On the consequences of getting dam breach analysis wrong
Marcelo Llano, Principal Geotechnical Engineer, Red Earth Engineering
Dam breach analysis (DBA) are an integral part of risk management of tailings facilities, they’re the process whereby a tailings facility is assumed to have breached based on a range of potential failure modes, and the outcomes are being studied to determine the potential consequences and impacts. The results from the DBA are then used to support the facility classification and design and develop emergency response plans. DBAs are required for a facility to comply with international standards (i.e. GISTM) and national standards (i.e. ANCOLD). More importantly than compliance they’re an excellent tool to help ensure safety of the facility and the people within the vicinity. Emergency preparedness in the event of a tailings facility failure requires knowledge of the impact resulting from said failure, such as the distance the tailings will travel, also known as runout distance.
The trigger mechanism regarding breach initiation is often handled distinctly and will not be covered within this article. Instead, we will focus on the limitations and challenges of the current DBA techniques being employed in the industry. Followed by touching on some of the future innovations in this space.
There are several methodologies to complete a DBA, each containing their own limitations. The first method typically employed is the use of guidelines and literature. They’ve been developed using empirical approaches based on historical data, one subset of this approach may be the use of imagery before and after an event to determine the extent of the runout. However, a key limitation to this method is that the more mature guidelines and literature techniques often focus, and were developed on water storage facilities and may only address specific scenarios like overtopping or piping.
Another commonly used technique would be to model the tailings as if it were a Newtonian fluid (i.e. water). It is known that the tailings behave differently due to the solids within the tailings acting in a viscous manner (toothpaste, jam and mayonnaise are examples of viscous fluids). Additionally, there is an unknown proportion of tailings that are mobilised. For example, in water storage facilities, the entire volume of water stored can be considered mobilised in a breach event. However, the simplicity of the water approaches makes them significantly easier to implement. Furthermore, with some specific failure modes, such as overtopping, the volume of water available becomes more relevant. It is important to note that for both approaches it is argued that the assumptions present very different results regarding the DBA. For example, it is falsely understood that treating tailings like water always result in more conservative scenarios. Water inundation maps for instance often show inundation depths of less than 1m and large footprints. The risk profile of such a situation is radically different to that of several meters of thickness of deposited mud and a reduced footprint.
MPM model of the Cadia failure in Orange, NSW (Pierce, 2021)
However, this does not always make use of guidelines and empirical techniques as the ideal approach, tailings are often stored in a loose and saturated or near-saturated state, making them susceptible to a process known as liquefaction. This process results in the sudden loss of strength of material and may lead to the tailings flowing as a slurry. The video capturing the moment of failure for the Brumadinho dam on the 25th of January 2019 showcases a terrifying example of what can go wrong when a failure occurs due to liquefaction.
Numerical methods are an important tool that can be used to assist DBAs to project the potential run-out occurring during a breach event and the volume of tailings mobilised.
To model a tailings facility breach it is necessary to capture significant amounts of deformation with the potential for the tailings to flow a distance reaching tens of kilometers (or hundreds of kilometers in extreme cases). The more commonly used finite element method (PLAXIS) and finite difference method (i.e. FLAC) model behaviour based on continuous mechanics and cannot capture large deformation problems. Recently developed numerical methods are more suitable to these scenarios and are capable of modelling large deformations of non-Newtonian fluids (i.e. mud-flows and landslides).
The material point method (MPM) is one technique utilising continuum mechanics that has been gaining traction within the tailings community and has been shown to successfully model run-out behaviour and volumes for a broad range of scenarios. Capable of capturing large deformations and incorporating actual tailings properties it can more accurately simulate the behaviour of the tailings during a failure. Previously MPM was more difficult to implement due to limitations of computation power, however this obstacle has largely been overcome in the recent decade due to computing advances.
By simulating the behaviour of tailings breach events more realistically the creation of more representative inundation maps can be accomplished, allowing for more informed decision making by the operator. The advantages of this cannot be overstated as emergency preparedness requires focusing limited site resources in a timely manner. With accurate information these resources can optimised for greater impact in the event of tailings facility failure. The application of MPM is one example whereby the tailings community has the opportunity to create safer and more sustainable tailings facilities leading into the future.