LEE, P. J., Institute of Sedimentary and Petroleum Geology, Geological Survey of Canada, Calgary, Alberta

ABSTRACT: Lognormal and Nonparametric Discovery Process Models - Reliable Tools in Petroleum Resource Assessment?

The objectives of this paper are: (1) to demonstrate the reliability of the lognormal (LDSCV) and nonparametric (NDSCV) discovery process models using data sets with known population parameters (number of pools in a play = 300); and (2) to test the adequacy of probability distributions used to represent oil or gas pool size distributions in petroleum plays.

Two known data sets are examined: one which obeys a lognormal probability distribution and the other, a Pareto distribution. The discovery processes were also simulated using an exploration efficiency = 0.6. The first 30 "discoveries" of the two data sets were analyzed using the two models. Both models suggest that the number of pools ranges from 250 to 300 or more. Given 300 as the number of pools, predictions are derived about pool size.

From these four analyses, it is possible to conclude that: (1) the nonparametric discovery process model can handle data sets that obey either the lognormal or the Pareto distribution; (2) the lognormal discovery process model can handle data sets that obey the lognormal distribution as well as those from a Pareto distribution; however, the model may underestimate the resource if the sample size is too small; (3) both models can handle "J"-shape distributions; and (4) as the sample size increases so does the prediction accuracy.

Data sets obtained from oil and gas plays of the Western Canada Sedimentary basin; the Permian sandstone play of the Cooper basin, Australia; the Middle Jurassic of the Paris basin, France; the Middle Jurassic play of the Viking Graben of the North Sea; the Pur-Neocomian gas play of the West Siberian basin, the Hackberry play of Louisiana; the Minnelusa play of the Powder River basin; and the Jurassic Smackover play stretching through Mississippi, Alabama, and Florida in the United States have been analyzed using the nonparametric discovery process model and quantile-quantile plots. The results show that lognormal or power normal distributions are the best choice for most petroleum plays. Occasionally, however, the Weibull distribution may prove superior to the lognormal distribution nd, generally speaking, the Pareto distribution may suffice for the largest few pools of a play.

AAPG Search and Discovery Article #90987©1993 AAPG Annual Convention, New Orleans, Louisiana, April 25-28, 1993.