Geostatistical tools have become widely used to construct reservoir models that realistically represent heterogeneity. Uncertainty is often calculated as a by-product of geostatistical modeling by simply generating multiple realizations. This uncertainty is much too small. There is significant uncertainty in the input parameters and the modeling procedure that must be accounted for. This paper discusses procedures for uncertainty quantification in presence of significant uncertainty in the modeling parameters and presents a study from Deep Panuke, east coast of Canada.
Uncertainty assessment requires much more than changing a random number seed and running multiple realizations. The space of uncertainty must be formulated to fairly address all key aspects of uncertainty. This space of uncertainty must be sampled. Each sample from the space of uncertainty is a complete specification of the reservoir geometry and internal structure, which can be visualized and evaluated for geological plausibility and all response variables such as in-place resources and recoverable reserves. Then, the results must be analyzed and presented to convey important results and sensitivities. We implement these three aspects of uncertainty management (1) problem formulation, (2) sampling uncertainty, and (3) analysis of results.
We calculate uncertainty in the gas resource of Deep Panuke off Canada’s East Coast. Seismic data, six wells, and well test data are used. The space of uncertainty is defined by the back reef margin, fore reef margin, top of structure, thickness, histogram of porosity, variogram of porosity, correlation of seismic-derived properties to porosity, water saturation, and formation volume factor. We explain the rationale for each source of uncertainty, the Monte Carlo simulation procedure, and the analysis of results.