How accurate are the performances of gridded precipitation data products over Northeast China?

The assessment of changes in the precipitation over a region is an important task that is significant for climate modeling and for understanding and monitoring natural hazards and hydrological cycle characteristics. However, the accuracies of gridded precipitation data sets should be evaluated and examined in reference to ground information and observations, as these data can be affected by systematic errors. Therefore, the current study was carried out to evaluate the performances of long-term gridded data sets (from the Global Precipitation Climatology Centre (GPCC), Climate Research Unit (CRU) and University of Delaware (UDEL)) with station based precipitation data obtained from the National Meteorological Administration (NMA) over Heilongjiang Province in Northeast China for the period from 1961 to 2005. For this purpose, this study adopted spatial comparisons among the data sets, the precipitation concentration index and drought frequency analysis based on the standardized precipitation index to evaluate the performances of these data sets in the entire study area. The results indicate that on spatial scales, the gridded data sets showed higher amounts of annual precipitation at most stations at the spatial scale than the NMA. The spatial correlations among the GPCC, UDEL and NMA were in good agreement compared with the CRU. The annual and seasonal precipitation concentration indexes suggested that the gridded data sets and station-based precipitation exhibited an irregular pattern, as 97% of the stations showed an extremely irregular pattern in the study area during the selected time period. The seasonal results of the precipitation concentration index showed that the gridded data products indicated that all of the stations exhibited irregular precipitation distributions (except during the summer), while the in situ station data indicated that most of the stations had a uniform pattern in the winter and summer. In the summer, the UDEL results were more similar to those of the NMA than those of the CRU and GPCC. The analysis based on gridded data sets revealed that 22-41% of the stations experienced a non-significant negative trend compared with the observed precipitation data sets. The drought frequency analysis based on in situ and gridded data was comparable. This study also revealed a new challenge regarding the utilization of gridded data sets due to discrepancies in the results. Therefore, it is very important to minimize these discrepancies by quantifying their uncertainties, identifying raw data systematic error structures, and applying satellite-based data to improve the quality of data in the region.