Probabilistic study of snow loads on flat roofs considering the effects of wind at representative sites in China

Abstract

Based on meteorological data, this paper simulates the evolution of ground and roof snow loads over time in winter. The snowmelt model is used to simulate the ground snow load, and the snow transport model is employed in cooperation with the snowmelt model to simulate the flat roof snow load. Through the simulations, samples of annual maximum ground snow loads, annual maximum roof snow loads, and ground-to-roof conversion factors for at least 60 years at 50 representative snowy sites in China are obtained. First, the statistical characteristics of the above samples are analysed. Then, the probabilistic models with different parameter estimation methods are used to fit the samples, and each model's goodness of fit is evaluated by using K-S test, Akaike information criterion (AIC), and Q-Q plot. The results show that the Generalized Extreme Value (GEV) distribution based on Maximum Likelihood Estimation (MLE) of parameters is the preferred probabilistic model for snow loads both on the ground and roofs. However, none of the models selected in this paper can accurately describe the ground-to-roof conversion factor. Therefore, this paper recommends using the R-year ground and roof snow loads obtained by the probabilistic model for calculating the ground-to-roof conversion factor. Finally, the ground-to-roof conversion factor is found to be quite scattered for the most of the studied sites, thus it could be more reasonable to use site-specific values estimated by the method proposed in this paper rather than a uniform value in consideration of safety and economy of the structure.