This paper reports on the analysis of historical wind data from 239 stations in the United States and 146 stations in Canada to derive design wind speeds (95%, 97.5%, and 99%) for the design of smoke control systems. As part of the analysis, the data were thoroughly checked for missing observations, internal consistency, and uniformity of location and measurement height. At each location, the design speeds were extracted from cumulative distributions of the wind speed, both by an interpolation method applied to binned data and by derivation from the two-parameter Weibull distribution using binned data. Analysis of the results of both methods showed that the interpolation method produced the most accurate and stable results for stations with long data sets. On the other hand, the Weibull fit could reasonably be used in locations for which long-term data were not available.

SI