Abstract: Manual observation is being replaced by automatic observation in China gradually. Till 2006, there are altogether 1904 automatic weather stations (AWS) in the country. Great changes have taken place in the observational rules and methods of AWS comparing with manual observations. Inhomogeneity of observational time series may be caused by many reasons and the change of observational instrument is an important one. The records of AWS and manual parallel observation in a certain period of time may be used in analyzing the observational differences between AWS and manual.Based on the data of automatic and manual parallel observations in 700 stations during 2001—2005, the differences of temperature, precipitation, pressure, relative humidity, wind speed, wind direction and ground surface temperature are calculated and the characteristic differences between automatic and manual observation in different areas are analyzed. Results show that certain differences exist in all the meteorological elements between automatic and manual observation, but in most areas the differences in temperature, precipitation, pressure, relative humidity, wind velocity and direction and ground surface temperature fall in the accepted error rang of AWS. In most stations in China the difference is of ±0.1℃ in the annual mean temperature. To be exact , the difference of annual mean maximum and minimum temperature is a little larger than annual mean temperature. More significant system error appears in the automatic observation of a few stations, and when the automatic observational data of temperature are used the instrument systematic error should be corrected. The difference of all elements between automatic and manual observation has different distribution features in China, and the same element may vary greatly in different climatic situations. In dry areas or in the seasons with little precipitation, there is a great differences between auto and manual precipitation observation. The areas with difference of 0.2—0.6 hPa in pressure are mainly distributed in the western areas with high sea level elevation. In the wet areas in the south of the Yangtze River, Tibetan Plateau, northeastern area and mountain stations of the northeast, big difference appears in the auto and manual observation of relative humidity. In the north during snowing period there is a significant difference between auto and manual observation of ground surface temperature. The use of automatic stations exerts certain influence on the continuity of annual mean temperature, so it is necessary to test whether the significant difference exists in the auto and manual observation time series if manual and auto observation data are used. Homogeneity correction is needed if the significant difference appears.The difference between auto and manual observation is unavoidable. Any instrument change in history will bring the data difference, especially the change from manual observation to automatic observation with quite different observational instrument and method. The time series of automatic observation are short in China and the replacement by automatic observation just starts, so with longer observational time series and more automatic stations further analysis is needed to know the causes for the difference between automatic and manual observation as well as correcting methods .