Wind energy is the fastest growing renewable energy source in the past decade. To estimate the wind energy potential for a specific site, the long-term wind data need to be analyzed and accurately modeled. Wind speed and air density are the two key parameters for wind energy potential calculation, and their characteristics determine the long-term wind energy estimation. In this paper, we analyze the wind speed and air density data obtained from two observation sites in North Dakota and Colorado, and the variations of wind speed and air density in long term are demonstrated. We obtain univariate statistical distributions for the two parameters respectively. Excellent fitting performance can be achieved for wind speed for both sites using conventional univariate probability distribution functions, but fitting air density distribution for the North Dakota site appears to be less accurate. Furthermore, we adopt Farlie-Gumbel-Morgenstern approach to construct joint bivariate distributions to describe wind speed and air density simultaneously. Overall, satisfactory goodness-of-fit values are achieved with the joint distribution models, but the fitting performance is slightly worse compared with the univariate distributions. Further research is needed to improve air density distribution model and the joint bivariate distribution model for wind speed and air density.