Abstract: A quality control experiment was performed for relative humidity (RH) fields ret rieved from multi spectral GMS-5 imagery by using an one-dimensional variatio nal (1-DVAR) analysis technique. First, the MM5 12 h forecast, satellite retri eval and assimilation analysis of RH fields were given respectively, from June 12 to June 29 in 1998, and then their error statistics (with respect to radiosond es) were analyzed in detail, including mean error, error covariance and RSME. Ba sed on the result, the GMS5 retrievals of RH field were assimilated from July 20 to July 30 1998, and an example of assimilation for 00:00Z 21 July 1998 wa s performed to compare the three types of moisture fields. The result indicates, as a whole, that the MM5 12h forecasts of RH are mainly wetter than the radiosondes observation in the lower troposphere (925-700 hPa) a nd drier in the upper/middle levels of the troposphere (above 500 hPa). At each level, the wet area and dry area are distinguished distinctly, but the dry area is mostly too dry and wet area too moist, so that the RMSE is more than 10 and the maximum 15.6. Though the retrievals of RH are generally rather close to the radiosondes, there is some wet in the dry area and some dry in the wet area. The satellite-retrieved fields are, on the whole, a little too wet, especially at t he middle and low levels. The RMSE is between 6.3 and 8.9. The RMSE of the assim ilation of humidity field that closes to the satellite retrievals is smaller tha n that of the GMS-5 retrievals of RH at the middle and low levels. The RMSE of the assimilation of moisture fields falls by 0.4 to 0.7, which is a reduction of 6% to 9% at the middle/low levels and is close to that of the GMS-5 retrievals at the upper levels. As for the 12h forecasts of RH, the reduction of RMSE wh en assimilating GMS-5 RH is between 1.4 and 9.0, which is a reduction of 17% to 59%. It is shown that the analyzed fields approach further to the actual atmosp heric moisture fields.