Simulation Models Of Cotton Canopy Characteristic Parameters Based On Hyperspectral Remote Sensing Data

Xinjiang as the largest base of cotton production, it is also suitable for application of hyperspectral remote sensing. For hyperspectral remote sensing, it can provide a method of quick obtaining cotton canopy information and data process. Establishing quantitative models of cotton canopy characteristic information, to enhance the management of cotton cultivation, regulation of cotton growing status and it is significant to carry out the precision agriculture in Xinjiang in future.Multi-temporal canopy hyperspectral data were recorded at key growing stages of cotton in a field experiment including seven cultivars, two cultivars with four level densities, and two cultivars planting four level patterns in north XinJiang. Utilizing Spectral Derivative, Spectral Position Analysis and Multivariate Regression Modelling, established the simulation models of cotton canopy absorbed photosynthetically active radiation(APAR) ,the fraction of photosynthetically active radiation(FAPAR), cotton aboveground net primary production (ANPP), leaf area index(LAI), mean foliage inclination angle(MFIA), extinction coefficient(K) and ground cover(GC) by using hyperspectral data respectively, and simulation models were tested, the results as follows:Analyzing the correlation between 8 vegetation indices ,3 edge area variables and APAR, FAPAR respectively, all correlation coefficients are highly positive (α=1%,n=130, except yellow edge area is negative). and the result shows that the NDVI (Normalized Difference Vegetation Index) against APAR, FAPAR, linear function has a higher correlation coefficient between estimating data and testing data. Based on NDVI retrieval of APAR, FAPAR, the correlation coefficients are 0.8179 and 0.8270, the regression function accuracy were 93.2%, 97.4% respectively.Highly correlated fit in with power function among 5 vegetation indices, red edge variable SDr against aboveground net primary production (ANPP), aboveground fresh biomass(AFBM) respectively, MSAVI2 power function fitting have a comparatively higher accuracy for estimating cotton ANPP, AFBM.Stepwise regression method is applied to analyze the correlation between reflectanceρ, reflectance reciprocal 1/ρ, logarithm log(ρ),the first derivative spectralρ',the second derivative spectralρ"and cotton Leaf area index(LAI), respectively, the highest correlation coefficient between the first derivative spectral data and LAI occurred at 734 nm band(R=0.6209^**). Reflectance reciprocal 1/ρagainst LAI the maximum correlations coefficient occurred at near infrared band 759nm, and red band 670nm respectively, combining (1/ρ)₇₅₉ and (1/ρ)₆₇₀ into vegetation index NDVI can also be used to estimate cotton canopy LAI(R=0.7315, RMSE=0.7075).Analyzing the correlation between 5 vegetation indices, red edge area variable and Mean foliage inclination angle (MFIA), extinction coefficient (K), respectively, all correlation coefficients reached 1% significant level. Cotton MFIA and extinction coefficient K were estimated by NDVI and MSAVI2 respectively, the results showed that there is 1% significant for testing measured value and estimated value.Established the models between 8 vegetation indices, red edge area, red edge slope variables and cotton ground cover (GC), all correlation coefficients are 1% significant level. stepwise regression method is applied to analyze the correlation between reflectanceρand the first derivative spectrumρ'and cotton ground cover (GC), respectively, the maximum correlation occurred at near infrared band 759nm and near infrared band 721 nm .Utilizing the reflectance at near infrared band 759 and first derivative spectral at near infrared band 721 to estimate cotton ground cover (GC) , respectively, their correlation coefficients are 1% significant linear positive between their estimating data and testing data.In conclusion, established simulation models of cotton canopy characteristic information, it provides an approach to analyze, simulate, evaluate, and predict the dimension of cotton canopy, and finally it can offer the theoretical foundation for monitoring the cotton canopy status by using hyperspectral remote sensing data in Xinjiang.