Total nitrogen content estimation models of summer maize leaves using hyperspectral remote sensing

Based on the treatments of five nitrogen fertilizer application amounts and two cultivars of summer maize , crop canopy spectral reflectance and total nitrogen content of maize leaves were measured at the jointing stage, huge bellbottom stage, tasseling stage, silking stage and milk stage. The canopy spectral reflectance in 470, 550, 620 and 720 nm wavelength of hyperspectral remote sensing were chosen to establish liner and nonlinear regression relationship between leaf total nitrogen content and canopy spectral parameters for each cultivar, which includes original spectral reflectance, first order differential transform, and part of hyperspectral characteristic parameters (i.e. spectrum position, area, characteristic parameters of vegetation index). Three models with high coefficients and F values of each cultivar at each growth stage were chosen to verify root mean square error and relative error with the second year data of spectral reflectance and total nitrogen content of two cultivars separately. The smallest root mean square error and relative error models were taken as the best models. The results show that: at the jointing stage, huge bellbottom stage, tasseling stage, silking stage and milk stage of maize，spectrum parameter for the best fitting regression relationship with leaf total nitrogen content was R720, DR720, SDb, DR550 and DR550, separately. And the best model to estimate total nitrogen content of maize leaf based on above best spectrum parameter of hyperspectral remote sensing in five growth stages was Y=5.129e-2.317x, Y=3.421－10.010x－477802.331x3, Y=4.070－2.304x－52.177x2, Y=－0.468－0.528lnx and Y=－2.390－0.793lnx, respectively.