Abstract:

Chlorophyll and nitrogen contents are important parameters as the indicators of crop photosynthesis productivity, state of growing and nutrition, and optimal diagnosis for crop nitrogenous fertilizer demands. In practical application process, testing nitrogen is more complicated than testing chlorophyll, and using chemicals are liable to pollute environment. Many researches show that chlorophyll content has a positive correlation with nitrogen content, so the status of nitrogen can be indicated by chlorophyll content or its elements. In the meantime, between chlorophyll content and hyperspectral characteristics, a positive correlation still exists, the status of nitrogen can be monitored by chlorophyll remote sensing. Traditionally, the status of chlorophyll density (CH.D) and leaf nitrogen accumulation (LNA) are studied by utilizing hyperspectral data, mainly focused on crops of wheat, rice and corn, mostly establishing a correlation model between hyperspectral data and one variable among CH.D and LNA. But it is scarce for combining CH.D and LNA together in the research, based on hyperspectral data monitoring state of crop canopy nutrition. This paper by utilizing non-imaging hyperspectral spectrometer, 8 cotton cultivars and two of them with 4 level densities planting in north XinJiang, and multivariate regression analysis method recorded multi-temporal hyperspectral data of canopy at cotton key growing stages and analyzed the correlation between reflectance and cotton canopy CH.D, LNA. The result showed that the maximum correlation coefficients between hyperspectral data and CH.D, LNA occurred at 762 and 763 nm (R_CH.D = 0.8845**, R_LNA = 0.787**, n = 47) respectively; the highest correlation coefficients between the first derivative spectral data and CH.D, LNA both occurred at 750 nm (R_CH.D = 0.9098**, R_LNA = 0.9164**, n = 47). Based on the first derivative data at 750 nm of modeling samples, we established the CH.D linear regression equation and estimated the CH.D of proving samples, then according to the model function between CH.D and LNA, estimated LNA of proving samples, correlation between tested LNA and estimated LNA was significant (R = 0.8982**, α = 1%, n = 94). The regression function accuracy was 86.2%, the RMSE was 1.0155, RE was 0.1380. The study shows that the status of cotton canopy leaf nitrogen accumulation can be monitored indirectly based on cotton chlorophyll density remote sensing.

Key words: Cotton, Hyperspectral, Chlorophyll density, Leaf nitrogen accumulation, Monitoring