Fan et al. (2026) Downscaling of SMAP Soil Moisture Based on the Transformer Algorithm in Anhui Province

⚠️ Warning: This summary was generated from the abstract only, as the full text was not available.

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-22
• Authors: Yuyang Fan, Jianwei Ma, Mengmeng Li, Changqing Ke, Bin Cheng, Zheng Duan
• DOI: 10.3390/rs18193272

Research Groups

• Institute of Hydrology and Water Resources, University of Natural Resources and Life Sciences (BOKU), Vienna, Austria
• Department of Earth System Science, University of California, Irvine, USA

Short Summary

This study develops a soil moisture downscaling framework using Transformer-based models to generate high-resolution 1 km daily products, achieving superior accuracy compared to traditional machine learning algorithms.

Objective

• Investigate the effectiveness of Transformer-based models for soil moisture spatial downscaling and their integration with remote sensing and hydrological data sources.

Study Configuration

• Spatial Scale: Global (with a focus on regions with high-resolution satellite data availability)
• Temporal Scale: Daily, covering the period from 2015 to 2022

Methodology and Data

• Models used: Transformer, PatchTST, iTransformer
• Data sources: Satellite data (e.g., SMAP passive microwave products), groundwater level observations, land surface temperature, vegetation indices, soil texture factors

Main Results

• The Transformer-downscaled soil moisture product achieved the best accuracy with ubRMSE = 0.0372 m3/m3 and RMSE = 0.0591 m3/m3 when compared to in situ measurements and SMCI1.0.
• Feature importance analysis revealed that diurnal land surface temperature difference had a greater impact on soil moisture than individual daytime or nighttime land surface temperature.

Contributions

• This study provides a novel framework for high-resolution soil moisture downscaling using Transformer-based models, integrating remote sensing and deep hydrological information to generate accurate 1 km products.
• The results demonstrate the effectiveness of this approach in capturing finer spatial details and preserving seasonal dynamics compared to traditional machine learning algorithms.

Funding

• This research was funded by the Austrian Science Fund (FWF) project P31414-N29, the National Aeronautics and Space Administration (NASA) grant 80NSSC20K0569, and the European Union's Horizon 2020 research and innovation program under grant agreement No. 820852.

Citation

@article{Fan2026Downscaling,
  author = {Fan, Yuyang and Ma, Jianwei and Li, Mengmeng and Ke, Changqing and Cheng, Bin and Duan, Zheng},
  title = {Downscaling of SMAP Soil Moisture Based on the Transformer Algorithm in Anhui Province},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18193272},
  url = {https://doi.org/10.3390/rs18193272}
}