Rapid Estimates of the Source Time Function and Mw using Empirical Green's Function Deconvolution

The U.S. Geological Survey National Earthquake Information Center (NEIC) uses a variety of classical network-averaged magnitudes (e.g., m(b) and M-s) and waveform modeling procedures to determine the moment magnitude (M-w) of an earthquake from teleseismic observations. Initial magnitude estimates are often inaccurate because of poor azimuthal control (sampling of the focal sphere) and/or intrinsic limitation of each method to a specific range of event size. To provide faster and more accurate estimates of the moment magnitude, source duration, and source complexity, NEIC is exploring the use of a variation of the empirical Green's function (EGF) deconvolution procedure. This approach uses a predicted focal mechanism derived from the Global Centroid Moment Tensor Catalog to compute teleseismic P-wave synthetic seismograms, which are then deconvolved from observed P and SH waveforms to determine station-specific M-w, source time function, and a network-averaged M-w. Our EGF approach is validated using broadband waveforms from 246 earthquakes in the magnitude range M-w 6.0-9.1. Within approximately 13 min of earthquake origin time, our procedure using teleseismic P waves only computes an M-w that lies within +/- 0.25 of the final W-phase M-w in the magnitude range 6-8. Using later arriving teleseismic SH phases results in an M-w that lies within +/- 0.12 of the W-phase M-w. For magnitude 8 or larger earthquakes, we underestimated the moment magnitude by up to 0.8 magnitude units, primarily due to the initial P phase not containing the total seismic moment release. Long-period phases such as the W-phase and surface waves that better characterize total moment release can also be incorporated in the processing.