Spectral Detection of Leaf Carbon and Nitrogen As a Proxy for Remote Assessment of Photosynthetic Capacity for Wheat and Maize under Nitrogen Stress

Zhenfeng Yang,Juncang Tian,Lixin Zhang,Ouyang Zan,Xinfang Yan,Kepeng Feng

DOI: https://doi.org/10.1016/j.compag.2024.109174

IF: 8.3

2024-01-01

Computers and Electronics in Agriculture

Abstract:Leaf carbon and nitrogen status has been shown to be cues for photosynthetic performance (e.g., Vcmax) in past studies. However, there is a theoretical gap in intuitive spectral evaluation for photosynthetic properties. Spectral assessment of carbon and nitrogen status provides a potential theoretical link to the remotely evaluated photosynthetic properties. In this study, the preferred leaf carbon and nitrogen spectral model was considered to infer photosynthetic traits in N-stressed crops with different photosynthetic pathways (wheat C3 and maize C4) from an active fluorescence perspective. Here, Different levels of N-stress field trials were executed at Ningxia University Irrigation Experimental Station in 2022.We screened the optimal spectral models for estimating leaf nitrogen content (LNC), carbon content (LCC), and the carbon-to-nitrogen ratio (LCC/LNC) to indirectly evaluate the initial quantum efficiency (alpha) (from the light response curve of the ETR). Leaf nitrogen and carbon contents of wheat and maize at different N application rates were collected from different phenological periods, and the corresponding leaf reflectance and ETR light response curve information were collected. We developed two types of partial least squares (PLS) models to evaluate LNC, LCC and LCC/LNC. One was based on raw reflectance at 400-1000 nm and full-band reflectance (350-2500 nm) and the other on newly calculated ratio vegetation indices (RVIs) (TOP10) and selected published vegetation indices (VIs). The optimal PLS model was used to evaluate alpha based on the quantified physiological associations of LNC, LCC and LCC/LNC with alpha. The results showed that samples from wheat, maize or a mixture of both had a nonnegligible contribution to the model performance. The PLS model (based on 400-1000 nm) had the optimal accuracy for mixed samples and was indirectly evaluated with respect to alpha (up to R2 = 0.73, LCC-PLS model). We also compared the direct evaluation of a PLS model constructed using the biochemical components of the radiative transfer model (RTM) inversion, the results of which suggested superior model robustness but reduced accuracy. Our study encourages the employment of spectral models of leaf carbon and nitrogen status to evaluate active fluorescence for extended scales.