Probabilistic rockfall hazard and risk analysis along the El Portal Road in Yosemite National Park (California, USA)

Abstract

Yosemite National Park attracts millions of visitors each year with its stunning landscape, characterized by 1000 m high granite cliffs that are highly susceptible to rockfalls. Between 2010 and 2020, more than 300 rockfalls affected the 12 km long El Portal Road, used by 30% of visitors to enter the park, causing road closures and fatalities. Since National Park policies limit engineering mitigation on natural slopes, risks along roads are managed through traffic practices based on local hazard evaluation. In this perspective we developed a probabilistic risk analysis workflow, aimed at estimating the annual probability of loss of life for people driving a vehicle along the road. The analysis was carried out for every 10-m-long segment of each travel lane, to and from Yosemite Valley. We based our analyses on 3D rockfall runout simulations performed with the Hy-STONE simulator, and on rockfall event (1857-2022) and vehicle traffic data collected by the U.S. National Park Service. Runout simulations were performed over 18 km2 with a spatial accuracy of 1 m. Simulation parameters were calibrated by back-analysis of past events and validated with field evidence. Fifteen million trajectories were simulated for five volume scenarios (0.01-100 m3), providing local information of transit frequency and kinetic energy. A probabilistic hazard analysis was developed using the probabilistic rockfall hazard analysis (PRHA) method (Lari et al, 2014), which calculates the kinetic energy that can be exceeded in N years for each road segment. The method integrates different rockfall volume scenarios, with specific return times, in a probabilistic framework accounting for modelling uncertainties. For each considered scenario, the annual rockfall onset frequency was derived by a magnitude-frequency (MF) curve, based on the Yosemite event data from 2010-2020 and combined with a field-based talus MF curve, to redistribute frequency among blocks disaggregated during runout. The annual rockfall frequency at each slope segment was then calculated by combining the onset frequency with the transit frequency provided by runout simulations. The exposure analysis, dependent on block size, vehicle size and speed, considered the probability of a vehicle being in the path of a falling block when it reaches each road segment. Since blocks coming from different sources may converge to a common location based on the 3D topography, we reconstructed the distribution of kinetic energy at each target road segment. The probability of exceeding specific energy values, combined with the annual frequency of rockfall occurrence, allowed deriving probabilistic hazard curves for each scenario and for the ensemble. Based on expected kinetic energy and considering the number of visitors passing along the road every day as well as assumptions on the vulnerability of vehicles, we calculated the possible annual number of casualties for each road segment and the entire road, to identify the road sectors with the highest risk. Computed risk varies in time with clear weekly and seasonal patterns depending on the number of daily visitors and the weather conditions. Our study will provide park managers with tools to make adaptive decisions for managing risk in dynamically changing conditions.