2019 AAPG Annual Convention and Exhibition:

Datapages, Inc.Print this page

Preservation of Structural Information Across Prospect Portfolios


Management of hydrocarbon prospect and asset portfolios is facilitated by subsurface maps that are stylistically consistent irrespective of structural geology and source data density. Map grids are usually modified by resampling and filtering during the mapping process to achieve required corporate standards; these procedures apply various degrees of smoothing. Mappers choose smoothing filters and parameters based on experience and instinct rather than guidelines that link the operations to structural information such as fold wavelength and fault geometry. Here we present the first comprehensive analysis of structural information attenuation under a wide range of smoothing options - resampling, mean, median, Gaussian and low-pass (fast Fourier transform, FFT) filtering. The results define optimized workflows for creating consistent, “clean” maps that retain key structural information.In addition to filter selection, at least three parameters are required in smoothing operations: grid spacing, filter width and number of iterations. Filters and their parameters are compared here using standard deviation and visual assessment of removed data. Resampling and filter width exert a first-order control on structural information retention governed by Nyquist-Shannon sampling theory. Additional results show that standard deviation of removed data varies with number of filter iterations according to power laws for mean and Gaussian filters. These filters continue to significantly attenuate structure with increasing number of iterations. On the other hand, the low-pass FFT filtering removes information in a pattern that varies according to logarithmic functions that level off after a certain number of iterations.These properties of smoothing filters are illustrated, explained and exploited to yield two workflows for preserving structural information across maps from diverse structural settings. One option is simply to resample at a grid spacing just below the critical structural wavelength. This is a straightforward approach, but with a disadvantage of it being awkward to obtain residual data and band-pass information at intermediate length scales. Alternatively, FFT can be applied without resampling and can yield low-pass structure (traps) and band-pass products at several length scales (e.g. faults).