A stochastic framework for mapping probable locations of discrete subsurface conductive pathways facilitating mass migration

Abstract

Discrete hydro(geo)logic features such as fractures and faults in subsurface formations critically influence the migration of dissolved-phase constituents, often serving as preferential pathways that complicate resource management planning and efforts. Robust characterization and representation of these features is crucial for designing effective management strategies and ensuring successful decision making. To address these challenges, we developed a novel stochastic modeling framework that couples groundwater flow and dissolved phase transport with a non-stationary categorical parameterization within an ensemble data assimilation scheme (PESTPP-iES), enabling improved characterization of such system controlling features. Using a hypothetical scenario involving an interbedded sandstone and fractured shale sequence, the framework was evaluated across various fracture configurations that represent differing levels of system property complexity and uncertainty. By creatively coupling distinct simulation models, the framework was able to capture highly localized, fine-scale fractures, while maintaining an efficient coarser scale representation of the primary aquifer units. Data assimilation involved conditioning the initial prior ensemble with geological information, followed by history matching with state (i.e., head and concentration) measurements from the aquifers. Results revealed up to 80 % reduction in spatial uncertainty of probable fracture locations, demonstrating the importance of assimilating all available information to aid in delineating probable damage zones. The framework is designed to be implemented within an iterative modeling workflow, whereby additional site data can be easily assimilated to progressively refine and constrain the probable locations of fractures through successive analyses.