Soden, B. J., and J. R. Lanzante, 1996: An assessment of satellite and radiosonde climatologies of upper-tropospheric water vapor. Journal of Climate, 9(6), 1235-1250.

Abstract: This study compares radiosonde and satellite climatologies of upper-tropospheric water vapor for the period 1979-1991. Comparison of the two climatologies reveals significant differences in the regional distribution of upper-tropospheric relative humidity. These discrepancies exhibit a distinct geopolitical dependence that is demonstrated to result from international differences in radiosonde instrumentation. Specifically, radiosondes equipped with goldbeater's skin humidity sensors (found primarily in the former Soviet Union, China, and eastern Europe) report a systematically moister upper troposphere relative to the satellite observations, whereas radiosondes equipped with capacitive or carbon hygristor sensors (found at most other locations) report a systematically drier upper troposphere. The bias between humidity sensors is roughly 15%-20% in terms of the relative humidity, being slightly greater during summer than during winter and greater in the upper troposphere than in the midtroposphere. However, once the instrumentation bias is accounted for, regional variations of satellite and radiosonde upper-tropospheric relative humidity are shown to be in good agreement. Additionally, temporal variations in radiosonde upper-tropospheric humidity agree reasonably well with the satellite observations and exhibit much less dependence upon instrumentation.
The impact that the limited spatial coverage of the radiosonde network has upon the moisture climatology is also examined and found to introduce systematic errors of 10%-20% relative humidity over data-sparse regions of the Tropics. It is further suggested that the present radiosonde network lacks sufficient coverage in the eastern tropical Pacific to adequately capture ENSO-related variations in upper-tropospheric moisture. Finally, we investigate the impact of the clear-sky sampling restriction upon the satellite moisture climatology. Comparison of clear-sky and total-sky radiosonde observations suggests the clear-sky sampling limitation introduces a modest dry bias (<10% relative humidity) in the satellite climatology.