China Meteorological Administration (CMA) commenced building surface automatic weather station (AWS) around 2000. More than 2400 national AWSs have been built to observe precipitation, air temperature, wind direction, wind speed, air pressure, air humidity and soil temperature, etc. Moreover, more than 55000 regional AWSs that are mainly applied to disaster prevention and reduction have been also set up to monitor precipitation, air temperature, wind direction, wind speed, air pressure or humidity. The data quality of the above large number of AWSs could not reach very correct degree. So, at the end of 2009 CMA launched a project named Construction of Quality Control and Comprehensive Evaluation System for Realtime Observation Data of AWS. (1) After analyzing various performance ways and causes of the erroneous meteorological data from AWS, the quality control technique was developed for realtime observed and transmitted hourly data of precipitation, air temperature, wind direction, wind speed, air pressure and air humidity. The quality control method includes climatic range check, internal consistency check, temporal consistency check, spatial consistency check, mancomputer interaction control, etc. Each kind of inspection process contains a number of control modes. (2) Based on the operational process of AWS observation data acquisition, transmission, collection, decode and realtime service, threestep quality control system of AWS realtime observation data was developed. In the system, the nationallevel system is mainly responsible for data quality evaluation, the provincial one including 31 provincial data management departments is in charge of data quality control, and more than 2400 national stations are responsible for data quality monitoring and erroneous data correction. The threestep quality control system operates based on national and provincial realtime databases, organized mainly by one realtime quality control software, mancompute interaction quality monitoring platform, data quality evaluation platform, nationprovincestation threestep quality automatic feedback process for suspect data information. (3) The threestep quality control system of AWS realtime observation data has been installed and operational applied at more than 2400 national AWSs, 31 provincial meteorological information departments and one national meteorological information center. Quality control, quality monitoring, quality feedback and quality evaluation are operated for meteorological data of realtime observation and transmission from above national and regional AWSs. The quality of all AWS observation data is greatly improved year by year. For example, data availability of hourly air temperature and precipitation realtime observed by above AWSs in main flood season increases from 88%, 83% respectively in 2009 to about 98% in 2012 and 2013.