Effect of the Number of Simulations on the Accuracy of a Rockfall Analysis

The reliability of the results of a probabilistic rockfall analysis mainly depends, among other factors, on the number of simulations of boulder trajectories along the slope. Depending on the input parameters of the analysis, the minimum number of simulations necessary to obtain reliable results in terms of boulder velocity, energy or rebound height can be very different. In this paper, a methodology is proposed to relate the expected error on the results of the analysis and the number of simulations. The methodology is exemplified with reference to the Rockyfor3D code. The results are statistically elaborated through the Kolmogorov-Smirnov method, in order to identify the minimum number of simulations required to maintain the error lower than a threshold value. A preliminary investigation on the possibility to relate this number to some input parameters of the analysis is finally presented and discussed.