Using Kappa value and taking out the impact of chance and random this paper measures the interagreement of predictive results of three different air quality forecasting methods. The result shows that, the predicting results of both numerical model forecasting and synthesis experience forecasting are of medium identities, and they are not casual. The results measured are clear, and the differences from subjective assessment can be avoided. Furthermore, this research would promote the understanding of different forecasting models and be important to the improvement of forecasting accuracy.