Based on satellite infrared cloud pictures and surface rainfallobservational data, the distribution features of autumn precipitation cloud species affecting the arid and semi-arid areas in Northwest china are analyzed, and regression equations are established for estimating thediurnal precipitation in September-November. These equations can beused to estimate daily and monthly precipitation for the areas short ofdata. Results show that when diurnal precipitation is within the range of 1.0-15.0mm, errors will be small, but the opposite is true when itis within the range of 0.1-1.0mm or≥15.0mm. The estimated monthly precipitation values are generally in agreement with the distribution trend of the observational data.