Integrated facies modeling in unconventional reservoirs using a frequentist approach: Example from a South Texas field

Abstract

Modeling of rock types or facies in unconventional reservoirs presents numerous challenges that are not encountered in conventional reservoirs. Because the exploitation of unconventional reservoirs frequently relies on the use of large numbers of long, data-poor horizontal wells, routine tasks in conventional reservoirs, such as petrophysical analyses and rock-physics diagnostics, become problematic in unconventional reservoirs due to insufficient data. Similarly, the inability to generate synthetic seismograms in the horizontal section of the wells makes seismic well ties and time-depth conversions more difficult. Our approach for facies modeling in unconventional reservoirs attempts to overcome these challenges by using the abundant log information available along the vertical pilot well to extract discrete facies indicators (facies flags) from logging-while-drilling information along the horizontals. After performing a time-depth conversion constrained by geosteering information, we estimate facies probabilities using facies flags and prestack elastic inversion results also extracted along the horizontals. As opposed to the Bayesian formalism commonly used for facies probabilities estimation, our frequentist approach estimates probabilities directly from proportions derived from crossplots of inverted elastic properties without using the Bayes formula. No prior information is required. The margin of error in the probabilities is also estimated by applying a well-known formula used to estimate the margin of error in opinion polls. Finally, we use seismic probability results to select only points with high probability and high reliability, which are treated as hard constraints for stochastic facies modeling. The selection of the seismic constraints is also weighted by the vertical facies trend expected for the area of interest. Our final facies model closely follows horizontal and vertical well facies data, vertical proportion curves, and selected seismic constraints. The application of the methodology is illustrated in a carbonate-rich, unconventional reservoir in South Texas.