Risk management of acid rock drainage under uncertainty
Acid rock drainage (ARD) is a major environmental problem that causes local and global pollution. ARD occurs when sulfide bearing materials are exposed to oxygen and water during mining activities. This reaction between sulfide and oxygen with the presence of water generates elevated metals and metalloids that may cause potential environmental and human health risks. The remediation costs of potentially acid-generating wastes at abandoned minesites are estimated to be over $20 billion in USA.
... 0 billion in USA. The major objective of this research is to propose a risk management framework for ARD that can improve the prediction of ARD chemistry, assess and manage environmental and human health risks to guide decision-making under uncertainty. The proposed framework consists of methodologies for filling in missing data, predict ARD chemistry, assess environmental risks, and manage risks of ARD. In the first methodology, missing values of ARD data are filled in using imputation algorithms that reduce loss of information and introduction of biases. After having the complete data, future ARD chemistry is predicted using machine learning techniques. The predictive uncertainty due to data, parameters and model is quantified using a probability bounds approach. Models are integrated using aggregation methods to reduce the uncertainty of the individual model. Case studies in minesites show that the developed methodology improves the prediction of future ARD chemistry under uncertainty. For ecological and human health risks assessment of ARD, two methodologies are developed based on the fugacity and PHREEQC approaches. The fugacity and PHREEQC approaches are applied in minesites with limited and adequate hydrogeological information, respectively. Case studies in minesites show that these methodologies are useful to quantify ecological and human health risks in the mining industry. In addition, they quantify the associated uncertainties in the risk assessments using the probability bounds and fuzzy-probabilistic approaches. For risk managem [...]