Diagnosis of Nitrogen Nutrition in Summer Maize based on Simulated Multispectral Data

Developing technology for monitoring nitrogen in the canopy of summer maize, particularly selecting sensitive monitoring wavelengths and establishing appropriate monitoring models, is beneficial for promoting the application of non-destructive monitoring of nitrogen concentration and accurate management of water and fertilizer based on hyperspectral technology in large-scale areas. The study was carried out in Daxing Experimental Research Base of Beijing Engineering and Technology Research Center of State Water-Saving Irrigation from June to September in 2017 and 2018. It focused on summer maize and involved small-scale lysimeter experiments and field plot trials. In view of the fact that the hand-held ground spectrometer can only observe the hyperspectral reflectance of crops within a small area over a certain period of time, this study used measured hyperspectral information of summer maize canopy to simulate the spectral reflectance of Landsat 8-OLI and GF-1 satellite sensors using sensor spectral response functions for large-scale observations. Based on sensitive wide-band and multi-band spectral indices, a model for estimating canopy nitrogen content was constructed, providing a theoretical basis for using satellite multispectral data for nitrogen diagnosis in summer maize. The results suggest that the prediction models constructed based on the reciprocal of simulated spectral reflectance in the red light band (1/R) and the reflectance at the center wavelength of 864 nm in the near-infrared band are suitable for estimating the canopy nitrogen content of summer maize throughout the growing season, and the determination coefficients of the models reach 0.621 and 0.613, respectively. A fitting model was constructed based on principal component analysis using 10 multi-band spectral indices calculated from simulated multispectral reflectance of summer maize throughout the growing season and the model achieved a determination coefficient of 0.692, indicating a high level of simulation accuracy.