Abstract:Based on the daily maximum and minimum temperature data of the national reference climate station (hereinafter referred to as the "national station") and 52 regional automatic observatories (hereinafter referred to as the "regional station") from December 2012 to August 2018 in Weining.The variation characteristics of the maximum and minimum temperature in recent years were analyzed by statistical method and weighted least square method. The results show that there are stable low-value centers and high-value centers in the lowest temperature during the four seasons, while there are only stable high-value centers in the highest temperature. The distribution of maximum and minimum temperature in spring, summer and autumn and minimum temperature in winter are basically consistent with that of topography.But the distribution of maximum temperature in winter is affected by topography and stationary front. According to the fitting equation, the accuracy of the temperature forecast value of each township is higher than that of the direct EC model, and the error value is less than | 2 |℃. With the decrease of the correlation coefficient of the equation, the prediction accuracy decreases correspondingly, and the error range increases correspondingly. The prediction accuracy is consistent with the variation of the correlation coefficient.