Hydraulic fracture mapping by locating microseismic events related to rock fracturing is used to evaluate the effectiveness of the stimulation in low-permeability reservoirs. The geometry of the events is used to infer fracture orientation, particularly in the case where events line up along an azimuth, or have a planar distribution in 3 dimensions. When the induced the microseismic events have a low signal-to-noise ratio (either due to low magnitude or propagation effects) their locations can have a high degree of uncertainty. Low signal-to-noise events are not as accurately located in the reservoir, or are not detected at all, so that the extent of fracture stimulated reservoir may be underestimated. In the Bakken Formation of the Williston Basin, we combine geological analysis with process-based and stochastic fracture modeling to build multiple possible fracture model realizations. Specific parameters in the models can be modified while honoring a realistic range for each parameter in order to explore the range of uncertainty. Fracture flow properties generated from the fracture models are validated via history matching iterations. The validated fracture models, in turn, provide a means to calculate a geometrically-constrained volume of rock and fracture permeability that can be used for estimating production. This paper presents a methodology for deciding which fracture parameters to vary (the high sensitivity parameters) in order to minimize the number of realizations that need to be generated during the model conditioning phase of a reservoir simulation program.