Introduction
Coal mining continues to play a central role in South Africa’s energy system, contributing significantly to national electricity generation. However, coal-based energy extraction is also associated with substantial environmental pressures, particularly land degradation, ecosystem disruption, and water resource stress. Understanding the spatiotemporal impacts of mining-induced land-use and land-cover (LULC) change is therefore essential for informing sustainable energy planning and environmental governance. Advances in satellite remote sensing, cloud computing, and artificial intelligence provide new opportunities for continuous, large-scale monitoring of energy-related environmental impacts. This study investigates LULC dynamics associated with coal mining activities around the Tendele Coal Mine in Somkhele, KwaZulu-Natal, South Africa, within the broader context of the energy–environment nexus.
Methods
Multi-temporal Landsat satellite imagery spanning the period 2008–2023 was processed using the Google Earth Engine (GEE) cloud-based geospatial analysis platform. Five primary LULC classes—forests, grasslands, built-up areas, water bodies, and mines and quarries—were mapped using a supervised machine-learning approach based on the Random Forest (RF) algorithm. Classification accuracy was evaluated using overall accuracy and Kappa statistics to ensure robust model performance. Post-classification change detection techniques were applied to quantify spatial and temporal transitions among LULC classes. To support future-oriented energy and environmental planning, a Cellular Automata–Markov (CA-Markov) model was implemented to project LULC changes to 2028 based on historical transition probabilities.
Results
The RF-based classification achieved reliable performance, with overall accuracies ranging from 73.62% in 2008 to 93.33% in 2020 and Kappa coefficients indicating substantial to near-perfect agreement, peaking at 0.9123. Change detection analysis revealed that coal mining activities significantly influenced regional land dynamics. The mining footprint expanded markedly between 2008 and 2014, stabilized during the 2017–2020 period, and exhibited a slight decline by 2023, suggesting the influence of regulatory controls and environmental management measures. Forest cover demonstrated an overall net increase by 2023, while grasslands showed an initial expansion followed by subsequent decline. Built-up areas expanded rapidly during periods of intensified mining activity before contracting after 2017. CA-Markov projections indicate continued growth in forest and built-up areas by 2028, alongside a modest expansion of mining and quarrying areas to approximately 7.19 km².
Conclusions
The findings highlight the complex and evolving environmental impacts of coal-based energy extraction on land systems. By integrating remote sensing, cloud computing, machine learning, and spatial modelling, this study demonstrates the effectiveness of AI-enabled geospatial frameworks for monitoring energy-related environmental change. The results provide decision-support insights for policymakers, environmental managers, and energy-sector stakeholders seeking to balance energy security with sustainable land management. Such approaches are essential for supporting environmentally responsible energy transitions in mining-dependent regions.
