An Evaluation of the Capability of Global Meteorological Datasets to Capture Drought Events in Xinjiang

With the accelerating pace of global warming, the imperative of selecting robust, long-term drought monitoring tools is becoming increasingly pronounced. In this study, we computed the Standardized Precipitation Evapotranspiration Index (SPEI) at both 3-month and 12-month temporal scales, utilizing observational data from 102 stations across Xinjiang and gridded observations spanning China. Our objective encompassed an assessment of the efficacy of three widely employed global meteorological estimation datasets (GMEs) in the context of drought monitoring across Xinjiang over the period of 1960–2020. Moreover, we conducted an in-depth examination into the origins of discrepancies observed within these GMEs. The findings of our analysis revealed a notable discrepancy in performance among the three GMEs, with CRU and ERA5 exhibiting significantly superior performance compared to NCEP-NCAR. Specifically, CRU (CC = 0.78, RMSE = 0.39 in northern Xinjiang) performed excellently in capturing regional wet–dry fluctuations and effectively monitoring the occurrence of droughts in northern Xinjiang. ERA5 (CC = 0.46, RMSE = 0.67 in southern Xinjiang) demonstrates a stronger capability to reflect the drought dynamics in the southern Xinjiang. Furthermore, the adequacy of these datasets in delineating the spatial distribution and severity of major drought events varied across different years of drought occurrence. While CRU and ERA5 displayed relatively accurate simulations of significant drought events in northern Xinjiang, all three GMEs exhibited substantial uncertainty when characterizing drought occurrences in southern Xinjiang. All three GMEs exhibited significant overestimation of the SPEI before 1990, and notable underestimation of this value thereafter, in Xinjiang. Discrepancies in potential evapotranspiration (PET) predominantly drove the disparities observed in CRU and ERA5, whereas both precipitation and PET influenced the discrepancies in NCEP-NCAR. The primary cause of PET differences stemmed from the reanalysis data’s inability to accurately simulate surface wind speed trends. Moreover, while reanalysis data effectively captured temperature, precipitation, and PET trends, numerical errors remained non-negligible. These findings offer crucial insights for dataset selection in drought research and drought risk management and provide foundational support for the refinement and enhancement of global meteorological estimation datasets.