• ERA5-Land and GLDAS-Noah better capture spatiotemporal patterns of snow density. • Reanalysis datasets fail to reproduce interannual trends in winter snow density. • Snow density significantly affects snow depth biases in reanalysis datasets. Snow density, as a key snow element, not only reflects the physical characteristics of snow cover but also plays a vital role in data assimilation, climate modeling, and hydrological cycle studies. However, systematic evaluations of snow density in reanalysis datasets are still lacking. In this study, we use 4,319 snow stations across major snow-covered regions in the Northern Hemisphere, to assess the applicability of snow density data from five widely used reanalysis datasets (ERA5-Land, GLDAS-Noah, GLDAS-CLSM, GLDAS-VIC, and JRA-3Q) during water year 2001–2023. Our results indicate that ERA5-Land and GLDAS-Noah better capture the spatial patterns and temporal variability of snow density across the study regions. Using these two datasets and in-situ observations, we analyze long-term trends of snow density in Canada, Russia, and the Western U.S. We find that, reanalysis datasets fail to reproduce the observed interannual trends. Reanalysis products tend to underestimate observed interannual changes of snow density in early winter or shift them to late winter months. Moreover, the impact of snow density biases on snow depth biases in reanalysis datasets varies by region and dataset through offsetting or amplifying snow water equivalent biases. This study provides a comprehensive evaluation of snow density accuracy in reanalysis datasets, and reveals the distinct spatiotemporal trends in snow density under global warming. Our results also highlight the divergent contribution of snow density biases to snow depth biases in reanalysis datasets across regions.
Li et al. (Tue,) studied this question.