Simulating flow in fractured porous media with spatially varying fracture statistics using a non-local kernel-based model

In this work we focus on expected flow in porous formations with highly conductive isolated fractures, which are of non-negligible length compared with the scales of interest. Accordingly, the definition of a representative elementary volume (REV) for flow and transport predictions may not be possible. Recently, a non-local kernel-based theory for flow in such formations has been proposed. There, fracture properties like their expected pressure are represented as field quantities. Unlike existing models, where fractures are assumed to be small compared with the scale of interest, a non-local kernel function is used to quantify the expected flow transfer between a point in the fracture domain and a potentially distant point in the matrix continuum. The transfer coefficient implied by the kernel is a function of the fracture characteristics that are in turn captured statistically. So far the model has successfully been applied for statistically homogeneous cases. In the present work we demonstrate the applicability for heterogeneous cases with spatially varying fracture statistics. Moreover, a scaling law is presented that relates the transfer coefficient to the fracture characteristics. Test cases involving discontinuously and continuously varying fracture statistics are presented, and the validity of the scaling law is demonstrated.