Abstract: Because the early observation equipment needs manual intervention, the historical observation data of wind and climate provided by the meteorological department are generally not continuous data. They are often observed every one or several hours, so data omission is inevitable, which may lead to deviation in the estimate of extreme wind speed. The continuous observation data of wind speed recorded by 11 urban meteorological stations in the United States are extracted into two minutes average wind speed series with three kinds of observation intervals of every 1 hour, every 6 hours and no interval. Combined with the improved independent storm method and the extreme value I-type distribution model, the extreme values of wind speed in given returned periods are obtained, and the relationship between them is analyzed. The results show that, the difference between the extreme wind speed corresponding to the continuous observation series and the extreme wind speed corresponding to the observation data series with the interval of one hour or six hours meets Rayleigh distribution. And the correction method is given by using the probability distribution characteristics of the wind speed difference. The modified data can give more accurate estimation results of extreme wind speed.