Comparison of monthly precipitation derived from high-resolution gridded datasets in arid Xinjiang, central Asia

Abstract

To investigate the reliability of global gridded precipitation datasets with horizontal resolution of 0.5° × 0.5° in arid Xinjiang, central Asia, the monthly precipitation datasets released by the Climatic Research Unit (CRU TS, version 3.21), the University of Delaware (Terrestrial Precipitation: 1900–2010 Gridded Monthly Time Series, version 3.01), the Global Precipitation Climatology Centre (GPCC Full Data Reanalysis Product, version 6) and the National Oceanic and Atmospheric Administration (NOAA's Precipitation Reconstruction over Land) during 1979–2010 were selected. A precipitation dataset released by the National Meteorological Information Center, China Meteorological Administration is applied as the observation series. The result indicates that changes in seasonal and annual precipitation can be simulated by the four gridded datasets in most regions of Xinjiang, but there are seasonal and regional discrepancies in the correlation coefficients (R), mean bias error (MBE) and root-mean-square error (RMSE). Generally speaking, the gridded datasets can be more effective in simulating precipitation in low-lying basins, that is, the elevation is less than 1500 m (mainly desert and scattered oasis), while RMSE in mountainous areas is much larger than that in low-lying areas. In Altay, Tianshan and Kunlun Mountains, the precipitation simulated by the gridded datasets is less than that of observation series. The accuracy of gridded precipitation datasets generally decreases with elevation. On a seasonal basis, the errors of gridded precipitation datasets in summer months are generally higher than those in other seasons, and RMSE usually shows low values during wintertime. Among the four gridded precipitation datasets, RMSE of CRU is slightly larger than other three datasets in most areas of Xinjiang. The wetting trend magnitude in this study area from 1979 to 2010 is generally underrated by all the gridded datasets, indicating that the long-term precipitation trend using gridded datasets should be treated with caution.