Hyperspectral Remote Sensing of Crop Canopy Chlorophyll and Nitrogen: The Relative Importance of Growth Stages

Remote sensing plays an important role in monitoring vegetation dynamics, and has been recognized as a reliable tool for monitoring biochemical and biophysical variations of agricultural crops, such as plant biomass, height, chlorophyll (Chl) and nitrogen (N). Nitrogen is one of the most essential elements in agro-ecosystems because of its direct role in determining crop yield and vegetation productivity, as well as its association with global N and carbon cycles. Canopy remote sensing of plant biochemical (e.g., N) and biophysical parameters (e.g., biomass) is often discussed separately. However, crop canopy structural characteristics and plant morphophysiological variations at different growth stages cause a confounding effect on the analysis and interpretation of the canopy spectral data. This study aimed to (1) understand the underlying mechanisms of canopy structural dynamics (mainly plant biomass and green leaf area) that impact the retrieval of canopy Chl and N at different growth stages, and (2) develop new algorithms and narrow band vegetation indices that may improve the estimation of Chl and N using hyperspectral data collected in the field and simulated by radiative transfer models (RTMs). To achieve the objectives, barley and rice experiments were conducted in Germany and China, respectively, from experimental plots to farmer fields; both empirical and physical models were employed but with an emphasis on the empirical methods. Results suggest that canopy hyperspectral data allow for the estimation of canopy Chl and N. However, with the advance of growth stages, plant growth rate is much faster than the rate at which N is accumulated in the plant mass until the stage of full heading (canopy closure), which results in a decrease of N concentration — the N dilution effect. Thus, growth stages have a significant effect on the correlation between the optical and biological traits of the crop canopy compared to the differences in crop cultivars and types. This effect is confirmed by five years of experimental data of barley and rice crops. Accordingly, empirical models based on different vegetation indices can be calibrated, before and after the canopy closure, which allows for the monitoring of canopy Chl and N status through the entire growing season. This study also suggests that multivariate models such as partial least squares (PLS) and support vector machines (SVM) are relatively resistant to the influence of growth stages and can be used to improve the estimation of canopy Chl and N compared to univariate models based on vegetation indices. To devise a simple approach for the estimation of canopy Chl and N status that is relatively insensitive to the confounding effect of canopy structural characteristics, new vegetation indices, the Ratio of Reflectance Difference Indices (RRDIs), were developed based on the multiple scatter correction (MSC) theory. This type of indices conceptually eliminates the linear influence caused by the confounding effect of multiple scattering and soil background as well as their interactions; therefore, RRDI weakens the effect of canopy structural variations on the analysis of canopy spectra when estimating biochemical variations. For example, the RRDI derived from the red edge (RRDIre) wavelengths proved to be a robust indicator of canopy Chl and N in both barley and rice crops with different cultivars and for the simulated data by RTMs. Therefore, the method is useful for improving the estimation of canopy biochemical parameters. This study improves the understanding of remote estimation of canopy Chl and N status by considering the dynamical co-variations between plant biomass and N across different growth stages and suggests the potential to improve the ability of canopy hyperspectral data to monitor the canopy biogeochemical cycles of agro-ecosystems using remote sensing. Additionally, this study indicates that hyperspectral vegetation indices based on water absorption bands are useful for the detection of crop diseases at the canopy level