Francois Noël
Co-Directors: Prof. Dr. Michel Jaboyedoff & Prof. Dr. Jacques Locat
Jury: Prof. Dr. Vincent Labiouse, Dr Axel Volkwein, Dr. Duncun C. Wyllie
This work was preceded by geotechnical engineering internships at the Québec Transport Ministry and related studies at Laval University involving fluid, soil, and rock mechanics, foundation designs, graduate advanced soil mechanics, and graduate in-situ soil testing. I then participated in natural hazard mapping tasks involving the complementary use of different remote sensing techniques like photogrammetry, airborne, and terrestrial laser scanning. The mapping also required efficiently ranking the risk from natural cliffs and rock cuts along linear infrastructures with limited access. Hundreds of kilometres of the Charlevoix railway and the ArcelorMittal’s Port- Cartier to Mont-Wright railway in Québec, Canada, were covered. With that background, my research focused on better observing and emulating the rockfall natural phenomena through a better understanding of its dynamic propagation and rebound behaviours and evaluating the performance of different modelling approaches at predicting quantities such as the bounce heights, velocities, and associated spatial extents. To meet these objectives related to natural hazard, risk, and mitigation assessments, my research has involved the following:
Advances in remote sensing: The advancements in the field can now be considered for rockfall studies, with the proposed algorithm to better objectively account for the geometrical effect of the terrain’s roughness. Precise empirical data acquisition: A proposed low-cost, flexible method was applied for reconstructing observed rockfall trajectories from several real-sized experiments and rockfall events with precise consideration of the incident impact geometry from detailed terrain models.
Observed rebound behaviours: Using the data collected from a wide range of conditions and energies, the
rockfall rebound dynamics were explored from different angles. A rolling friction rebound model that doesn’t require subjective adjustments of terrain material parameters was proposed and compared to the observations among other rebound models.
Limited predictive performance of existing models: With some exceptions, the process-based “dynamic” or “trajectographic” models tested gave better predictions. They are thus better suited than the geometric methods for most assessment types, but often limited by requiring per-site adjustments. Otherwise, the 𝛼-𝛽 geometric method is preferable.
Better rockfall predictions: From a collection of extensively mapped rockfalls at a dozen sites, “blind”
simulations from the proposed Rolling friction model utilizing detailed terrain models provided up to 6× better runout predictions and were 2× to 3× more precise than all tested alternatives. At the same time, the predicted bounce heights and velocities were very close to the observations. Although constant terrain material properties were used, the better objective consideration of the impact geometry greatly improved the predictions.