Abstract: The radiosonde sounding is a major tool for measuring the vertical structure of atmospheric variables. Accurate monitoring of water vapor is vital for numerical weather prediction and climatic change assessment. The radiosonde sounding system in China begins upgrading to L-band radar digital radiosonde system from 59-701 system since 2002 and completes in 2011, and the sensor for humidity measurement is changed from goldbeater's skin to a carbon hygristor. These changes may result in discontinuity of radiosonde observation series.In order to determine the observational bias brought by the upgrade, comparative analysis of relative humidity (RH) measurement between pre-change and post-change of the system is made using radiosonde humidity observations. 98 radiosonde stations out of 120 stations in China are selected to analyze differences between 3 pre-change years and 3 post-change years. Results show that RH after changing the sensor has a significant dry bias. There are statistically significant dry bias in almost all selected stations at 200 hPa, and in 75% and 54% stations at 500 hPa and 850 hPa, respectively. RH dry biases increase with height in troposphere, with values of 14.6%, 8.3% and 5.3% at 200 hPa, 500 hPa and 850 hPa, respectively. There are more dry biases during daytime than nighttime due to impacts of solar radiation. The probability distribution of relative humidity after the system upgrading has a significant shift comparing to pre-upgrading. The occurrence frequency of RH with the value less than 20% after system change is much higher than that during pre-change, which are 53% vs 10% at 200 hPa. The frequency of RH with the value less than or equal to 3% are 16.2%, 9.9% and 2.2% on 200 hPa, 500 hPa and 850 hPa after system upgrading, but that is nearly 0 at three levels before upgrading. Causes of these dry biases and biases between actual value and the observation will be further studied. Also, methods for correcting the bias should be developed.