Effects of Downsampling on the Prediction of the Italian Extreme Winds

Abstract

In assessing the design wind speed, downsampling has the effect of underestimating the annual maximum values. This study aims at investigating this effect for the Italian territory, using a correction method for synoptic events recently proposed by the authors. The methodology is briefly recalled and then applied to a dataset consisting of 26 SYNOP and 5 METAR records from Italian weather stations. The observation period varies from station to station, and years lacking 10% of data or more are removed from the datasets. Annual maximum wind speeds were first transformed into standard conditions only considering the effects of anemometer height. Then, the correction for downsampling was made. The Ordinary Least Square fitting method with Gringorten plotting positions was finally applied to both the uncorrected and corrected annual maxima to derive the parameters of the Extreme Value Type I distribution function. The results reveal that there is a significant amount of underestimation when downsampled datasets are used. For SYNOP records, the level of underestimation ranges between $10.6\%$ to $27.6\%$ with a mean of $14\%$ and a coefficient of variation equal to $4.3\% .$ Whereas, for METAR records, the range is obtained between $5.6\%$ to $9.7\%$ with a mean of $7.6\%$ and a coefficient of variation equal to $1.4\%$.