Exploring structures of the Rochefort Cave (Belgium) with 3D models from LIDAR scans and UAV photoscans.

Amongst today's techniques that are able to produce 3D point clouds, LIDAR and UAV (Unmanned Aerial Vehicle) photogrammetry are probably the most commonly used. Both methods have their own advantages and limitations. LIDAR scans create high resolution and high precision 3D point clouds, but such methods are generally costly, especially for sporadic surveys. Compared to LIDAR, UAV (e.g. drones) are cheap and flexible to use in different types of environments. Moreover, the photogrammetric processing workflow of digital images taken with UAV becomes easier with the rise of many affordable software packages (e.g., Agisoft PhotoScan, MicMac, VisualSFM). In this canvas, we present a challenging study made at the Rochefort Cave Laboratory (South Belgium) comprising surface and underground surveys. The main chamber of the cave ( 10000 m³) was the principal target of the study. A LIDAR scan and an UAV photoscan were acquired underground, producing respective 3D models. An additional 3D photoscan was performed at the surface, in the sinkhole in direct connection with the main chamber. The main goal of the project is to combine this different datasets for quantifying the orientation of inaccessible geological structures (e.g. faults, tectonic and gravitational joints, and sediments bedding), and for comparing them to structural data surveyed on the field. To go through structural interpretations, we used a subsampling method merging neighboured model polygons that have similar orientations, allowing statistical analyses of polygons spatial distribution. The benefit of this method is to verify the spatial continuity of in-situ structural measurements to larger scale. Roughness and colorimetric/spectral analyses may also be of great interest for several geosciences purposes by discriminating different facies among the geological beddings. Amongst others, this study was helpful to precise the local petrophysical properties associated with particular geological layers, what improved interpreting results from an ERT monitoring of the karst hydrological processes in terms of groundwater content.