Estimation of source function and medium response function by autoregressive method

By modelling seismograms as “low” and “high” order autoregressive (AR) processes, the source function and the medium response function are separated from a single channel seismogram. Akaike's final prediction error is used as a statistic to select the appropriate “low” and “high” AR order of the process. Case studies of synthetic data show that the recovered source and reflectivity functions compare very well with the input functions. Using this method, arrivals of the surface reflected P phases of five explosions from the Soviet region and of two earthquakes from Kamchatka, recorded at Gauribidanur Seismic Array (GBA), India, are identified. Certain features of the source and source region of these events are also inferred.