Adaptive least-squares RTM with applications to subsalt imaging

Abstract

Least-squares reverse time migration (LSRTM) refines the seismic image toward true reflectivity by inversion. Its iterative nature and modeling capability enable the use of synthetic data to guide the preconditioning of input data. When the velocity contains errors, dynamic warping can be used to shift the input data and force the traveltime to be consistent with the imperfect migration velocity. A crosscorrelation-based confidence level is introduced to control the quality of dynamic warping for field data. The confidence level also is used as an inverse weighting to adaptively precondition the data residual. The adaptive preconditioning automatically balances data fitting in the shallow and deep and speeds up convergence in subsalt. Both synthetic and field data experiments based in the Gulf of Mexico show that the adaptive LSRTM can improve the image quality in subsalt effectively and efficiently. Within only a few iterations, the adaptive LSRTM suppresses the salt halo artifacts and increases the signal-to-noise ratio in poorly illuminated areas. It also improves the termination of sediments against salt boundaries and enhances subsalt image coherency. Compared with conventional RTM, the adaptive LSRTM image is more favorable to geologic interpretation.