High spatial resolution snow water equivalent (SWE) is critical for hydrological, ecological, and disaster research. However, passive microwave SWE products (10/25 km) with coarse spatial resolution can no longer meet modern demands for high precision and fine resolution. This study integrated newly calibrated enhanced-resolution brightness temperature data with optical snow area fraction and snow cover days, employing the deep learning FT-Transformer model to retrieve daily snow depth data at 5 km spatial resolution during the snow cover period (October to April) in the Three-River Source Region. The snow depth was subsequently converted into 5 km spatial resolution SWE data using monthly averaged snow density. This work establishes a robust data foundation for snow resource monitoring in the Three-River Source Region.
（1）The Calibrated Enhanced Resolution Brightness Temperature (CETB) data are provided by the National Snow and Ice Data Center (https://nsidc.org/data/NSIDC-0630/versions/1). The Calibrated Enhanced Resolution Brightness Temperature (CETB) data is provided by the National Snow and Ice Data Center. This data covers observed bright temperature data from different satellites since 1978 with a temporal resolution of 1d and a spatial resolution of 6.25 km/3.125 km.
（2）The snow area ratio data were obtained from the literature https://www.sciencedirect.com/science/article/pii/ S0924271624003265, which has a temporal resolution of 1 d and a spatial resolution of 5 km. Snow cover days data were obtained from the National Cryosphere Desert Data Center (http://www.ncdc.ac.cn), which has a temporal resolution of 1 d and a spatial resolution of 500 m.
（3）The DEM data were obtained from the National Geospatial-Intelligence Agency (NGA) and the National Aeronautics and Space Administration (NASA). ) and the Shuttle Radar Topography Mapping Mission (SRTM) operated by the National Aeronautics and Space Administration (NASA) (http://srtm.csi.cgiar.org/SELECTION/inputCoord.asp), with a spatial resolution of 90 m.
（4）Land-use type data were derived from the MCD12Q1 V061 dataset ( https://earthexplorer.usgs.gov/), using the annual land cover types of its IGBP classification standard, which has a temporal resolution of 1 yr and a spatial resolution of 500 m. The land use type data are available from the MCD12Q1 V061 dataset (https://earthexplorer.usgs.gov/).
（5） Snow density data sourced from the National Cryosphere Desert Data Center（ http://www.ncdc.ac.cn ）The grid data set of monthly and multi-year average snow cover density on the Qinghai Tibet Plateau includes 12 snow cover density maps of 5 km per month.
(1) Using the python platform to unify the batch processing of various data sources with a spatial and temporal resolution of 5 km day by day as a way to construct data inputs for snow depth retrieval; (2) Implementing a snow depth retrieval model with multiple data fusion through deep learning model training and parameter optimization; (3) Estimating snow depth data using the training-saved model for the Three-River Source Region; (4) Mask the water body and then fill the orbital gap of the passive microwave radiometer by averaging the day before and after; (5) Multiply snow depth data by snow density to obtain snow water equivalent.
The data quality is good. Since the snow water equivalent mainly comes from the conversion of snow depth through the monthly average snow density, the accuracy of snow depth has been verified to be good: the root-mean-square error (RMSE), the mean absolute error (MAE) and the correlation coefficient (R) are used to represent the snow depth error, and the ground measured snow depths from 2000 to 2020 are used for evaluation. The verification results show that the RMSE of 5 km snow depth in the Sanjiangyuan area is located at 8 - 8.5 cm, and the MAE is located at 5.6 - 6.5 cm, and R is greater than 0.7. Compared with the long-time series snow depth data (25 km) in China (RMSE is located at 10 - 11.5 cm, MAE is located at 7.5 - 8.3 cm, and R is greater than 0.45), the accuracy is better. The results show that the retrieved snow depth of 5 km has good accuracy in the Sanjiangyuan area. Therefore, the transformed snow-water equivalent data set can serve as a reliable data basis for evaluating snow cover resources in this area.
| # | number | name | type |
| 1 | 2023YFC3206300 | 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.NIEER-SNOW.DB6815.2025
10.12072/ncdc.nieer-snow.db6815.1
