Quantitive Remote Sensing Of Vegetation Biochemical Parameters By Hyperspectral LiDAR

Vegetation accounts for about 70% of the earth's land surface.As the hub of material circulation and energy flow on the earth,it is one of the most important components of the ecosystem.Vegetation biochemical parameters,including chlorophyll,water,nitrogen and other constituents in plants,are closely related to many important ecological processes and are powerful indicators for characterizing vegetation stress and ecosystem function.Therefore,it is particularly important to estimate the vegetation biochemical parameters quickly and accurately.Remote sensing technology has the advantages of conducting rapid,large-scale and high-precision measurements,and has an irreplaceable role in obtaining parameters of ecological and environmental factors.Hyperspectral Li DAR(HSL)can acquire accurate three-dimensional spation information and rich spectral information simultaneously.Since its birth,it has attracted wide interest from researchers in many countries.As a new type of sensor,HSL provides new possibilities for vegetation growth status and biochemical parameter monitoring.With great application potential,HSL is becoming an important monitoring means of vegetation growth status and biochemical parameters.However,the empirical model for detecting biochemical parameters of vegetation by HSL is still blank.In terms of physical model inversion,it's difficult to detect vegetation transmittance spectra through HSL,while standard PROSPECT model inversion requires both reflectance and transmittance.In addition,the physical model is built based on direction-hemispherical passive spectrum in the 400-2500 nm range,while HSL detection can only detect direction-directional light at a dozen channels.Therefore,the possibility and method of biochemical parameter inversion based on PROSPECT model for HSL need to be studied.It is also necessary to determine the most efficient wavelengths in the numerical inversion of PROSPECT model to guide the design and application of HSL.In this thesis,the remote sensing inversion of vegetation biochemical parameters by HSL is studied.Empirical models and physical models based on HSL system are used to quantitatively monitor various vegetation biochemical parameters.The main research contents and innovation work are as follows:1)The characteristics,development and current situation of HSL system are sysmtematically summarized.The characteristics of vegetation leaf reflectance spectrum are analyzed based on the structure and biochemical characteristics of vegetation.The research status and basic theoretical methods of the two fundamental methods for quantitative detection of vegetation biochemical parameters are introduced in detail: empirical model inversion and physical model inversion.2)Taking the inversion of leaf nitrogen content as an example,the empirical models of HSL detection of vegetation biochemical parameters are studied.Based on the observation experiment of nitrogen content in rice leaves for two years,regression models of nitrogen content are established by using a variety of linear and nonlinear machine learning algorithms.In addition to the HSL system,comparative observation experiment on rice leaves was conducted simultaneously using ASD passive spectrometer and four-wavelength multispectral Li DAR.Research is carried out in two dimensions:(1)comparison of the performance of empirical models using reflectance spectra from different active and passive remote sensing detectors,and(2)comparison of the performance of a variety of linear and nonlinear regression empirical models on the estimation of leaf nitrogen content.An empirical model estimating nitrogen content in rice leaves with good performance both for active and passive reflectance is proposed.3)The PROSPECT physical model inversion method based solely on reflectance spectrum is proposed to solve the problem that standard PROSPECT model inversion requires both reflectance and transmittance spectra,while it remains difficult for most remote sensing methods including HSL to detect the vegetation transmittance spectrum.By constructing different cost functions,the performance of three inversion strategies were studied:(1)based solely on reflectance spectrum,(2)based solely on transmittance spectrum,and(3)based on both reflectance and transmittance spectra.It is proved that leaf chlorophyll content and water content can be retrieved with high precision based solely on reflectance spectrum,with accuracy similar to using both reflectance and transmittance spectra,which lays a foundation for the physical model inversion of the HSL data.4)The chlorophyll content inversion verification of HSL based on PROSPECT model is carried out.The number of detection channels for HSL is limited and cannot meet the wide spectrum requirements(400-2500 nm)of standard PROSPECT model inversion.A specific cost function is constructed based on the sensitivity analysis of PROSPECT model and the detection channels of the HSL system to retrieve leaf chlorophyll content.HSL measurements are used to prove the system's ability to retrieve leaf chlorophyll content based on PROSPECT model.The retrieval performance is better than the general empirical model.This provides experimental support for the physical model inversion of HSL data.5)Based on the PROSPECT physical model,a method for selecting the optimal wavelengths towards estimating the vegetation biochemical parameters by HSL is proposed.Based on the sensitivity analysis of different versions of PROSPECT model and the band spatial-autocorrelation analysis of vegetation hyperspectral data,an automatic wavelength selection algorithm for HSL with a physical basis is proposed.Ultimately,the optimal wavelength combination is determined: 680,716,1104,1882 and 1920 nm,in order to realize a high-precision retrieval of chlorophyll content and water content through the PROSPECT model inversion.This provides a theoretical guide for the design and manufacturing of new low-cost,highly-efficient HSL system.