Supporting rockfall risk management along roadways in Yosemite National Park, California (USA) by field-constrained high-resolution 3D modeling

Yosemite National Park is a major natural asset of the USA and attracts millions of visitors each year. Its geology and geomorphology make it particularly susceptible to rockfalls, with tens of kilometers of granite cliffs up to 1000 m in height. Between 2010 and 2020, 640 rockfalls were recorded; almost half of these caused damage to the road network somewhere within the park. Approximately 300 rockfalls affected the Merced River corridor, which contains the El Portal Road, the entranceway preferred by about 30% of the visitors. In addition to causing road damage and temporary road closures, rockfalls have also caused fatalities along roadways. Because National Park policies generally preclude mitigations on natural slopes, rockfall risks along roads are mitigated through traffic management practices based on the evaluation of local hazard conditions. Due to the widespread occurrence of rockfalls and the variability of geological conditions, implementing these practices remains challenging and requires a distributed yet accurate quantitative rockfall analysis approach. We performed high-resolution 3D rockfall simulations using the Hy-Stone rockfall runout model over an area about 18 km² in size that contributes to rockfall hazards along two sections of roadway within the park, including the El Portal Road.

We set up our models using existing datasets (1m LiDAR DEM, canopy height, geological and vegetation maps), a database of Yosemite rockfall events (1857-2020), and new field surveys of infrastructure, rockfall paths and deposits, and visible damage caused by previous rockfalls. We identified rockfall sources using a morphometric approach refined by mapping rockfall evidence and additional unstable areas. Sources were classified into “cliff” and “roadcut” (engineered) categories. We mapped Quaternary deposits at the scale of consideration, reclassified vegetation types in categories relevant to rockfall interactions, and produced a unique condition map for model parametrization.

We calibrated Hy-Stone parameters (initial velocity, impact restitution, and rolling friction coefficients) by the back analysis of occurred rockfalls, for which field-based evidence was collected by NPS and USGS. We used post-event aerial pictures of the 2017 Parkline rockfall to map the location and size of 4700 blocks, producing a reference block size distribution for the simulations. Model parameters were calibrated by optimizing the fit between simulated and observed arrest locations and volumes.

We performed forward simulations over the study area considering “cliff” rockfall sources and two different block volume scenarios: a) realistic, stochastically variable volumes; b) worst-case, constant volume (100 m³). An additional simulation considered roadcut sources with variable block volumes. Results were extracted as raster maps of block frequency, velocity, energy, and height and validated against the historical and field databases, making it possible to perform a quantitative evaluation of rockfall susceptibility using the Rockfall Hazard Vector (RHV) method.

Our models combine robust 3D simulations with detailed field data, allowing the characterization of rockfall susceptibility over a large area with the spatial accuracy typical of site-specific studies. This provides robust inputs to quantitative risk analysis that will allow optimizing risk management and granting safer access to the park.