Regularising time-domain multi-dimensional deconvolution withoffset-directional derivatives

ABSTRACT

Multi-dimensional deconvolution (MDD) is a technique used at different stages of the seismic processing and imaging value chain to suppress overburden effects by deconvolving the up- and downgoing components of a given wavefield at a target of interest. Whilst the time-domain implementation has recently been identified as the de-facto solution for 2D applications, owing to its stability and ability to include physics-based preconditioners, the extension to large-scale 3D datasets is still in its infancy and may require some compromise.

For example, to use a reciprocity preconditioner, one is required to solve the MDD problem for all virtual sources at once, a prohibitive scenario for 3D applications. In this work, we present a simple strategy to regularise the solution of time-domain MDD that leverages the similarity between wavefields from adjacent virtual sources. The proposed approach requires one to solve the MDD problem only for a group of virtual sources simultaneously, and therefore is amenable to 3D applications