Abstract: Based on the numerical climate prediction data of 1982-2016, we studied the 500 hPa geopotential height of the northern hemisphere and the surface air temperature and precipitation for the next 45 days in Qinghai plateau, and constructed the prediction model and its operation system platform by applying downscaling method and PP, MOS forecasting method. The results showed that: the system platform construction must take into consideration the advanced system structure and function, the efficiency of data transmission, the storage backup and recovery, and the maintainability of system management, as well as the foresight, openness, stability, security and expansion. System module was prepared by using front-end web pages, back-end programs, model-view-controller mode and common communication protocol technologies. The downscaling methods such as the Tchebycheff polynomial, empirical orthogonal function factorization and predictand division are applicable to the interpretation of the climate model in Qinghai plateau. The average test score of monthly precipitation for 35 samples and the temperature for 51 samples in Qinghai in 2012-2016 is 64% and 75%, respectively. The point to get a higher score in the forecasting is to select the best factors’ construction plan and set up the regression prediction model.