Basin Analysis and Artificial Intelligence--Now That Our Basin is Classified, What do We do with It?

Jay E. Leonard, Douglas W. Waples

Basin analysis is perhaps the most integrated of disciplines in the geologic sciences. It is seemingly impossible for one individual to fully command a working knowledge of all the separate subdisciplines affecting basin evolution and possible hydrocarbon generation, migration, and entrapment. As a result of this dilemma, expert systems have been developed to aid the explorationist in the classification and subsequent analogous description of virgin basins. This aspect is indeed useful; however, it may only represent 10% of the basin evaluation process. Both deterministic and probabilistic methods must be utilized to provide an adequate and reasonable solution to basin problems.

Cutting-edge research conducted by commercial and oil company laboratories and a few academic institutions have directed basin analysis to a high degree of quantitative sophistication. These deterministic models have proven to provide a good correlation between predicted and observed results. The selection of methods (i.e., Lopatin versus Tissot and Espitalie) requires some a priori knowledge of basin history. Moreover, the selection of algorithms needed to solve the seemingly unsolvable equations can be optimized by an understanding of the basin.

Probabilistic approaches to the basin evaluation process are also extremely important. Although their approach is radically different than deterministic methods, they do provide a different insight to the evaluation process. For example, to efficaciously utilize a Monte Carlo approach, some knowledge of data distribution is necessary. This knowledge can be provided by an expert system.

At the time of this writing, we believe that the state of artificial intelligence in basin analysis is somewhat akin to the "Guess the Animal" game, which was pervasive in computer labs during the mid-'70s. What is needed is a stronger dependency upon the quantitative approaches which respond to a heuristic concept of basin information. We should now look toward the time of "artificial intelligence transparency" in the basin modeling disciplines, whereby adequate parameters can be supplied to drive a model to the best possible solution.

AAPG Search and Discovery Article #91038©1987 AAPG Annual Convention, Los Angeles, California, June 7-10, 1987.