Quantification and dynamic monitoring of nitrogen utilization efficiency in summer maize with hyperspectral technique considering a non-uniform vertical distribution at whole growth stage

As one of the most mobile indicators, nitrogen utilization (assimilation) efficiency (NUtE) exhibits a pronounced heterogeneity in its vertical distribution in summer maize canopies. Although the vertical heterogeneity of summer maize NUtE has been recognized, the vertical NUtE gradient in canopies has not been considered in canopy hyperspectral remote sensing (CHRS) so far. The main goal of this study was to quantitatively define the effects of the interactive changes in N nutrition and growth stages on the vertical distribution of NUtE, identify the sensitive leaf layers and effective wavelengths, and develop a monitoring model considering the vertical NUtE distribution using canopy hyperspectral data. Four field experiments were conducted for three consecutive years (2018–2020) to demonstrate how the maize canopies influences the CHRS estimation of NUtE distribution in the vertical canopy at different growth stages. Canopies of each treatment were divided into three layers of equal vertical at elongation stage (V6) and flare opening stage (V12) (i.e., 1st layer, 2nd layer and 3rd layer), and four layers at silking stage (R1), filling stage (R2) and milk stage (R3) (i.e., 1st layer, 2nd layer, 3rd layer and 4th layer). Continuous wavelet transform (CWT) was ufsed to process the collected spectral reflectance; partial least square (PLS) and lambda-lambda r2 (LL r2) models were applied to analyze the relationships between NUtE in different layer and the spectral reflectance. Results showed that a vertical distribution pattern of NUtE existed, presenting an evident increase characteristics from the upper to lower layer. CWT technique can significantly improve the monitoring accuracy of summer maize NUtE in different layers among various growth stages, and the best decomposition scales are CWT-4, CWT-5, and CWT-6. The PLS model for NUtE prediction in the three decomposition scales yielded a relatively higher accuracy compared to the canopy raw spectra based on the full range hyperspectra, however, the prediction accuracy varied greatly in different layers and growth stages, the effect of the 2nd layer and 3rd layer and earlier stages were the best. The effective wavelengths for NUtE estimation were exhibited significant differences among the different layers and growing stages. Compared with the upper layer and V6-V12 stage, a strong “red shift” phenomenon towards the longer wavelengths was observed at lower layer and R1-R3 stage. Additionally, the newly developed PLS model using the effective wavelengths at different growth stages for NUtE prediction in various leaf layers also performed well (RPD>1.40). These findings will provide theoretical basis and effective methods for applying CHRS technology to real-time detection of NUtE and high yield and high efficiency of crop precision management for summer maize.