Investigating the ability of multiple reanalysis datasets to simulate snow depth variability over mainland China from 1981 to 2018

Abstract

AbstractThough the use of reanalysis datasets to analyze snow changes is increasingly popular, the snow depth variability in China simulated by multiple reanalysis datasets has not been well evaluated. Also, the extent of regional snow depth variability and its driving mechanisms are still unknown. In this study, monthly snow depth observations from 325 stations during the period of 1981–2018 were taken to evaluate the ability of five reanalysis datasets (JRA55, MERRA2, GLDAS2, ERA5, and ERA5L) to simulate the spatial and temporal variability of snow depth in China. The evaluation results indicate that MERRA2 has the lowest root-mean-square deviation of snow depth and a high spatial correlation coefficient with observations. This may be partly related to the high accuracy of precipitation and temperature in MERRA2. Also, the 31 combinations of the five reanalysis datasets do not yield better accuracy in snow depth than MERRA2 alone. This is because the other four datasets have larger uncertainty. Based on MERRA2, four hotspot regions with significant snow depth changes from 1981–2018 were identified, including the central Xinjiang (XJ-C), the southern part of the Northeastern Plain and Mountain (NPM-S), and the southwestern (TP-SW) and southeastern (TP-SE) of the Tibetan Plateau. Snow depth changes mostly occurred in spring in TP-SW and winter in XJ-C, NPM-S, and TP-SE. The snow depth increase in XJ-C, NPM-S, and TP-SW is mainly caused by increased seasonal precipitation, while the snow depth decrease in TP-SE is attributed to the combined effects of decreased precipitation and warming temperature in winter.