Ge et al. (2026) Optimizing Priestley–Taylor Model Based on Machine Learning Algorithms to Simulate Tomato Evapotranspiration in Chinese Greenhouse

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

Identification

• Journal: Horticulturae
• Year: 2026
• Date: 2026-01-14
• Authors: Jiankun Ge, Jiaxu Du, Xuewen Gong, Quan Zhou, Guoyong Yang, Yanbin Li, Huanhuan Li, Jiumao Cai, Hanmi Zhou, Mingze Yao, Xinguang Wei, Weiwei Xu
• DOI: 10.3390/horticulturae12010089

Research Groups

Not explicitly mentioned in the provided text.

Short Summary

This study proposes three methods to revise the Priestley–Taylor (PT) model coefficient α for improved greenhouse crop evapotranspiration (ET) prediction under varying irrigation conditions. The machine learning-optimized PT-M(XGB) model demonstrated significantly higher accuracy for greenhouse tomato ET, providing a robust reference for precise water management.

Objective

• To further improve the prediction accuracy for greenhouse crop evapotranspiration (ET) under different irrigation conditions and enhance irrigation water use efficiency.

Study Configuration

• Spatial Scale: Greenhouse scale (specific to tomato cultivation).
• Temporal Scale: Two full growing seasons (2020 and 2021), covering different growth stages (seedling, flowering and fruiting, fruit enlargement, harvest).

Methodology and Data

• Models used: Priestley–Taylor (PT) model, MPT model (PT model with leaf senescence coefficient fS, plant temperature constraint parameter ft, and soil water stress index fsw), PT-M model (PT model combined with inverse calculation from Penman–Monteith model), PT-M(XGB) model (PT-M model optimized using XGBoost algorithm), Penman–Monteith (PM) model.
• Data sources: Field measurements of meteorological data inside the greenhouse, tomato physiological and ecological indices, observed evapotranspiration (ET), cumulative evaporation (Ep) from a 0.2 m standard evaporation pan. Two irrigation treatments were applied: K0.9 (0.9Ep) and K0.5 (0.5Ep).

Main Results

• The MPT model mean coefficient α for the entire growth stage was 1.27 (K0.9) and 1.26 (K0.5). The PT-M model mean coefficient α was 1.31 (K0.9) and 1.30 (K0.5).
• For both MPT and PT-M models, α was significantly lower than the conventional value of 1.26 during the seedling and flowering/fruiting stages, rapidly increased during the fruit enlargement stage, and then gradually declined towards 1.26 during the harvest stage.
• The PT-M model underestimated observed ET (ETm) by 8.71% to 16.01% during the seedling and harvest stages, and overestimated by 1.62% to 6.15% during the flowering/fruiting and fruit enlargement stages. Overall errors compared to ETm were 0.1% to 3.3% with an R² of 0.92 to 0.96 across both irrigation treatments and years.
• The PT-M(XGB) model achieved the highest prediction accuracy, with errors compared to ETm of 0.35% to 0.65% and an R² above 0.98 across both irrigation treatments and years.

Contributions

• Proposes three novel methods (MPT, PT-M, PT-M(XGB)) to dynamically revise the Priestley–Taylor model coefficient α for greenhouse crop ET calculation.
• Demonstrates that the PT-M(XGB) model, integrating the XGBoost algorithm, significantly improves the prediction accuracy of greenhouse tomato ET.
• Provides a valuable reference for the precise calculation of greenhouse tomato ET, contributing to enhanced irrigation water use efficiency.

Funding

Not explicitly mentioned in the provided text.

Citation

@article{Ge2026Optimizing,
  author = {Ge, Jiankun and Du, Jiaxu and Gong, Xuewen and Zhou, Quan and Yang, Guoyong and Li, Yanbin and Li, Huanhuan and Cai, Jiumao and Zhou, Hanmi and Yao, Mingze and Wei, Xinguang and Xu, Weiwei},
  title = {Optimizing Priestley–Taylor Model Based on Machine Learning Algorithms to Simulate Tomato Evapotranspiration in Chinese Greenhouse},
  journal = {Horticulturae},
  year = {2026},
  doi = {10.3390/horticulturae12010089},
  url = {https://doi.org/10.3390/horticulturae12010089}
}