Sensitivity Analysis And Optimum Bandwidth Selection Analysis Of LAI Inversion Using Vegetation Index

Leaf area index(LAI)is an important parameter to characterizing physiological and biochemical properties.It determines the size of plant atmosphere interface.It is the main variable used to simulate the process of canopy photosynthesis and evapotranspiration,which can effectively reflect the growth status of crops.The application of satellite remote sensing provides a feasible and economic method for large-scale LAI research.How to replace LAI by expressing data more accurately has always been a hot topic for new understandingAt present,in order to improve the accuracy and universality of the inversion model of LAI,on the one hand,researchers constantly optimize and improve strategies and algorithms to reduce model errors;on the other hand,they are committed to analyzing the role of soil background,soil type,observation geometry,hot spot effect and other factors in the inversion process of LAI to find out the corresponding methods to reduce or eliminate the influence of interference factors.However,There are few studies about the influence of bandwidth on LAI inversion.The PROSAIL model was used to simulate 20000 canopy spectral reflectance(400 nm?2500 nm)to calculate the vegetation index under different bandwidths(5 nm?80 nm),and establishes the LAI empirical inversion model.At the same time,in order to select the vegetation index which is the most sensitive to LAI and the least sensitive to interference parameter,the vegetation index screening method based on sensitivity analysis(weight method)were constructed.Firstly,the global sensitivity of the input parameters of PROSAIL model,such as leaf area index,leaf structure parameter,leaf chlorophyll,equivalent water thickness,dry matter content,hot spot parameter and soil brightness parameter,is analyzed quantitatively.Then,the new input parameter data combination is generated by resampling the sensitive parameters while the insensitive parameters are fixed.Based on the combination of these parameters,35 vegetation indexes,such as SR[800680]and OS AVI,were calculated.In order to select the vegetation index suitable for LAI inversion,the global sensitivity coefficient output by PROSAIL model is taken as its mean value and homogenized,which is used as the weight of sensitivity score calculation to participate in the selection of vegetation index,and verified based on the measured wheat spectral data.The main conclusions of this paper are as follows:(1)Bandwidth is one of the important factors that affect the estimation of LAI,and different vegetation indices have the best bandwidth for inversion.According to the change trend of coefficient of determination(R2),the indices can be divided into three categories:1)narrow-band vegetation index(MTVI2,OSAVI,RDVI,TVI,Dattl,NDCI,NVI,SPVI,SR[752,690],SR[800,680],TCARI and TCARI2):R2 of inversion model decreases with the increase of bandwidth;2)middle-band vegetation index(SR[700,670]?Carte5?SR[675,700]):R2 first rises and then falls with the increase of bandwidth,and the change curve has obvious peak value;3)broad-band vegetation index(Carte3,Carte4,Datt2,Datt3,MCAR],mSR705,MTVI1,NDVI705,RI1dB,SR[750,700],SR[750,550],SR[750,710],YOG1,VOG3,VOG2,OSAVI2,GNDVI,mNDVI705 and SPVI2):R2 increases with the increase of bandwidth.Therefore,the purpose of this paper is based on the PROSAIL model simulating canopy spectral data set.In order to provide some reference for different remote sensing data to select the appropriate vegetation index,and improve the accuracy of empirical inversion LAIThe study of the influence of bandwidth on the inversion accuracy of LAI to determine the most appropriate bandwidth of vegetation index.(2)The global sensitivity analysis of PROSAIL model shows that hspot,skyl,tts,tto,psi,N and Car are insensitive parameters in the simulation of canopy spectral reflectance by PROSAIL model.Therefore,when using PROSAIL model to simulate spectral reflectance,these parameters can be fixed to improve the efficiency of model.LAI,Cw,ALA,Cm,psoil and Cab are sensitive parameters.Therefore,when using PROSAIL model to simulate canopy spectrum,the input range of sensitive parameters should be determined with prior knowledge.(3)Through the comprehensive analysis of the optimum bandwidth and global sensitivity of vegetation index,this paper attempts to compare the accuracy of LAI estimated by the empirical model method and neural network method under the optimal bandwidth,and use the measured data to verify.The results show that the weight method proposed in this paper has a high reliability in the selection of vegetation index,which provides a new way for the determination and selection of vegetation index in the inversion of LAI by empirical model method.