Development of a global historic monthly mean precipitation dataset

Global historic precipitation dataset is the base for climate and water cycle research. There have been several global historic land surface precipitation datasets developed by international data centers such as NCDC (National Climatic Data Center), ECA&D (European Climate Assessment & Dataset) project team, and Met Office etc., but so far there are no such datasets developed by any research institute in China. In addition, each dataset has its own focus of study region, and the existing global precipitation datasets only contain sparse observational stations over China, which may result in uncertainties in East Asian precipitation studies. In order to take into account comprehensive historic information, users might need to employ two or more datasets. However, the non-uniform data formats, data units, station ids and so on add extra difficulties for users to exploit these datasets. For this reason, a complete historic precipitation dataset that takes advantages of various datasets has been developed and produced in the National Meteorological Information Center (NMIC) of China. Precipitation observations from 12 sources are aggregated, and the data formats, data units, station ids are uniformed. Duplicated stations with the same identifications are identified with duplicated observations removed. Consistency test, correlation coefficient test, significance t-test at the 95% confidence level, and significance F-test at the 95% confidence level are conducted first to ensure the data reliability. Only those datasets that satisfy all the above four criteria are integrated to produce the China Meteorological Administration (CMA) Global Precipitation (CGP) historic precipitation dataset version 1.0. It contains observations at 31 thousand stations with 1.87×107 data records, among which 4152 time series of precipitation are longer than 100 a. This dataset plays a critical role in climate research due to its advantages in large data volume and high station network density compared to other datasets. Using PMT (Penalized Maximal t-test) method, significant Inhomogeneity has been detected in historic precipitation datasets at 340 stations. The ratio method is then applied to effectively remove these remarkable change points. Global precipitation analysis based on CGP v1.0 shows that rainfall has been increasing during the period of 1901 to 2013 with an increasing rate 3.52±0.5 mm/(10 a), slightly higher than that in the NCDC data. Analysis also reveals distinguished long-term changing trends at different latitude zones.