Expanding
Uncertainty: Predictive Distributions for Undiscovered Oil and Gas Pools in a
Play
Kaufman, Gordon M.1, John H.
Schuenemeyer2 (1) MIT,
The U.S. Minerals Management Service (MMS)
is responsible for assessing magnitudes of undiscovered oil and gas in offshore
Federal waters. They have long used modified versions of a Canadian system
called PETRIMES. While this system provides critical information, a peer review
suggested that PETRIMES under-represents the degree of uncertainty that should
be attributed to undiscovered oil and gas pools.
In response and with the support of MMS,
we outline an alternative discovery process modeling approach based on an
algorithm that drives PETRIMES. A play is assumed to possess N pools whose
magnitudes are generated by independent sampling from a Lognormal
distribution. Then pools are discovered by sampling proportional to magnitude
and without replacement from this finite population.
Given discovery data and expert
probability judgments about key parameters, we compute (Bayesian) predictive
probability distributions of properties of undiscovered oil and gas pools;
i.e., the distributions unconditional with respect to prior uncertainty about
the number N of pools in the play, pool size distributions at various
fractiles, and about Lognormal (super-population) parameters.
Our approach relies heavily on
computational schemes that were not in current use when PETRIMES was created
such as Importance Sampling, Acceptance-Rejection Sampling and Markov Chain
Monte Carlo. Discovery data either in order of discovery or unordered can be
handled.
We will illustrate with an application to
a typical petroleum play.
AAPG Search and Discover Article #90063©2007 AAPG Annual Convention, Long Beach, California