A computer procedure to analyze seismic data to estimate outcome probabilities in oil exploration, with an initial application in the tabasco region of southeastern Mexico

Abstract

The Tabasco region contains a number of major oilfields, including some of the emerging “giant” oil fields which have received extensive publicity. Fields in the Tabasco region are associated with large geologic structures which are detected readily by seismic surveys. The structures seem to be associated with deepseated movement of salt, and they are complexly faulted. Some structures have as much as 1000 milliseconds relief of seismic lines. A study, interpreting the structure of the area, used initially only a fraction of the total seismic lines That part of Tabasco region that has been studied was surveyed with a close-spaced rectilinear network of seismic lines. A, interpreting the structure of the area, used initially only a fraction of the total seismic data available. The purpose was to compare “predictions” of reflection time based on widely spaced seismic lines, with “results” obtained along more closely spaced lines. This process of comparison simulates the sequence of events in which a reconnaissance network of seismic lines is used to guide a succession of progressively more closely spaced lines. A square gridwork was established with lines spaced at 10 km intervals, and using machine contour maps, compared the results with those obtained with seismic grids employing spacings of 5 and 2.5 km respectively. The comparisons of predictions based on widely spaced lines with observations along closely spaced lines provide information by which an error function can be established. The error at any point can be defined as the difference between the predicted value for that point, and the subsequently observed value at that point. Residuals obtained by fitting third-degree polynomial trend surfaces were used for comparison. The root mean square of the error measurement, (expressed in seconds or milliseconds reflection time) was found to increase more or less linearly with distance from the nearest seismic point. Oil-occurrence probabilities were established on the basis of frequency distributions of trend-surface residuals obtained by fitting and subtracting polynomial trend surfaces from the machine-contoured reflection time maps. We found that there is a strong preferential relationship between the occurrence of petroleum (i.e. its presence versus absence) and particular ranges of trend-surface residual values. An estimate of the probability of oil occurring at any particular geographic point can be calculated on the basis of the estimated trend-surface residual value. This estimate, however, must be tempered by the probable error in the estimate of the residual value provided by the error function. The result, we believe, is a simple but effective procedure for estimating exploration outcome probabilities where seismic data provide the principal form of information in advance of drilling. Implicit in this approach is the comparison between a maturely explored area, for which both seismic and production data are available, and which serves as a statistical “training area”, with the “target” area which is undergoing exploration and for which probability forecasts are to be calculated.