Multi-indicator integrated irrigation assessment for solar greenhouse grapevines using coupled CWSI and LAI ranges

Refining canopy parameter benchmarks for solar greenhouse grapevines in cold Northeast China may support robust vine growth and efficient water use. A three-year experiment (2023–2025) evaluated three deficit-irrigation treatments (I1: 65%–85% θ f , I2: 60%–80% θ f , I3: 55%–75% θ f ) and a fully irrigated control (CK: 70%–90% θ f ). Relationships between canopy parameters, including crop water stress index (CWSI) and leaf area index (LAI), and production indicators were analyzed. A comprehensive production score (D value) was constructed using principal component and membership function analyses. Three machine-learning models (support vector regression (SVR), k-nearest neighbor (KNN), and Gaussian process regression (GPR)) were used to characterize canopy–production relationships and identify favorable CWSI–LAI response regions. These regions were then compared with observed canopy ranges under the irrigation treatments to derive practical recommendations. Results showed that CWSI negatively affected aboveground dry matter accumulation, evapotranspiration, water-status indicators, and yield, whereas LAI showed positive effects. Both were also associated with fruit quality. SVR showed slightly more favorable overall test performance, with a test-set correlation of 0.91, the highest R 2 (0.83) and the lowest NRMSE (24.19%). The model-indicated favorable regions showed the closest overall agreement with canopy conditions under I1. I1 also provided favorable water-saving and production performance. Its recommended stage-specific CWSI ranges were 0.14–0.24, 0.14–0.23, 0.15–0.23, and 0.22–0.29, corresponding to LAI ranges of 0.02–1.35, 1.52–2.16, 2.73–4.70, and 4.71–4.99 m 2 m –2 . Considering inherent modeling and measurement uncertainties, these ranges serve as candidate management references rather than absolute thresholds, requiring further independent validation.