Gis-based gully erosion monitoring and simulation

Abstract

Soil erosion, especially of valuable arable land, is a growing global issue, exacerbated by climate change. A major concern is the irreversibility of soil erosion, with annual soil loss often outpacing new soil formation. While sheet erosion leads to significant topsoil loss, gully erosion is particularly concerning due to its severe impact on landscapes. To mitigate gully erosion, identifying the most vulnerable areas is crucial. In this study, machine learning methods such as Random Forest, XGBoost, and SVM were used to map gully erosion susceptibility in Nepal. Various factors, including land use, soil, geology, and slope, were considered. The model performance was assessed using the Area Under the Curve (AUC) score, with initial results showing scores between 76% and 84%, indicating good predictive power across all methods. Beyond susceptibility mapping, simulating and predicting future gully development is essential for long-term erosion management. A QGIS plugin was developed to model gully erosion over time, requiring a digital elevation model (DEM) and soil parameters as inputs. The plugin predicts potential gullies by extrapolating drainage patterns from the DEM and calculates changes in gully depth, width, and volume. The results are then integrated into the DEM for 3D visualization. Validation of the tool on gullies in South Africa showed good alignment between predicted and observed gully dimensions, though some dynamic aspects varied. This approach provides a valuable tool for identifying at-risk areas and simulating future gully erosion, aiding in the development of effective mitigation strategies.