Abstract: Based on the classical extreme-value theory, an estimatingmethod for expected extreme-value of wind pressure with non-Gaussianprobability distribution is proposed with a sample whose length is just onestandard observation interval. At first, the wind tunnel test and test dataprocess of this study are introduced in detail. A method to calculateexpected extreme values of time series with a long time interval is thenproposed with its sub-sections based on the classical Gumbel theory forextreme values and the independence of observed extreme values. At last, theextreme values of the wind pressure coefficients of the present wind tunneltest are calculated with the proposed method and methods used widely atpresent, such as peak factor method, improved peak factor method andSadek-Simiu method. The results indicate that the probatilistic parametersof extreme values of sub-sections of a non-Gaussian wind pressure timeseries can be used to estimate accurately expected extreme value of theirparent section with the proposed method. The length of the sub-section canbe determined with auto-correlation analysis on the parent section.Comparison shows that the proposed method can estimate extreme values ofnon-Gaussian wind pressure more accurately than other methods used widely atpresent.