中文摘要：

本文采用一维变分法，并将MM5模式的相对湿度6小时预报作为背景信息，对1998年6-7月两种卫星反演地球大气相对湿度资料进行偏差校正试验，以提高其精度.这两种反演资料分别是用统计反演法反演GMS-5静止气象卫星多通道遥感信息得到的GMS反演地球大气湿度，和用同步物理反演法反演NOAA-14极轨卫星的TOVS晴空测值所得的TOVS反演湿度.校正试验结果表明，一维变分法能有效减小两种卫星反演湿度相对于匹配的NCEP/DOE再分析相对湿度的平均偏差和均方根偏差：850-300 hPa GMS反演湿度的绝对平均偏差下降了0.59%-2.87%；各层GMS反演湿度的均方根偏差的减少量为3.26%-7.49%，其中925-400 hPa从11%-14%降为6%-9%，300-200 hPa从20%-24%降为13%-18%；500-300 hPa TOVS反演湿度的绝对平均偏差从7%-13%降至1%-5%；各层TOVS反演湿度的均方根偏差减少了12.61%-15.1%，其中1000-500 hPa从21%-24%降至8%-10%，400-300 hPa从25%-29%降至11%-14%.校正分别使925-400 hPa GMS反演湿度和1000-500 hPa TOVS反演湿度的均方根偏差降至10％以下，达到了WMO对卫星资料反演地球大气湿度垂直分布的可用精度要求.

英文摘要：

By using a one-dimensional variational （1D-Var） method, relative humidity （RH） derived from GMS-5 multi-channel satellite images and from NOAA-14′s TIROS Operational Vertical Sounder （TOVS） radiances were corrected by 6-h RH forecasts of the MM5 meso-scale NWP model. Errors with respect to the collocated NCEP/DOE reanalysis （R-2） were assessed. It shows that the root mean square differences （RMSD） of the GMS RH retrievals are 11%-14% in 925-400 hPa and 20%-24% in 300-200 hPa, while the TOVS RH retrievals have RMSD of 21%-24% in 1000-500 hPa and 25%-29% in 400-300 hPa, respectively. After 1D-Var correction, the absolute mean biases of the resulting GMS RH retrievals in 850-300 hPa are reduced by 0.59%-2.87%, and that of the resulting TOVS RH retrievals in 500-300 hPa are reduced from 7%-13% to 1%-5%. The reduction in RMSD of the corrected RH retrieved from GMS and from TOVS is 3.26%-7.49% and 12.61%-15.1% respectively. The RMSD of the corrected RH retrieved from GMS in 925-400 hPa and from TOVS in l000-500 hPa are dropped to below 10%, the accuracy requirement for satellite RH retrievals determined by the World Meteorological Organization （WMO）.