Modelling the Transport of Tailings Over Erodible Bed During a Tailings Dam Failure

Abstract

Tailings and muddy slurries are highly concentrated, and the fluidized material can consist of a mixture of water and several solid phases. The bulk flow properties of viscous fluids are very complex and conventional flood routing models are not dependable for predicting tailings dam breach inundation extent as they do not consider the non-Newtonian visco-plastic nature of tailings. Tailing's transport and floods following a tailings dam failure are also not just confined to the hyper-concentrated fluidized material released but there is broader morphodynamic influence. During tailings flow, entrainment of additional solids can cause dramatic surge in sediment supply and significant changes in the morphology of channels. Also, tailings dams (TD) are usually located in valleys that are not easily accessible and hence one global digital elevation model (DEM) is used in most of the TD breach outflow modelling studies but the challenges this impose when calibrating visco-plastic rheological models has not been fully investigated. Implications of using relatively low resolution (18 m) GeoBase DEMs for model calibration is studied by comparing it with a model that is calibrated using a relatively high resolution (2.5 m) merged DEM generated from multiple products. The catastrophic breach of the Mount Polley tailings dam is studied to assess the model performance. The influence of rheological parameters, roughness coefficient and sediment concentration (Cᵥ) on simulated variables is studied. It was concluded that when GeoBase DEM was used for back analysis of the event, a physically meaningful set of rheology values could not be assigned. Flow behaviour was strongly influenced by Cᵥ and least influenced by Manning's values. Runout distance was more sensitive to viscosity than yield stress. Mudflow depth and arrival time increased with increase in viscosity and Cᵥ. In this study, a robust non-Newtonian numerical model that can simulate coupled mudflow and sediment transport processes is also used to model downstream geomorphic changes during a tailings dam failure. Spatial and temporal variation of flow properties like density, viscosity, yield stress, sediment size fractions, and friction are considered in the model. Square mesh is used for modelling single phase tailings flow over fixed bed. Flexible triangular mesh will be adopted for quasi two-phase tailings flow as this helps reduce the computational time significantly compared with structured grid with the same average resolution where additional unnecessary grid points are added for local mesh refinement. High resolution multi-temporal Digital Elevation Models (DEMs) are used for the detection of geomorphic changes and estimation of erosion and deposition volumes. The results indicate that the final distribution of tailings deposits obtained from the model results is in good agreement with that obtained from topographic analysis. The model predicted the total volume of erosion and deposition with almost 93% accuracy. The tailings characteristics followed the full Bingham model during the propagation. Different runout scenarios obtained from the mudflow-morphodynamic model, and a mudflow-fixed bed model are analyzed, and it was concluded that maximum mudflow depth of the flood wave is higher when bed entrainment is considered. Sediment concentration is higher in the wavefront and middle regions and lesser in the tail zone. The mudflow fixed bed model could underestimate the downstream flood arrival time. The results also confirm that the adopted approach can efficiently model the sediment budget in the study site and the volume of tailings and bed materials released to the downstream lake.