Abstract: The radiosonde observation data of L-band has been one kind of the most basic and important data used in either the weather forecast or the numerical prediction, and its quality for humidity observations has a direct impact on the precipitation forecast. In this study, by calibrating the humidity data from the radiosonde observations with those derived from other type observations and comparing to the background fields based on the NCEP, GRAPES and EC modeling analyses, the qualities for the humidity data from radiosonde observation of L-band in China were assessed. So we could have a new understanding about the quality of the radiosonde observation, which would provide basis information for better use of the data.It was found that these sounding humidity tended to be drier than the analyses, thus the piecewise functions were applied for the humidity data correction. When the humidity was more than 60% in the background field, the deviation of the radiosonde observations for humidity would be significant, and the correction effect would become very prominent. The results of case studies and continuous experiments showed that the forecasting capabilities of precipitation intensity could be significantly improved after applying the bias correction for sounding humidity observations. Meanwhile, in term of precipitation forecast test score (TS) and forecasting bias (B), the humidity correction could obviously have positive effects on both the moderate and torrential rain forecasts.