Wu et al. (2026) An RTE-guided mixture-of-experts framework for AMSR2 passive microwave soil moisture retrieval

Identification

• Journal: International Journal of Applied Earth Observation and Geoinformation
• Year: 2026
• Date: 2026-09-16
• Authors: Kebiao Mao, Jiancheng Shi, Zhonghua Guo, Sayed M. Bateni
• DOI: 10.1016/j.jag.2026.105597

Research Groups

• School of Electrical and Electronic-Engineering, Ningxia University, Yinchuan 750021, China
• State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China, Institute of Agricultural Resources and Regional Planning, Chinese Academy of Agricultural Sciences, Beijing 100081, China
• National Space Science Center, Chinese Academy of Sciences, Beijing 100190, China
• Department of Civil and Environmental Engineering and Water Resources Research Center, University of Hawaii at Manoa, Honolulu, HI 96822, USA
• UNESCO-UNISA Africa Chair in Nanoscience and Nanotechnology College of Graduates Studies, University of South Africa, Muckleneuk Ridge, Pretoria 392, South Africa

Short Summary

This study developed an RTE-guided mixture-of-experts (MoE) framework for AMSR2 soil moisture retrieval by combining iterative proxy-label refinement with a gated end-to-end MoE network trained on 14-channel brightness temperatures. The proposed method improved agreement with in situ observations while maintaining strong proxy-reference consistency.

Objective

• To develop an accurate and robust method for retrieving global soil moisture from AMSR2 multi-frequency brightness temperature observations.
• To improve the accuracy of soil moisture retrieval, especially in densely vegetated and arid regions where traditional methods often fail.

Study Configuration

• Spatial Scale: Global scale with a spatial resolution of approximately 10 km (0.1◦).
• Temporal Scale: Daily temporal resolution for model training and evaluation.

Methodology and Data

• Models used: Mixture-of-experts (MoE) framework, including an FCNN-based iterative proxy-label refinement module and a gated MoE backbone.
• Data sources: AMSR2 Level-3 brightness temperature and soil moisture datasets from the JAXA G-Portal database, LPRM AMSR2 soil moisture product, and ISMN in situ observations.

Main Results

• The proposed method reduced MAE from 0.048 to 0.034 m3/m3 and RMSE from 0.068 to 0.046 m3/m3 relative to the JAXA product.
• The refined-label MoE outperformed the MoE trained directly on JAXA-initialized labels, reducing MAE from 0.039 to 0.034 m3/m3 and RMSE from 0.055 to 0.046 m3/m3.

Contributions

• This study provides a novel RTE-guided MoE framework for global AMSR2 soil moisture retrieval.
• The proposed method improves agreement with in situ observations while maintaining strong proxy-reference consistency.

Funding

• This research was supported by the National Natural Science Foundation of China (Grant No. 42171319).
• The authors also acknowledge funding from the State Key Laboratory of Efficient Utilization of Arid and Semi-arid Arable Land in Northern China.
• Additional support came from the University of Hawaii at Manoa and the UNESCO-UNISA Africa Chair in Nanoscience and Nanotechnology.

Citation

@article{Wu2026RTEguided,
  author = {Wu, Yurong and Mao, Kebiao and Shi, Jiancheng and Guo, Zhonghua and Bateni, Sayed M.},
  title = {An RTE-guided mixture-of-experts framework for AMSR2 passive microwave soil moisture retrieval},
  journal = {International Journal of Applied Earth Observation and Geoinformation},
  year = {2026},
  doi = {10.1016/j.jag.2026.105597},
  url = {https://doi.org/10.1016/j.jag.2026.105597}
}