Estimation of the Source Time Function Based on Blind Deconvolution with Gaussian Mixtures

It is demonstrated that the blind deconvolution method is fully capable of recovering the unknown Green’s function and of estimating the source time functions from observed seismic data of small earthquakes. Based on the assumption of the Gaussian-mixture model of the Green’s function, the newly-formulated algorithm is evaluated using synthetic seismic data along with those of the May 8, 1996 Mexico earthquake (Mc = 4.6). Since the estimated results closely match the theoretical input very well, the method is then employed to analyze the source time functions of the July 7, 1995 Pu-Li, Taiwan earthquake (ML = 5.3). The stations triggered by this event were azimuthally well covered. Using the estimated source time functions, information pertaining to the directivity effect is readily obtained, and the actual fault plane of this event is identified, thus clearly indicating that this method provides a most efficient way to estimate the source time function of a small earthquake.