Spatial assessment of soil degradation under intensive agricultural pressure in the central loess plateau

Abstract

Soil degradation (SD) threatens the sustainability of semi-arid agricultural systems in the Central Loess Plateau, where intensive cultivation and repeated land-use cycles accelerate soil deterioration. This study developed a spatially explicit and uncertainty-aware framework for assessing and mapping SD by integrating soil quality indicators (SQIs), environmental data, remote sensing, and Random Forest (RF) modelling. A total of 350 surface-soil samples were collected and analyzed. Principal component analysis and the minimum data set (MDS) approach were used to identify representative soil-quality indicators. The final MDS included clay, soil bulk density, pH, soil organic carbon, calcium carbonate equivalent, cation exchange capacity, potassium, and sodium adsorption ratio. Among these indicators, pH, SAR, CEC, SOC, K, and clay received the greatest weights, highlighting the importance of soil chemical conditions, sodicity, nutrient retention, organic matter, and texture in soil quality assessment. Soil degradation was quantified by comparing SQI values of agricultural soils with reference values from natural ecosystems. Environmental predictors were then incorporated into an RF model to estimate the spatial distribution of SD and assess predictor importance. The model showed good predictive performance, with R 2 values of 0.83 for training and 0.76 for validation, and RMSE values ranging from 0.031 to 0.034. Severe degradation was mainly concentrated in the central and southwestern regions, reflecting the cumulative effects of intensive cultivation, nutrient depletion, and SOC loss. Prediction-interval analysis identified areas with greater uncertainty and priority for additional sampling. The proposed framework provides a practical basis for spatially targeted soil conservation, sustainable agricultural management, and long-term ecosystem resilience. Future research should incorporate biological indicators and temporal monitoring to improve SD prediction under changing climatic and land-use conditions.