On the basis of the idea that averaged wind direction calculation must reflect the correlation between former and latter wind directions. It is suggested that the trapezoidal mean method is better in calculating the averaged wind direction than the arithmetical average method. The analysis indicates that the trapezoidal mean method is equivalent with the arithmetical average method in dealing with observation errors， but better in dealing with fitting errors. The statistics show that the trapezoidal mean method and arithmetical average method both pass the WatsonWilliams test， but the trapezoidal mean method is more accurate in terms of average deviation and reliability.