Rockaval - Propagation of rock mass falls using machine learning
Rockaval tries to find out whether statistical modelling based on machine learning, which exploits the information power of a large number of rockfall events recorded worldwide, offers a new solution for predicting the runout of rockfall.
Steckbrief
- Beteiligte Departemente Hochschule für Agrar-, Forst- und Lebensmittelwissenschaften
- Institut(e) Multifunktionale Waldwirtschaft
- Forschungseinheit(en) Gebirgswald und Naturgefahren
- Förderorganisation SNF
- Laufzeit (geplant) 01.03.2025 - 28.02.2029
- Projektleitung Prof. Dr. Luuk Dorren
-
Projektmitarbeitende
Prof. Dr. Luuk Dorren
Alexander Dominik Ezuma May
Christoph Schaller -
Partner
Université de Lausanne
Institut national de recherche pour l'agriculture l'alimentation et l'environnement INRAE
Université Grenoble Alpes - Schlüsselwörter Rockfall, Bergsturz, runout, propagation, modelling, hazard assessment
Ausgangslage
Worldwide, residential areas and infrastructure that are built downslope from steep rock faces are potentially threatened by rockfall hazards. This can due to a direct impact, as was feared in Brienz/Brinzauls (Grisons, Switzerland) between April and mid June 2023. Or, it can be due to rockfall initiated cascading hazards, mostly due to debris flows, as described by and seen in Bondo (Switzerland) in August 2017 following the rockfall at Piz Cengalo or to a lesser extent at the Mont Granier (France) in 2016. To come up with realistic predictions of the areas that are potentially endangered by rockfall hazards, modelling rockfall propagation, trajectories and/or runout zones is a method that provides important information. This research project proposes the development and testing of an advanced predictive model using machine learning that integrates geological and topographical data to accurately predict the runout distance of rockfalls, rock mass falls and rock avalanches.
Vorgehen
RockAval project consists of 5 work packages. The first work packages (WP1) deals with all the management tasks in the project. In WP2, we will carry out the literature review including the assessment of existing Open Data relevant for this project. In WP3 we will compile the rockfall events database. In WP4, we will use machine learning to develop models for simulating rockfall runout zones and test these using 100 well documented rockfall events ranging from rock particle fall to rock avalanches. In WP5 we will apply and validate the best performing model for rock particle fall and blockfall in an area covering approx. 2500 km2 both in France and in Switzerland.
