Real-time centroid moment tensor determination for large earthquakes from local and regional displacement records

We present an algorithm to rapidly determine the moment tensor and centroid location for large earthquakes employing local and regional real-time high-rate displacement records from GPS. The algorithm extracts the coseismic offset from the displacement waveforms and uses the information to invert for the moment tensor. The Green's functions for a layered earth are obtained numerically from open source code EDGRN. To determine the centroid, multiple inversions are simultaneously performed within a grid of inversion nodes, and the node with the smallest misfit is then assigned the centroid location. We show results for two large earthquakes replayed in simulated real-time mode using recorded 1 Hz GPS displacements: the 2003 Mw 8.3 Tokachi-oki and the 2010 Mw 7.2 El Mayor-Cucapah earthquakes. We demonstrate that it is feasible to obtain accurate CMT solutions within the first 23 min after rupture initiation without any prior assumptions on fault characteristics, demonstrating an order of magnitude improvement in latency compared to existing seismic methods for the two earthquakes studied. This methodology is useful for rapid earthquake response, tsunami prediction and as a starting point for rapid finite fault modelling.