A main challenge for wildcat risk and resource assessment of prospects with complex, multiple layers and compartments is handling explicitly and consistently both risk and volume dependencies. This paper compares and contrasts three main alternative approaches to modeling risk dependencies: A complete enumeration of outcome probabilities, a bottom-up approach that aggregates risks using explicit shared risk estimates, and a top-down approach that generalizes the 2-level marginal play, conditional prospect risk model to a 3-level model with a conditional segment level risk. We consider assumptions, data needs and ease of use of the different approaches using illustrative cases and applications.


Petroleum exploration is increasingly targeting complex prospects with multiple pay zones and compartments. Equally important, targets in demanding settings such as deep-water Gulf of Mexico require accurate assessment of risks and volumes for all zones. It is often no longer sufficient to consider solely the largest zone or compartment. In order to pass economic hurdles, the project evaluation must include all possible hydrocarbon-bearing layers and compartments.

Each zone and compartment could be treated as independent prospects and aggregated accordingly. Or all zones and compartments could be treated as a single contiguous prospect (full dependence). However, assuming either full independence or full dependence will in general not give an unbiased and accurate estimate of either the aggregate potential of the prospect or the likelihood that the prospect is dry. As shown by Murtha1, the expected aggregate resources are not affected by the degree of dependency. However, the variance in the expectation clearly is. Similarly, the expected number of successful layers and compartments is not affected by the degree of dependency, but the probability of a dry hole clearly is.

Dependencies can apply to both risks and volumes. We can say that risks are conditioned by risk outcomes across layers and compartments while volume parameters are correlated across layers and compartments.

In this paper we compare and contrast three alternative approaches to modeling risk dependencies. The first enumerates all outcome probabilities. The second is a bottom-up approach that aggregates risks using explicit dependency estimates, while the third is a top-down approach that generalizes the 2-level marginal play, conditional prospect risk model to a 3-level model with a conditional segment level risk.

We use the term segment to describe a discrete component of a prospect. The key requirement is that a segment cannot become a proven petroleum accumulation unless it is explicitly targeted. Thus, segments can be different pay zones. Segments can also be components of the same pay zone that are separated by faults or other compartmentalizing mechanisms.

The rest of the paper is organized as follows. The next section presents the three alternative approaches to modeling dependencies and uses a simple 2-segment case to show how the 3 approaches formally interrelate. The 3-segment case presented by Murtha1 is used to show how the methods differ, where the full enumeration approach provides a complete information benchmark and the two other approaches are approximations. We also investigate conditions when the top-down and bottom-up approximations are more or less valid. The discussion section reviews the strengths and weaknesses of the alternative approaches.

Alternative approaches to modeling risk dependencies

The probability of success for a single segment prospect, P(S), is usually modeled by considering key geological factors, estimating the likelihood that they exist or are adequate, and then deriving an overall estimate of P(S) as the product of the probability of success of each factor. Thus, for example, source, migration, timing, reservoir, trap and seal might be the risk factors considered. The probability of success of the prospect, P(S), is then P(Source)*P(Migration)*P(Timing)*P(Reservoir)*P(Trap)*P(Seal).

This content is only available via PDF.
You can access this article if you purchase or spend a download.