Quality Detection of Surface Historical Basic Meteorological Data

Surface basic meteorological data are observational values from surface stations, including hourly values, daily extreme and cumulative values. Chinese surface historical basic meteorological data consists of observations from more than 2400 national stations since as early as 1951, including 20 kinds of elements, such as air temperature, air pressure, humidity, wind, precipitation, evaporation and so on. These data are the basis of regional and global climate change researches and predictions, synoptic dynamic analysis and public meteorological services. The digitization of these data is started by China Meteorological Adminiatration in 1979. High quality digital data should be faithfulness to the paper reports. But incorrect data and data missing problems caused by digitization and restoring have been found, besides those caused by error observations. Resolving these problems will contribute to the improvement of operational application accuracy, scientific researches and data processing.In order to identify quality problems of Chinese surface basic meteorological data and improve the data quality, several methods are applied to detect the data quality of air temperature, pressure, humility, wind and precipitation observed by 2474 national surface stations in China from 1951 to 2009. The above data are collected based on National Meteorological Information Center (NMIC) and provincial archived electronic data files. Results show that a large number of error data different from actual measurements are stored in both NMIC and provincial archived data files, which are caused by incorrect digitization. For example, some electronic data files are replaced by other station measurements, some meteorological elements aren't observed for a period, and some data are miss-typed. Some flaws between NMIC and provincial archived data files also shows, including the data file differences in time span and various data source (automatic or manual observation), and data difference induced by asynchronous correction of wrong data. It is impossible to detect all the above data problems using conventional data quality control methods, so some special methods are proposed for each kind of problem. The proportion of incorrect measurements of the whole data is low, though its number is huge. For example, for more than 2400 national stations, only 0.06 percent of data are wrong copied, and for more than 700 baseline stations, only 0.34 percent of precipitation data are missing. The data quality detecting method mainly focuses on temperature, pressure, humidity, wind and precipitation data from about 700 baseline stations, and data problems are verified by checking the monthly paper reports but not corrected. With this work experience, thorough data problem detections and corrections on the whole elements from historical electronic data files are carried out by China Meteorological Administration. A review of this program and the results will be published soon.