Extracting Red Edge Position Parameters from Ground- and Space-Based Hyperspectral Data for Estimation of Canopy Leaf Nitrogen Concentration in Rice

To realize non-destructive and real-time monitoring of crop nitrogen status for precision management in rice fertilization, we characterized the reflectance spectra in the red edge area from ground- and space-based data, and quantified the relationships between the red edge position (REP) derived from different algorithms and canopy leaf nitrogen concentrations (LNC) at various nitrogen rates in four seasons in various cultivars of field-grown rice (Oryza sativa L.). The results showed that the spectrum in the red edge area was significantly affected by the nitrogen level and cultivar type, and "three-peak" feature could be observed in the first derivative spectrum in this area. Traditional REP (maximum value of the first derivative spectra in red-edge region) was not sensitive to canopy LNC because of the three-peak property, but the REPs based on inverted Gaussian fitting, linear four-point interpolation, linear extrapolation and adjusted linear extrapolation generated continuous REP values, and could be used for estimating canopy LNC. REP from a three-point Lagrangian interpolation with three first-derivative bands (690 nm, 700 nm and 705 nm) had a good relationship with canopy LNC. Among the six REP approaches, REP based on adjusted linear extrapolation algorithm was found to have the best relations with canopy LNCs in Hyperion image data. Comparison of the different REPs with both ground-based and space-borne hyperspectral data indicated that the adjusted linear extrapolation method (755FD(730) + 675FD(700))/(FD730 + FD730) proposed here gave the best prediction of canopy LNC. This simple and reliable REP approach to monitoring canopy LNC in rice requires further verification with other hyperspectral sensors and crop types.