The Limitations of Lognormal Distributions: Using Subsurface Data to Make More Accurate Resource Estimations

Abstract

The outcomes produced by multiplying two independent variables are lognormally distributed (e.g. the product of two die or area). Exploration risk and uncertainty pioneer, E. C. Capen, first advocated using lognormal distributions to estimate petroleum reserves in an AAPG short course titled Evaluating and Managing Petroleum Risk in 1984. Subsequently, working together with R. E. Megill and P. R. Rose, Capen offered this course more than 50 times in the succeeding years, and the use of lognormal distributions became the industry standard when describing a range of potential outcomes for everything from EUR to gross rock volume. Lognormal distributions give more accurate reserve estimates but have one inherent flaw—they start at zero and extend to infinity. Capen addressed this issue in 1992 by explaining that one has to “sense check” the high side outputs and truncate appropriately. However, this upper truncation is affected by its own uncertainty. How big is “too big”? Building on this previous work, a new workflow has been designed that reduces the uncertainty in predrill resource estimates and constrains high-side estimates to geologically reasonable values. “Full field” uncertainty analysis allows for stochastic Monte Carlo simulation, while accounting for potential variance in the mapped horizon.

AAPG Datapages/Search and Discovery Article #90259 ©2016 AAPG Annual Convention and Exhibition, Calgary, Alberta, Canada, June 19-22, 2016