This report makes part of the KNMI HYDRA project and it gives an overview of the reproduction and analvsis of and adiustments to the so called "A Compound Weibull Model For The Description Of Surface Wind Velocity Distributions" (Rijkoort, 1983). This model is used to calculate return levels of hourly mean wind speeds with high return periods (up to 10^4 years) at several stations in the Netherlands.
It turned out that the persistence correction used in the model causes unreliable results with respect to the return levels of interest, especially when the correction is used in combination with the occurrence of relatively many low wind speed values.
In order to use the model properly, adjustments have to be made. It makes sense to use a threshold to prevent the impact of low wind speed values and to determine the persistence correction based on physical grounds.
Alternative extreme wind models also have been examined. The combined GEV distribution results in unreliable return levels due to the split up into seasons and wind direction groups. The one-step Markov chain model has to be extended substantially before results of interest can be obtained. Finally, the analysis of separate storms instead of the parent distribution of wind speed or the distribution of annual maxima of wind speed is discussed briefly.
Ilja Smits. Analysis of the Rijkoort-Weibull model
KNMI number: TR-232, Year: 2001, Pages: 73