High-throughput photosynthetic phenotyping using hyperspectral reflectance in paddy rice (Oryza sativa L.) under field conditions

• Hyperspectral reflectance enabled fast, simultaneous prediction of photosynthetic traits. • A PLSR framework combining spectra preprocessing, data splitting, and rdCV was developed. • SG preprocessing with stratified random sampling yielded the most accurate and stable prediction models. • Traits achieved moderate to high accuracy, with test R² values from 0.44 to 0.89. Accurate and high-throughput assessment of photosynthetic phenotyping traits is a major bottleneck in advancing crop breeding and precision agriculture. This study evaluated the feasibility of using leaf hyperspectral reflectance to estimate photosynthetic phenotyping traits in rice, including leaf nitrogen (N) content and four traits—PNmax, Isat, Icomp, and Rd—derived from in situ light response curves (LRCs) under field conditions. We developed a comprehensive Partial Least Squares Regression (PLSR) framework integrating six spectral preprocessing methods, two data splitting approaches for model training and test, and repeated double cross-validation (rdCV) for optimal PLS component selection. The combination of Savitzky-Golay (SG) preprocessing and stratified random sampling (data splitting Approach 1) consistently yielded the most accurate and stable model performance, except for Isat, where the Variable Sorting Normalization (VSN) and Approach 1 combination yielded better results. The models best predicted N content (R² = 0.89, RMSE = 0.25 %), followed by Isat (R² = 0.85, RMSE =128.2 (μmol (photon) m –2 s –1 ), PNmax (R² = 0.68, RMSE = 2.95 μmol (CO 2 ) m −2 s −1 ), Icomp (R² = 0.68, RMSE = 12.01 μmol (photon) m –2 s –1 ), and Rd (R² = 0.44, RMSE = 0.61 μmol (CO 2 ) m −2 s −1 ). For Isat, the model performance was obtained by excluding high-light-acclimated leaf samples, for which actual values could not be accurately observed. These findings highlight the capability of using a single leaf hyperspectral reflectance measurement for non-destructive, high-throughput, and simultaneous estimation of multiple traits associated with photosynthetic acclimation to light.