An important task in constructing digital reservoir models is the capture of geological detail with sufficient resolution to reduce the uncertainty inherent in interpolation between discrete data points. Even the most sophisticated mathematical or statistical reservoir modeling algorithm, however; remains ignorant of the processes by which reservoir strata are deposited. In a test of process-oriented modeling, which places lithofacies in a reservoir body according to geological rules, we constructed a series of conceptual fluvial models by stochastically placing multiple channel-fill models within the model volume. The desired outcome was a digital fluvial reservoir model incorporating the small-scale, local lithologic variation generated by channel switching and stream aggradation within a broad meander belt.
Process-oriented modeling software was used to generate multiple model realizations ranging from 50% to 70% sandstone by volume. Geometric parameters (channel azimuth, meander wavelength and amplitude, and channel width and depth, etc.) were defined in terms of mean and standard deviation. The software also allows explicit definition of these values. Up to 100 channels were placed in each model volume via random processing. As in the case of geometric parameters, channels may also be deterministically placed by digitization of a channel axis or axes.
Fluvial channels generated by the modeling display geologically accurate internal geometry. The model exhibits cross-cutting relationships between channels and isolated remnants of overbank and abandoned-channel deposits that would form local permeability barriers. The resultant distribution of lithofacies differs from geostatistically-generated models in its preservation of small-scale variations in lithology, and in rule-based lateral and vertical variation and sharp contacts among lithofacies. The use of process-oriented modeling yields facies geometries that honor the distribution of lithologies and sedimentary structures observed in the rock record. This geology-based rock distribution allows improved accuracy in population of a reservoir model, especially where reservoir parameters are facies-dependent.
Rex Knepp, Geomodeling.