Using the PMTred method developed by the Environment Canada, ERAinterim reanalysis data as reference series and the detailed metadata of each radiosonde station in China, the inhomogeneity test and corrections are carried out on the monthly relative humidity data at mandatory levels at 125 radiosonde stations in China. Combining with the detailed metadata we analyze the main causes for the inhomogeneity problem of the radiosonde relative humidity data in China. The results show that the shifts of instruments, the changes of data calculating methods and radiation correction method in sounding system and the update of sounding systems are the main reasons for the inhomogeneity in Chinese sounding relative humidity data. The earlier relative humidity data are obviously discontinuous because the 59701 radiosonde sensors carry more water vapor when they pass through clouds. Therefore, the changes for Lband radarsounding system after 2002 lead to the big discontinuous problem in our radiosonde relative humidity data. With the height increasing, both the number of the inhomogeneous stations and break points increase significantly. Also, the mean magnitude of correction increases with height and the correction values are negative at all mandatory levels, which indicates the radiosonde relative humidity data in China are systematically higher. The raw monthly mean relative humidity series of the 125 radiosonde stations in 850-300 hPa mandatory levels from 1979 to 2012 present decreasing trends. After the adjustments, the trends change to insignificant trends.