Accurate estimation of the 50-year return extreme wind speed (Vref) based on sufficiently long duration wind data is important for accurate characterization of wind turbine storm loads. Often times though there is only a limited amount of site measurement data available, typically in the order of one year of data, leading to unsatisfactory levels of accuracy in the estimated Vref. GE developed a new method to overcome this problem by combining on-site wind speed measurements in combination with more long-term regional wind data in a Bayesian framework. Two inputs are used for the new Bayesian Vref estimation method: 1.A Vref distribution using wind velocity measurements from the actual site. 2.Prior belief: Vref distribution derived from longer duration data for a nearby site which is climatologically similar to the one under investigation. This dataset will serve as long-term correction. The Bayesian algorithm combines both inputs and computes the posterior Vref distribution at the site, a long-term corrected probabilistic determination of Vref. Results show that the method reduces the uncertainties in Vref estimation which in turn reduces uncertainty in the wind turbine site suitability assessment.