Characterization of complex fluvial architecture through outcrop studies – dealing with intrinsic data bias at multiple scales in the pursuit of a representative geomodel

Abstract

Abstract The practice of building analog models and training images from outcrop exposures is an important tool in better predicting subsurface facies distribution in the petroleum industry. As with subsurface data, however, incomplete information and data bias can lead to inaccurate characterization of outcrop geology at multiple scales. Cretaceous fluvial strata of Wyoming offers excellent exposure of two systems — the sand-rich and highly amalgamated Trail Member of the Ericson Sandstone and the sand-poor, isolated channels of the Dry Hollow Member of the Frontier Formation. For each system, multiple outcrops were characterized through the traditional means of stratigraphic column measurement, as well as through photogrammetric survey acquisition and interpretation. We saw in both studies that, despite an effort to measure sections that were representative of the entire outcrop, measured sections consistently overestimated the reservoir proportions. Ten measured sections within the Trail Member show a Net-to-Gross (NTG) ranging from 50–80% sandstone, with an average of 72%. A more complete spatial characterization of the entire outcrop through photogrammetric interpretation suggests a much lower NTG of 53%. Similarly, for the Dry Hollow Member fluvial strata, measured sections show NTG ranges of 8–50% with an average of 37% sandstone, while the photogrammetric model shows a NTG of only 16%. These differences are significant and lead to very different reservoir models. Further, the assumption is commonly made that the outcrop, if well characterized, is representative of the formation at a larger scale. Models of the Dry Hollow Member at Cumberland Gap show that this is a tenuous assumption and can lead to models that are not representative of the system. Outcrops of the Dry Hollow are sparse and often discontinuous, and extrapolation of calculated facies proportions between two well-exposed outcrops at Cumberland Gap led to significant placement of sands between the outcrops, where the lack of exposure leads to a lack of control data in the model. This resulted in increased reservoir connectivity that is not representative of the system, and shows that even on a sub-kilometer scale, the extrapolation of detailed, quantitative facies proportions can be inappropriate, and if done blindly can lead to an inaccurate characterization of the system. Through detailed characterization of the Trail and Dry Hollow fluvial systems, it is shown that building quantitative geomodels from outcrop exposures, even using modern techniques such as photogrammetric analysis, can be subject to significant bias and mischaracterization at multiple scales and for multiple reasons if care is not taken.