In this study, a machine learning algorithm was introduced to fuse gridded snow depth datasets. The input variables of the machine learning method included geolocation (latitude and longitude), topographic data (elevation), gridded snow depth datasets and in situ observations. A total of 29,565 in situ observations were used to train and optimize the machine learning algorithm. A total of five gridded snow depth datasets—Advanced Microwave Scanning Radiometer for the Earth Observing System (AMSR-E) snow depth, Global Snow Monitoring for Climate Research (GlobSnow) snow depth, Long time series of daily snow depth over the Northern Hemisphere (NHSD) snow depth, ERA-Interim snow depth and Modern-Era Retrospective Analysis for Research and Applications, version 2 (MERRA-2) snow depth—were used as input variables. The first three snow depth datasets are retrieved from passive microwave brightness temperature or assimilation with in situ observations, while the last two are snow depth datasets obtained from meteorological reanalysis data with a land surface model and data assimilation system. Then, three machine learning methods, i.e., Artificial Neural Networks (ANN), Support Vector Regression (SVR), and Random Forest Regression (RFR), were used to produce a fused snow depth dataset from 2002 to 2004. The RFR model performed best and was thus used to produce a new snow depth product from the fusion of the five snow depth datasets and auxiliary data over the Northern Hemisphere from 2002 to 2011. The fused snow-depth product was verified at five well-known snow observation sites. The R² of Sodankylä, Old Aspen, and Reynolds Mountains East were 0.88, 0.69, and 0.63, respectively. At the Swamp Angel Study Plot and Weissfluhjoch observation sites, which have an average snow depth exceeding 200 cm, the fused snow depth did not perform well. The spatial patterns of the average snow depth were analyzed seasonally, and the average snow depths of autumn, winter, and spring were 5.7, 25.8, and 21.5 cm, respectively. In the future, random forest regression will be used to produce a long time series of a fused snow depth dataset over the Northern Hemisphere or other specific regions.

The spatiotemporal heterogeneity of snowpack properties—particularly snow grain size, density, and liquid water content—combined with the attenuating and radiating effects of forest canopy, continues to pose critical challenges that limit the accuracy of snow depth retrieval from passive microwave remote sensing. To address these issues, this study introduces the Equivalent Volume Scattering Index (EVSI), a physically informed metric designed to isolate the radiative contributions of non-snow-depth factors in microwave signal propagation. The EVSI is defined as the ratio of the differential brightness temperature between high-frequency (e.g., 37 GHz) and low-frequency (e.g., 19 GHz) passive microwave channels to in situ ground-based snow depth observations. Leveraging the joint spatiotemporal patterns of in situ snow depth and EVSI, we first classified Northern Hemisphere snowpack into seven distinct snow types via unsupervised cluster analysis. This typology captures dominant regimes characterized by unique combinations of microphysical and environmental conditions. For each snow type, we then developed a dynamic, regionally adaptive, and partially non-resetting EVSI-based snow depth retrieval model. The “partially non-resetting” design preserves key snow state variables across time steps—such as grain size evolution and liquid water retention—while allowing radiative transfer parameters to adapt dynamically to evolving snow and canopy conditions. In contrast to conventional passive microwave snow depth algorithms, the proposed framework not only ensures physical interpretability through its foundation in microwave radiative transfer theory but also prioritizes operational feasibility by relying exclusively on readily accessible inputs, including daily air temperature, daily precipitation, and daily brightness temperatures from both low- and high-frequency microwave channels. Consequently, the algorithm simultaneously achieves higher retrieval accuracy and enhanced spatiotemporal generalizability, demonstrating robust performance across diverse climatic zones and seasonal cycles—thereby advancing both scientific understanding and practical applicability in global snow monitoring.