Assessing Soil Erosion Potential At Harvest Sites Using High-Resolution Drone Imagery, Deep Learning, and Geographic Information Systems

Conventional methods for estimating soil erosion from a harvest site involve on-the-ground surveys that are time-consuming and labor-intensive. With the advancement of technology in forestry, we aimed to use drone technology to estimate soil erosion from a harvest site. This study presents an innovative approach to estimating soil erosion within harvest sites through the integration of high-resolution imagery, deep learning techniques, and geographic information systems (GIS). Conducted across ten distinct harvest sites in Alabama, using deep learning, we first performed a drone orthomosaic map classification of harvest sites into five categories, including roads, skid trails, landings, Streamside Management Zones (SMZs), and clear-cut areas, to assess soil erosion potential respective to each category. Rainfall data were downloaded from the National Oceanic and Atmospheric Administration (NOAA) Climate Data Online website, soil erodibility factors data were assessed from the UC Davis soil survey website, the slope data were assessed from drone-derived Digital Elevation Models (DEMs), and crop management factors for each site were derived from the field survey in line with the Universal Soil Loss Equation (USLE) Forest guidelines. The potential soil erosion rate from each site was estimated using Revised USLE (RUSLE) in ArcGIS Pro. The study highlights the effectiveness of combining high-resolution drone imagery with advanced computational techniques to enhance the accuracy and efficiency of soil erosion estimation in forestry practices.