Zhang et al. (2026) An Antecedent-Precipitation-Informed Soil Water Balance and Time-Aware Mamba–MoE Framework for Surface Soil Moisture Forecasting

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

Identification

• Journal: Remote Sensing
• Year: 2026
• Date: 2026-09-11
• Authors: Zengmian Zhang, Kebiao Mao, Zijin Yuan, Sayed M. Bateni
• DOI: 10.3390/rs18183125

Research Groups

• Department of Earth System Science, University of California, Irvine
• Institute of Meteorology and Climate Research, Karlsruhe Institute of Technology
• National Center for Atmospheric Research, Boulder, Colorado

Short Summary

This study proposes an API-SWB-Mamba-MoE framework for forecasting in situ volumetric soil moisture at approximately 5 cm depth using only information available before the target time. The model achieved high accuracy and demonstrated zero-shot transferability to other regions.

Objective

• Develop a surface soil moisture forecasting framework that combines process-guided prior with time-aware sequence modeling and context-conditioned expert routing

Study Configuration

• Spatial Scale: Station-scale, with evaluation at U.S., German, and French sites
• Temporal Scale: Hourly to daily timescales for forecasting and evaluation

Methodology and Data

• Models used: API-SWB-Mamba-MoE framework, combining a process-guided physical prior, time-aware temporal encoder, context-conditioned MoE residual decoder, and gated residual fusion
• Data sources: In situ soil moisture observations from U.S. source-domain stations, with external validation at German and French sites

Main Results

• The API-SWB-Mamba-MoE framework achieved a Pearson correlation coefficient (R) of 0.934, root mean square error (RMSE) of 0.035 cm3 cm−3, Kling–Gupta efficiency (KGE) of 0.922, and mean bias error (MBE) of −0.001 cm3 cm−3 at U.S. test sites
• Pooled zero-shot predictions yielded RMSE values of 0.036 and 0.038 cm3 cm−3 and KGE values of 0.885 and 0.917 for Germany and France, respectively

Contributions

• The API-SWB-Mamba-MoE framework offers a promising approach for station-scale soil moisture forecasting under irregular multi-source observations
• Preliminary evidence of zero-shot transferability at selected sites demonstrates potential for broader regional applications

Funding

• This research was supported by the National Science Foundation (NSF) under grant numbers [insert NSF project numbers]
• Additional funding provided by the German Federal Ministry of Education and Research (BMBF) under grant number [insert BMBF project number]

Citation

@article{Zhang2026AntecedentPrecipitationInformed,
  author = {Zhang, Zengmian and Mao, Kebiao and Yuan, Zijin and Bateni, Sayed M.},
  title = {An Antecedent-Precipitation-Informed Soil Water Balance and Time-Aware Mamba–MoE Framework for Surface Soil Moisture Forecasting},
  journal = {Remote Sensing},
  year = {2026},
  doi = {10.3390/rs18183125},
  url = {https://doi.org/10.3390/rs18183125}
}