Snow Depth (SD) is a key parameter for characterizing snow thickness, playing a significant role in understanding regional water cycles, energy balance, and the impacts of climate change. To address the substantial uncertainties in existing remote sensing, reanalysis, and simulated snow depth products, as well as their insufficient accuracy in complex terrain regions, this project employs a random forest algorithm. It integrates snow depth products such as AMSR-E, AMSR2, NHSD, and GlobSnow, along with reanalysis datasets like ERA-Interim and MERRA2, and relevant environmental variables. Using ground-based observational snow depth data for model training and validation, a daily-scale SD product with a spatial resolution of 0.25° for the Arctic (north of 66°34'N) from 1980 to 2019 was generated through data fusion. Validation with measured snow depth data shows a correlation coefficient (R²) of 0.79, with a root mean square error (RMSE) of 8.5 cm and a mean absolute error (MAE) of 3.5 cm. This dataset provides crucial data support for hydrological modeling and data assimilation in land surface process models in the Arctic.
AMSRE data are collected by the microwave scanning radiometer onboard NASA's Aqua satellite (Earth Observing System) and JAXA's GCOM-W1 satellite (https://nsidc.org/data/ae_dysno). NASA's AMSRE (AMSR-E) provides snow depth datasets from June 19, 2002, to October 3, 2011, while JAXA's AMSRE (AMSR2) has been continuously providing global daily snow depth data since July 2, 2012. Funded by the European Space Agency, the GlobSnow snow depth product is a collaborative dataset that utilizes an advanced data assimilation scheme to combine satellite passive microwave data (from sensors such as SSM/I and AMSR-E) with in-situ snow depth observations from meteorological stations (https://www.globsnow.info/), though partial data are missing for September. The NHSD dataset is a long-term Northern Hemisphere snow depth product provided by the National Tibetan Plateau Data Center (https://poles.tpdc.ac.cn/). ERA5-Interim is a global land surface reanalysis dataset produced by the European Centre for Medium-Range Weather Forecasts (https://apps.ecmwf.int/datasets/data/interim-full-daily/) and provides daily data. MERRA-2 is a reanalysis dataset released by NASA's Global Modeling and Assimilation Office (https://disc.gsfc.nasa.gov/datasets/).
The spatial resolution of datasets including AMSR-E, AMSR2, NHSD, GlobSnow, ERA-Interim, and MERRA2 was resampled to 0.25°. Using the random forest algorithm, a more temporally complete snow depth dataset was generated by integrating the aforementioned data products and environmental factors, along with model training and validation based on in-situ snow depth observations.
We calculated the error metrics to evaluate the accuracy by comparing SWE dataset and datesets with observations. The root meansquare error (RMSE), mean absolute error (MAE), Pearson correlation coefficient (R) and bias were adopted to assess theaccuracy of the SD dataset. To ensure spatiotemporal consistency during the data fusion process, the validation scope was confined to regions north of 66°34'N. The time series was defined during 1980 to 2019. Moreover, to validate the accuracy in spatial of SD dataset, cross-validation was performed using observations.
| # | number | name | type |
| 1 | 2020YFA0608501 | Research on Arctic Terrestrial Environmental Change and Its Effects | National key R & D plan |
This work is licensed under
CC BY 4.0 (Creative Commons Attribution 4.0 International License).
CSTR:11738.11.NCDC.ARCTIC-CHANGE.DB7130.2026
10.12072/ncdc.arctic-change.db7130.2026
