The application of curvature attributes on seismic horizons or 3D seismic volumes has been discussed in the literature in several ways. Such discussion largely ignores the detail of parameter selection that must be made by the working interpreter or the expert processor. Parameter selection such as window size and filtering methods for seismic curvature estimates have not been extensively compared in the literature and have never been validated using quantitative ground truthing to log or drilling data. Of even greater relevance to the interpreter is the lack of discussion of curvature parameters as they are relevant to interpretive and operational concerns. We focus on the seismic most-positive curvature attribute, its parameterization, and filtering for the overpressured tight sand target in the Falher F formation of the deep basin of Alberta, Canada. This sand has numerous natural fractures that constitute an occasional drilling hazard due to mud losses. Various parameterizations on horizon- and volume-based curvature extractions are made and examined in the context of the drilling results of four horizontal wells, one of which has image log fracture density along the lateral portion of the well. We compared different lateral (and vertical where applicable) window sizes in the initial curvature estimates, as well as different postcurvature filtering approaches including unfiltered, Gaussian-filtered, and Fourier-filtered products. The different curvature attribute estimates have been evaluated by way of map comparisons, cross-section seismic line comparisons, and correlations with the upscaled fracture density log data. We found that our horizon-based estimates of positive curvature suffered from mechanical artifacts related to the horizon picking process, and the volume-based methods were generally superior. Of the volume-based methods, we found that the Fourier-filtered curvature estimates were the most stable through smaller analysis windows. Gaussian-filtering methods on volumetric curvature gave results of varying quality. Unfiltered volumetric curvature estimates were only stable when very large time windows were used, which affected the time localization of the estimate. The comparisons give qualitative and quantitative perspective regarding the best parameters of curvature to predict the key properties of geologic target, which in this case are the potentially hazardous natural fractures within the overpressured Falher F sandstone.