Comparison of the Weibull Model with Measured Wind Speed Distributions

Wind is the principal driver of the Wind Erosion Prediction System (WEPS), which is a process−based computer model for the simulation of wind−blown sediment loss from a field. WEPS generates wind using a stochastic wind generator. The objectives of this study were to improve the stochastic generation of wind speed and direction and to update the wind statistics used by the generator with statistics derived from more recent, quality−controlled data for the 48 contiguous states of the U.S. Erosive wind power density (WPD) was chosen to evaluate how well wind is generated, since it is proportional to sediment transport by wind. It is important that WPD calculated from stochastically generated data (WPDg) closely reproduces WPD calculated from the underlying measured data (WPDm). The commonly used two−parameter Weibull model did not fit wind speed distributions well enough for application in wind erosion models. WPDg deviated more than 20% from WPDm for 168 out of the 332 stations having WPDm > 5 W m−2. Fitting the model to the high wind speeds only, with the expectation of a better curve fit, resulted in some generated wind speeds exceeding 100 m s−1, which is unacceptable. A more direct method uses the wind speed distributions themselves instead of the Weibull model that describes them. Wind speeds are then generated directly from the distributions using linear interpolation between data points. With this more robust direct approach, there was only one station (down from 168 stations) where WPDg deviated more than 20% from WPDm. The direct method of wind speed generation reproduces wind speeds more accurately than the Weibull model, which is important for wind erosion prediction and may be important for other applications as well.