Estimation of Leaf Nitrogen Content of Winter Wheat Based on Akaike's Information Criterion

Nitrogen is one of the important indices for evaluation of crop growth and output quality. At present, there are a lot of researches on the estimation of crop nitrogen content, but most of the studies do not consider whether the model established by vegetation index and crop nitrogen content is the best. The purpose of this study was to estimate the nitrogen content of wheat leaves and to establish a method and the optimal model for monitoring nitrogen content in wheat leaves. Spectral reflectance of leaves and concurrent leaf nitrogen content parameters of samples were acquired in during 2013 and 2014 wheat growth season, in Beijing Academy of Agriculture and Forestry Sciences. 17 vegetation indices related to nitrogen content were chosen, and the relationship between related vegetation indices and leaf nitrogen content were built for screening vegetation indices with variable importance projection (VIP). Choosing first 10 different vegetation indices after ranking with VIP value as the independent variable for estimating nitrogen content of leaf in wheat. And the number of vegetation indices was gradually increased from top 4 to 10. The leaf nitrogen content estimation model with different vegetation indices can be built using the integrated model of variable importance projection (VIP) - partial least squares (PLS). At the same time, Akaike's Information Criterion (AIC) value was calculated in different estimation model, and the AIC value of 7 different estimation model was compared. Then the optimal model with 5 vegetation indices was selected, which AIC value is the lowest. The optimal model was validated by leave one out cross-validation method. The result, (1) the comprehensive interpretation ability of the first 10 spectral indices on the nitrogen content of winter wheat leaves was PSSRc, GMI-2, SR705, RI-half, ZM, GMI-1, PSSRb, RI-3 dB, VOGc and CIred edge. (2) The optimal model with 5 vegetation indices was selected from 7 models. (3) The decision coefficient (R-2) and root-mean-square error (RMSE) of the optimal model respectively were 0.73 and 0.33. The R-2 and RMSE of wheat by validating were 0.73 and 0.33, respectively. The study showed: (1) The VIP-PLS model has higher ability to estimate the nitrogen content of leaf in wheat, which laying an important foundation for improving the precision of forecasting winter wheat leaf nitrogen content with remote sensing. (2) The AIC method can be used to select the optimal model, and the selected model has the higher predictive ability. And the optimal estimation model of wheat LNC can be obtained based on AIC.