A multidisciplinary team, composed of stratigraphers, petrophysicists, reservoir engineers, and geophysicists, studied a portion of Boonsville gas field in the Fort Worth Basin of North‐Central Texas to determine how modern geophysical, geological, and engineering techniques could be combined to understand the mechanisms by which fluvio‐deltaic depositional processes create reservoir compartmentalization in a low‐ to moderate‐accommodation basin. An extensive database involving well logs, cores, production, and pressure data from 200‐plus wells, [Formula: see text] [Formula: see text] of 3-D seismic data, vertical seismic profiles (VSPs), and checkshots was assembled to support this investigation. The reservoir system we studied was the Bend Conglomerate, a productive series of gas reservoirs composed of Middle Pennsylvanian fluvio‐deltaic clastics 900 to 1300 ft (275 to 400 m) thick in our project area. We were particularly interested in this reservoir system because evidence suggested that many of the sequences in this stratigraphic interval were deposited in low‐accommodation conditions (that is, in an environment where there was limited vertical space available for sediment accumulation), and our objective was to investigate how fluvio‐deltaic reservoirs were compartmentalized by low‐accommodation depositional processes. Using an extensive well log database (200 plus wells) and a core‐calibrated calculation of rock facies derived from these logs, we divided the Bend Conglomerate interval into ten genetic sequences, with each sequence being approximately 100 ft (30 m) thick. We then used local VSP and checkshot control to transform log‐measured depths of each sequence boundary to seismic two‐way time coordinates and identified narrow seismic data windows encompassing each sequence across the [Formula: see text] [Formula: see text] 3-D seismic grid. A series of seismic attributes was calculated in these carefully defined data windows to determine which attributes were reliable indicators of the presence of productive reservoir facies and which attributes could, therefore, reveal distinct reservoir compartments and potentially show where infield wells should be drilled to reach previously uncontacted gas reservoirs. Our best success was the seismic attribute correlations we found in the Upper and Lower Caddo sequences, at the top of the Bend Conglomerate. These sequences were deposited in a low‐accommodation setting, relative to other Boonsville sequences, and we found that reflection amplitude and instantaneous frequency, respectively, were reliable indicators of the areal distribution of reservoir facies in these low‐accommodation sequences.
A committee neural network for prediction of normalized oil content from well log data: An example from South Pars Gas Field, Persian Gulf
