Abstract: Aim: In recent years, the ever increasing greenhouse gas (GHG) emissions and extreme weather events had significantly impacted plant photosynthesis. Photosynthesis is closely related to plant growth and development, with photosynthetic rate is a key indicator of plant health, which has been widely used in predicting global carbon cycle dynamics. The net photosynthesis rate is also an important parameter in calibrating agriculture facility. Therefore, accurate prediction of plant photosynthesis rate is important to agriculture, forestry and grassland research. Methods: We used photosynthesis meter to obtain photosynthetic data of common reed (Phragmites australis) and smooth cordgrass (Spartina alterniflora) under different environmental conditions, we then fitted seven single-factor photosynthetic response models, and developed a multifactorial photosynthetic rate prediction model using theoretically-guided neural network (TgNN). Important findings: The multifactor photosynthetic model outperformed single-factor photosynthetic response models. All existing single-factor light response models show good performances in fitting photosynthetic data but lack theoretical research value; whereas the multifactor photosynthetic model based on TgNN shows good predictability and generality. Our contribution provides a new avenue for calibrating more accurate and reliable models for predicting plant photosynthetic rate.

Key words: photosynthetic prediction model, neural network, photoresponse model, grey box model