Li et al. (2026) Integrated evaluation of snow density reanalysis products in the Northern Hemisphere

Identification

• Journal: Journal of Hydrology
• Year: 2026
• Date: 2026-02-03
• Authors: Yizhuo Li, Xin Miao, Xinyun Hu, Le Wang, Xueliang Zhang, Pengfeng Xiao, Weidong Guo
• DOI: 10.1016/j.jhydrol.2026.135071

Research Groups

• School of Atmospheric Sciences, Nanjing University, Nanjing, China
• Jiangsu Provincial Key Laboratory of Geographic Information Science and Technology, Key Laboratory for Land Satellite Remote Sensing Applications of Ministry of Natural Resources, School of Geography and Ocean Science, Nanjing University, Nanjing, China

Short Summary

This study systematically evaluated the accuracy and applicability of snow density from five reanalysis datasets (ERA5-Land, GLDAS-Noah, GLDAS-CLSM, GLDAS-VIC, JRA-3Q) across the Northern Hemisphere using 4,319 in-situ stations, finding that while ERA5-Land and GLDAS-Noah performed best in spatial and temporal representation, all reanalysis products failed to accurately reproduce observed long-term interannual snow density trends.

Objective

• To systematically evaluate the accuracy and applicability of snow density data from five widely used reanalysis datasets (ERA5-Land, GLDAS-Noah, GLDAS-CLSM, GLDAS-VIC, and JRA-3Q) across major snow-covered regions in the Northern Hemisphere, covering spatial patterns, monthly variations, annual trends, and snow depth-dependent biases.
• To analyze long-term snow density variations and the impact of snow density biases on snow depth estimates.

Study Configuration

• Spatial Scale: Northern Hemisphere, specifically Canada, Russia, and the Western United States (WUS). All reanalysis data were remapped to a uniform 0.1° × 0.1° grid.
• Temporal Scale: Water year (WY) 2001–2023 for primary evaluation (November to April). Long-term trend analysis for Canada (WY 1986–2014), Russia (WY 1967–2014), and WUS (WY 2004–2023).

Methodology and Data

• Models used:
  • ERA5-Land (HTESSEL model)
  • GLDAS-Noah (Noah Land Surface Model (LSM))
  • GLDAS-CLSM (Catchment LSM)
  • GLDAS-VIC (Variable Infiltration Capacity (VIC) model)
  • JRA-3Q (JMA Simple Biosphere (SiB) model)
• Data sources:
  • In-situ observations: Integrated data from 4,319 snow stations across the Northern Hemisphere.
    • Canada: Canadian historical Snow Water Equivalent dataset (CanSWE, 1928–2023), 2921 stations.
    • Russia: Russian Research Institute for Hydrometeorological Information–World Data Center, 517 stations (since 1966).
    • United States (Western US and Alaska): Snowpack Telemetry (SNOTEL) stations, 881 stations (since 1994).
  • Reanalysis datasets:
    • ERA5-Land (European Centre for Medium-Range Weather Forecasts (ECMWF))
    • GLDAS-2.1 (Noah, CLSM, VIC)
    • GLDAS-2.0 (Noah) for extended long-term analysis
    • JRA-3Q (Japan Meteorological Agency (JMA))

Main Results

• ERA5-Land and GLDAS-Noah consistently outperformed other datasets in both spatial and temporal accuracy. Spatially, they exhibited lower Root Mean Square Error (RMSE) (57.9 kg/m³ and 49.6 kg/m³, respectively), higher spatial correlation coefficients (0.58 and 0.65), and higher Taylor Skill Scores (0.35 and 0.46) across the Northern Hemisphere.
• Reanalysis performance varied regionally, being best in Russia and worst in the Western United States.
• Temporally, ERA5-Land and GLDAS-Noah showed high consistency with in-situ observations for annual means and multi-year monthly averaged snow density, with higher Consistency Index (CI) values (0.92 and 0.88).
• Reanalysis snow density biases for ERA5-Land and GLDAS-Noah showed a clear snow depth (SD) dependence, with the bias range narrowing as SD increased.
• Long-term trend analysis revealed that reanalysis datasets generally failed to accurately reproduce observed interannual trends:
  • Canada (WY 1986–2014): Observed a non-significant decline (-2.41 kg/m³ per decade), not captured by reanalysis.
  • Russia (WY 1967–2014): Observed a significant decrease (-4.19 kg/m³ per decade), primarily in early winter (November–January), which reanalysis datasets failed to capture (GLDAS-Noah (v2.0) showed an opposite significant increase).
  • Western US (WY 2004–2023): Observed a significant increase (6.85 kg/m³ per decade), while reanalysis datasets showed no significant decreasing trend.
• Reanalysis datasets tended to underrepresent early-winter (November–January) interannual variability and, in some cases, exaggerated or misrepresented changes in late winter (February–April).
• Snow density biases significantly influenced snow depth (SD) biases in reanalysis datasets. Underestimated snow density could partially offset underestimated Snow Water Equivalent (SWE) biases, leading to improved (or "false") SD accuracy, or exacerbate SD biases when SWE was overestimated, with these interactions varying by region and dataset.

Contributions

• Provided the first systematic and comprehensive evaluation of snow density accuracy from five widely used reanalysis datasets across major snow-covered regions of the Northern Hemisphere using an extensive integrated in-situ dataset.
• Revealed pronounced spatial heterogeneity and monthly variations in observed snow density trends, highlighting distinct regional patterns under global warming.
• Identified significant limitations of current reanalysis products in accurately capturing long-term trends and early-winter interannual variability of snow density.
• Demonstrated the complex, region- and dataset-dependent contribution of snow density biases to snow depth biases, including instances of "false accuracy" in snow depth estimates.
• Offered insights into the potential sources of reanalysis snow density biases, linking them to differences in meteorological forcing and land surface model snow parameterization schemes.

Funding

• National Natural Science Foundation of China (42305033)
• Second Tibetan Plateau Scientific Expedition and Research Program (STEP) (2019QZKK0103)

Citation

@article{Li2026Integrated,
  author = {Li, Yizhuo and Miao, Xin and Hu, Xinyun and Wang, Le and Zhang, Xueliang and Xiao, Pengfeng and Guo, Weidong},
  title = {Integrated evaluation of snow density reanalysis products in the Northern Hemisphere},
  journal = {Journal of Hydrology},
  year = {2026},
  doi = {10.1016/j.jhydrol.2026.135071},
  url = {https://doi.org/10.1016/j.jhydrol.2026.135071}
}