Li et al. (2026) Estimating reference evapotranspiration using a BiLSTM model incorporating spatiotemporal features in Southwestern China

Identification

• Journal: Agricultural Water Management
• Year: 2026
• Date: 2026-08-24
• Authors: Zongyang Li, Zhengxin Zhao, Liwen Xing, Ningbo Cui, Huanjie Cai, Zongjun Wu, Shenglin Wen
• DOI: 10.1016/j.agwat.2026.110708

Research Groups

• College of Water Resources and Architectural Engineering, Northwest A & F University
• Institute of Water-saving Agriculture in Arid Areas of China, Northwest A & F University
• College of Water Resource and Hydropower, Sichuan University
• Chengdu Agricultural Science and Technology Center

Short Summary

This study develops a hybrid deep learning framework integrating a Convolutional Neural Network (CNN), Temporal Position Attention (TPA), Bidirectional Long Short–Term Memory network (BiLSTM), and Adaptive Boosting (Adaboost) to enhance ETo estimation accuracy from both temporal and spatial perspectives under limited data scenarios in Southwestern China.

Objective

• To develop a hybrid deep learning framework for accurate and transferable ETo estimation under diverse meteorological input conditions.
• To evaluate the applicability and simulation performance of empirical models and various deep learning models (BiLSTM, CNN–BiLSTM, TPA–CNN–BiLSTM, and TPA–CNN–BiLSTM–Adaboost) under different meteorological input combinations across southwestern China.

Study Configuration

• Spatial Scale: Southwestern China with complex terrain and diverse climatic conditions.
• Temporal Scale: Daily meteorological data spanning 60 years (1961–2020).

Methodology and Data

• Models used:
  • BiLSTM model
  • TPA–CNN–BiLSTM–Adaboost model
  • Empirical models (Priestley–Taylor, Hargreaves–Samani, Brockamp–Wenner)
• Data sources: Daily meteorological data from 95 stations in Southwestern China.

Main Results

• The proposed TPA–CNN–BiLSTM–Adaboost framework consistently achieved the highest predictive accuracy across all climatic zones and validation schemes.
• The Rn based configuration achieved the best overall performance, especially in reducing error metrics compared with the T–based and RH–based configurations.
• The model demonstrated smaller performance discrepancies between internal and external validation schemes compared with other models.

Contributions

• This study establishes an advanced BiLSTM-type ETo estimation framework capable of accommodating diverse meteorological input scenarios in Southwestern China.
• The proposed framework provides effective technical support for agricultural water management and irrigation decision-making in climatically complex regions.

Funding

• This research was funded by the [project name], [reference code].

Citation

@article{Li2026Estimating,
  author = {Li, Zongyang and Zhao, Zhengxin and Xing, Liwen and Cui, Ningbo and Cai, Huanjie and Wu, Zongjun and Wen, Shenglin},
  title = {Estimating reference evapotranspiration using a BiLSTM model incorporating spatiotemporal features in Southwestern China},
  journal = {Agricultural Water Management},
  year = {2026},
  doi = {10.1016/j.agwat.2026.110708},
  url = {https://doi.org/10.1016/j.agwat.2026.110708}
}