In addition to total undiscovered resource, predicting the number and size of future hydrocarbon discoveries is a critical part of resource assessment. Economic analyses are often highly sensitive to accumulation size, thereby influencing when and at what cost a potential resource will become available. For established plays, it may be expedient to base predictions of future field sizes on previous discoveries. Having a better understanding of the underlying parent distribution would be invaluable in this process.
Many previous works have shown that size distributions for discovered fields within a play or basin often appear to be lognormal. In most cases, the historical size distribution is inappropriate to apply to the remaining potential within an established play because the largest accumulations tend to be discovered early in the exploration process. This intentional sampling bias must be accounted for when projecting sizes of future discoveries, as does the tendency to underreport the ultimate recoverable volume of hydrocarbons early in the life of a field. Both of these factors can be influenced by levels of geologic knowledge and technical sophistication at the time of the analysis.
Field size data from the Gulf of Mexico shelf were used to explore the relationship between discovered and future field sizes. Using generalized Pareto distributions as assumed parent populations, it is possible to simulate discovery sequences that closely match the historical development of the basin. The selected parent distributions are of a class suggested in the literature as more reflective of the natural endowment and have empirical support. A discovery creaming model that explicitly incorporated the notion of sampling bias closely matched the overall Gulf of Mexico results as well as intermediate results at several points in the exploration history. The results can be approximated using lognormal distributions across multiple stages of exploration and over a wide distribution of discovery sizes. These observations suggest that one cannot impute the shape of the underlying parent distribution from the size distribution of past discoveries. Simulations suggest that future discoveries will follow lognormal distributions over moderate sample sizes, but eventually will reflect the Pareto nature of the underlying parent population. Pareto distributions imply a larger number of small fields than lognormal, and thus may significantly impact economic analyses. Economic censoring may provide a reasonable explanation of the observed lognormality in the size distribution of discovered oil and gas accumulations.