Photosynthetic rate prediction and synergistic environmental optimization for cucumber based on PSO-XGBoost

This is the authors' abstract. We don't add key points for this paper.

Protected agriculture relies on the precise regulation of environmental factors such as temperature, light, gas, water, and nutrients. The net photosynthetic rate (Pn) of crops is a crucial physiological indicator that characterizes the effectiveness of environmental control and growth potential. To investigate the interactive effects of nutrient solution electrical conductivity (EC), photosynthetic photon flux density (PPFD), CO2 concentration, and temperature (T) on cucumber Pn, and to provide a basis for precise environmental regulation, cucumber plants were used as the experimental material. Based on Pn data obtained from various multi-level combinations of EC, CO2 concentration, PPFD, and T, prediction models including support vector regression, random forest, and extreme gradient boosting (XGBoost) were constructed. The particle swarm optimization (PSO) algorithm was applied for hyperparameter tuning. Furthermore, SHapley Additive exPlanations (SHAP) was integrated to analyze the contribution of each environmental factor, exploring their non-linear impacts on Pn prediction and their potential interactions. The results demonstrated that the PSO-XGBoost model exhibited the optimal comprehensive performance, achieving a coefficient of determination of 0.997 and a root mean square error of 0.627 μmol m-2 s-1 on the test set. The relative importance of the environmental factors to Pn followed the order: PPFD > CO2 > T > EC. Notably, EC exhibited a significant threshold effect: under the current experimental conditions, an EC level of 3.0 mS cm-1 was optimal for maintaining elevated Pn. These results provide theoretical and practical insights for synergistic environmental and nutrient regulation in greenhouse cucumber production.