Exploring the relationship between annual soil loss and formation rate in different land use scenarios using support vector machine (SVM) learning models in Tigray Highlands

Extensive soil degradation in the Tigray Highlands, Ethiopia, threatens agricultural sustainability. This study quantifies the critical imbalance between soil loss and formation in the severely affected Kola Embahasti Watershed. An integrated modeling approach was employed to assess the watershed's soil budget. Soil erosion was estimated using the Revised Universal Soil Loss Equation (RUSLE), with the land use/land cover (LULC) C-factor classified using a Support Vector Machine (SVM) model trained on field and satellite data (Sentinel-2B). Soil formation rates were calculated using the Arrhenius equation to model weathering processes. Statistical analysis was used to determine the relationship between erosion and formation rates. The results reveal a severe soil degradation, with a mean annual soil loss of 61.29 t ha⁻¹ yr⁻¹, far exceeding the mean soil formation rate of 2.45 t ha⁻¹ yr⁻¹. This yields a net annual soil loss of 58.84 t ha⁻¹ yr⁻¹, indicating an unsustainable rate of degradation. A strong negative relationship was found between soil loss and formation, heavily influenced by land cover. Bare land exhibited the highest erosion (94 t ha⁻¹ yr⁻¹) and lowest formation (0.98 t ha⁻¹ yr⁻¹), while dense forest demonstrated the lowest erosion (5.14 t ha⁻¹ yr⁻¹) and highest formation (4.9 t ha⁻¹ yr⁻¹). The RUSLE model, physiochemical soil formation model, and SVM algorithm are integrated for innovative conservation action, recommending targeted interventions e.g., afforestation and sustainable land management.