Mechanistic models supporting uncertainty quantification of water quality predictions in heterogeneous mining waste rocks: a review

Polluted drainage from weathering of sulfide-rich waste rock deposits can cause long-term impairment to waterways and biodiversity near mining sites. Mechanistic models represent established tools to support the predictions of the quantity and quality of waste rock drainage, and their associated risks. Yet, model-based predictions in typical waste rock systems are ubiquitously uncertain because of the strongly heterogeneous nature of these waste deposits. Embedding heterogeneity within predictive modeling is complicated by the magnitude and level of knowledge of the waste rock heterogeneity, and the large number of scale-dependent parameters feeding the model equations. This review encompasses deterministic and stochastic modeling approaches that emphasize consolidated tools and emerging modeling solutions to deal with heterogeneity for the modeling of waste rocks. Physical (e.g., variability of texture, hydraulic and pneumatic properties), geochemical (e.g., variability of mineralogy and kinetic parameters), and thermal heterogeneities are evaluated. The review points out the importance of stochastic modeling as a fundamental approach to embed uncertainty in long-term model-based decisions. Regulators and decision makers must be convinced of the benefit of using stochastic modeling, which is still considered to belong mainly to the academic sphere.

Automated summary

This article discusses the potential harm of polluted drainage from sulfide-rich waste rock deposits to waterways and biodiversity near mining sites, introduces the application of mechanistic models in predicting the quantity and quality of waste rock drainage, and emphasizes the impact of heterogeneity on model predictions. It reviews the evaluation of physical, geochemical, and thermal heterogeneities, and emphasizes the importance of stochastic modeling as a fundamental approach to embed uncertainty in long-term model-based decisions.