Abstract
Tailings dam breach models are commonly used for both dam classification and emergency response planning. While many models are well executed and meet all the requirements of recent dam breach guidance, there are still significant limitations in the current state of practice. When model predictions are used for dam classification, limitations can be addressed by simply being more conservative in applying dam classification rules. However, when dam breach modelling is used as a basis for emergency response planning, limitations in predictions can put additional lives at risk.
This paper summarizes the limitations observed in a review of 18 dam breach studies from 23 high or extreme risk tailings facilities. The review included studies by 5 consulting companies, from sites on 3 continents with widely different climate and topography, and models spanning the range of behaviours from thin slurry flows to non-Newtonian “mud-flows” to frictional flow “runouts”. Examples are used to show how the observed limitations can lead to underestimated or overlooked impacts. Importantly, many of the observed limitations were found to arise from incomplete communication of model uncertainty, rather than weaknesses in the currently available models. A revised approach to dam breach studies is recommended and includes a much richer interface between the modellers, site staff, and emergency planners to ensure all potential impacts are considered, even those beyond the capabilities of currently available models.
