Abstract: Joint distribution of wind speed and direction is widely used in project planning, structural design and assessment of wind energy. Wind roses, a common measure to analyses wind situation, reflect frequency distributions of only 16 wind directions at different wind speed levels. Based on a continuous angular-linear distribution, an approach was proposed to establish the joint probability distribution of wind speed and direction by using the data of wind rose. An original graph reflecting wind frequency with wind speed and direction was also presented. The proposed approach was applied to construct the joint probabilistic distribution of wind speed and direction at Changxingdao Observation Station in Liaodong Bay. The contours of joint distribution for wind speed and direction were illustrated. Calculation results show that:1) The presented statistical model has a higher reliability and shows a strong correlation with their original wind roses; 2) Compared with the model in which the distribution of wind direction is discrete, the novel model offers a more reliable modeling of wind speed distribution in adjacent directions; 3) The new graph of joint distribution reflects much more accurate information of the wind frequency distribution without the restrict of directions, which will make up the shortage of traditional wind roses.