| Forecasts of the volumes of hydrocarbons to be discovered through the exploration process mostly rely on the probabilistic drilling model developed by Barouch and Kaufman (1976). This model has been termed the "non-central, multivariate, hypergeometric" (NMH) distribution. There are no easy-to-evaluate formulas for the expected value and standard deviation of discovery volumes, as functions of well number, and the only practical method of estimation is through lengthy simulations. However, rather than running lengthy simulations, Fuller et al. (1990) suggested a new, very rapid method for approximating the means and standard deviations. The method was first developed by Manly (1974) in the context of biometric experiments. Fuller et al. (1990) applied this approximation to hydrocarbon exploration and examined the accuracy of Manly's approximation. However, this work considered only a specific data set, and the results cannot be generalized and applied with confidence by oil companies. Hence, the issue of accuracy of the approximation method needs to be investigated further.;In addition to the hypothetical data, real data are selected and tested to assess the accuracy of the approximation method. The regression models are used to adjust the results from the approximation method close to the results from the simulation method.;This research assesses the credibility of the use of the approximation method in the context of hydrocarbon exploration. However, the use of the approximation method is not without drawbacks. The usefulness of the approximation method depends on the trade-off between the computation time saved compared to the simulation method, and the accuracy of the approximation method.;In this research, the accuracy of the approximation method is examined for different types of initial pool size distributions. The different types of distributions used are: Lognormal, Exponential, Weibull, and Gamma distributions. Hypothetical data are generated from these distributions and the accuracy of the approximation method is tested by comparing with the simulation method. The comparison results are summarized statistically in terms of mean value and standard deviation value differences between the approximation and the simulation methods. Regression models are developed from these differences so that for a given data set, the accuracy of the approximation method can be predicted. |
