Main Physiological Indicators Of Hyperspectral Inversion In Alfalfa Seed Production

Using a SVC HR-768portable spectrometer, the spectral reflectance of alfalfa canopy with differentIrrigation and nitrogen application rates from budding to mature stage were studied in the Grasslandresources and ecology experiment stations in Hutu bi County of Xinjiang Agricultural University. Westudied spectral reflectance in different transform formats, i.e.mathematical methods, features bandextraction, stepwise regression analysis method, etc. Alfalfa canopy spectral characteristics are analyzedand characteristic waveband are ascertain，established alfalfa water content, nitrogen, biomass, andother physiological parameters spectral inversion model. The results showed that,（1）with the increases of irrigation and nitrogen fertilizer amount, the canopy spectral reflectance inthe near-infrared bands rised gradually.（2）the spectral reflectance of alfalfa canopy had the same changing trend in different growth stages.Along with the growth of alfalfa, the canopy reflectance of alfalfa firstly increased and then decreased, inthe flowering reached the maximum.（3）first derivative spectra at1450nm wave band model is best spectral inversion model of alfalfaleaf water content in squaring period, which was Y=0.076+6.474x1450(adjusted R2was9.03, and theaverage relative error was9.6%).（4）significant correlation relationships existed between the first derivative spectra and theaboveground biomass in budding, and flowering. But the model in budding stage was better than that inflowering stage, the model was Y=20.447x978-1.538, and the adjusted R2of the model was0.923, theaverage relative error was7.1%. good correlation relationship existed between dry matter accumulation ofbud leaf and first derivative spectra in the squaring period, and the established model was Y=38.407x1741-2.648. among all the models established with the first derivative spectra and the dry matteraccumulation amount of ground organs, the spectral prediction model with dry matter accumulation ofleaves in budding stage was the best, which was Y=-105.487x825+2.701, with an adjusted R2of0.871andan average relative error of8.30%.（5） when using vegetation indices (RVI, DVI, NDVI and RDVI) in visible light (510nm,560nm, 610nm,680nm,710nm,760nm) and near-infrared (810nm,870nm,950nm) to predict nitrogen content ofaboveground organs, the predictive model of DVI (760,810) and the plant nitrogen content in the buddingstage was the best, with adjusted R2and the average relative error were0.921%and8.7%, respectively.