Time-lapse seismic data reconstruction using compressive sensing
Mengli Zhang, David Lumley
Abstract
ABSTRACT The time-lapse seismic method plays a critical role in reservoir monitoring and characterization. However, time-lapse data acquisitions are costly. Sparse acquisitions combined with postacquisition data reconstruction could reduce costs and facilitate more frequent applications of the time-lapse seismic monitoring. We have developed a sparse time-lapse seismic data reconstruction methodology based on compressive sensing. The method works with a hybrid of repeated and nonrepeated sample locations. To make use of the additional information from nonrepeated locations, we develop a view that nonrepeated samples in space are equivalent to irregular samples in calendar time. Therefore, we use these irregular samples in time coming from nonrepeated samples in space to improve the performance of compressive sensing reconstruction. The tests on synthetic and field data sets indicate that our method can achieve a sufficiently accurate reconstruction by using as few as 10% of the receivers or traces. The method not only works with spatially irregular sampling for dealing with the land accessibility problem and for reducing the number of nodal sensors, but it also uses the nonrepeated measurements to improve reconstruction accuracy.