Soil moisture is one of the important climate variables that controls the exchange of water, carbon, and energy between land and atmosphere. Accurate and detailed understanding of the spatiotemporal distribution of soil moisture is crucial for various applications in the Earth system. As part of the Global Land Surface Satellite (GLASS) product suite, a long-term global 1km daily surface soil moisture product was generated from 2000 to 2020. This product (GLASS SM) is mainly generated from GLASS albedo, LST and LAI products, ERA5 land reanalysis soil moisture products, and auxiliary datasets based on ensemble machine learning models. The data values contained in GLASS SM products represent the volumetric moisture content of the topmost soil layer (0-5 cm). The file is stored in sine projection format and provided in Geo Tiff format. The value of 'Node Data' is set to -9999
This model was developed by integrating multiple datasets, including albedo, surface temperature, and leaf area index products from the Global Land Surface Satellite (GLASS) product suite, as well as soil moisture products from the European Reanalysis (ERA5 Land), in-situ soil moisture datasets from the International Soil Moisture Network (ISMN), and auxiliary datasets (Multi Error Deletion Improved Terrain DEM and SoilGrids).
The integrated learning model (eXtreme Gradient Boosting XGBoost) was used to generate daily spatiotemporal continuous soil moisture products (GLASS SM) for 1 km worldwide from 2000 to 2020.
To comprehensively evaluate the performance of the model, three validation strategies were explored: random, site independent, and year independent. The results showed that for random test samples, the XGBoost model trained using the TC method to select representative sites had the highest accuracy, with an overall correlation coefficient (R) of 0.941 and a root mean square error (RMSE) of 0.038 m3 m-3; For test samples that are independent of the site and year, although the overall performance of the model is relatively low, training the model with representative sites can still greatly improve its overall accuracy. Meanwhile, compared with the model without station screening, the model trained with representative stations showed significant improvement in validation accuracy on most stations. The median R and unbiased RMSE (ubRMSE) of each station model increased from 0.64 to 0.74, and decreased from 0.055 to 0.052 m3 m-3, respectively. Further validation of the GLASS SM product in four independent soil moisture networks showed its ability to capture the temporal dynamics of measured soil moisture (R=0.69-0.89; ubRMSE=0.033-0.048 m3 m-3). Finally, the mutual comparison between the GLASS SM product and two global microwave soil moisture datasets -1 kilometer soil moisture active passive/Santiano-1 L2 radiometer/radar soil moisture product and the European Space Agency Climate Change Initiative 0.25 ° combination soil moisture product - showed that the derivative product maintained more complete spatial coverage and exhibited high spatiotemporal consistency with these two soil moisture products.
This work is licensed under a
Creative
Commons Attribution 4.0 International License.
CSTR:11738.11.NCDC.ZENODO.DB6552.2024
10.5281/zenodo.7172664