Quantifying rock fall probabilities and their temporal distribution associated with weather seasonality

Abstract

Quantifying rock fall hazards requires information about their frequency and volumes. Previous studies have focused on quantifying rock fall volume–frequency relationships or the weather conditions antecedent to rock fall occurrences, and their potential use as prediction tools. This paper is focused on quantifying rock fall occurrence probabilities and presents approaches for quantifying rock fall temporal distributions. In particular, von Mises distributions allow direct correlation between seasonal weather variations and rock fall occurrences. The approaches are illustrated using a rock fall database along a railway corridor in the Canadian Cordillera, in which rock fall occurrences were correlated to the morphology and lithology. A Binomial probability distribution applied to the annual rock fall frequency suggests an average daily rock fall probability of 1 × 10−2 across the study area. However, circular (von Mises) distributions associated with weather trends in the area, and fitted to monthly rock fall records, allow estimation of daily rock fall probabilities in different months. This approach allows a direct correlation between rock fall frequencies and seasonal variations in weather conditions. The results suggest daily rock fall probabilities between 4 × 10−3 and 8 × 10−3 for April through July and up to 2.1 × 10−2 in October. Moreover, local peaks in rock fall monthly records are quantitatively explained through the seasonality of weather conditions. Similar values are obtained when applying the Binomial distribution to monthly records. However, this last approach does not show strong distribution fits and does not allow a correlation between rock fall frequencies and seasonal weather variations.