“We needed to know how far and how fast the event could travel”
In brief
- The village of Brienz/Brinzauls in the Canton of Graubünden was evacuated for the second time in November 2024 because of the threat of another landslide.
- Jordan Aaron, an expert in engineering geology at ETH Zurich, used a model he developed to perform thousands of simulations to assess the potential destructiveness of the event.
- The results of this analysis were incorporated into the risk assessment that provided the basis for the evacuation decision.
ETH News: You provided advice to an expert group that supports the Commune of Albula in the event of the Brienz landslide. What was your contribution?
Jordan Aaron: My role was to assess how far and how fast the possible landslide could travel. At the beginning of November, I started to simulate the potential runout of a future event using a computer model I developed. I then provided the simulation results and a decision framework to a group of geological experts, who assigned probabilities to the various scenarios in order to categoris e which of them were more and less likely.
When were you brought in to assess the landslide?
I have had research and scientific service contracts with the Canton of Graubünden since autumn 2023. We had initial discussions about this project before the collapse in June 2023. Following this event, the authorities wanted to know what the remaining risk to the village was. My group contributed to this risk assessment.
How many simulations for Brienz did you calculate with this model?
I have created several thousand simulations to map various scenarios. This is because, as mentioned, it is very uncertain what might happen. The results depend, for example, on how the moving material fails, and what the climatic conditions are at that time. Including all these factors in our simulations is quite challenging, so the simulations have some uncertainties. We can assess some of these by running many different potential scenarios. However, all results have to be additionally assessed and evaluated by experts.
What was the most unfavourable scenario that emerged from your simulations?
We took a closer look at two different scenarios that could have very serious consequences for the village, adjacent cantonal road and RhB railway. In the first scenario, the failed material moves over a water-saturated subsoil. In this case, the frictional resistance would be very low, and the material would travel a long distance with high velocity. Another disastrous scenario could occur if the moving material itself became saturated with water, which would severely weaken it. In the worst case, the two factors of a water-saturated subsoil and water-saturated debris could occur simultaneously. In this scenario, some of our simulations show that the runout could extend to the Albula river, which could have devastating effects for the nearby road and railway. However, it is very unlikely that this case will occur.
Which scenario is most likely?
The most likely scenario is that the material will continue to creep downhill as it has done so far, i.e. at a speed of a few dozen centimetres per day. This is a comparatively fast slide, but not that dangerous for a village a few hundred metres away.
But not as close to the village as in the island event in 2023?
The travel distance could be similar; however, the speed would be much slower. For comparison, we looked at another slide on this mountain, the Igl Rutsch. Back then, the material came down quite close to the village. But from the time it started moving until it came to a stop, three years passed. It was a slow motion landslide that had similar speed to the currently moving material. This is another possibility that we have to consider. The next most likely event is that, as in the 2023 collapse, material breaks off within minutes and slides rapidly downhill. In my simulation, the material comes close to the village but misses it.
But you wouldn’t question the decision to evacuate the village, which is partly based on these simulations.
No. I’m a scientific advisor. Politicians and decision-makers make the decisions. But if one of the two dire scenarios were to become a reality, the simulations show that it would have devastating consequences for the village. And because there was a high enough probability that these critical scenarios could occur, it means that the villagers are exposed to high risk.
What is special about your model?
It exploits advances in computing power. Modern graphics cards, called GPUs, are relatively inexpensive, but if a program is written in a certain way, they can provide these speed-ups. This is what AI algorithms do, and this is what my model does. That’s why it’s over 200 times faster than comparable models.
How long does a normal desktop computer need to calculate a simulation using your model?
Between three and ten seconds. The previous version takes considerably longer, around 40 minutes. This low runtime time allows me to do so many individual simulations and to explore a variety of potential scenarios without taking months to complete the simulations.
But how do you find the realistic ones in thousands of simulations?
That is indeed a big challenge. I did my PhD in exactly this field: predicting how far and how fast landslides run out. For this, I compiled a database of comparative cases from around the world. In the case of Brienz, I can now use this database and compare it with similar cases from other places in the world. This way, we can compare the simulations with real similar cases and evaluate which scenarios could potentially happen. They are not perfect matches, but they are pretty good analogies of what could happen.
Do you also use your model in other cases in Switzerland?
Yes, I have already applied it to other cases in Switzerland, including for the Canton of Graubünden.
What does this mean for your research? Do you still have time for research outside of these hazard assessments?
Brienz is an example of how my group’s research is useful to society, as our findings and tools can help the authorities to make scientifically sound decisions that protect the population. My vision as the Chair of Engineering Geology at ETH Zurich is to perform innovative research that increases our knowledge of geological processes such as landslides and to transfer this new understanding into practical tools and advice that can be used by practitioners and decision makers. The ultimate goal is to enable society to make informed decisions that protect people and infrastructure. This is the overall context with which we conduct fundamental research in my group.
Can you improve the model?
Yes, certainly, and we will continue to improve it. To do that we need a better understanding of the mechanisms underlying such natural disasters. My group has a number of research projects going in this direction, and we have excellent people working on them. I’m optimistic that we will be able to develop the new tools that are urgently needed to face society’s future challenges, at least where they relate to engineering geology.
