Seismic similarity analysis through self-organising maps

Abstract

In this work we present a methodology for extracting valuable information from several seismic attributes by computing seismic similarity values through a pattern recognition approach, which is based on self-organising maps. This methodology allows for identifying regions with seismic properties similar to a pre-defined reference location of interest, for instance, a good well producer or a dry well; and it can be used during the exploratory phase when only limited and scarce well information is available. The methodology we propose improves the classical seismic similarity analysis and can be used on two-dimensional seismic maps or three-dimensional seismic volumes for frontier exploration, i.e. where there are scarce or limited well data but much seismic information. Using two case studies, we show how the proposed method constitutes a valuable tool for exploration geophysics and reservoir characterisation.