Based on in suit observed dataset from 194 stations in China, NCEP/NCAR (NIT), NCEP/DOE(N2T), ERA-Interim (ET) and JRA-25 (JT) reanalysis dataset, the intercomparison of winter minimum temperature (WTMIN) between observed dataset and reanalysis datasets in China is performed over the 1979-2012 by means of EOF decomposition, variance analysis, linear analysis, the cumulative departure, M-K mutation test, standard Morlet wavelet analysis as well as correlation coefficient. Results suggest that:(1) In the aspect of annual mean, four reanalysis datasets have comparable reappearance ability in the regions south to the Yangtze river, and ET’s reproduce ability is optimal in Tibet Plateau and Xinjiang while NIT and N2T have a better reappearance ability in NortheastChinaregion.(2) Except the north and southeast of Xinjiang and the north of northeast of China the OBT decreased during 1979-2012, other regions increased, NIT is pretty better in depicting the linear trend of the OBT than other 3 reanalysis WTMIN.(3) Annual and interdecadal variances of observed WTMIN are larger in Northern China and Tibetan plateau than that on southern China, the 4 reanalysis WTMIN show different depiction in different area, in general the ET is better.(4) The 4 reanalysis WTMIN are equal in reproducing the features of interdecadal of OBT, especially in mid and east of China and the north of Xinjiang, ET is better; The two reanalysis WTMIN of NCEP have a better ability to represent the annual variability of OBT in northeast of China, ET is better in represent the features of 3~4 periods before 1990 and after 2005 than the other 3 reanalysis data.(5) The number of FD in NCEP reanalysis datasets is smaller. ET and JT have a better ability to represent the annual mean and long-term change features. All FDs show a downward trend with a smaller difference between ET and OBT. The regression coefficient of NIT is most close to the observational data. FD’s decadal change is greater than annual variation. FD in the Yangtze River region shows the earliest abrupt change in five datasets.(6) Climatic mean TNn is high in South China while the value in North China is low. ET has a better ability to reproduce the mean TNn field, and JT can reproduce the long-term trends ofTNn primely. In the aspect of regional averageTNn, N2T and JT’s retrieval ability are better. |