Using the sounding observation data (SO), raw sounding observation data (RSO), GPS/MET data, microwave radiometer data (MWR), GFS reanalyzed data and the mesoscale area limited model WRF data, the precipitable water vapor (PWV) was estimated. Through the error analysis of different data, it is found that the SO PWV and RSO PWV are the same basically. The correlation coefficient of the GPS/MET PWV, the MWR PWV, the GPS PWV, the WRF PWV and the SO PWV are 0.94, 0.92, 0.93, 0.80， respectively. The correlation coefficients of the GPS/MET PWV, the MWR PWV and the SO PWV reduce to 0.85 and 0.81 seperately when raining, and the error distribution of the GPS/MET PWV is concentrated, while the error distribution of the MWR PWV increases significantly, because the water vapor density error increases abnormally in the part 1-2 km. The PWV is generally low, except for the MWR PWV and the GPS/MET PWV at Yichang. The GPS/MET PWV is averagely 3 mm lower than the SO PWV at Yichang and Enshi, the GFS PWV at Wuhan and Enshi is respectively 1 mm and 7 mm, the WRF PWV at Enshi is averagely 2 mm and 6-8 mm at Wuhan and Yichang. The surface pressure of GFS is lower than the observation at Enshi is the reason for the GFS PWV errors, but the dew point temperature of GFS is higher than SO. The dew point temperature analysis shows that the GFS at Wuhan, the GFS at Yichang, and WRF at three stations are lower below 850 hPa and higher above 850 hPa than SO, except for the GFS at Enshi. The correlation coefficient and the error distribution of the WRF PWV and the SO PWV analysis show that the 12 h forecasting is better than 24 h.