Snow loads for structural design are calculated using the product of the ground snow load and several coefficients that transform the ground load to a roof load. The basis for the ground snow load is a statistical analysis of the annual extreme snow-water equivalents. This statistical analysis traditionally uses the lognormal distribution and is based on an initial analysis of 76 weather stations in the northeastern United States between 1952 and 1980. Since the original study by Ellingwood and Redfield, additional data have been collected, thus enabling a reanalysis of ground snow load data for 113 weather stations with at least 40 years of data. In reevaluating the data, the authors first conducted a statistical analysis for trends or nonstationarity at all stations. The ground snow load data were then evaluated using multiple extreme-value distributions at each station. An Anderson-Darling statistical goodness-of-fit test was applied to each analysis at each station, and the results were compiled to determine which extreme-value distribution is most appropriate for analyzing snow loads. Conclusions showed that 90 percent of the stations analyzed display no significant temporal trend in ground snow load over the four-decade period of record and confirm the findings of Ellingwood and Redfield that the lognormal distribution is a reasonable model for estimating ground snow load for structural design.
展开▼
