Facies-guided 3-dimensional seismic modeling and reservoir characterization

A set of 3-D seismic models has been generated for a lower Miocene progradational micro-tidal shore-zone system at Powderhorn field, Calhoun County, Texas. These models are based on detailed mapping of facies and rock-properties that involves 250 m of stratigraphic section over an area of 13 by 9 km. Examination of logs from 115 wells shows that fourteen sandy depositional units averaging 3-30 m in thickness are encased in 15 shale units. Effective porosity, P-wave velocity and bulk density are calculated from suites of wireline logs in five recent wells, and then expanded to older wells through regression with SP and resistivity logs. Property mapping in two dimensions is accomplished by employing the directional weighted average method (DWA), which simulates the hand mapping of a geoscientist by incorporating the directional and geometrical information gained from facies analysis into gridding. This biased mapping approach produces a striking visual impact compared to unbiased computer mapping. A predictogram analysis also confirmed that DWA is better than other computer algorithms for generating sandstone thickness maps, though rock-property pedictions were not improved. The predictogram measures information content of both wells and seismic data, and can help to explain the mistie between synthetic and real seismic traces at well locations. Idealized 3-dimensional seismic models are generated by convolving a Ricker wavelet with the primary reflection coefficient series from stacked 2-D P-wave velocity and impedance models. The idealized stratal slices show that to achieve the best image of depositional facies, seismic frequency should be tuned to the maximum sandstone thickness. Framework sand bodies surrounded by muddy deposits are well illustrated. Clean sandstone and pure shale have similar impedance, creating ambiguous amplitude patterns. The stratal resolution of seismic imagery depends not only on wavelet frequency, but also on the magnitude of geological interference. This 3-D model also provides a realistic basis for evaluating the expected hydrocarbon detectability for the reservoirs in this evironment. Examples illustrate that the presence of gas should produce stratal slice amplitude anomalies that are very large compared to the facies-related amplitude variations. Seismic events do not necessarily follow geological time lines. The correlation error of peak picking is often large enough to distort reservoir facies image on stratal slices. In some cases, event crossing connects sand bodies from different units to form a new event. An event tying as many as six depositional units over 50 m has been observed. Phantom mapping produces truer stratal slices, making it possible to map detailed depositional facies within reservoir sequences from 3-D seismic dataGeological Science