Xu et al. (2026) Water status diagnosis in greenhouse drip-irrigated tomato and celery using leaf turgor dynamics and machine learning

Identification

• Journal: Frontiers in Plant Science
• Year: 2026
• Date: 2026-01-16
• Authors: Quanyue Xu, Ruixia Chen, X. Li, Hongxiang Wu, Juanjuan Ma, Lijian Zheng
• DOI: 10.3389/fpls.2025.1743809

Research Groups

• College of Water Resource Science and Engineering, Taiyuan University of Technology, Taiyuan, China
• Shanxi Key Laboratory of Collaborative Utilization of River Basin Water Resources, Taiyuan University of Technology, Taiyuan, China

Short Summary

This study developed a non-invasive method using leaf patch clamp pressure (LPCP) probes and machine learning to diagnose water status in greenhouse drip-irrigated tomato and celery, identifying distinct diurnal turgor patterns linked to soil water content and achieving high prediction accuracy for precision irrigation.

Objective

• To examine the transient and interday characteristics of tomato and celery leaf turgor pressure under various moisture conditions.
• To explore the response relationship between leaf turgor pressure and environmental factors.
• To construct a long-time series prediction model for greenhouse tomato and celery leaf turgor pressure in drip irrigation using Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), and Random Forest (RF).

Study Configuration

• Spatial Scale: Experiments conducted in a naturally ventilated solar greenhouse (60 m × 11 m) in Liujiabao Village, Taiyuan City, Shanxi Province, China (112°29′E 37°39′N, altitude 766 m).
• Temporal Scale:
  • Tomato: May 17, 2021, to September 19, 2021, and May 25, 2022, to September 30, 2022.
  • Celery: November 1, 2021, to January 19, 2022, and November 15, 2022, to February 11, 2023.

Methodology and Data

• Models used: Support Vector Machine (SVM), Extreme Gradient Boosting (XGBoost), Random Forest (RF).
• Data sources:
  • Leaf Patch Clamp Pressure (LPCP) probes: Output parameter (P_p) measured every 5 minutes.
  • Automated weather station: Net solar radiation (Rs, W·m⁻²), atmospheric temperature (T, °C), relative humidity (RH, %), wind speed (WS, m·s⁻¹), saturated water vapor pressure deficit (VPD, kPa) measured every 5 minutes.
  • Time-Domain Reflectometry (TDR): Soil water content (SWC, %) at 0–60 cm depth, measured approximately every 7 days and before/after irrigation.

Main Results

• Diurnal P_p patterns exhibited two distinct states: State I (unimodal, no or mild stress) and State II (troughed, severe stress).
• Soil water content (SWC) thresholds for State I were > 20% (tomato) and > 19% (celery); for State II, they were < 18% (tomato) and < 16% (celery).
• Non-full irrigation treatments significantly increased the proportion of State II and resulted in higher Pp,max and Pp,min values (tomato: 15.39%–138.39% increase; celery: 3.44%–94.02% increase) compared to full irrigation.
• In State I, P_p was positively associated with solar radiation and negatively associated with SWC (tomato) and wind speed (celery). Correlations were weaker in State II.
• The Random Forest model, integrating substate P_p prediction based on meteorological factors and SWC (Combination 4), achieved the highest accuracy (R² = 0.995, MSE = 2.419, RMSE = 1.540, MAE = 0.531).
• SWC was identified as the most influential feature parameter in the optimal prediction model.

Contributions

• First comprehensive characterization of leaf turgor pressure (P_p) diurnal patterns (unimodal and troughed) and their corresponding soil water content thresholds for greenhouse drip-irrigated tomato and celery.
• Development of a highly accurate, non-invasive, and continuous predictive model for leaf turgor pressure using machine learning (Random Forest), which integrates both meteorological factors and soil moisture content, significantly outperforming models without SWC or state differentiation.
• Provides a scientific foundation for optimizing precision irrigation scheduling in greenhouse vegetables, offering a cost-effective alternative to extensive probe deployment.

Funding

• National Natural Science Foundation of China (52109061)
• Natural Science Foundation of Shanxi Province, China (202403021211047)

Citation

@article{Xu2026Water,
  author = {Xu, Quanyue and Chen, Ruixia and Li, X. and Wu, Hongxiang and Ma, Juanjuan and Zheng, Lijian},
  title = {Water status diagnosis in greenhouse drip-irrigated tomato and celery using leaf turgor dynamics and machine learning},
  journal = {Frontiers in Plant Science},
  year = {2026},
  doi = {10.3389/fpls.2025.1743809},
  url = {https://doi.org/10.3389/fpls.2025.1743809}
}