Modeling Reservoir Heterogeneity in a Fluvial Sandstone: The Gypsy Sandstone of Oklahoma

Reservoir simulation models based on detailed outcrop description provide a means to evaluate the importance of various types of heterogeneity on reservoir performance. This study models tracer flow in a multistoried fluvial sequence, the Pennsylvanian Gypsy sandstone. To better understand the effects of heterogeneity on a subsurface fluvial reservoir, three-dimensional reservoir simulation models were constructed incorporating information from correlative rocks which outcrop fifteen miles from the subsurface site. The heterogeneity considered in this study is the higher-order architecture of the Gypsy i.e. its channel sandbody configuration.

At the five acre subsurface site, six wells penetrate the Gypsy from 890 ft to 1000 ft measured depth and encounter a depositionally similar fluvial sequence as at outcrop. The wells comprise an inverted five spot with one observation well. Within this site, Gypsy net sandstone varies from 27 ft to 60 ft. Cores indicate that the number of channel sandbodies varies from three to as many as six. Gypsy channel sandbodies can be subdivided into three intervals primarily on correlation of laterally continuous mudstone and siltstone bodies. Within these intervals, correlation of individual channel sandbodies is uncertain.

Due to uncertainty in channel sandbody distribution, the general approach is to develop multiple reservoir models. Each of these models honors well data but differ in configuration of channel sandbody orientation. However, for each, interpretation of channel sandbody configuration is constrained by information on channel patterns from the outcrop and by local net sandstone trends surrounding the subsurface site.

The tracer test simulations constitute a portion of planning and design for a contrasting salinity flood of the Gypsy at the subsurface site. The geological models used for this simulation study are based on outcrop studies. Because the present focus is on the effects of channel architecture, permeabilities and porosities were modeled as constant within channels. Channel bases were modeled as partial barriers for which permeabilities were determined from a summary of core measurements. The objectives of this simulation study are to determine whether differences in channel configurations are detectable by a tracer response and to assess the magnitude of the impact of differing channel configurations on tracer response.

The tracer test simulation entails using a particle tracking method to control numerical dispersion. A slug of tracer is represented by a cloud of particles introduced around an injection well during a period of tracer injection. Each particle is characterized by its initial (x, y, z) coordinates. To simulate flow of the tracer, each particle is moved in small steps according to the local fluid velocity and the porosity. The grid size used to model the reservoir descriptions was 25 × 25 × 27 which is comprised of 30 ft × 30 ft × 3 ft blocks.

Results of the simulations indicate that channel configuration plays a major role in affecting tracer response and breakthrough times. Differences in channel sandbody orientation not only impact the signature of tracer response curves but also impact tracer breakthrough times by as much as 33% of the average breakthrough time (138 days). Even minor variation in channel configuration may affect tracer response by changing flow capacities in the flooded area. These results demonstrate that the path of fluid flow from the injector to producer is significantly affected by differences in channel orientations. Examples of differences in tracer responses and breakthrough times between different channel configurations are shown.

Background information for preparation of the models was obtained from the outcrop by mapping the fluvial architecture of six channel sandbodies on the parallel walls of a 1000 ft roadcut using photomosaics. Macroforms identified within sandbodies are lateral accretion deposits modified by chute channels. Channel sandbodies decrease in size upward from 600 ft wide and 30 ft thick to 200 ft wide and 10 ft thick. Erosional surfaces comprising the bases of the channels could be traced longitudinally 1100 ft into a grid of 22 shallow boreholes; lower-order features could not be correlated. Laterally, the channel sandbodies thin into floodplain mudstones and siltstones.

Outcrop mapping of features within channel sandbodies provides a background for further assessment of the impact of smaller scale heterogeneity. A detailed two-dimensional description of the exposure also provides a basis for measuring the spatial variability of reservoir properties and for assessing and calibrating geostatistical methods which have been proposed for use in developing reservoir models. Each channel-fill sequence is comprised of up to 5 sandstone facies in addition to laterally and vertically adjacent mudstones and siltstones. All of these facies were extensively sampled to determine their porosity and permeability characteristics. Spatial distribution of these facies is variable, and their lateral extent is commonly less than the spacing between borings. Spatial distribution of porosity and permeability differs by facies.