Summary Simple but geologically reasonable and calibrated 2D stochastic models are useful to quantify significant risks and uncertainties associated with alternative-development-well trajectories, particularly when statistical relationships can be established to help quantify those risks and uncertainties, and when the geologic features that create the risks and uncertainties are not adequately addressed within reservoir-flow models. Our example stochastic 2D model considered the naturally fractured depositional-slope region of an isolated carbonate buildup, and the model was populated with relevant features including distributions and geometric details of natural fractures, natural-fracture clustering, and intraformational slope clinoforms that define a mechanically layered sequence. The model was calibrated by use of well-production results and production-logging data so that it reproduced observed well results for cases where the lower sequence boundary does not occur above the oil/water contact (OWC), adding confidence that the model could be used to represent the statistical effect of various alternative trajectories for future wells. Experimental design (ED) was used to determine the significant uncertainties and well-path decisions. Heel and toe elevation and the number of clinoforms encountered by the well were the only significant variables for modeling the frequency of water production. For modeling the frequency of direct well communication to the gas cap, the same variables were significant, in addition to well direction, completion length, and fracture density. The amount of fracture clustering applied in the model was also significant. For our example case, changing the well-elevation profile was effective in managing gas or water risks; however, tradeoffs were evident—and quantified—in attempting to simultaneously address both risks. Minimizing drawdown was not an effective strategy because productivity was low and rarely resulted in economic water-free production if any open fracture connected the well with the aquifer.

This paper illustrates a practical method to incorporate geologically reasonable information and develop a simple but calibrated stochastic model that can be used to quantify some of the significant risks associated with alternative development well trajectories. The example in the paper describes a stochastic, 2D model of the naturally fractured depositional slope region of an isolated carbonate buildup with both gas-oil and oil-water fluid contacts, and illustrates using the model to evaluate the risks of rapid gas or water breakthrough in a horizontal well using observed fracture densities and well production histories to calibrate the model. The model uses sequence boundaries and geologically reasonable distributions of natural fractures, natural fracture clustering, and intraformational, slope clinoforms that define a mechanically layered sequence. Data are deterministic or stochastic, depending on their uncertainty ranges. The model uses an exponential fracture height distribution, and fracture spacing is proportional to fracture height. Natural fractures and clinoforms are distributed independently of well path considerations. Fractures are distributed between the sequence boundaries until the fracture lateral location exceeds model length. Next, the well and fracture geometry are used to determine whether fractures communicate between the well and the gas cap or the aquifer. The number of trials needed for modeling various scenarios was determined using an error minimization analysis. The model was successfully calibrated so that it reproduced observed field results for cases where the lower sequence boundary does not occur above the oil-water contact, adding confidence that the model could be used to represent the statistical impact of various alternative trajectories for future wells. Experimental design was used to determine the significant uncertainties and well path decisions. Heel and toe elevation, and number of clinoforms encountered were the only variables significant for modeling the frequency of water production. For modeling the frequency of direct well communication to the gas cap, the same variables were significant, in addition to well direction, completion length, and fracture density. The amount of fracture clustering applied in the model was also significant. For the example case, minimizing drawdown was not an effective strategy because productivity was low and would rarely lead to water-free production if any open fracture connected the well with the aquifer.