Predicting rice (Oryza sativa L.) canopy temperature difference and estimating its environmental response in two rice cultivars, ‘Koshihikari’ and ‘Takanari’, based on a neural network

Abstract

ABSTRACT Canopy photosynthesis is an important component of biomass production in field-grown rice (Oryza sativa L.). Although canopy temperature differences (CTD) provide important information for evaluating canopy photosynthesis, the measurement of CTD is still a labor-intensive task. Therefore, we designed this study to establish a model for predicting CTD under different field conditions using meteorological data and evaluated the environmental response of CTD using the established model. Our study collected 2,056,264 CTD data points from two rice cultivars having different photosynthetic capacities, ‘Koshihikari’ and ‘Takanari’, and then used these data to create a novel model using a neural network (NN). The input variables were limited to meteorological data, and the output variable was set to CTD. The established NN model produced a prediction accuracy of R2 = 0.792 and RMSE = 0.605°C. We then used this NN model to simulate the CTD response of the Koshihikari and Takanari cultivars in response to various environmental changes. These predictions revealed that Takanari had a lower CTD than Koshihikari when exposed to high relative humidity (RH) or low to moderate solar radiation (Rs ). In contrast, the CTD of Koshihikari tended to be lower than that of Takanari under lower RH or higher Rs . This result implies that the advantages of the single-leaf gas exchange system in Takanari can be mitigated under extremely high-VPD conditions. Thus, our new method may provide a powerful tool to gain a better understanding of gas exchange, growth processes, and varietal differences in rice cultivated under field conditions.