The paper examines the variability of wind power generated by wind farms that are geographically dispersed in a region. The reduced variability of wind power generation with increasing region size and number of wind farms is referred to as the smoothing effect. In this paper, the smoothing effect is quantified by changes in standard deviation, skewness and kurtosis of hourly wind power generation and its temporal gradients. General approximative equations are derived that allow the calculation of these parameters in dependence of the region size and the mean generation. It is shown from which point on the smoothing effect depends only on the region size and no longer on the number of wind farms. Upscaling can then be an appropriate method to simulate future wind power generation as smoothing effects are no longer neglected. This paper also shows how approximative duration curves of wind power generation or its temporal gradients can be simulated by means of the calculated parameters. Hence, the analysis can be used to assess unmeasured or potential wind power generation i n a region. The presented approach and equations are verified by German and Irish generation data. Copyright © 2010 John Wiley & Sons, Ltd.