Closure Confidence: How Big is That Field? A Case Study

Abstract

In many cases, the areal extent of closure is one of the biggest uncertainties in determining the in-place hydrocarbon volumes of a field. In this paper, I use the technique of cross-validation to rank depth maps over one particular field. The depth maps are generated by massive numbers of samples from a wide range of possible velocity models. Cross-validation drops out each well in turn, and predicts the top reservoir depth from the missing wells. The RMS prediction error is a measure of the quality of the velocity model. The best 100 velocity models (those with the lowest RMS) give 100 best depth maps. I assume that the best depth maps are equally probable because each has a low cross-validation RMS value. An automatic spillpoint-finding algorithm is then used to find the area of closure in each of the best maps. Based on these maps, statistics are computed about the likely size of a field, and the uncertainty thereof. I conclude that there is a chance that the field size is significantly different to that given by the initial deterministic ``best technical' depth maps.