Passive reflectance sensing using optimized two- and three-band spectral indices for quantifying the total nitrogen yield of maize

Precision nitrogen (N) management requires accurate and effective quantification of the total nitrogen yield (TNY) crops. Thus, this study aimed at establishing the robust prediction model to quantify the TNY of maize plants across growth stages, cultivars and years through optimizing two-band spectral indices, i.e. the normalized difference spectral index (NDSI) and the ratio spectral index (RSI), and three-band spectral indices, i.e. the canopy chlorophyll content index (CCCI) by re-determining the central wavelength and bandwidth. Field experiments with three maize cultivars and five N treatments were carried out in the North China Plain during 2011, 2012, and 2013. The results showed that the optimized band ranges of NDSI and RSI were mainly located within 720-760 nm (the red-edge domain) and 750-900 nm (the near-infrared domain). The central wavelengths of the NDSI were 768 and 740 nm, whereas those of the RSI were 756 and 744 nm. The most suitable band domains of the CCCI were 720-760, 500-600 and 740-800 nm, and their central wave-lengths were 766, 738 and 548 nm. Our study found that the optimized spectral indices could predict the TNY of maize accurately and robustly compared with existing spectral indices. The relationships between the optimized NDSI and RSI and maize TNY reached a high coefficient of determination (R-2 = 0.83). However, the prediction accuracy of TNY using the NDSI and RSI was gradually decreased with an increase in bandwidth, i.e., the bandwidths with central wavelengths of 740 and 768 nm of the NDSI were 13 and 21 nm, respectively. For RSI, the bandwidths with central wavelengths of 744 and 756 nm were 29 and 17 nm, respectively. The results also demonstrated that an optimized narrow and broadband CCCI was significantly linear in relation to the TNY of maize (R-2 = 0.85), suggesting an optimized CCCI may be used to quantify TNY of maize plants with higher accuracy by avoiding the saturation effect and improving sensitivity.