Source-Depth Estimation by Prediction Error Filtering of Synthetic Explosion Signals

摘要

For detecting the depth phases, a parametric model for P signal was adopted. In other words, the primary P pulse is characterised by numerical values of a few parameters (usually 10-20). As the reflected P phase will be almost identical to the P phase, the same model will be valid for such phases also. For seismic signals, a special type of parametric model called autoregressive (AR) model is found useful. It is shown that the prediction error filter (PEF) for a given order as computer by AR model of the time series is same for the primary P phase as well as for the composite wave comprising the primary P phase and several reflected and refracted phases. Therefore, when the composite wave is convolved with its PEF, there will be large local errors around the instants when the reflected or refracted phases start entering the seismogram, thereby facilitating the identification of such phases. It is shown that using any order of the PEF, not necessarily the optimum, the depth phases can be extracted from the seismograms. However, in order to detect the depth phases unambiguously, the delay time of the depth phases w.r.t. the P phase should be greater than or equal to 0.3s. The method could unambiguously detect the depth phases from the signals having signal-to-noise ratio greater than or equal to 4. For weaker signals, array beams may be used to extract the depth phases. It was also demonstrated that T/sup */ (travel time to quality factor ratio)= 0.6s tends to give better results than T/sup */ = 1.0s. 40 references, 19 figures, 3 tables. (ERA citation 13:019098)