Prediction of extreme 3-sec. gusts accounting for seasonal effects

Abstract In conventional statistical analysis of extreme 3-sec. gusts, extremal distributions are often used to fit the annual maximum value obtained from available records, invoking the assumption of stationarity and statistical independence. This paper presents an alternative approach in the form of an exponential distribution for the magnitude of the gust above a pre-determined base level compounded with a nonhomogenous Poisson model for the occurrence of the exceedances. The base level of 13.5 m/s is found to be appropriate. The prediction of the extreme gusts using this method agrees well with a modified conventional extreme value analysis which accounts for the seasonal variations within a year through the monthly extremal distribution parameters. The compound distribution approach is preferred over the conventional method since it does not rely solely upon the maximum value but rather utilizes directly more of the available information.