Analysis of Uncertainty and Sensitivity in Tailings Dam Breach-Runout Numerical Modelling.

Tailings dam breaches (TDBs) and subsequent flows can pose significant risk to public safety, the environment, and the economy. Numerical runout models are used to simulate potential tailings flows and understand their downstream impacts. Due to the complex nature of the breach-runout processes, the mobility and downstream impacts of these types of failures are highly uncertain. We applied the first-order second-moment (FOSM) methodology to a database of 11 back-analyzed historical tailings flows to evaluate uncertainties in TDB runout modelling and conducted a sensitivity analysis to identify key factors contributing to the variability of the HEC-RAS model output, including at different locations along the runout path. The results indicate that prioritizing resources toward advancements in estimating the values of primary contributors to the sensitivity of the selected model outputs is necessary for more reliable model results. We found that the total released volume is among the top contributors to the sensitivity of modelled inundation area and maximum flow depth, while surface roughness is among the top contributors to the sensitivity of modelled maximum flow velocity and flow front arrival time. However, the primary contributors to the sensitivity of the model outputs varied depending on the case study; therefore, the selection of appropriate rheological models and consideration of site-specific conditions are crucial for accurate predictions. The study proposes and demonstrates the FOSM methodology as an approximate probabilistic approach to model-based tailings flow runout prediction, which can help improve the accuracy of risk assessments and emergency response plans. Supplementary Information: The online version contains supplementary material available at 10.1007/s10230-024-00970-w.

Las roturas de presas de relaves (TDBs) y los flujos subsiguientes pueden suponer un riesgo significativo para la seguridad pública, el medio ambiente y la economía. Los modelos numéricos de desbordamiento se utilizan para simular posibles flujos de relaves y comprender su impacto aguas abajo. Debido a la naturaleza compleja de los procesos de rotura-desbordamiento, la movilidad y los impactos aguas abajo de este tipo de fallos tienen mucha incertidumbre. Se aplicó la metodología del segundo-momento de primer-orden (FOSM) a una base de datos de 11 flujos históricos de relaves analizados retrospectivamente para evaluar las incertidumbres en la modelización del desbordamiento de TDB y se realizó un análisis de sensibilidad para identificar los factores clave que contribuyen a la variabilidad de los resultados del modelo HEC-RAS, incluso en diferentes ubicaciones a lo largo de la trayectoria de fuga. Los resultados indican que es necesario priorizar los recursos hacia avances en la estimación de los valores de los principales contribuyentes a la sensibilidad de los resultados del modelo seleccionado para obtener resultados más fiables del modelo. El volumen total liberado se encuentra entre los principales contribuyentes a la sensibilidad del área de inundación modelizada y la profundidad máxima del flujo, mientras que la rugosidad de la superficie se encuentra entre los principales contribuyentes a la sensibilidad de la velocidad máxima del flujo modelizado y el tiempo de llegada del frente de flujo. Sin embargo, los principales factores que contribuyen a la sensibilidad de los resultados del modelo varían dependiendo del caso de estudio; por lo tanto, la selección de modelos reológicos apropiados y la consideración de las condiciones específicas del emplazamiento son cruciales para obtener predicciones precisas. El estudio propone y muestra la metodología FOSM como un enfoque probabilístico aproximado para la predicción de la extensión de flujos de relaves basada en modelos, que puede ayudar a mejorar la precisión de las evaluaciones de riesgos y los planes de respuesta a emergencias.

Environmental pollution and human health risk due to tailings storage facilities in China.

Tailings storage facilities (TSFs) represent an anthropogenic source of pollution, resulting in potential risks to both environmental integrity and human health. To date, the environmental and human health risks from TSFs in China have been under-researched. This study attempts to address this gap by developing, and geo-statistically analyzing two comprehensive databases. The first database (I) focuses on failed TSFs; we supply the statistics of environmental damages from 143 TSF failure incidents. Notably, approximately 75 % of the failure incidents involved tailings flows released into water bodies, resulting in a significant exacerbation of environmental pollution. To better inform ecological and human health risks, we present another database (II) for 147 non-failed TSFs to investigate the soil heavy metal contamination, considering 8 heavy metals. The findings reveal that (i) Cd, Pb, and Hg are the prominent pollutants across the non-failed TSF sites in China; (ii) lead­zinc and tungsten mine tailings storage sites exhibit the most severe pollution; (iii) Pb, Cd, and Ni present noteworthy non-carcinogenic risks to human health; (iv) >85 % of TSF sites pose carcinogenic risks associated with arsenic; and (v) health risks resulting from dermal absorption surpass ingestion for the majority of heavy metals, with the exception of Pb, where ingestion presents a more pronounced route of exposure. Our study presents a comprehensive evaluation of environmental and human health risks due to TSFs, highlighting the necessity for risk assessment of >14,000 existing TSFs in China.

A benchmarking study of four numerical runout models for the simulation of tailings flows.

Numerical runout models are important tools for predicting the potential downstream impacts of tailings dam breaches that generate tailings flows, which is a crucial step in emergency preparedness and planning, and risk management. Most existing runout models were originally developed for the analysis of water floods or flow-like landslides (e.g. debris flows). In this study, we back-analyze two well-documented historical tailings dam breaches (1985 Stava, Italy and 1994 Merriespruit, South Africa) using four numerical models (DAN3D, MADflow, FLO-2D and FLOW-3D). The main objective of this multi-case, multi-model benchmarking exercise is to identify collective opportunities to adapt these types of models and associated modelling methods to tailings dam breach runout applications. Comparing best-fit simulation results, we find that all four models are capable of reproducing the bulk behaviour of the real events; however, (i) multiple sets of rheological parameters may produce very similar output results, (ii) the best-fit input parameter combinations are non-transferable between models and inconsistent with independently measured rheological properties of stored tailings, and (iii) choosing an appropriate set requires sufficient understanding of material rheological properties and expert judgment. Using a systematic sensitivity analysis with the First-Order Second-Moment (FOSM) approach, we also find that each model is sensitive to different input parameters, although the total released volume is among the main high-influence parameters in every scenario. We conclude that more case study back-analyses are needed to enhance our understanding of these sensitivities and develop better guidance on the use of these types of numerical models for tailings flow runout prediction.