The principle of Kalman filter (KF) and the recursive algorithm of KF are introduced from the view point of meteorological applications. In order to describe the application of the recursive algorithm of KF, a daily minimum temperature up to 36h ahead has been made for Beijing city for the period from 1 to 30 Jan. 1993. The results are satisfactory and approve that the method of KF is mostly practical. The advantages of KF compared to that of MOS method are that one do not have to collect large historical data. So KF will adapt easily to numerical weather prediction model changes.