Evaluating the performance of snow depth reanalysis products in the arid region of Central Asia

Central Asia (CA) faces water scarcity issues and heavily relies on snowmelt; however, the limited number of monitoring stations cannot meet snow monitoring needs. Reanalysis data could fill this gap, but their accuracy in CA remains uncertain. This study evaluates snow depth (SD) products from ERA5, ERA5-Land, JRA-55, MERRA-2, and GLDAS in CA using in situ data and the three-cornered hat (TCH) method. In situ data evaluations indicate that JRA-55 outperforms the other SD products overall. However, each of the other four SD products demonstrates unique strengths under different conditions. The TCH method indicates that the ERA5, JRA-55, MERRA-2 and GLDAS SD products have low uncertainty, with regions of uncertainty less than 1.0 cm covering more than 70% of CA, whereas ERA5-Land displays relatively high uncertainty. Both in situ data and TCH indicate poor performance for all five SD products in Tajikistan and in high-altitude areas. Errors in the five SD reanalysis datasets in certain regions may stem from inaccuracies in precipitation and air temperature forcing data. The use of a multidataset ensemble average SD product significantly enhances the ability to capture SDs in CA. Our study provides reliable data support for SD monitoring in the CA.