Abstract: The Use of Categorical Indicator Geostatistics for Modeling Facies in Sand-Rich Turbidite Systems: An example from the Deep-Water Gulf of Mexico

David J. Goggin, Douglas L. Jordan

The Green Canyon 161/205 prospect, located 150 miles offshore Louisiana in the central Gulf of Mexico at 2500-3000 ft water depths, has between 350-425 million barrels of original oil in place that is stratigraphically and structurally trapped within lower Pleistocene and upper Pliocene turbidites. Overall sand content within the major pay zones is greater than 80%. The vertical lithofacies variations, seismic facies maps, and architecture of the primary reservoirs in the prospect are characterized using data available from conventional well logs, whole core, pressure transient tests, and a 3D seismic survey. Due to the high costs associated with developing such a prospect, an extensive 3D reservoir simulation evaluation program was employed. In this study we outline the ole of geostatistical methods in the creation of these 3D simulation models.

A two-stage approach was used to capture the effects of geological heterogeneity on fluid flow. In the first stage, large-scale seismic facies variations, which could be identified through correlations between log interval averages and seismic amplitudes, were first identified and mapped for each reservoir zone. Inherent facies map uncertainty was modeled using a new geostatistical "facies indicator" simulation method (i.e., PDF or probability density function simulation). The seismic facies maps were first sampled as point values at pseudo-well locations. Information from these selected points assisted in creating "facies indicator variograms" for each facies identified in the seismic maps. Next, the pseudo-wells, along with the sparse actual wells, were treated as "hard" data points in geostatistical simulations of equiprobable facies scenerios.

The second stage of model construction involved translating fine-scale heterogeneities observed in core and logs into synthetically-generated traces of turbidite sedimentation. These synthetic traces were then combined with lateral correlation length scales obtained from outcrop studies to condition geostatistical cross-sections of porosity and permeability. This procedure was performed for each seismic facies type, and flow simulations were used to scale-up effective flow properties for the full-field 3D simulations. Scaled-up absolute permeabilities were derived from single-phase flow simulations, and scaled-up, two-phase pseudo-relative permeability curves were generated from two-phase flow simulations by the Kyte and Berry (1975) method.

AAPG Search and Discovery Article #90956©1995 AAPG International Convention and Exposition Meeting, Nice, France