Abstract:Currently the method which GRAPES 3DVar (global\regional assimilation and prediction system) used can only detect the data of clear FOV, the whole channel data needs to be removed if the FOV is polluted by cloud. In fact, studies have shown that the data above cloud top in cloud FOV is more important to numerical prediction. Therefore, refering to McNally and Watts cloud detection schemes and, combining GRAPES 3DVar system and instrument characteristics, we built a detection scheme which is suitable for GRAPES 3DVar mode. This scheme can not only detect the field data of clear sky, but also detect the channel data above cloud top which is not influenced by cloud in cloud FOV. Besides, it can calculate the height of cloud top, and judge the cloud base of high, middle and low cloud. The detection scheme for the field view of clear sky cloud and the detection scheme for clear channel cloud are used to detect the AIRS observational data. The results show that the clear data detected by the field view of clear sky scheme stands for a 9.14% of the total and those by the clear channel cloud detection scheme the rate reaches 34.86%, 3.8 times more than the traditional cloud detection scheme.