Uncertainties of Simulating Rockfalls and Debris Flows Using RAMMS

The threat of mass movements has always existed in Norway, but due to the climate change, an increase in mass movement events is expected. Today, there are over 150.000 people living in areas threatened by mass movements, and the need to map hazard zones is rising. The use of the numerical software RAMMS (RApid Mass Movement Simulation) as a tool in the hazard mapping is increasing. With it comes uncertain- ties in regard to field measurements, parameter choices, and the model structure. In this thesis we investigate the two latter using the modules RAMMS::ROCKFALL and RAMMS::DEBRISFLOW by analysing three events from the south of Norway: one rockfall event in Kvam, a debris avalanche in Oldedalen, and a debris flow in Skjeldvik. With the rockfall module we conduct a back-analysis of the event, followed by a sensitivity analysis in regard to rock shape, terrain type, and forest density. With the debris flow module we perform a sensitivity analysis, where our main focus is the module’s sensitivity in regard to the friction parameters μ and ξ, the initiation area, and entrainment. In our studies we found that the RAMMS::ROCKFALL module is highly sensitive to the rock shape, volume, and the terrain hardness. The RAMMS::DEBRISFLOW module is controlled by the friction parameters, and there is a compensation effect between the two parameters. We can thus achieve the same result with different friction parameter pairs. We also found that the flow heights is more sensitive to changes in μ, while the velocities and pressures are more controlled by ξ. The uncertainties in regard to the choice of friction para- meters can be limited by the use of multi-variable analysis, and with the same analyses for other events, we may find a standard range of friction parameters for debris avalanches in Norway. The size of the initiation area definitively has an effect on the results, however, further investigations are needed. Lack of entrainment in the model leads to a significant decrease in all measured variables; simulations run without entrainment in hazard analyses may considerably underestimate the results