Local Enhancement of First Arrivals Embedded in Background Noise by Double Stacking: Application to Automatic Picking

Abstract:

The high noise level in the seismic records can significantly decline the efficiency of the automatic picking algorithms, particularly in the case of weak arrivals. Traditional methods used for improving the quality of arrivals are commonly based on pure spectral filtering. The problems related to the later arise when the signal and noise share the same band frequency. Furthermore, some of these filters may distort the key parameters of automatic picking (onset time and polarity). In the near surface surveys, the stacking addresses these problems thanks to data redundancy. To improve the signal-to-noise ratio of first arrivals, and thereby increase their detectability, we propose an optimal alternative: the double stacking in the time domain. The first weighted stack is based on local similarity of the common offsets traces for a few adjacent shots. This stack approach is more tolerant to significant misalignment signals, and locally emphasizes the buried signals. The second stack is applied to a few neighboring traces (within moving spatial window), in the same shot gather, after correcting the delay times via cross correlation tools. In cases of very noisy data, the waveform stacking and the cross-correlation analysis are less efficient at higher frequencies, due to the limited coherence of short-period waveform. De-noising using wavelet transform was carried out as an auxiliary procedure before the second stack. To validate our approach, a synthetic profile was generated from a realistic model via finite difference modeling, and considerable noise levels were added to the simulated data. The weak events are remarkably enhanced, and are more identifiable on the stacked sections. We perform additionally a comparative study of the performance of automatic picking methods, using adaptive multi-method algorithm on noisy real data. The results demonstrate the feasibility of applying the double stacking before the automatic picking.