Machine learning models for net photosynthetic rate prediction using poplar leaf phenotype data.

Area of Science:

• Environmental Science
• Plant Science
• Machine Learning

Background:

• Tree planting is crucial for reducing atmospheric CO2 and meeting global emission reduction goals.
• High net photosynthetic rates (Pn) in fast-growing trees are key to efficient CO2 sequestration.
• Predicting Pn using leaf traits can inform tree planting strategies for climate change mitigation.

Purpose of the Study:

• To develop a predictive model for net photosynthetic rate (Pn) in Populus simonii (P. simonii) using leaf phenotype data.
• To assess the effectiveness of machine learning algorithms in predicting Pn from sparse, noisy, and correlated leaf trait data.
• To guide tree planting policies by establishing a predictive relationship between leaf characteristics and Pn in woody plants.

Main Methods:

• Collected leaf phenotype and photosynthetic rate data from P. simonii across 23 artificial forests in northern China.
• Applied data preprocessing techniques including outlier removal and cluster analysis.
• Evaluated four regression methods: extreme gradient boosting (XGBoost), support vector machine (SVM), random forest (RF), and generalized additive model (GAM), with cross-validation and regularization.

Main Results:

• XGBoost achieved the highest accuracy, with a 0.77 correlation between predicted and actual net photosynthetic rates.
• XGBoost demonstrated a 35% reduction in root mean square error (RMSE) compared to the data's standard deviation.
• The study confirmed machine learning's capability to predict Pn accurately from noisy leaf phenotype data.

Conclusions:

• Machine learning models, particularly XGBoost, can effectively predict net photosynthetic rates in woody plants like P. simonii using leaf phenotype data.
• This predictive capability offers significant guidance for plant and environmental science, especially for optimizing tree planting strategies in northern China.
• The findings support the use of leaf traits and machine learning for informed decision-making in climate change mitigation efforts through afforestation.