Digital Analogues to Assess Uncertainty of Layering Methods in Reservoir Modelling

Abstract

Abstract Standard reservoir modeling workflows requires the definition of geostatistical parameters, such as variograms or vertical proportion curves, to guide the definition of facies and other properties away from wells. These parameters can be calculated from local data or be defined based on data collected from digital analogs, outcrops or better understood subsurface assets. The well markers and defined layers in the model grid are the backbone of both well analysis for calculating geostatistical properties and posterior population of properties in 3D. We analyze the pitfalls of this methodology and how the proper use of digital analogs can help us reduce bias. Digital analogs of generic reservoirs were generated using stratigraphic forward simulators. These models, which include temporal relationships of facies distributions in 3D space, can be used to benchmark other approaches to modeling. In our tests, we simulate the deposits of fluvial meandering and progradational margin systems. The layers boundaries from these models represent chronostratigraphic horizons. Extracting synthetic wells from the digital analogs allows the effect of different layering, including proportional and conformable, on the calculation of typical parameters used in reservoir modeling to be quantified. As expected, layering selection can play an important role in the definition of geostatistical parameters, introducing significant bias in the way 3D reservoir models are built. The density of synthetic well data versus the relative size of the geological features represented in the digital analogs is another major factor in the ability to properly sample and parametrize the model. Even when using different layering approaches, the calculated parameters are different from the ones calculated on the digital analogs, since standard methods of layering do not necessarily consider temporal correlations. This raises the question of whether it is possible to correctly assess subsurface uncertainty by performing the standard layering workflow, even when different layering scenarios are considered. Geologically realistic layering, derived from plausible temporal well correlations are required. In this study, we applied a new method for assessing uncertainty in reservoir modeling. Incorporating the right digital analogs can be a first step to improve the way well properties are correlated and the rest of the modeling steps that cascade from this.