**6. Conclusions**

In this paper, we propose a downscaling method for CYGNSS SM based on the XGBoost algorithm, using high-resolution CYGNSS observables and auxiliary variables as input data to improve the spatial resolution of GNSS-R technique retrieval of SM to 3 km. The method selects common downscaled variables such as DEM, land cover, NDVI, and rainfall. We enhance and improve the polynomial-based downscaling regression model by incorporating parameters of SR, SNR, LES, and TES from CYGNSS. Experiments were conducted using data covering the southern United States, and the results were validated by 78 in situ sites. The results show that the downscaled SM achieves *R*, RMSE, and MAE

of 0.712, 0.068, and 0.058, respectively, compared with the in situ SM observations. Spatial analysis using MODIS EVI and MODIS ET products shows that the spatial distribution and temporal variation of the downscaled CYGNSS SM products are more consistent with the EVI and ET products. The feasibility of the method is proved. Additionally, we discuss a number of problems that came up throughout the downscaling and validation process.

Overall, the findings of this study offer valuable insights for enhancing SM downscaling methods. This is crucial for advancing high-resolution SM retrieval. In future research, it will be key to develop gap-filling methods to address missing remote sensing data and refine the downscaling model. Additionally, researchers could consider using satellite SM products from various sources (e.g., SMAP, SMOS, AMSR-E, NASA-USDA, etc.) as reference values for downscaling models. This could aid in determining the most efficient downscaled SM products that are best suited to the particular conditions of the selected study area. Additionally, future research could consider a gradual downscaling approach (for instance, downsizing from 36 km to 9 km, followed by a reduction from 9 km to 3 km), as opposed to an immediate downscaling from 36 km to 3 km.

**Author Contributions:** Q.L.: Methodology, Writing—original draft, Writing—review and editing. Y.L.: Methodology, Writing—original draft, Writing—review and editing. Y.G.: Visualization, Validation. X.L.: Supervision. C.R.: Project administration. W.Y.: Software. B.Z.: Data curation. X.J.: Investigation. All authors have read and agreed to the published version of the manuscript.

**Funding:** This work was supported by the National Natural Science Foundation of China (No. 41901409); the Natural Science Foundation of Guangxi (No. 2021GXNSFBA220046); the National Natural Science Foundation of China (No. 42064003).

**Data Availability Statement:** CYGNSS data can be downloaded from the Physical Oceanography Distributed Active Archive Center (PO.DAAC, https://podaac.jpl.nasa.gov, accessed on 1 April 2023). SMAP data can be obtained from the National Snow and Ice Data Center (NSIDC, https://nsidc.org, accessed on 15 April 2023). The in situ soil moisture data can be accessed publicly (http://ismn.geo. tuwien.ac.at, accessed on 20 April 2023).

**Acknowledgments:** The authors are grateful to NASA EOSDIS Physical Oceanography Distributed Active Archive Center (DAAC), Jet Propulsion Laboratory, Pasadena, CA, USA, for making the CYGNSS data available, at https://www.esrl.noaa.gov/psd/ (accessed on 1 April 2023).

**Conflicts of Interest:** The authors declare no conflict of interest.
