THE CHALLENGE OF PREDICTING ROCKFALL RUNOUTS IN U-SHAPED GLACIAL VALLEYS (Invited Presentation)

Glacially carved valleys composed of hard rock tend to have a “U” shaped profile with vertical cliffs that are prone to rockfalls. Predicting rockfall bounce heights, energies and runouts in U-shaped valleys is a challenging task for rockfall simulation models, primarily because empirically based models, such as those that use restitution coefficients to control the returned velocity of a rock projectile after impact, struggle to simulate the high velocities and energies reached after long vertical freefalling phases. Such models often use high velocity damping factors on the velocity component normal to the terrain to artificially maintain the simulation in a lower velocity domain and to limit bounce heights. This can, however, underestimate or exaggerate simulated runouts in U-shaped valleys when calibrated for V-shaped valleys. Empirical damping factors like restitution coefficients then need to be adjusted based on the geometry of the terrain and predicted reached velocities, and not just because of the encountered slope-forming materials. This introduces subjectivity to the trajectographic simulation models, leaving room for potential misuse and misinterpretation. We developed a new rockfall simulation model to help address these shortcomings, whether the valley is U-shaped or V-shaped, that requires as few parameters as possible. The model uses the ground roughness from detailed 3D terrain models and a simple rolling friction impact law instead of empirical damping factors to slow down the rock projectiles. Subjectivity is minimized by using fixed parameters for the friction law that are empirically calibrated for conservative results, that is, they tend to reproduce the longest observed rockfall runouts and extend slightly beyond for a safety margin. The calibration was made from a database of more than four hundred reconstructed 3D rockfall impacts, with energies ranging from kilojoules to hundreds of megajoules. Because the model is built and calibrated on the rock projectile behavior at impact where most of the energy changes happen and not on energy lines from observed runouts, it can reproduce rockfall events across many terrain geometries. The model shows good results for recent rockfalls from the steep cliffs of Yosemite Valley.