Correction of historical records to improve the reliability of design wind speeds

It is not uncommon that design wind speeds are derived from measured data for which the sampling period is longer than the averaging time, e.g. 10-min averages measured once every hour; we call these downsampled data. Using downsampled data has the effect of shifting the calibrated distribution of the yearly maxima to the left, thus reducing the predicted return wind speed. The origin of this has probably to be found in the fact that a large amount of research on the assessment of design wind speeds was developed in the UK, where hourly wind speed averages were available, measured twenty-four times a day, i.e. contiguously. Procedures were then applied also in the case in which downsampled data were available, apparently disregarding the bias this brings. Indeed, even contiguous measurement of mean wind speeds is insufficient, as for the proper evaluation of the return wind speed continuous measurements would be needed. In this paper, the difference between continuous, contiguous disjunct and downsampled disjunct data measurements is pointed out, and the effects of downsampling on the statistics of the extreme wind speed are quantified. A technique for correction of disjunct measurements is proposed and validated in order to remove the bias deriving from downsampling and to improve the reliability of design wind speeds and velocity pressures. Finally, the procedure is applied to datasets from four Italian meteorological stations showing an increase of the return velocity pressure after correction of about 10% to 30%.