Abstract: Based on an observational data serial of ground-based GPS Zenith Total Delay (ZTD) in North China for 2013 and 2014, a quality control scheme for the ZTD observation is developed. The quality control scheme can check and mark out those data that make the statistic characteristic of the data serial to deviate from Gaussion distribution by taking account of the observation site stability, accuracy, outliers, systematic bias and standard deviation. The statistic characteristics of the data serial after quality control can better meet the hypothesis of variational assimilation system, which confirms the validity of the quality control scheme. Analysis of verification scores of continuous simulation experiments shows that the results of experiments with assimilation of quality controlled ZTD are better than those of experiments with assimilation of uncontrolled ZTD and precipitable water (PW). According to the above analysis, it is clear that the assimilation of ZTD can efficiently improve the precipitation forecast skill. It is feasible to replace PW by ZTD in operational numerical weather forecast. In summary, this study on quality control and assimilation of ZTD reveals some meaningful results, which would support and further improve the application of ZTD in numerical weather forecast.