The Challenge
The legacy of gold mining in Central Victoria has left behind large amounts of waste materials which can threaten the environment and human health due to high levels of arsenic, mercury, lead and other metals. These wastes’ location, extent and composition are often unknown or poorly documented. Current soil sampling methodologies, while valuable, provide only a fragmented view of contamination, leaving critical gaps in understanding the full extent of arsenic pollution. In some areas, arsenic concentrations have been found to be up to 10,000 times above safe levels for residential land use, particularly near waste deposits. This incomplete data hinders the EPA’s ability to assess risks effectively and plan for appropriate remediation efforts, delaying informed land-use decisions.
The challenge lies in developing a comprehensive approach to map and assess soil contamination across impacted areas to provide a clear understanding of the risks associated with legacy mining activities. This is essential to support practical decision-making for land use and to safeguard both the environment and public health in the region.
Partners
The project was a collaboration between the Environment Protection Authority Victoria, Senversa, and FrontierSI.
The Solution
This project successfully implemented a machine learning model that uses multispectral satellite observations to predict the potential locations of surface mine wastes.
Through the project, a map was developed that shows potential locations of surface mine waste, particularly arsenic sands, in two different regions in Victoria. A data model was used to analyse key environmental indicators, successfully reducing false positives in areas with no mining activity. The project also introduced a classification system with three zones (A, B, C) to indicate the likelihood of mine waste being present, with Zone A having the highest probability. The model had moderate success, correctly identifying mine waste in 40.8% of cases while maintaining a high level of accuracy (99.9%) in identifying areas without mine waste. A process was also developed to fine-tune the prediction boundaries for better results.
Impact
This project developed a cost-effective method for preliminary screening of potential contamination, allowing for more targeted use of EPA resources in field investigations. Such capability enhances EPA Victoria’s capacity to efficiently assess and manage risks associated with legacy mining sites across large areas.
The project demonstrates the potential of combining machine learning, remote sensing, and environmental data to efficiently map potential mine waste sites across large areas. Significant opportunities exist to build on this project’s success to develop a suite of powerful tools for environmental management, revolutionising how we monitor, predict, and respond to environmental challenges across diverse regions.
Contact
To learn more, contact FrontierSI at contact@frontiersi.com.au.