Abstract: The data of daily minimum temperature and corresponding climatic background at 6 artificial stations in Xuzhou city were collected. Taking Xuzhou station as the standard station, the error distribution between stations was analyzed, and the data were classified into two types： Two-terminal type, eastern type, western type and whole type. By using the methods of orthogonal function and statistical analysis, the correlation characteristics of the main meteorological factors which cause the great difference of temperature between stations, including wind direction, wind speed, humidity, cold air, etc., were analyzed and the correction methods were refined. The results showed that Feng county was the most prone to appear obvious errors; When the average wind speed was less than 0.5 m/s, the influence of wind direction was not considered. When the average wind speed was more than 0.5 m/s, Xuzhou city was not conducive to cooling when it was affected by southerly wind. Pei county was not conducive to cooling when it was affected by east to southeast wind and Suining county was affected by south to southwest wind. The characteristics of wind direction in Feng county, Pizhou city and Xinyi city were not obvious; The occurrence frequency of large-value errors was highest in autumn and winter in the west, and highest in spring and least in summer in the east; When there was no cold advection, the influence of local wind direction and wind speed environment was dominant, the two-terminal type should be considered, and when there was cold advection, the western type and eastern typeshould be paid more attention. Through subjective revision, the error of site could be reduced and the level of fine forecasting could be further improved.

Key words: low temperature, data analysis, meteorological factors, forecast accuracy