A combined algorithm of direct sampling method with considering fluid flow in probabilistic simulation of a fractured aquifer

Having a representative fracture network model is a crucial step in the effective simulation of the flow and transport process in a discrete fracture-matrix system. The imprecise modeling of spatial characteristics of fracture networks leads to unacceptable fluid flow simulation results. This is why traditional fracture network modeling methods can only describe the geometry of fractures by statistical probability distributions and ignore spatial variability. In this study, a novel iterative combined approach is implemented, which integrates the direct sampling (DS) method with the fluid flow simulation technique to handle spatial variability reproduction in modeling. The implemented approach is applied to a real micro-scale physical model with a complex structure to verify the efficiency of the applied approach. The novelty lies in optimizing hydrogeological parameters (i.e., calculating hydraulic conductivity) and geostatistical constraints (i.e., training images with different sizes and patterns) simultaneously. The result indicates the effectiveness and stability of the implemented method in the generation of the desired fracture network with representative spatial parameters. In other words, the fulfilled approach satisfies both statistical and spatial variability in fracture network modeling.