Optimized Spectral Indices Based Estimation Of Canopy Nitrogen Content In Wheat And Maize

Timely and accurate quantification of canopy nitrogen content at the critical growth stage of crops is important for the rationally applying nitrogen fertilizer and reducing environmental risk. The current research aimed to identify optimized spectral indices for remotely estimate canopy nitrogen content in winter wheat and summer maize. Experiments with different N rates for different cultivars of wheat and maize were conducted in different area from 2008 to 2011. The spectral bands of different models were optimized in this paper. First, based on the concept model, band combinations were optimized to derived canopy N content in wheat and maize. Second, using an optimum algorithm based on empirical model, three-band spectral indices were developed. The optimized three-band spectral indices significantly increased predictive power in estimating canopy N content of wheat and maize compared to traditional spectral indices. Finally, the optimized spectral indices developed by using concept model and empirical model were further validated with the theory of PRO SAIL model. The results showed that, compared with traditional red light based spectral indices, optimized spectral indices significantly improved the prediction power, overcoming saturation problem in deriving canopy nitrogen content of wheat and maize. The best performing spectral index of SI was R766/R738-1 for maize, R796/R760-1 for wheat and Rg76/R730-1 for wheat and maize combination. The best performing spectral empirical NPDI index was (R768/R740-1)/((R-768-R548)/(R768+R548)) for maize, (R876/R74o-1)/((R876-R55o)/(R876+R550)) for wheat and (R848/R732-1)/((R848-R536)/(R848+R536)) for wheat and maize combination. Band combination of optimized indices varied with the difference in species and canopy structure of crops. The optimal wavelength range mainly concentrated in the green side (530-550 nm), red (730-760 nm) and red edge (760-880 nm) transitional region to near infrared. Optimization results show that empirical model had higher predictive power compared to conceptual mode. The two bands optimum algorithms were validated by PROSAIL theory model. The validation results further confirmed that the optimized spectral indices can estimate the nitrogen content of different crops, had the lowest predictive error, indicating that optimized spectral indices could estimate canopy nitrogen content of crops. In conclusion, the band optimum of spectral indices is a promising approach to derive canopy N content in wheat and maize. The findings from this study may be useful for designing improved nitrogen diagnosis systems and for enhancing the applications of satellite-based sensors.