Estimates of the accuracy of up to 100 year return period wind velocity calculation via Weibull probability distribution approximation are analysed. Contribution of possible observation, probability, and approximation errors are shown. Also, dependence of the accuracy upon histogram partition is analysed and recommendations for decreasing errors are suggested.
Multidecade return period wind velocity values are of great interest for shelf engineering project implementations. It is one of the main environmental characteristics which determines both construction design and material selection. As a rule wind velocity values are calculated for return period of up to 100 years, meanwhile the basis series are some one to three decades. Russian polar and offshore hydrometeorological stations feature especially short observation data series, moreover, computer available data cover only the latest 15 to 20 years. Also, newly founded site located stations have short time series of observations. The shortness of data series does not make those data unuseful. They could be used in environment studies, but it is necessary to realise the degree of confidence of those data. So that, all multidecade return period wind velocity values should be accompanied with error bars. Possible errors which arise during real procedure of long return period wind velocity calculations are analysed in the presented paper. The technique to get multidecade return period wind velocity value is standard. Empirical data probability distribution is simulated with any appropriate theoretical probability distribution model, and required wind velocity values are simulated using probability values corresponding to a required return period. For the wind velocity probability distribution simulation, the most appropriate and the most usable model is the two parameter Weibull distribution. So, the error bars can be expressed through the error bars of distribution parameters.