In this study, we develop a model discrete fracture network (DFN) for the unconventional, naturally fractured Tensleep Sandstone oil reservoir at Teapot Dome, Wyoming. Reservoir characterization is based on three-dimensional (3D) seismic data, fracture image logs from Teapot Dome, and field observations of the Tensleep exposure in the Alcova anticline and Fremont Canyon areas. Image logs reveal that the dominant reservoir fracture set trends parallel to the present-day maximum horizontal compressive stress (S-Hmax) inferred from drilling induced fractures. Analog field studies of the Alcova anticline and Fremont Canyon suggest fracture heights and lengths are power-law distributed, while the fracture spacing distribution is best described as log-normal. Image-log-derived fracture apertures are also log-normally distributed. These properties are incorporated into a model DFN. We assume subseismic folds, faults, and fracture zones control fracture intensity distribution and use composite 3D seismic attributes to locate subtle changes in seismic response interpreted to result from subseismic structure. Directional curvature defines aperture-opening strain normal to the dominant reservoir fracture set. Seismic attributes are scaled and combined to control fracture intensity variations in the model. Grid-cell porosity and permeability distributions derived from the DFN suggest the presence of northeast-southwest-trending reservoir compartments. We suggest that enhanced oil recovery operations may be optimized using lateral CO2 injection and production wells oriented along interpreted compartment boundaries at high angles to S-Hmax. This combination of CO2 injection and production laterals could help maximize CO2 storage and hydrocarbon recovery in depleted reservoirs and in down-dip residual oil zones.