DANIEL C. ZWEIDLER, Shell Deepwater Development Inc.

Abstract: Challenges of Wildcat Risk Assessment in Turbidite Systems, Deepwater Gulf of Mexico.

The main challenge of wildcat risk assessment in turbidite systems lies in the inherent difficulty to rank amplitude supported opportunities together with more conventional opportunities. The level of dependence between stacked objectives moreover exacerbates the different levels of risk and uncertainty associated with the two types of opportunities.

Turbidite rock properties in the Gulf of Mexico often cause hydrocarbon-bearing reservoirs to have significant acoustic impedance contrast relative to encasing rocks, though this varies with age, location, and other factors. Seismically these impedance contrasts are expressed as amplitude anomalies, those allow for visual discrimination of hydrocarbon bearing reservoirs from wet sands. This discrimination constrains both risk and uncertainty about the number of stacked events and their volume distribution in a prospect. In some geological settings, however, imaging problems prevent amplitude anomalies from being detected (e.g. subsalt). In other areas, rock properties do not allow for discrimination between wet and hydrocarbon bearing sands altogether. Risking of these non-amplitude constraint plays has to rely on geological knowledge rather than seismic attribute information.

Prospects in deepwater Gulf of Mexico comprise on average 3-4 stacked economic reservoirs with as many as 14. This stacking phenomenon helps compensate for the rather limited extent of the traps, which rarely exceeds 25 km 2 . In assessing risks of prospects with multiple stacked objectives, it is essential to define if the risks described is specific to one particular event, or represents the whole prospect. For amplitude supported prospects, the former often applies, but for conventional opportunities, risks are often given for the total prospect. If risks describe the individual objectives then the total probability of success has to be calculated by stochastic addition of risks handling partial dependencies. Various mathematical models with different approaches to this complex problem are suggested. One of the more common pitfalls is to represent a prospect with multiple objectives through one single expectation curve. This expectation curve does not carry the information on the possible combinations of events and their associated volume distributions. Only a subsurface scenario approach retains the level of granularity necessary to ensure proper economic evaluation of an opportunity.

The poster illustrates the following observations and conclusions with data from the deepwater Gulf of Mexico.

1) Amplitude supported prospects HC columns are constrained by the extent of the amplitude anomaly. Low variance and skewness characterize simulated volume distribution.

The predicted number of stacked objectives is derived from direct observation of.seismic data. The number of possible stacked objectives is given as a single number.

The distribution of total volumes for any given combination of partially dependent stacked objectives is strongly constrained by the seismic attributes. A limited number of development scenarios are necessary to accommodate the subsurface uncertainties.

2) Non-amplitude supported prospects

HC column length is defined by a probabilistic assessment of potential leak points along closure. High variance and skewness characterize simulated volume distributions.

Individual objectives are difficult or impossible to identify The number of possible stacked objectives are derived by inference from prior probabilities and given in the form of a range of values.

The number of combinations for an uncertain number of partially dependent stacked event is potentially very large A detailed scenario approach is often unrealistic due to the range of possibilities.

3) General

Prospects in turbidite systems tend to have multiple stacked objectives.When adding the individual risk per objective, partial dependence has to be taken into account in the mathematical model. The model affects the prospect P(s), and a development scenario approach is recommended.

AAPG Search and Discovery Article #90923@1999 International Conference and Exhibition, Birmingham, England