Extreme winds derived from simulations using mesoscale models are underestimated due to the effective spatial and temporal resolutions. This is reflected in the spectral domain as an energy deficit in the mesoscale range. The energy deficit implies smaller spectral moments and thus underestimation in the extreme winds. We have developed two approaches for correcting the smoothing effect resulting from the mesoscale model resolution on the extreme wind estimation by taking into account the difference between the modeled and measured spectra in the high frequency range. Both approaches give estimates of the smoothing effect in good agreement with measurements from several sites in Denmark and Germany.