Integrating deterministic lithostratigraphic models in stochastic realizations of subsurface heterogeneity. Impact on predictions of lithology, hydraulic heads and groundwater fluxes

摘要

Realistic representations of geological complexity are important to address several engineering and environmental challenges. The spatial distribution of properties controlling physical and geochemical processes can be effectively described by the geological structure of the subsurface. In this work, we present an approach to account for geological structure in geostatistical simulations of categorical variables. The approach is based on the extraction of information from a deterministic conceptualization of the subsurface, which is then used in the geostatistical analysis for the development of models of spatial correlation and as soft conditioning data. The approach was tested to simulate the distribution of four lithofacies in highly heterolithic Quaternary deposits. A transition probability-based stochastic model was implemented using hard borehole data and soft data extracted from a 3-D deterministic lithostratigraphic model. Simulated lithofacies distributions were also used as input in a flow model for numerical simulation of hydraulic head and groundwater flux. The outputs from these models were compared to corresponding values from models based exclusively on borehole data. Results show that soft lithostratigraphic information increases the accuracy and reduces the uncertainty of these predictions. The representation of the geological structure also allows a more precise definition of the spatial distribution of prediction uncertainty, here quantified with a metric based on Shannon information entropy. Correlations between prediction uncertainties for lithofacies, hydraulic heads and groundwater fluxes were also investigated. The results from this analysis provide useful insights about the incorporation of soft geological data into stochastic realizations of subsurface heterogeneity, and emphasize the critical importance of this type of information for reducing the uncertainty of simulations considering flux-dependent processes. (C) 2015 Published by Elsevier B.V.