Abstract
Hyperspectral remote sensing (RS) has demonstrated to be useful for estimating vegetation photosynthetic traits, such as the maximum carboxylation rate of Rubisco (Vcmax) and the electron transport rate (Jmax). However, the spectral ranges used by RS models for predicting photosynthetic traits vary, and their predictive performance across different crop varieties is poor. This study aimed to investigate whether augmenting the modeling dataset's variability through various nitrogen experiments on tea chrysanthemum could improve RS-based photosynthetic trait models' applicability. Leaf-level measurements linked high-throughput spectral reflectance observations with photosynthetic traits obtained via a portable photosynthesis system. Results revealed strong correlations between the green and red edge bands and photosynthetic traits. Among newly developed vegetation indices, the structure insensitive pigment index [SIPI(850,691,476)] and SIPI(850, 699, 579) effectively predicted Vcmax and Jmax, respectively. In contrast, partial least squares regression (PLSR) modeling combined with reflectance from 400 to 1000 nm outperformed in estimating photosynthetic traits of tea chrysanthemum and exhibited excellent performance in a multi-variety validation dataset, indicating applicability across different varieties. Our findings suggest that increasing the modeling dataset's variability enhances the universality of RS estimation models for photosynthetic traits, providing a valuable tool for breeders to efficiently collect photosynthetic trait information.