Assessing climate change influence on soil erosion by utilizing geospatial techniques in Sekhukhune District, South Africa

Abstract

Soil erosion remains a critical environmental issue that undermines food security, biodiversity, and ecosystem stability, particularly in semi-arid regions. This study assessed the influence of climate change on soil erosion in the Greater Sekhukhune District, South Africa. The analysis was conducted during the period from 1994 to 2050 using geospatial and machine learning techniques. The objectives of the study were: to quantify historical soil erosion change (1994–2024) using remote sensing, to analyse long-term rainfall and wind trends and their influence on erosion, and to project future erosion intensity (2026–2050) utilizing the Revised Universal Soil Loss Equation (RUSLE). A multi-method geospatial approach integrating remote sensing, climate trend analysis, and predictive modelling was employed. Landsat imagery was classified using the Random Forest algorithm to map land-cover dynamics and extract eroded surfaces, supported by Normalized Vegetation Index (NDVI) and Bare Soil Index (BSI). Rainfall, Climate Hazards Group Infrared Precipitation with Stations Coupled Model Intercomparison Project (CHIRPS), and wind speed, European Centre for Medium-Range Weather Forecasts Reanalysis Version 5 (ERA5) were examined using the Sequential and Spatial Mann-Kendall tests with Sen’s slope to quantify temporal and spatial trends. Future soil loss was simulated using a RUSLE-based modelling system in Google Earth Engine, combined with CA-Markov land-cover prediction and Random Forest variable-importance assessment under SSP2-4.5 and SSP5-8.5 scenarios. Results revealed a slight but non-significant increase in rainfall (3.3 mm/year) and wind speed (0.004 m/s/year), indicating gradual climate shifts. Eroded areas expanded by 18.69% between 1994 and 2024, driven by vegetation loss, land conversion, and rainfall variability. A strong positive correlation (r = 0.828) between rainfall and erosion confirmed precipitation as the primary driver of soil loss. Future projections using an integrated RUSLE, CA-Markov, and Random Forest model within Google Earth Engine predicted a decline in average soil loss between 2026 and 2050 but an expansion of high-risk erosion zones due to vegetation degradation and urbanization. The cover management factor (C) emerged as the dominant predictor of future erosion. Overall, soil erosion in Sekhukhune, driven by climate and human activities, can be effectively predicted using advanced geospatial tools to guide sustainable land management toward SDG 15.3.