GMM-SWE v1.0: A GRACE–Meteorology-Constrained Dataset of Snow Water Equivalent, Snow Depth, and Snow Density over the Northern Hemisphere, 2003–2022

Abstract. Seasonal snow constitutes one of the largest and most dynamic freshwater reservoirs in the Northern Hemisphere, yet reliable large-scale estimates of Snow Water Equivalent (SWE), snow depth, and snow density remain highly uncertain due to observational limitations and simplified model representations. Here, we present the Gravity–Meteorology Coupled Model (GMM-SWE), a physics-constrained data assimilation framework designed to reconstruct spatially continuous and physically consistent estimates of SWE, snow depth, and snow density across the Northern Hemisphere (35°–85° N) for the period 2003–2022. GMM-SWE integrates satellite gravimetry from GRACE and GRACE-FO with high-resolution meteorological forcing from ERA5-Land within an adaptive Kalman filtering scheme. Snow mass anomalies are isolated from total terrestrial water storage by explicitly subtracting non-snow components derived from an ensemble of land surface and hydrological models. Elevation-dependent melt dynamics and temperature thresholds are incorporated using global topographic information, enhancing realism in complex mountainous terrain. The assimilated snow mass is subsequently refined using a Rauch–Tung–Striebel (RTS) smoother, reducing temporal noise and ensuring physically coherent seasonal evolution. A key innovation of GMM-SWE is the dynamic decoupling of snow mass and volume. Snow depth and density are retrieved by coupling the smoothed SWE estimates with a physics-based SNOW-17 compaction scheme, resolved at sub-daily time steps, enabling explicit representation of snowpack densification and metamorphic processes without reliance on static density assumptions. Validation against an extensive network of 13,135 in-situ stations, as well as ERA5-Land, GLDAS, WGHM, and the GlobSnow product, demonstrates consistently strong performance across a wide range of climatic and topographic regimes. The framework shows particular skill in high-latitude continental interiors and substantially improves SWE representation in complex mountain regions where passive microwave retrievals are severely limited. Sensitivity analyses indicate that large-scale SWE patterns and interannual variability are primarily governed by physical mass-balance constraints rather than fine-tuned parameter choices. Overall, GMM-SWE provides a unified, physically informed approach for hemispheric-scale estimation of SWE, snow depth, and snow density, offering a robust dataset for hydrological applications and climate-scale assessments of cryospheric change.