Double-wavelet double-difference time-lapse waveform inversion

Xin Fu, Sergio J. Romahn, Kristopher A. Innanen

Time-lapse seismic data are widely used to monitor reservoir changes. And time-lapse waveform inversion is a valuable tool for seismic exploration. A popular time-lapse wave-form inversion strategy is the double-difference time-lapse waveform inversion (DDWI)(inversion of the differential data starting from the reverted baseline model). It is an effective way to solve the problem that baseline and monitor inversions of time-lapse waveform inversion are easily at different convergences, and it results in coherent model error in time-lapse inversion. Nevertheless, the double-difference method (DDWI) demands an almost perfect repeatability between the two baseline and monitor surveys, which is the most challenging for DDWI. Specially, when sources wavelets for the two data sets are different, the results of DDWI are seriously impacted. To solve this problem, we propose a double-wavelet double-difference time-lapse waveform inversion method (DWDDWI). This works because the data difference caused by wavelet difference is eliminated. DWDDWI is developed based on the convolution relationship between the shot gather and Green’s function. And its premise is that the wavelets for both baseline and monitor data sets are known. To test the feasibility of this method, a numerical example is used.