Improving three-dimensional high-order seismic-stratigraphic interpretation for reservoir model construction: An example of geostatistical and seismic forward modeling of Permian San Andres shelf–Grayburg platform mixed clastic–carbonate strata

Abstract

The three-dimensionally complex, highly progradational mixed siliciclastic–carbonate strata of the San Andres and Grayburg Formations have long been the backbone of conventional hydrocarbon reservoir production from the Permian Basin, and significant recovery continues via waterflooding and CO2 injection. Besides, nonreservoir equivalents of these formations have recently taken increasing significance as produced water disposal targets. However, seismic-stratigraphic interpretations are challenged by complex internal shelfal-stratal geometries and numerous laterally continuous but vertically thin fluid barriers in overlying platforms. We built a three-dimensional (3-D) geocellular model of Guadalupian 8–13 high-frequency sequences (G8–G13 HFSs) and then conducted forward seismic modeling (35-Hz 0° phase). This allows investigations on the validity of applying conventional reflection-geometry–based interpretation to delineate the G9 HFS top and base, which can potentially serve as bounding/constraining surfaces for upper San Andres shelf–Grayburg platform reservoirs. This study contributes to 3-D modeling methodologies by introducing a query tree to select geostatistical methods for modeling dual-scale heterogeneities and by integrating data from diverse sources for seamless and realistic 3-D models. Our seismic-stratigraphic evaluation demonstrates that conventional reflection–geometry-based interpretation does not adequately resolve the G9 top and base; deviations from the geocellular model reach up to 80 m (260 ft) and are thus well beyond the maximum acceptable error limits of ±0.5 wavelength. We suggest improving conventional interpretations of the G9 base by selective interpolation or mixed-polarity event picking near the error-prone shelf margin and upper slope. Besides, instead of picking the highly discontinuous seismic peak as G9 top, bulk-shifting of a shallower trough horizon near actual G10 top should deliver a more accurate surface representing G9 top.