Estimating photosynthetic capacities from reflectance spectra: techniques and scales

Abstract

Accurate measures of photosynthetic capacity are crucially important for terrestrial earth system models to improve understanding of magnitudes and trends of global land carbon fluxes. Estimation of photosynthetic capacities in a high-throughput manner holds merits for selecting for and/or redesigning improved photosynthetic pathways – promising solution to increase crop yield to satisfy rising demands for food and fuel incurred by a growing human population. However, spatially and temporally resolved measurements of photosynthetic traits are still not readily available partly due to technical and scale challenges. Built upon a three-year field experiment (2017-2019), our study synthesized machine learning-based (e.g., support vector machine regression and partial least squares regression) and remote sensing-based (e.g., spectral indices and inversion of a remote sensing radiative transfer model) approaches to estimate photosynthetic capacities at leaf and canopy scales. Eleven tobacco genotypes including both genetically modified and wild types together exhibiting a wide range of photosynthetic capacity were used in the field experiments. Reflectance spectra of these plants were measured at the leaf level using a portable spectroradiometer and at the canopy level using a hyperspectral camera mounted on a mobile phenotyping platform. Ground-truth photosynthetic parameters were provided by a portable leaf gas exchange system. Results show that the performance of these developed approaches to characterize photosynthetic variations is comparable across scales. Our study holds important implications for broad-scale mapping of photosynthesis and acceleration of plant breeding process to develop crop cultivars with improved photosynthesis.