Optimizing UAV-SfM photogrammetry for efficient monitoring of gully erosion in high-relief terrains

Abstract

High-resolution 3D topographic surveys are crucial for detecting topographic changes and combating gully erosion, with Unoccupied Aerial Vehicle-Structure-from-Motion (UAV-SfM) photogrammetry offers a cost-effective solution. However, the influence of key controllable factors-including image collection strategies, ground control points (GCPs), and position and orientation system (POS) constraints-on the accuracy of gully models and monitoring precision remains insufficiently understood. This study evaluates multiple UAV-SfM scenarios to elucidate the factors influencing modeling accuracy and incorporates these insights into a gully erosion monitoring framework. Results indicate that flight altitude and camera angle have minimal impact on model accuracy, with an altitude of approximately 100 m achieving a ∼0.03 m resolution. While oblique imagery is more laborious than nadir imagery, it remains preferred for monitoring internal gully changes as it enhances detail capture. The quantity of GCPs exerts a stronger influence on bundle adjustment than their distribution, with two GCPs per hectare achieving centimeter-level positional accuracy (0.026 ± 0.012 m) but having minimal impact on volume errors. Notably, high-precision POS data impose stronger constraints than GCPs and image collection strategies. In POS-georeferenced scenarios, gully volume error remains lower and stable (0.87 % ± 0.81 %) compared to GCP-only georeferencing and decreases further without GCPs (0.42 % ± 0.40 %). Furthermore, an erosion change detection threshold of 0.086 ± 0.019 m was established as a practical benchmark. These findings validate the feasibility of high-precision POS-georeferenced UAV imagery for gully erosion monitoring, enabling accurate assessments without GCPs or complex image collection strategies and supporting more efficient UAV-based monitoring.