Quality Check and Analysis on the Historic Global Surface Synoptic Observations

The primary purpose of the quality check is to verify the data validity and correct the wrong ones on time. Global surface synoptic data are one of the largest quantities and the most frequently used data in meteorological observations. As the climatic data received and saved in the Climatic Data Office of NMIC (National Meteorological Information Center) in China, the global surface synoptic data are made up of 45 meteorological variables, 4 times observations in term hour daily from approximately 6000 stations around the world. The data have to go through visual observation, instrument reading, recording, encoding transmission by GTS, and decoding etc. It is generally agreed that error data may come into being as a result of each of the above processing stages. Data message format detection and error station number correction are processed to decrease error data induced by data transmission prior to the quality check. The quality control (QC) regulations of global surface synoptic data have been developed in China. The QC procedures in the regulations have been implemented in the operation on global surface synoptic observations' real-time receiving and saving. Using the QC procedures, the historic global surface synoptic dataset stored in Climatic Data Office of NMIC is quality-checked. At the same time, large amounts of data decoded by errors are corrected and then historic global surface synoptic dataset version 2 with quality flags is made. Also, the result of quality check and quality analysis of historic global surface synoptic data from 1980 to 2003 is presented.A large number of error data detected at the stages of data message format check, error station number check and data quality check are corrected or deleted. The quality of the historic global surface synoptic dataset version 2 has been improved greatly. By quality check, the suspect data are 20.3‰, error data is 2.0‰ in historic global surface synoptic dataset version 2. Both the total number of doubtful messages and that of error data of various main variables decrease gradually from 1980 to 2003, which shows their qualities in the dataset are improved. The main variables are those whose percentage received during the hour following the time of observation in term hour is more than 80%. After quality check and some error data processing, the suspect data and the error data of dataset version 2 from 1982 to 1999 are about 18.3‰ and 1.8‰ respectively. Due to the wrong decoding of air temperature and dew point temperature, the error data of dataset version 2 are up to 3.5‰ or so from 1980 to 1981.The percentage of suspect data after 1999 increases to about 30.2‰ as a result of inconsistency between precipitation amount collected in various time intervals, inconsistency between minimum air temperatures and ground surface status.