A method for quantifying 3D variation in photosynthetic ability in maize canopies

The 3D heterogeneity in nitrogen content and temperature within the canopy affects canopy photosynthesis. Currently, there are no methods for efficiently assessing the heterogeneous 3D-distribution of leaf nitrogen content and leaf temperature and integrating that information into a 3D model of canopy photosynthesis. We therefore developed a high-throughput pipeline for collecting canopy photosynthesis parameters in maize (Zea mays) by combining several innovations. First, we used readily obtained SPAD502Plus meter readings to infer local leaf nitrogen content. Second, a Bayesian inference method allowed us to parameterize a C4 leaf photosynthesis model. Third, we used neural radiance fields (NeRFs) to recreate 3D plant architecture and SPAD distribution. Finally, we developed an indoor ray tracing and energy balance model to estimate local light distribution and leaf temperature within a canopy. SPAD values showed a distinct 3D pattern, suggesting within-canopy variation in photosynthesis. Bayesian inference efficiently parameterized the C4 leaf photosynthesis model, with estimated parameter values correlating well with SPAD values. In addition, NeRF more accurately reconstructed 3D architecture and estimated 3D SPAD distribution than traditional methods. This resulted in calculated leaf temperatures being similar to measured values. Different model assumptions can cause significant differences in simulated canopy photosynthetic rate. Omitting 3D SPAD heterogeneity alone produced a 1% to 8% difference in simulated canopy photosynthetic rate. Ignoring leaf temperature heterogeneity led to a difference in the calculated canopy photosynthetic rate of only 1% to 3% near the optimal temperature, but of up to 38% at 35 °C. This pipeline can be realized by high-throughput phenotyping platforms, making it suitable for exploring genetic differences and optimizing ideotype design for improved canopy photosynthesis.