Suppression of the acquisition footprint for seismic sequence attribute mapping

Abstract

Seismic coherency has proven to be very effective in delineating geologic faults as well as considerably more subtle stratigraphic features, including channels, canyons, slumps, levees, glacial gouges, dewatering patterns and pinnacle reefs. Unfortunately, seismic coherency estimates, which quantitatively measure the similarity or dissimilarity of adjacent traces in 3-D, are particularly sensitive to coherent noise that passes through the acquisition and processing flow. They are also sensitive to dissimilarities in fold, offset, and azimuth distribution introduced through the 3-D acquisition and binning processes. Edge enhancement algorithms further exacerbate these linear artifacts. We define the acquisition footprint to be any pattern of noise that is highly correlated to the geometric distribution of sources and receivers on the earth’s surface. While the strong acquisition footprints such as those caused by normal moveout (NMO) stretch on vintage single‐fold data have been largely ameliorated by modern 2-D multifold recording, we see acquisition 3-D footprint in the low‐fold shallow section and throughout the entire section when recording sparse 3-D land surveys that often result in only six- to sevenfold data. One may partially suppress the acquisition footprint on the seismic coherency time or depth slices using conventional 2-D image processing. Unfortunately, such filtering is inappropriate for dip/azimuth maps, cluster analysis maps, and other maps that may not be continuous real variables or with maps that have cyclic values, such as the wavelet phase. We show that simple 3-D true‐amplitude dip filtering of the input 3-D (t, x, y) time- or depth‐migrated seismic data volume can be quite effective in minimizing the detrimental effect of the acquisition footprint on conventional 3-D seismic attributes for both marine and land data acquisition geometries. However, 3-D dip filtering of migrated data will often eliminate the fault‐plane reflections necessary for sharp coherency images. We therefore recommend that, whenever possible, suppression of acquisition footprint be performed before 3-D migration on the stacked data volume, where sharp fault truncations in depth are represented by smoothly varying diffractions in time.