to obtain fault slip and moment release during medium-to-large earthquakes by almost an order of magnitude. We investigate gains made through combined analysis of GPS and seismic instruments (accelerometers) within rapid modeling versus GPS-only modeling with respect to the aforementioned methodologies as well as peak ground displacement scaling relationships.