Randomized sampling without repetition in time-lapse seismic surveys

Abstract

Vouching for higher levels of repeatability in acquisition and processing of time-lapse (4D) seismic data has become the standard with oil and gas contractor companies, with signifi-cant investment in the design of acquisition systems and processing algorithms that attempt to address some of the current 4D challenges, in particular, imaging weak 4D signals. Recent developments from the field of compressive sensing have shown the benefits of variants of randomized sampling in marine seismic acquisition and its impact for the future of seismic exploration. Following these developments, we show that the requirement for accurate survey repetition in time-lapse seismic data acquisition can be waived provided we solve a sparsity-promoting convex optimization program that makes use of the shared component between the baseline and monitor data. By setting up a framework for inversion of the stacked sections of a time-lapse data, given the pre-stack data volumes, we are able to extract 4D signals with relatively high-fidelity from significant subsamplings. Our formulation is applied to time-lapse data that has been acquired with different source/receiver geometries, paving the way for an efficient approach to dealing with time-lapse data acquired with initially poor repeatability levels, provided the survey geometry details are known afterwards.