The Endicott Field, located on the North Slope of Alaska, is a mature field with a wealth of production data. The upper subzones of the Endicott reservoir, 3C and 3B, comprise a low net/gross siliciclastic succession with an estimated 300+ MMB original oil in place. At present, in comparison to the rest of the field, recovery from this interval is relatively low. Previous to the current description, these subzones were divided into 8 subunits based on a net to gross layering scheme. These sequences have proven to be inadequate as predictive layers for either fluid movement or sand presence due to the complex heterogeneous nature of the reservoir. To reduce infill drilling risks, enhance reservoir management, increase recovery, and optimize a potential EOR implementation, an integrated reservoir description/modeling study was initiated in 1995.
The approach taken was to describe and model the reservoir using sequence stratigraphic concepts and to incorporate as much production data in developing a layering scheme. The main features of the reservoir description study include: core description for depositional environment, sandbody geometry, paleodirection and lithotype identification; reservoir quality calibration & poroperm analysis; geochemistry, tar history and presence prediction; flow unit identification using tracer data, RFT barriers, production log and well test data; fault conductivity and subseismic fault population; Spearman Rank analysis for fluid flow directionality: and a reinterpreted 3D seismic study.
One of the processes that was developed in building the geological model was to identify field-wide flow units in the upper subzones based only on production data. Data that was used included waterflood tracers, production log analysis for fluid flow fractions, RFT and static pressure data, and well tests. In parallel with the flow unit identification, a sequence stratigraphic model of the upper subzones was developed through the use of core data, wireline logs, and geological analogues. Detailed interpretation of the core produced a revised sedimentological model of the reservoir. In previous models, reservoir quality sandstones were interpreted as fluvial channels. The reevaluation of the core data, which focused on ichnology, suggested that many of these sandstones were mouth bars and that the bulk of the channels were tidally influenced. Careful choice of datums allowed the correlation of flooding surfaces defined by coals and bay silts. Some of these flooding surfaces are underlain by multistorey channel sandstones that can be mapped out into broad belts and are interpreted as valleys.
After merging the flow units and sequence stratigraphic layering, it was seen that the flow barriers were recognizable as flooding surfaces or sequence boundaries. This integrated description is currently being incorporated into a conditioned stochastic object model. A lithotype model has been constructed from the well data and core description work to model the various types of sandbody geometries (mouthbar, valley fill, distribuartary channels). While the valley fill channels would be modeled as objects, the mouth bars are distributed using variograms and a gaussian method of distribution. The model is flexible enough that the channel sandstones can be modeled both as valley fills or as unconfined distributary channels.