The modelling of dambreaks and tailings runout is not a trivial matter. The way tailings behaves once it leaves a Tailings Storage Facility (TSF) is impacted by many factors that are hard to represent accurately in any model.
We've chosen to focus on a specific type of dambreak scenario in Muk3D since it matches well with the existing modelling approach for tailings. When we talk about modelling dambreaks, we are referring to what are called 'sunny day' failures, or where the dam fails and the tailings flows out under gravity, without a large pond mobilising tailings. If there is a significant volume of water and this becomes the driver for the runout, it becomes more of a hydraulic/debris flow modelling problem and the approach used in Muk3D is not well suited.
If you're looking at modelling dambreaks for TSFs with a large volume of water (i.e. volume of tailings mobilised by water is the primary driver for the dambreak) then the tools in Muk3D are not appropriate and should not be used to identify the impact of a potential tailings failure.
If you're considering TSF failures due to overtopping, we'd encourage you to look at the Muk3D Flood Routing tools to help understand the likelihood of overtopping in flood scenarios for different design flood/spillway options.
Estimating the volume of tailings released in a failure
Accurately predicting the volume of tailings released in the event of a failure is not a trivial task. It depends on many factors including the mechanism of failure, the volume of water behind the dam, the breach elevation and width, the strength of the tailings in the TSF, and the geometry of the TSF. This is not an exhaustive list by any means, but its important to understand that any answer for runout volume using any modelling technique is going to have a significant margin of error.
There have been empirical relationships established based on the release volumes from various failures, however since they don't consider the geometry of the TSF, they can't be relied upon and if used, should come with an acknowledgement of the potential margin of error.
Our approach for estimating release volumes is based on the following factors:
- Breach elevation
- Breach width
- Back-failure slope within the tailings
- Assumption of there being a small volume of water in the TSF that mobilises a small volume of tailings compared to the volume mobilised by gravity.
- Failure slope within the dam itself.
- Breach location.
Through the dam, the breach is modeled as a trapezoidal section, as shown below.
From the breach location, an inverted cone is projected back into the TSF at a user nominated slope until it either daylights through the tailings surface, or hits ground. The volume of tailings above the original ground, and above the inverted cone is then taken as the potential release volume.
To estimate the release volumes, the following surfaces (grids) are needed:
- Original ground. This is the pre-development surface and the assumption is that none of the original ground is mobilised during a dambreak.
- Grid + dam. This is a grid that represents the dam fill. For downstream dams, its simply the final centerline dam structure merged into the Original ground. For upstream or centerline construction, it will be a dam shell that approximates the final dam and so may have a near-vertical face on the upstream side. This doesn't exactly represent the dam shell, but will give a reasonable approximation.
- Tailings surface at the time of failure.
Using these three surfaces and the breach parameters, its possible to estimate the mobilised volume of tailings and dam fill material. We do this by taking the breach location, creating a cut based on the breach width and elevation and estimating the volume of dam material lost to create said cut using parameters. We then create a cone at the upstream side of the breach, at the back-failure slope, and determine the volume of tailings material that can be mobilised.
Dealing with uncertainty
Its important to recognise that we can't predict the breach elevation, width, or failure slopes. And so, instead of running the model once and using that as the output, we encourage users to look at ranges of values.
Instead of entering a single breach elevation, or failure slope, its possible to enter ranges. By doing this, the model will run multiple times and give a range of runout volumes.
Estimating the runout surface post-failure
It's not possible to accurately predict what tailings will do once its left a TSF and is on the downstream side of it. Depending on the amount of water in the TSF at the time of failure, breach elevation, width, strength of the tailings, etc, the distance that tailings will flow, and the extent it will impact is highly variable.
Again, it's important to recognise that all assumptions have some degree of uncertainty and that its better to look at a range of outcomes rather than a deterministic one.
To model the tailings runout, the following is needed:
- A surface (grid) that represents the dam + tailings. You could also just run this on a bare earth topo, or one with the dam and you should get pretty much the same answer. The tailings surface isn't considered by this command.
- Released volume of tailings. This can be a range instead of a single value. This will let you evaluate the impact of different release volumes.
- Residual slope of the runout tailings. Again, since this is an uncertain parameter, this is a range that will let you see the impact of different slopes.
Tailings can then be poured downstream of the TSF and a range of outcomes will be shown.
.
Again, since the inputs are highly variable, great care must be taken when interpreting these results.
Consider this scenario. There is a town adjacent to a TSF. The question that could be asked will be is will this locale be impacted by a failure from the nearby TSF. If a range of scenarios are run, and none of them impact this area, then the TSF exhibits a lower risk than one which impacts the area in many of the scenarios run. If every/many failure scenarios negatively impact a town, then this is a good indication that more intensive and rigorous modelling is needed to better characterise the risk to the community.
Sharing results
Once some runout modelling has been done, aside from sharing results in Muk3D and through screencaptures, its possible to export the runout surfaces to Google Earth as KML files.
Scenarios are organised by runout volume and runout slope and include the tailings contours and the footprint.
Runout images can also be coloured by thickness of the final deposit of the tailings.