Research In Estimation Of Wheat Resource Use Effciency Based On Hyperspectral Remote Sensing

On the basis of stabilizing crop yields,improving the utilization efficiency of farmland water,fertilizer and light energy resources has always been a focus of attention in the development of modern agriculture.Precision agriculture is an important direction for thedevelopment of modern agriculture.The use of hyperspectral remote sensing technology to monitor and evaluate the resource utilization efficiency of crop production is an important research direction for precision agriculture.Based on field experiments of different years,locations,varieties,irrigation frequency,and nitrogen fertilizer treatment,this study comprehensively uses hyperspectral remote sensing technology,crop physiology and resource efficiency determination,and modern data analysis methods to analyze wheat under different experimental conditions.The quantitative relationship between the canopy spectrum and ihe growth efficiency index,through the design and calculation of spectral parameters,constructs the sensitive parameters and estimation models of the wheat nitrogen use cfilciency(NUE),light use efficiency(RUE) and water use efficiency(WUE);at the same time,adopts Partial Least squares method(PLSR),neural network(BPNN),support vector machine(SVM)and other methods are used to estimate and evaluate NUH.RUE and WUE respectively,and establish the best estimation method.The research results provide ihcorclical guidance and technical support for the precise cultivation and regulation of agriculmral production and the cultivation and selection of high-yield and efneient varieties.1.Remotely Assessing Photosynthetic Nitrogen Use Efficiency with Hyperspectral Remote Sensing in Winter WheatIn order to obtain wheat NUE quickly and accurately,the quantitative relationship between photosynthctic nitrogen use efficiency(PNUE)and NUE of wheat at different growth periods was first analyzed.The R were higher than 0.676 in a single growth period.Further examining the relationship between common vegetation indices and PNUE.(he data sets of vegetative growth period and reproductive growth period cannot be consistent with each other.mSR(760.850,680)is obtained by increasing the indicator band R₆₈₀ of net photosynthetic rate(P_N)on the basis of SR(760,850),and then introducing a floating coefficient(1,8+R₆₈₀/R₈₅₀)on the mSR(760,850,680)frame,thereby reducing the impact of soil background exposure and leaf area during reproductive growth period.This parameter was named as the Nitrogen Efficiency Index(NEI).The data set of integrated synergistic vegetation stage and reproduction growth stage and the PNUE fitting equation have a good effect,the coefficient of determination R²=0.765,and the root mean square error RMSE=0.636.The independent year test was used to test the model,and the applicability of the model was good.This shows that the new parameter NEI can explain the changes of PNUE well and can quickly indicate wheat NUE.2.Remotely Assessing Light Use Efficiency with Hyperspectral Remote Sensing InWinter Wheat.The efficiency of light use effciency is an important determinant of carbon fixation in terrestrial ecosystems.The Photochemical Reflectance Vegetation Index(PRI)was first proposed based on the lutein cycle to estimate RUE,but the relationship between PRI and RUE is different under environmental conditions.In order to improve the accuracy of using PRI to estimate RUE,systematically analyze the relationship between field test spectral data and physiological indicators,and found that the carotenoid(Car)to chlorophyll(Chl)ratio is more consistent with the change in RUE during the whole growth period.Car/Chl was converted into PRI by screening the spectral parameters sensitive to the Car/Chl reaction,thereby obtaining the modified PRI(mPRI).At the same time,using multi-angle spectral data,the relationship between the common vegetation index and mPRI was tested and analyzed under 13 observation angles and its angle range.Regardless of the forward or backward observation angle,mPRI is superior to other common spectral parameters.The best observation angle is -10°,and the most suitable angle range is -20-10°.In addition,the estimation model can also be applied to MODIS data,which shows that by optimizing the PRI,it can well indicate the change of RUE under non-uniform conditions.3.Remotely Assessing Water Use Efficiency With Hyperspectral Remote Sensing in Winter WheatReal-time,fast and non-destructive acquisition of water use efficiency is an important measure for reasonable and optimal allocation of resources in wheat production practice.Comprehensive use of the three wheat varieties under different water and nitrogen treatments in field test data for many years,analyze and calculate the quantitative relationship between the canopy spectral reflectance and related physiological indicators during the key growth period of wheat,and establish a quantitative estimation model of wheat WUE.The results show that the backward observation angle of all candidate spectral parameters is better than the forward observation angle.The relationship between Lo and NDDA and WUE in the common vegetation index is better,but saturation occurs when the WUE is large.In order to alleviate the saturation phenomenon estimated by the model,a water efficiency index(WEI)was proposed based on the study of agronomic mechanism,which can establish a unified model within the range of -20-10°,and the model R² and RMSE are 0.623 and 0.406,respectively.Using open independent test data for inspection,R²,RMSE and relative error(RE)were 0.685,0.473 and 11.847%,respectively.The aboveresults show that the constructed water efficiency index is more sensitive to WUE changes than the common vegetation indices,and the monitoring model has a wider angle of application,which can provide a reference for the parameter design of professional sensors.4.Comparison of Modeling Methods Based on Remote Sensing Estimation Efficiency IndexIn order to make full use of more comprehensive monitoring information,it is necessary to analyze the full spectrum.In addition to the common vegetation indices analysis,three methods,PLSR,BPNN and SVM,are used for full spectra analysis of resource utilization efficiency.The SVM method uses three kernel functions: linear,polynomial and Gaussian functions for calculation.The prediction results of the three ctficiency indicators are consistent,and the polynomial function results are better than theGaussian function than the linear function.Comparing a variety of mathematical statistics methods,the model prediction accuracy is shown as SVM(polynomial kernel function)>PLSR>BPNN>VIs.PLSR is the most common method in multivariate analysis.It can filter out sensitive bands by reducing uncorrelated latent variables and maximizing(he covariance of related variables.However,each coefficient in the BPNN model has a root mean square error,which is susceptible to deviation and it aficcted by outliers.Based on this,this study inputs the sensitive band selected in the PLSR method as a variable into the BPNN model.Combining the advantages of the two,the established PLSR-BPNN joint model has a prediction accuracy of 0.806,0.753 and 0.810 for NUE,RUE and WUE,respectively.Crop resource utilization monitoring and evaluation provide method reference.