The effect of input data sampling on prestack interpolation efficacy: Lessons learned from a sparsely shot and heavily structured 3D data set

Abstract

Two different prestack interpolation methods, 5D interpolation by Fourier reconstruction and dip-scan-based data synthesis, are contrasted and compared on both synthetic data and also on a sparsely acquired and heavily structured 3D land real data set. In particular, the Fourier reconstruction algorithm’s limitation in performing regular upsampling across the cmp coordinates is studied for the case of a generic upsampling problem on a well-sampled orthogonal geometry, and we attempt to gain additional insight into this upsampling limitation by computing and analyzing four-dimensional sampling operator spectra. Systematic real data testing reveals that cascading the two different interpolation methods gives good results which provide a combination of (i) regular upsampling along the crossline midpoint coordinate and (ii) gap-filling along certain shot and receiver lines which were truncated in the field because of the rugged terrain.