Application of telemetric sensors in mining to holistically manage vulnerability and risk in watersheds


Gabriel Castillo, The University of British Columbia / Aquatic Life Ltd.; Mehrdad Mehrjoo, University of Manitoba / Aquatic Life Ltd.

Access to water resources is a constant challenge in the mining industry worldwide, with different issues in different regions. Seasonal variability of hydrologic regimes in watersheds often generate risks for different users of the water, communities, the ecosystem, the mining operations and other industries. Therefore, to successfully manage withdrawals from the watershed, it is critical to consider inter-annual variability, climate change and increases to water demand. The ability to understand and identify the spatial distribution of the most vulnerable areas in watersheds supports water risk management, gaining public acceptance and ultimately funding. Hazard-specific vulnerability mapping in a watershed aims to characterize areas most vulnerable to diminished water quality and quantity. Consequently, mapping of vulnerability requires integrated spatial datasets to assess susceptibility and the range of hazard threats. Actually, machine learning techniques are able to predict some events in the near future. However, it is a relatively new idea to use these technologies in water monitoring systems. Some machine learning methods can predict different sensing features to set water risk alerts with good accuracy. Being able to predict the occurrence of these risks in advance will help the mining companies prepare to handle the situation and support good water stewardship within the watershed. Through a literature review, use of the Support Vector Machine (SVM) regression method and recurrent neural network methods for time series, this paper focuses on understanding how these technologies may be used as a tool to support the goal of understanding the watershed vulnerabilities and the balancing between environmental flow needs (EFNs) with the mining, communities and other industries water demands.
Keywords: Water management; Machine learning; Real-time monitoring