Using UAV and LiDAR data for gully geomorphic changes monitoring

Gully formation and evolution represent important aspects not only of landform evolution, but also of practical interest regarding hydrology and agriculture. The classic methodology of assessing the intensity of these erosion processes, and the volumes of sediments involved, is to use field measurements or classical/digital photogrammetry. These methods were recently completed by the use of high-resolution digital elevation models (DEMs) derived from multitemporal LiDAR data and UAV images. In the Moldavian Plateau (northeastern Romania), gullies are common landforms due to geologic, topographic, climatic, and anthropic factors. Their episodic development and the relationship with high rainfall and/or snowmelt events constitute a key point in the deciphering of the gully evolution. For this work we have chosen the case of four gullies developed in the lacustrine deposits of abandoned anthropic reservoirs which presented an obvious dynamic in the last two decades on remote sensing images. A DJI Phantom 4 Pro UAV was flown over the study case areas and acquired images with 80% side and forward overlap at 20 MP resolution. The UAV point cloud was obtained using the structure from motion (SfM) technique in VisualSFM open source software from overlapping images and was georeferenced with ground control points. Georeferenced LiDAR point clouds acquired in winter 2012 were used as a reference dataset. The filtering of the point clouds for obtaining bare ground points was performed with the multiscale curvature classification algorithm. The point cloud ground data for both the sources and periods were used to interpolate a 0.25-m resolution bare earth DEM for every gully. This LiDAR reference DEM was used together with the UAV SfM DEM for deriving the DEM of differences (DoDs) using the geomorphic change detection (GCD) technique of Wheaton et al. (2010) implemented in SAGA GIS and R stat. GCD was applied with both uniform and spatially variable thresholding, the threshold errors being derived from GCPs and from co-registration of DEMs. Geomorphological mapping was performed for establishing the spatially variable thresholds and for assessing the sediment budget. The results highlighted the areas that were affected by erosion and deposition and allowed us to evaluate the process rate for each studied gully, each gully element, and to derive a raw sediment budget, showing that LiDAR, UAV SfM, and DoD are useful methods in geomorphological mapping and rate of process studies.