Low temperature event is one of many meteorological disasters,which always causes much loss in agriculture,transportation,electricity and so on.The analysis of extreme low temperature can improve our understanding.As a powerful complement to observation data,the reanalysis data play an important role in related researches.Hence,the quality of reanalysis data is related to the accuracy of research results.Therefore,it is necessary to explore the performance of reanalysis data on low temperature filed.Based on the daily minimum temperature data of 192 gauge stations in China and the reanalysis minimum temperature data of NCEP/NCAR,NCEP/DOE,JRA-55,ERA-Interim from 1979 to 2013,and the homogenization temperature data from 1979 to 2004,this study calculates low temperature threshold,and explores the features and differences between reanalysis data and observation data on climatology,interannual and interdecadal variations,and long-term trend.The following conclusions are obtained by calculating correlation coefficient and linear trend,and by using EOF analysis and other methods.Results show that:(1) Compared with the observations,the thresholds of homogenization data are lower in Northeast China,western Nei Mongol,Guangdong and Guangxi while higher in the east side of the Tibetan Plateau,northern Xinjiang and the middle and lower reaches of the Yellow River,but the linear trends are the opposite.(2) The thresholds of reanalysis data are lower in the south of China while higher in the northeast.The correlation coefficients display that the credibility is better in the east of China than that in the west,which may be related to the terrain and the number of gauge stations.(3) The reanalysis data can capture the features of cooling trend over the central and western Nei Mongol and warming trend over the Tibetan Plateau,but there are differences in values and ranges,and the data underestimate the significant warming trend in North China Plain based on the observations.(4) The thresholds derived from the four reanalysis data can depict the observed characteristics of nationwide consistent and reverse between the northeast and other regions.The features of interannual variation can be inferred from the four reanalysis data,but only JRA-55 and ERA-Interim can reappear the interdecadal variation of the low temperature thresholds.(5) Counting the number of low temperature days by threshold value,the results of reanalysis data are generally less than those of the observation data.From the results,we can conclude that the applicability of reanalysis data is better in eastern China,and the effect of ERA and JRA is better than that of NCEP products.