Abstract: Based on the daily wind speed data recorded during 1971-2017 in Dali, China, the multiplication rule and angular-linear (AL) model are employed to construct the joint probability density function (JPDF) of wind speed and direction. Firstly, the optimal probability distribution of wind speed in all directions is determined by comparing the regression results of different distributions. Secondly, the harmonic function and the finite mixture of von Mises distribution are applied to describe the probability density function of wind direction. Furthermore, based on multiplication rules and continuous AL model, two different JPDFs of wind speed and direction are constructed. Finally, the extreme wind speed associated with 50-year return period is calculated for Dali using the established JPDF. The analysis results show that the Gumbel model is better to describe the distribution of wind speed in Dali. The AL model can provide a better fit for wind speed and direction than the approach based on multiplication rule and harmonic function. More importantly, the extreme wind speed corresponding to the 50-year return period in Dali could be significantly overestimated when the effect of wind direction is neglected.