Hyperspectral Estimation Of Physical And Chemical Parameters Of Rice In The Irrigation Area Of Ningxia

Rice is the major food crop in China,and its steady production plays a fundamental role in ensuring national food security.Accurate estimation of physiological and biochemical parameters of rice by hyperspectral remote sensing technology is an effective means to monitor the growth and development of rice in real time,and it is of great significance to realize the sustainable development of rice.Taking rice in yellow river irrigation area of Ningxia as the research object,through field experiments,the physiological and biochemical parameters and hyperspectral data of rice in heading stage,milk stage and ripening stage were observed,the change law of rice leaf and canopy spectrum at different growth stages and different nitrogen levels was explored,and the response characteristics and correlation of rice physiological and biochemical parameters and spectral data were analyzed.Using general regression analysis method and random forest algorithm,the rice physiological and biochemical parameters estimation models based on characteristic band,hyperspectral characteristic parameters and the best vegetation index were established,and SPAD value of leaf and canopy,LAI and LNC at different growth stages were estimated.The research results can provide theoretical basis and technical support for remote sensing monitoring of rice in yellow river irrigation area of Ningxia.The main conclusions are as follows:(1)The spectra of rice leave and canopy at different growth stages and different nitrogen levels showed the same change rule in visible and near infrared band.From heading stage to ripening stage,the spectral reflectance in visible light band gradually increased,while that in near infrared band gradually decreased,and the red edge position moved to shorter wavelength.With the increase of nitrogen application rate,the spectral reflectance in visible light band decreased gradually,and that in near infrared band increased,and the red edge position moved to longer wavelength.(2)The spectra of leave and canopy at different SPAD levels showed significant differences.The larger SPAD value is,the lower the spectral reflectance in the visible band is,and the higher the spectral reflectance in the near infrared band is.The canopy spectra of different LAI levels have no obvious difference in visible light band,and in near infrared band,the spectral reflectance increases significantly with the increase of LAI.The change law of canopy spectrum with LNC is as follows: the larger LNC,the lower spectral reflectivity in visible light band and the higher spectral reflectivity in near infrared band.(3)In different growth stages,the parameters with the strongest correlation with SPAD,LAI and LNC of rice are the best vegetation index,and optimal univariate estimation model is a nonlinear model constructed with the best vegetation index as an independent variable.Using random forest algorithm to establish multivariate estimation model,the prediction accuracy is obviously improved.(4)The first derivative spectrum has a better reflection on physiological and biochemical parameters than the original spectrum in some bands.Modeling in different growth stages can improve the accuracy of estimating physiological and biochemical parameters,and it is necessary to accurately estimate the SPAD value,LAI and LNC of rice canopy in different growth stages.