Rockfall hazard assessment and mitigation design relies heavily on three-dimensional trajectory modeling, in which the coefficient of restitution (COR) is a governing parameter controlling rebound, energy dissipation, and runout distance. In practice, COR values are commonly selected from generalized tables based on slope material type, introducing significant epistemic uncertainty and limiting predictive accuracy. This study presents a comparative evaluation of large-scale field rockfall experiments and 3-D numerical simulations conducted at a former aggregate quarry in Boise, Idaho, to assess the performance of tabulated restitution coefficients. Concrete blocks of controlled shape (spheres, cubes, and rectangular prisms) and mass (17–68 kg) were instrumented with inertial sensors and released from two slope configurations. High-resolution UAV-based LiDAR was used to reconstruct slope geometry, while dynamic cone penetrometer and friction tests were performed to characterize spatial variability in slope material stiffness. These data were incorporated into RocFall3 to simulate block trajectories using spatially varying COR values. Initial models assuming zero rotational velocity and tabulated COR ranges failed to reproduce observed runout distances, dispersion patterns, and modes of motion, particularly for non-spherical blocks. Incorporating field-measured initial rotational velocities significantly improved agreement between modeled and observed trajectories, by correcting the unrealistic sliding mode of motion previously observed. However, quantitative discrepancies in deposition and dispersion persisted, highlighting limitations associated with simplified slope geometry and the loss of small-scale surface features during LiDAR surface reconstruction. The results demonstrate that restitution behavior is strongly shape-dependent and that realistic initial conditions are essential for physically meaningful simulations. The findings underscore the need for site-specific, material-informed approaches to COR estimation and for improved integration of high-fidelity field data into physics-based rockfall models.
Goertzen et al. (Wed,) studied this question.