Reducing source wavelet non-repeatability for time-lapse shot gathers

Xin Fu, Kristopher A. Innanen

In time-lapse seismic surveys, the effective seismic signals produced by property changes of subsurface rocks are often ruined by non-repeatability noises. In this paper, starting from the wave equation, based on the relationship between wavefields and Green's functions, we propose two frequency-domain matching filters to reduce source wavelet non-repeatability for time-lapse shot gathers, one is the spectrum ratio of the baseline and monitoring wavelets, the other one is the average spectrum ratio of the baseline and monitoring traces. The former requires the wavelets information of baseline and monitoring data, and the latter is source-independent. After reducing the source wavelet non-repeatability, we employ time-shifts corrections by a fast local cross-correlations algorithm to further reduce non-repeatability errors in the difference data (monitoring data minus baseline data) that caused by time shifts between monitoring and baseline data. And then, a reverse time migration (RTM) in depth with a Poynting Vector imaging condition used to reduce the remaining errors arising from the inaccuracy of the source-independent matching filter is carried out. The feasibility of our methods is demonstrated by the synthetic noise-free and noisy data tests. The spectrum ratio of the baseline and monitoring wavelets can effectively solve the source wavelet non-repeatability issue. The source-independent filter also shows good performance in these tests.