Abstract: The atmospheric humidity profiles simultaneously observed by the Vaisala RS92 radiosonde (RS92) and the Cryogenic Frost-point Hygrometer (CFH) online operated by the iMet radiosonde (iMet) in Lhasa (daytime) and Nyingchi (nighttime), Xizang autonomous region, are comparatively analyzed for better understanding of accurate vertical distribution of water vapor over the Qing-zang plateau. With highly accurate atmospheric pressure ( p ) and temperature ( T ) measurements, RS92 indicates that CFH online water vapor volume mixing ratio ( \chi _\mathrmV ) is lower than 100 μL/L (about 2%) in the troposphere and generally is less than 1% above 10 km, while the CFH online relative humidity (RH, on the liquid surface) is lower by 3.2% in Lhasa and 4.1% in Nyingchi at the altitudes of 16—18 km. When iMet T is used, CFH online RH is about 10% higher and times of water vapor supersaturation about 50% higher in clouds at 6—8 km height. Using the humidity profiles of RH and \chi _\mathrmV recalculated offline based on RS92 T and CFH dew (frost) point temperature as references, the RS92 humidity data are evaluated. In Nyingchi, the corrected RS92 RH and \chi _\mathrmV with the consideration of the radiosonde humidity sensor time lagging and solar heating effect are respectively 0.4±2.8 percentage point (0.7%±6.3%) and 172±332 μL/L(1.8%±5.2%) lower than that of their references below 10 km height; in Lhasa, the uncorrected RS92 RH and \chi _\mathrmV are respectively 4±7.4 percentage point (−5.3%±10.4%) and 539±866 μL/L(−2.7%±15.6%) lower, the corrected and uncorrected \mathrm\chi _\mathrmV values in Nyingchi and Lhasa are 13±21 μL/L(16%±25%) and 19±88 μL/L(5%±33%) lower respectively from 10 to 16 km height, while spurious moistening occurs above 16 km. These results indicate that the fine vertical structure of the humidity profile can not be displayed in the corrcted RS92 RH and \chi _\mathrmV data. It is suggested that the atmospheric humidity profile observation should be conducted through the radiosonde platform with high accuracy of p and T measurements and the influence of solar heating effect on the radiosonde sensors should be avoided. RS92 humidity profile below 10 km is sufficiently accurate as that of CFH, however, either the corrected or uncorrected RS92 humidity data from 10 to 16 km are significantly lower and should not be used for long-term trend analysis.