Seismic source mapping by surface wave time reversal: application to the great 2004 Sumatra earthquake

Abstract

Different approaches to map seismic rupture in space and time often lead to incoherent results for the same event. Building on earlier work by our team, we ‘time-reverse’ and ‘backpropagate’ seismic surface wave recordings to study the focusing of the time-reversed field at the seismic source. Currently used source-imaging methods relying on seismic recordings neglect the information carried by surface waves, and mostly focus on the P-wave arrival alone. Our new method combines seismic time reversal approach with a surface wave ray-tracing algorithm based on a generalized spherical-harmonic parametrization of surface wave phase velocity, accounting for azimuthal anisotropy. It is applied to surface wave signal filtered within narrow-frequency bands, so that the inherently 3-D problem of simulating surface wave propagation is separated into a suite of 2-D problems, each of relatively limited computational cost. We validate our method through a number of synthetic tests, then apply it to the great 2004 Sumatra–Andaman earthquake, characterized by the extremely large extent of the ruptured fault. Many studies have estimated its rupture characteristics from seismological data (e.g. Lomax, Ni et al., Guilbert et al., Ishii et al., Krüger & Ohrnberger, Jaffe et al.) and geodetic data (e.g. Banerjee et al., Catherine et al., Vigny et al., Hashimoto et al., Bletery et al.). Applying our technique to recordings from only 89 stations of the Global Seismographic Network (GSN) and bandpass filtering the corresponding surface wave signal around 80-to-120, 50-to-110 and 40-to-90 s, we reproduce the findings of earlier studies, including in particular the northward direction of rupture propagation, its approximate spatial extent and duration, and the locations of the areas where most energy appears to be released.