Machine Contouring: Reading Between the Lines
Jeffrey M. Yarus, Aaron Brown
The wide availability of microcomputers and inexpensive software has encouraged geologists to use contouring programs. These programs are generally based on interpolation techniques that include inverse distance, gradient projection, polynomial fit, and triangulation. Most programs produce in contour patterns that are highly at variance to the real data and can lead to very serious and expensive mistakes. As an example, algorithms were tested on simple known geometric surfaces: inclined, unimodal, bimodal, and faulted. Surfaces were sampled using different sampling techniques: equally spaced, linear point, random, and clustered.
In most cases, results showed false structural variations such as closure, nosing, and structure orientations which were erroneous. Better maps were produced with extensive manipulations of key parameters including grid size and weighting exponents.
When a priori knowledge of the surface is not known, as with real data, results indicate that in some cases satisfactory contouring can be generated using the above-described surfaces as surrogate mapping horizons sampled by the real data array. In this way, values for key parameters can be estimated prior to mapping the real data. Although this method offers no guarantees, it does tend to eliminate artifacts imposed on a surface by the algorithm.
Unless extreme caution is exercised when using machine contouring algorithms or better algorithms become available, one should rely on the hand contouring methods of an experienced geologist when important economic decisions are at stake.
AAPG Search and Discovery Article #91038©1987 AAPG Annual Convention, Los Angeles, California, June 7-10, 1987.