Hyperspectral estimation of net photosynthetic rate of winter wheat under different water and nitrogen supplies

Abstract

Photosynthesis is a vital physiological activity in winter wheat that directly influences the production and accumulation of biomass. The net photosynthetic rate is a key indicator of photosynthetic capacity. Measuring the net photosynthetic rate using traditional methods can be challenging for high-throughput real-time monitoring. Reflectance spectroscopy has been shown to predict the physiological activities of crops and can track the physiological traits. This study focused on using leaf hyperspectral reflectance to estimate the net photosynthetic rate of winter wheat under different water and nitrogen supplies. At first, we transformed the raw spectral reflectance into relevant vegetation indices and extracted sensitive features using Competitive Adaptive Reweighted Sampling (CARS) and Successive Projections Algorithm (SPA). Then, estimation models for the net photosynthetic rate were constructed using Random Forest (RF) and Partial Least Squares Regression (PLSR) methods. Finally, the performance of the eight estimation models was compared using coefficient of determination (R²) and Root Mean Square Error (RMSE). The results showed that transforming raw spectral reflectance into vegetation indices significantly improved model performance. RF showed notably higher accuracy than PLSR. The VI-SPA-RF model was most accurate, with an R² of 0.9429 for the training set and 0.7784 for the validation set. Therefore, the leaf hyperspectral data can be used for nondestructive monitoring of the net photosynthetic rate of winter wheat in real-time.