It has important significance if the surface automatic weather stations (AWS) data with high spatial and temporal resolution can be fully applied in the weather forecast, but it is hard to ensure the quality of data for various reasons. AWS data from 151 national AWS and 2 600 regional AWS in Jiangsu and Anhui Provinces are selected to discuss the quality of all kinds of AWS data and the quality control(QC) scheme. Based on the AWS data from 2012 to 2014, the missing rates are estimated respectively. A systematic and sophisticated QC scheme containing the missing data statistics, the climate limit check, the climate extreme value check, the internal consistency check, the second iterated space consistency check, the time consistency check, the continuous check and the comprehensive decision-making algorithm is designed for selecting out accurate information and rejecting abnormal information. What's more, the suspicious stations are marked according to the results of QC scheme. It turns out that not only the missing rates of national AWS data of various meteorological elements are apparently lower than the regional ones according to the statistics, but also the quality of national AWS data is obviously better than the regional ones in general. Among various elements, the quality of temperature and pressure data is best, the next is the quality of relative humidity data, and the quality of wind data is worst. The fail rates of all elements of regional AWS data are much higher than the national ones in the QC scheme, except for the wind field data in the second iterated space consistency check, which has little difference between the regional and national AWS data. If the results of QC scheme, especially the information of error data and suspicious stations can be provided to the corresponding stations in time, it is beneficial to the improvement of the real-time data quality and the maintenance and correction of corresponding instruments.