By comparing the ERAInterim temperature reanalysis data and observations at 08:00 BT and 20:00 BT over Jianghuai Region from 1992 to 2016, we found the deviations between them decrease first and then increase vertically from low level to top level. The spatial differences are obvious in lower tropos phere and tend to be consistent in middle and higher levels. The reanalysis data at 500 hPa and above have transition from lower (before the year 2000) to higher (after the year 2000) deviations. Except for surface, the average absolute deviations of these two kinds of data show significant decreasing tendencies at other pressure levels. In monthly distribution of deviations, reanalysis data at surface and above 500 hPa are generally higher than sounding data, and the absolute deviations at all pressure levels are the smallest around August and September. Further researches showed that under the weather phenomena of snow, sleet, freezing rain and ice particles, deviations of reanalysis data from surface to 1000 hPa and 850 hPa are generally higher than sounding data. However, the higher amplitudes of reanalysis data minus sounding data at 1000 hPa are significantly larger than surface in foggy weather, and under all abovementioned phenomena deviations of reanalysis data are higher above 500 hPa. So, the 850 hPa temperature data should be used carefully when using ERAInterim to distinguish precipitation phase. The false temperature inversion in the nearsurface layer will have a great influence on the identification of fog.