Estimation Of Wheat Growth And Nitrogen Use Efficiency Based On Hyperspectral Remote Sensing

The use of hyperspectral remote sensing technology to monitor and estimate crop growth and predict crop yield and quality is an urgent need for modern precision agriculture,and is also an important research content of agricultural remote sensing.Based on wheat field trials with different varieties,different nitrogen rates and different locations,the wheat growth parameters(leaf area index,leaf dry weight and shoot biomass)and nitrogen nutrition parameters(leaf nitrogen content)were systematically analyzed.Leaf nitrogen accumulation,nitrogen content in shoots and nitrogen accumulation in shoots,grain yield,grain protein content and nitrogen use efficiency parameters(seed production efficiency,plant nitrogen use efficiency,nitrogen fertilizer partial productivity,and nitrogen fertilizer physiological utilization efficiency),using five modeling methods(spectral index method,partial least squares regression,lasso regression,support vector machine)Regression and Gradient Enhancement Algorithm),respectively,based on the original canopy spectrum(band range: 400nm-900nm)and the continuum removal of the converted red edge spectrum(band range: 550nm-750nm),wheat growth status and grain yield and quality The hyperspectral estimation model is designed to achieve the estimation accuracy of the original canopy spectral model using the red-edge band model with less input.The main results were summarized as follows:(1)The leaf area index,leaf dry weight,leaf nitrogen content and leaf nitrogen accumulation of wheat showed a trend of increasing first and then decreasing with the growth of the growth stage,which reached the maximum at the flowering stage and then decreased during the filling stage.The plant biomass,the nitrogen content of the aboveground plants and the nitrogen accumulation of the aboveground plants showed a tendency to increase with the advancement of the growth period.It indicated that during the vegetative growth period,the nutrients absorbed by wheat were mainly stored in vegetative organs,namely roots,stems and leaves.Therefore,in the vegetative growth stage,leaf area index,leaf dry weight,leaf nitrogen content and leaf nitrogen accumulation of wheat were observed.The amount is increasing continuously until it reaches a peak at the flowering stage,and after the flowering stage,the wheat enters the reproductive growth stage,and the nutrients in the vegetative organs are gradually transferred to the reproductive organs,ie,the ears and the grains,and the leaves gradually turn yellow and senescence,and then fall off.Leaf area index,leaf dry weight,leaf nitrogen content and leaf nitrogen accumulation decreased to a low point,but at this time,the total amount of nutrients in wheat plants did not decrease,but the storage position shifted,so until the filling stage,wheat shoots Plant biomass,nitrogen content in shoots and nitrogen accumulation in shoots still increased.(2)Continuum removal treatment can enhance the correlation between canopy spectra and wheat growth parameters.The continuum removal of the spectral index can better estimate the wheat leaf area index and leaf dry weight.The estimation accuracy is similar to that of the original canopy spectral canopy index,but it is obviously inferior to the original canopy spectrum in estimating the aboveground biomass of wheat.Canopy index.The estimation model of wheat growth parameters constructed by partial least squares regression,lasso regression,support vector machine regression and gradient lifting algorithm is superior to spectral index accuracy.Among them,the estimation model constructed by gradient lifting algorithm performs best.The modeling decision coefficients are basically above 0.9.The wheat leaf area index and leaf dry weight estimation model based on continuum removal spectra were not significantly different from the original canopy spectra in the modeling data set,but were superior to the original canopy spectrum in the test data set.In the aboveground plant biomass of wheat,there was no significant difference between the continuum removal spectrum and the original canopy spectrum in the modeling data set and the test data set,which indicated that in the wheat leaf area index and leaf dry weight estimation,continuum removal The spectrum has a higher signal-to-noise ratio and better stability.In the estimation of wheat aboveground biomass,the continuum removal spectrum may only have some information missing due to only intercepting the red edge band,thus reducing the estimation.Measurement accuracy.(3)Continuum removal treatment can improve the correlation between canopy spectra and wheat nitrogen nutrition parameters.The continuum removal of the spectral index can better estimate the nitrogen nutrition parameters of wheat,and is superior to the original canopy spectral parameters in estimating the nitrogen accumulation of wheat aboveground plants.The wheat nitrogen nutrition parameter estimation model constructed by partial least squares regression,lasso regression,support vector machine regression and gradient lifting algorithm is superior to the spectral index accuracy,and the estimated model performance constructed by gradient lifting algorithm is used.Preferably,the modeling decision coefficients are all above 0.91.The estimation model of nitrogen content and nitrogen accumulation of wheat plants based on continuum removal spectra was not significantly different from the original canopy spectra in the modeling data set,but superior to the original canopy spectrum in the test data set.In the estimation of nitrogen accumulation in wheat leaves and nitrogen content in aboveground plants,there was no significant difference between the continuum removal spectrum and the original canopy spectrum in the modeling data set and the test data set.(4)Analysis of the correlation between wheat grain yield and protein content and growth period and nitrogen nutrition parameters.It was found that the nitrogen nutrition parameters in flowering stage had a good correlation with wheat grain yield and protein content,among which flowering leaves The correlation coefficient between nitrogen content and wheat grain yield and protein content was the highest;the multiple regression prediction of wheat grain yield and protein content based on the original canopy spectral index and continuum spectral index for estimating growth period and nitrogen nutrition parameters were constructed respectively.The model found that the gradient-elevation prediction model based on the spectral index of flowering wheat showed the best performance,far superior to other prediction models.The accuracy of the model for predicting wheat grain yield by wheat is better than that of wheat grain,and the accuracy of the spectral prediction model of wheat grain protein content needs to be further improved.(5)The correlation between wheat nitrogen use parameters and growth period and nitrogen nutrition parameters was analyzed.It was found that the correlation between nitrogen nutrition parameters with wheat grain production efficiency and plant nitrogen use efficiency were good.The spectral parameters of the nitrogen content in the flowering period and the estimated nitrogen content were used to predict the three nitrogen utilization parameters.The original canopy spectral index based on estimating the growth period and nitrogen nutrition parameters of different growth stages were constructed respectively.The multivariate regression prediction model of wheat nitrogen utilization parameters with continuum removal of spectral index showed that the gradient elevation prediction model based on the flowering index of wheat was the best,far superior to other prediction models.Because the nitrogen use efficiency of wheat and the growth factors of different growth stages and nitrogen nutrition parameters are relatively low,it is difficult to estimate the nitrogen use efficiency more accurately.Therefore,it is decided to classify different wheat varieties into high nitrogen use efficiency based on previous studies.Variety(abbreviation: nitrogen efficient variety)and low nitrogen use efficiency variety(nitrogen inefficient variety),and then based on the classification of physiological parameters and spectral parameters,found that the classification based on single physiological parameters and spectral parameters The accuracy of the model is above 0.7,and the performance is better.The classification accuracy of the classification model based on multiple physiological parameters and multi-spectral parameters can reach above 0.8,with good performance and excellent application potential.