Integrating leaf scaled spectra and machine learning for rapid estimation of photosynthetic phenotypes across soybean genotypes

Photosynthesis is central to crop productivity and global carbon dynamics, yet traditional methods for evaluating photosynthetic capacity are slow, low-throughput, and difficult to scale. This study presents a novel, rapid, nondestructive approach to estimate soybean leaf photosynthetic phenotypes using hyperspectral reflectance and machine learning. A new index, the comprehensive photosynthetic phenotype indicator (CPPI), was developed using principal component analysis to integrate gas exchange, chlorophyll fluorescence, and pigment traits into a single predictive variable. To achieve rapid estimation of soybean leaf photosynthetic phenotypes, leaf-scaled hyperspectral data along with photosynthetic indicators, fluorescence parameters, and pigments across 62 soybean genotypes over two growing seasons and three growth stages were collected. Three machine learning methods combined with transformed spectra were used to test the predictive performance for soybean photosynthetic phenotypes. The results showed that the developed CPPI achieved strong predictive performance (R2v= 0.704, RMSE = 0.779, RPD = 1.838), comparable to the best single trait of Trmmol (R2 v = 0.793, RMSE = 1.827, RPD = 2.012). These key spectral regions or wavelengths of 350–380, 550, 680–750, 860, 1390, 1660, and 1710 nm contributed to the spectral prediction as they are highly and secondarily related to leaf photosynthetic phenotypes. Moreover, it was found that spectral pretreatment offered limited improvement to the spectral model performance, while integrating partial least squares (PLS) regression with selected informative wavelengths enhanced prediction stability and robustness. Our results demonstrate the potential of CPPI as a scalable, integrative trait for high-throughput phenotyping and breeding applications. This approach allows for rapid screening of high-efficiency photosynthetic genotypes under field conditions, offering a practical tool for improving crop productivity through spectral selection.