SLOPE MONITORING AND CHANGE DETECTION WITH TIME SERIES REMOTE SENSING DATA FOR AN ACTIVE SOIL LANDSLIDE ALONG I-40 IN EASTERN TENNESSEE

Terrain modeling with unmanned aerial systems (UAS) is an essential tool for slope stability research. UAS data can produce low-cost, high-accuracy digital elevation models (DEMs), orthoimagery products, and raster derivatives to detect landslide features and slope changes. Time series data collected at user-defined intervals can be compared to produce a DEM of difference (DoD) raster of relative change in slope elevation, but multi-source data from differing aerial LiDAR and UAS methods are often misaligned, making change detection ineffective. In this study, we monitored slope change at a 57-acre slow-moving rotational landslide site along I-40 north of Rockwood, TN. Sliding was initially encountered at the site during interstate construction in 1967. In-ground monitoring indicated persistent movement despite remediation attempts. For change detection, a 2015 LiDAR DEM was compared to two UAS datasets collected roughly one year apart during leaf-off. Photogrammetry images were collected in spring 2023 with a DJI Mavic 3 Pro. In winter 2024, UAS LiDAR data were collected with a DJI Matrice L1. Images were processed in Agisoft Metashape and LiDAR data were processed in DJI Terra, as DEMs and orthomosaics. Spatial alignment was corrected between all datasets in CloudCompare, enabling precise comparison of DEMS. Slope elevation profiles were compared along a transect of concern, and a geologic cross-section was created based on available subsurface data for limit equilibrium analysis in RocScience Slide2. Inspection of 3D models, DoD rasters, and profiles indicated areas of active sliding across the site, including a shallow slump. Limit equilibrium analysis found a deep-seated rotational failure surface with Factor of Safety of 1.17 and confirmed localized areas of slumping with factor of safety less than 1. This UAS slope monitoring procedure can be used to perform accurate change detection for past and future remote sensing datasets acquired from diverse sources.