Observation data of AWS-city from May 2001 to March 2002 and AWS-milos500 from January to may 2003 have been respectively contrasted with synchro conventional observation data. The cause analysis of the observation difference is made from their difference in sensor pieces and detecting principle, observation time, weather condition etc. Statistics analysis of the temperature error of AWS-city shows that the error is related with the weather factors. Based on that, the correction equations are given. Statistical test and actual verification proved that it had a better effect. The analysis shows that the observation differences of the AWS ere in or a little over the allowed scope, even thought they had many faults. AWS are superior to man-observed station at the aspect of the unifying of observation time.