Using machine learning to identify the spatial extent of acid mine drainage from gold mining activities in Gauteng, South Africa

Abstract

Acid mine drainage (AMD) poses a significant environmental and public health risk, particularly in mining-intensive regions. This study examines the spatial extent of AMD contamination in the Bloubankspruit and Crocodile River systems in Gauteng, South Africa, by analysing heavy metal concentrations in surface soils along a 20 km stretch downstream of the abandoned Kromdraai gold mine. Soil samples were collected at six locations at 4 km intervals and analysed using X-ray fluorescence (XRF) spectroscopy. The results indicate that AMD contamination is most severe within the first 4 km, with sulphur concentrations reaching 6 911 mg/kg, exceeding the acceptable threshold of 4 000 mg/kg by 72,8%. Cobalt levels were elevated across all sample sites, with the highest recorded concentration at 49.2 mg/kg—over eight times the permissible limit of 6 mg/kg. Chromium and copper concentrations also surpassed regulatory standards at 40,6 mg/kg and 32,6 mg/kg, respectively. A statistical analysis of variance (ANOVA) confirmed significant variation in contamination levels across sites (F = 30,23, p < 0,05), with AMD pollutants decreasing exponentially downstream due to natural dilution and sedimentation processes. The study further demonstrates that the gentle topographical gradient (0,0097) of the region limits surface runoff, constraining AMD migration beyond the immediate vicinity of the mine. The findings emphasize the urgent need for targeted remediation efforts, such as bioremediation and passive treatment systems, to mitigate the long-term environmental impact of AMD. This research provides critical insights into the transport dynamics of mining-related pollutants and supports the development of more effective water quality management strategies in affected river systems.