Past trends and future projections: a spatio-temporal analysis of land cover change in the uMngeni catchment using remote sensing and CMIP6 scenarios

Understanding long-term land-cover dynamics in the uMngeni catchment is critical for safeguarding ecosystem integrity and sustaining biodiversity under increasing anthropogenic and climatic pressures. However, spatially explicit assessments of these transformations remain limited, particularly under future climate change. This study analysed land-cover changes from 1991 to 2021 using Landsat (5, 7, and 8) imagery and random forest classification. Future land-cover dynamics were simulated for 2031 by integrating CMIP6-derived climate projections with a Cellular Automata–Artificial Neural Network model under SSP126, SSP245, and SSP585 scenarios. The random forest classification achieved overall accuracies of 81.7%, 83.1%, 82.2%, and 88.0% for 1991, 2001, 2011, and 2021, respectively, with corresponding kappa coefficients of 0.77, 0.79, 0.78, and 0.85. Between 1991 and 2021, grassland and forest cover declined by 17.9% and 6.6%, respectively, while cultivated and built-up areas increased by 6.9% and 12.4%. Projections indicate continued losses of grasslands, wetlands, and natural vegetation under all scenarios, with greater declines under higher-emission pathways. Cultivated and built-up areas expanded across scenarios, with the largest increases under SSP585. These findings highlight increasing pressure on ecosystem stability and provide spatially explicit evidence to support integrated land-use planning, biodiversity conservation, climate-resilient management, and sustainable resource management in the uMngeni catchment.