A canopy photosynthesis model based on a highly generalizable artificial neural network incorporated with a mechanistic understanding of single-leaf photosynthesis

Abstract

• A canopy photosynthesis model was built based on an artificial neural network (ANN) • The model was combined with a process-based single-leaf photosynthesis model • The model could predict the canopy photosynthetic rate under various environments • The model was highly generalizable even when training data were scarce Crop productivity is largely dependent on canopy photosynthesis, which is difficult to measure at farming sites. Therefore, real-time estimation of the canopy photosynthetic rate ( A c ) is expected to facilitate effective farm management. For the estimation of A c , two types of mathematical models ( i.e ., process-based models and empirical models) have been used, although both types have their own weaknesses. Process-based models inevitably require many model parameters that are difficult to identify, while empirical models, including artificial neural network (ANN) models, have a low predictive ability outside of the range of training datasets. To overcome these weaknesses, we developed a hybrid canopy photosynthesis model that included components of both process-based models and ANN models. In this hybrid model, the single-leaf photosynthetic rate ( A L ) and leaf area index (LAI) were first estimated from information easily obtainable at farming sites: A L was estimated by the process-based model of A L ( i.e ., the biochemical photosynthesis model of Farquhar et al . (1980)) from environmental data (photosynthetic photon flux density (PPFD), air temperature ( T a ), humidity, and atmospheric CO 2 concentration ( C a )), and the LAI was estimated by an analysis of crop canopy imagery. As highly explainable information for A c , the estimated A L and LAI were input into the ANN model to estimate A c . As such, the ANN model learned the logical relationships between the inputs ( A L and LAI) and the output ( A c ). Detailed validation analysis using nine spinach A c datasets revealed that the hybrid ANN model can estimate A c accurately throughout the whole growth period, even when training and test datasets were obtained in different seasons under different CO 2 concentrations and based on training datasets of only three days. This study highlights the high generalizability of the hybrid ANN model, which is a prerequisite for practical application in environmentally controlled crop production.