Stochastic Modeling of Snow Loads Using a Filtered Poisson Process

Abstract

The Bernoulli pulse process has been used in the past for modeling snow loads. However, it is not an appropriate model for heavy snow-load areas as the snow accumulation cannot be simulated, which may lead to an underconservative assessment of buildings in such areas. In this study, a filtered Poisson process (FPP) is investigated and demonstrated to be an effective stochastic model capable of simulating snow loads with or without accumulation. Weather records obtained from the National Climatic Data Center are used to calibrate the simulated ground snow-load records using the FPP model. A genetic algorithm is employed to determine the parameters of the FPP model. Illustrated by three selected sites in the United States, the annual maximum and daily ground snow-load characteristics are well captured by the FPP model. Potential applications of the model in reliability analysis and risk assessment are discussed.