Abstract: To master the temperature forecast capabilities of NMC, local subjective and objective products in temperature forecasting is of great significance to improve the quality of urban weather forecasting in Ningxia and its national ranking in weather forecasting. Based on the two model products of T639 and ECMWF with the temperature data of 20 national observatories in Ningxia from January 2009 to December 2014, the highest and lowest temperatures of the coming 168 hours were predicted by adopting two MOS methods which grouped the data by month or by season．The verification was done on the forecasted results for the time period from January 2015 to June 2016 using the current inspection and evaluation method adopted in Ningxia，and the comparisons with YBY and NMC were also conducted. The results show that the whole effect of the MOS method is superior to the NMC, and its forecast effect with monthly grouped data is superior to that when the data was grouped in season. The T639MOS by month and ECMOS by month is the optimal objective products of the highest and lowest temperature. For predicting the lowest temperature, compared with the optimal objective product，The forecast by YBY is better than that of T639MOS by month. For predicting the highest temperature, the forecast of ECMOS by month is better than that by YBY. The physical quantities of the two best objective predictive products are the equivalent temperature, or the potential pseudo-equivalent temperature, the temperature, the wet potential vorticity and the height, which are concentrated in the middle and lower layers or the ground．However，there are some factors that play a key role in the forecast of the highest and lowest temperatures．

Key words: MOS method, different data partition, inspection and evaluation