Wu et al. (2026) A hybrid machine learning and data assimilation framework for penman–monteith evapotranspiration estimation over China

Identification

• Journal: Remote Sensing of Environment
• Year: 2026
• Date: 2026-07-14
• Authors: Xiaowen Wu, Xiang Li, Jianli Ding, Shaomin Liu, Tongren Xu, Lisheng Song, Sayed Mohyeddin Bateni, Ziwei Xu, Dongxin Wu, Xinlei He, Jinjie Wang, Rui Liu, Yaxuan Guo
• DOI: 10.1016/j.rse.2026.115565

Research Groups

• College of Geography and Remote Sensing Science, Xinjiang University (China)
• State Key Laboratory of Earth Surface Processes and Disaster Risk Reduction, Faculty of Geographical Science, Beijing Normal University (China)
• Key Laboratory of Earth Surface Processes and Regional Response in the Yangtze-Huaihe River Basin, School of Geography and Tourism, Anhui Normal University (China)
• Department of Civil, Environmental and Construction Engineering, and Water Resources Research Center, University of Hawaii at Manoa (USA)
• School of Resource and Environmental Sciences, Wuhan University (China)
• Institute of Loess Plateau, Shanxi University (China)
• Institute of Urban Study, School of Environmental and Geographical Sciences, Shanghai Normal University (China)

Short Summary

This study presents an integrated machine-learning and data-assimilation framework for improving Penman-Monteith-based evapotranspiration estimates over China. The proposed framework achieves the best site-scale performance among four hybrid frameworks evaluated.

Objective

• Investigate the feasibility of integrating machine learning and data assimilation techniques to improve Penman-Monteith model-based evapotranspiration estimates over China.

Study Configuration

• Spatial Scale: National scale (China)
• Temporal Scale: Long-term (interannual variability, seasonal cycles)

Methodology and Data

• Models used:
  • Ensemble Kalman filter (EnKF) scheme
  • Four-dimensional variational assimilation (4DVar) scheme
  • Machine learning methods:
    • Long short-term memory (LSTM)
    • Artificial neural network (ANN)
    • Extreme gradient boosting (XGB)
• Data sources:
  • Satellite data (e.g., Landsat, MODIS)
  • Observation data from flux sites
  • Reanalysis data (e.g., ERA5-Land)

Main Results

• The proposed framework (4DVar-XGB-PM) achieved the best site-scale performance among four hybrid frameworks evaluated.
• Key findings:
  • Reduced positive bias at low-to-moderate evapotranspiration rates
  • Decreased absolute errors at high evapotranspiration rates
  • Retained physical basis of Penman-Monteith model

Contributions

• The study provides a novel framework for improving Penman-Monteith-based evapotranspiration estimates over China.
• The proposed framework demonstrates the potential to reduce biases and improve accuracy in evapotranspiration estimation.

Funding

• This research was supported by the National Natural Science Foundation of China (Grant No. 42161148003)
• The State Key Laboratory of Earth Surface Processes and Disaster Risk Reduction, Beijing Normal University (Grant No. SKLESPDRR2023001)

Citation

@article{Wu2026hybrid,
  author = {Wu, Xiaowen and Li, Xiang and Ding, Jianli and Liu, Shaomin and Xu, Tongren and Song, Lisheng and Bateni, Sayed M. and Xu, Ziwei and Wu, Dongxin and He, Xinlei and Wang, Jinjie and Liu, Rui and Guo, Yaxuan},
  title = {A hybrid machine learning and data assimilation framework for penman–monteith evapotranspiration estimation over China},
  journal = {Remote Sensing of Environment},
  year = {2026},
  doi = {10.1016/j.rse.2026.115565},
  url = {https://doi.org/10.1016/j.rse.2026.115565}
}