Modeling dolomitized carbonate-ramp reservoirs: A case study of the Seminole San Andres unit - Part II, Seismic modeling, reservoir geostatistics, and reservoir simulation

In part I of this paper, we discussed the rock-fabric/petrophysical classes for dolomitized carbonate-ramp rocks, the effects of rock fabric and pore type on petrophysical properties, petrophysical models for analyzing wireli ne logs, the critical scales for defining geologic framework, and 3-D geolo gic modeling. Part II focuses on geophysical and engineering characterizati ons, including seismic modeling, reservoir geostatistics, stochastic modeli ng, and reservoir simulation. Synthetic seismograms of 30 to 200 Hz were generated to study the level of seismic resolution required to capture the high-frequency geologic features in dolomitized carbonate-ramp reservoirs. At frequencies <70 Hz, neither t he high-frequency cycles nor the rock-fabric units can be identified in sei smic data because the tuning thickness of seismic data is much greater than the average thickness of high-frequency cycles of 6 m. At frequencies >100 Hz, major high-porosity and dense mudstone units can be better differentia ted, while the rock-fabric units within high-frequency cycles can be captur ed at frequencies higher than 200 Hz. Seismic inversion was performed on the 30- to 200-Hz synthetic seismograms to investigate the level of seismic resolution required to recover the high -resolution inverted impedance logs. When seismic data were noise free, wav elets were known and sampling rates were high; deconvolution techniques yie lded perfect inversion results. When the seismic data were noisy, the inver ted reflectivity profiles were poor and complicated by numerous high-freque ncy spikes, which can be significantly removed using the moving averaging t echniques. When wavelets were not known, the predictive deconvolution gave satisfactory inversion results. These results suggest that interwell inform ation required for reservoir characterization can be recovered from low-fre quency seismic data by inversion. Outcrop data were collected to investigate effects of sampling interval and scale-up of block size on geostatistical parameters. Semivariogram analysi s of outcrop data showed that the sill of log permeability decreases and th e correlation length increases with an increase of horizontal block size. P ermeability models were generated using conventional linear interpolation, stochastic realizations without stratigraphic constraints, and stochastic r ealizations with stratigraphic constraints. The stratigraphic feature of up ward-shoaling sequences can be modeled in stochastic realizations constrain ed by the high-frequency cycles and rock-fabric flow units. Simulations of a fine-scale Lawyer Canyon outcrop model were used to study the factors affecting waterflooding performance. Simulation results show th at waterflooding performance depends strongly on the geometry and stacking pattern of the rock-fabric units and on the location of production and inje ction wells.