Search

Soil erosion poses a significant threat to sustainable land management and agricultural productivity. Addressing this issue requires advanced predictive models that can accurately identify areas at risk and inform soil conservation strategies. This study focuses on the development and interpretation of well-optimized deep learning (DL) models to predict soil erosion probability, aiming to enhance decision-making in land management. Utilizing the Revised Universal Soil Loss Equation (RUSLE) in conjunction with ground-truthing, we identified critical erosion-prone areas. To predict soil erosion probability, we employed Bayesian optimization to fine-tune Deep Neural Network (DNN), Convolutional Neural Network (CNN), Fully Connected Neural Network (FCNN), and DNN-CNN hybrid models. These DL models were verified using a set of metrics. SHAP value analysis as a means of explainable artificial intelligence (XAI) was used to interpret these DL models for better decision-making. RUSLE estimations and ground truthing highlight that Soil erosion rates in the northeastern and northwestern regions are nearing the highest observed at 25 tonnes per hectare annually, largely due to steep slopes and limited vegetation. In contrast, the southern and southeastern areas have lower erosion rates, due to denser vegetation and gentler slopes. Deep learning models, optimized using Bayesian methods, demonstrate high performance in spatially modeling soil erosion probability. The DNN model achieved an accuracy of 0.93, a precision of 0.92, and an F1-score of 0.94, identifying 222.73 sq. km as highly susceptible to erosion, which indicates its strong ability to detect true erosion events. The CNN model identified 49.68% of the study area (503.30 sq. km) as high-risk, with an accuracy of 0.90 and a precision of 0.91. The FCNN model showed a balanced risk distribution, indicating 37.04% of the land (375.25 sq. km) as very low risk and 36.43% (369.11 sq. km) as very high risk, with an accuracy of 0.91. The DNN-CNN hybrid model highlighted 41.58% of the area (421.20 sq. km) as high risk, demonstrating its effectiveness in capturing spatial patterns of erosion susceptibility. SHAP value analysis indicates that land use and soil type (LULC and K-factor) are crucial in erosion predictions, with LULC having a significant predictive influence in the DNN model. These insights facilitate the prioritization of soil conservation measures, enabling decision-makers to focus on the most impactful factors for mitigating soil erosion.

Soil erosion is a major problem in arid regions, including the Abha-Khamis watershed in Saudi Arabia. This research aimed to identify the soil erosional probability using various soil erodibility indices, including clay ratio (CR), modified clay ratio (MCR), Critical Level of Soil Organic Matter (CLOM), and principle component analysis based soil erodibility index (SEI). To achieve these objectives, the study used t-tests and an artificial neural network (ANN) model to identify the best SEI model for soil erosion management. The performance of the models were then evaluated using R2, Root Mean Squared Error (RMSE), Mean Squared Error (MSE), and Mean Absolute Error (MAE), with CLOM identified as the best model for predicting soil erodibility. Additionally, the study used Shapley additive explanations (SHAP) values to identify influential parameters for soil erosion, including sand, clay, silt, soil organic carbon (SOC), moisture, and void ratio. This information can help to develop management strategies oriented to these parameters, which will help prevent soil erosion. The research showed notable distinctions between CR and CLOM, where the 25–27% contribution explained over 89% of the overall diversity. The MCR indicated that 70% of the study area had low erodibility, while 20% had moderate and 10% had high erodibility. CLOM showed a range from low to high erodibility, with 40% of soil showing low CLOM, 40% moderate, and 20% high. Based on the T-test results, CR is significantly different from CLOM, MCR, and principal component analysis (PCA), while CLOM is significantly different from MCR and PCA, and MCR is significantly different from PCA. The ANN implementation demonstrated that the CLOM model had the highest accuracy (R2 of 0.95 for training and 0.92 for testing) for predicting soil erodibility, with SOC, sand, moisture, and void ratio being the most important variables. The SHAP analysis confirmed the importance of these variables for each of the four ANN models. This research provides valuable information for soil erosion management in arid regions. The identification of soil erosional probability and influential parameters will help to develop effective management strategies to prevent soil erosion and promote agricultural production. This research can be used by policymakers and stakeholders to make informed decisions to manage and prevent soil erosion.