A semi-automatic workflow for structural interpretation of large point-cloud Digital Outcrop Models on complex fractured metamorphic rocks (Aosta Valley, Italy)

Characterization of fracture networks, both in fault zones and in the less-fractured background, is essential for the analysis and modelling of mechanical and hydraulic properties of the rock mass (i.e. rock plus fractures). Here we present our experience in characterizing fracture networks and other structural features on large outcrops of different basement and metamorphic cover units in the Penninic, Austroalpine and Helvetic units of the Aosta Valley. These units show a variety of lithological, mechanical, and rheological characteristics and were subjected to different ductile and brittle tectonic evolution, resulting in complex combinations of compositional layering, metamorphic schistosity, and fracture networks.

Our methodology is based on a combination of traditional field surveys and remote-sensing techniques such as ground-based and UAS photogrammetric surveys, and terrestrial or helicopter laser scanning. The first task, whose importance is too often overlooked, is represented by selecting outcrops that are representative in terms of structural and lithological properties of a larger rock volume, based on a thorough knowledge of regional structural geology and tectonics. The field survey is carried out with traditional techniques, paying attention to the kinematics, relative chronology, and mineralization (e.g. veins or mineral coatings) of structures. These features, that are often overlooked in fracture studies, are fundamental to frame the evolution of a complex schistosity and fracture network, to separate tectonic fractures with respect to those related to slope dynamics, and to develop predictive models of fracturing at depth (where slope-related fracture will not be present). At the same time, remote-sensing datasets are collected. The choice of the survey technique (terrestrial vs. aerial, photogrammetry vs. laser scanning) depends on various conditions, but in all cases the output is a point cloud DOM, colorized with RGB values, that should have a density (points/area) sufficient to characterize the smallest relevant structural features. From this, also textured surface DOMs and/or DEM plus orthophotos (for almost flat outcrops) can be obtained.

The first step of DOM analysis is carried out “manually”, selecting facets and traces with suitable software tools (e.g. Compass plugin in CloudCompare). This allows selecting different sets of structures, characterizing their orientation statistics, and assigning them to sets defined in the field (with kinematics, chronology, etc.). This step also allows understanding how well the structural features recognized in the field are represented in the DOM. The second step of DOM analysis consists in an automatic segmentation (in case of a point cloud) or tracing (in case of a DEM of triangulated surface textured with images) with algorithms calibrated with results of the manual interpretation. Overall, this results in a supervised semi-automatic workflow, allowing to extract huge structural datasets in a reasonable time, maintaining the connection with kinematic and chronological observations carried out in the field.

The fracture datasets can be eventually characterized with tools allowing to measure statistical distributions of different parameters of the fracture sets using virtual scanlines and/or scanareas, and these distributions can be used to model different properties of the fracture networks or generate stochastic DFN models.