Abstract: Automatic Surface Observation System (ASOS) is set up by China Meteorological Administration (CMA) about 2000. Main surface meteorological parameters start to be automatically measured at about 600 base stations, 1200 ordinary stations in CMA by the end of 2005. Since both field intercomparison experiment results and parallel comparison observation results show that the difference is much great between the amount of rainfall measured by commonly used rain gauges (manual observation) and that measured by tipping-bucket rain gauges (automatic recording), commonly used rain gauges are still kept on rainfall measurement manually twice a day in China surface meteorological observation operation, though automatic measurements of rainfall have been implemented to record hourly precipitation totals at a large number of stations at present. What the difference and relativity are between the amount of rainfall measured by automatic recording and by manual observation in observational operation remains a question to many meteorologists. Based on 55641 operational rainfall observations respectively by automatic way and by manual way at 627 national base stations in 2005, various analyses are carried out on the rainfall observation difference, the difference and relativity between above two observation methods.In the rainy season, when there is no rain by manual observation, there is no rain by automatic recording at each station. The average daily rain amount observed by automatic way is 0.12 mm (or 1.42%) more than that by manual way. The standard deviation of the difference is 0.70 mm. In the 55641 rainfall comparison observations, 26.2% daily data observed in the two ways is identical, 30.8% daily data observed by automatic way are less, 43.0% daily data observed by automatic way are more than the manual way. The annual rainfall measured by automatic recording is more than that by manual observation at 70% base stations. There are 80% stations whose difference values in annual rainfall respectively observed in the two ways are less than 5%, and the difference values are above 10% at 4% stations. Comparing to field experiment results in the two observation methods in early 1990s, the difference of operational annual rainfall observations respectively in the two ways is obviously improved at the 627 base stations in 2005. In general, the more the difference value in annual rainfall, the less the annual rainfall in a station. The distribution difference of rainfall in space, the difference in measurement time of the two observation ways at 20:00 and other urgent events can result in the rainfall difference in daily data, but they have no systematic effect on accumulative total. Different observation instruments and likely low quality can lead to the systematic deviation in above comparing observations. The daily rainfall by automatic recording is obviously linearly related with that of manual observation and the correlation coefficient is 0.9988.