Integrated Workflow for Parametrization of Fracture Networks in Digital Outcrop Models: Focus on Directional Topology and H/L ratio calculation

Mesoscale fractures, with lengths between meters and hundreds of meters, cannot be effectively characterized in the subsurface, due to limitations of borehole and geophysical datasets. However, large quantitative structural datasets can be collected by combining field and remote sensing techniques in digital outcrop models (DOMs). These data can be then used to constrain stochastic models of subsurface fracture networks with the outcrop analogue approach. However, to date a methodology optimized to characterize all parameters of a fracture network remains elusive. In this contribution we present a workflow that leverages digital outcrop models including both pavement and wall exposures, allowing for a three-dimensional analysis. Different parameters are calculated starting from different types of support, so we collect orientation data on point cloud DOMs (PC-DOMs) of vertical outcrops with semi-automatic methods. These data are classified based on field observations and segmented using a k-medoid approach. The goodness-of-fit to orientation distributions is tested with a proper statistical treatment. Topological parameters are measured on the fracture network digitalized from textured surface DOMs (TS-DOM). Standard topological analysis only provides averaged information on the whole fracture network. In this contribution a novel approach called directional topology is presented, in which every node retains information about the branches that generated it. This not only provides a more comprehensive understanding of the network's connectivity but also allows for the extraction of quantitative parameters about the degree of abutting of a specific fracture set on another (on horizontal outcrops) and on the extent to which a set is stratabound (on vertical outcrops). The trace length, height and spacing distributions are measured with a robust innovative approach, accounting for the censoring bias with survival/reliability analysis. P21 data are collected distributing several grids of scan area with increasing edge length, and the representative elementary area is qualitatively defined. A particular focus will be placed on the calculation of the H/L ratio, often overlooked but of fundamental importance, as it is responsible of the jump in dimensionality in 3D stochastic models.