Hyperspectral Lidar-Based Estimation Of Plant 3-D Biochemical And Structural Parameters

Plant structural and biochemical properties generally present non-uniform distribution in three-dimensional（3-D）space,and monitoring the heterogeneity of vegetation attributes is conducive to managing vegetation ecosystems and evaluating ecosystem evolution.Insufficient extraction of vegetation vertical information leads to the fact that 3-D vegetation can only be expressed in the two-dimensional plane,thus losing the 3-D information used for judging the growth status of vegetation.The accuracy of the reconstructed radiative transfer model may also reduce due to the lacking of vertical information.Therefore,it is necessary to explore new remote sensing technologies to accurately extract the 3-D distribution information of vegetation structural and biochemical parameters.Passive remote sensing is widely used in plant properties by constructing models between spectral information and target traits.However,it is easily confounded by other complicated factors（e.g,illumination,canopy structure,and backgrounds）,and its obtained 2-D dimensionless images have limited ability in characterizing plant structural traits in 3-D space.Light detection and ranging（lidar）,as an active remote sensing technique,can provide the 3-D point cloud with high accuracy,and has been widely utilized for deriving plant structural traits to serve precision agriculture.Traditional lidar systems only employ a single wavelength,thus having limited ability in detecting plant spectral information.In contrast,the novel hyperspectral lidar（HSL）system designed based on the supercontinuum laser source combines the advantages of both passive remote sensing and traditional lidar.The hyperspectral point cloud derived from hyperspectral lidar full-waveform data can be used for the simultaneous extraction of plant structural and biochemical properties.Therefore,it has great potential in monitoring the 3-D characteristics of plants.This study aims to explore the ability of hyperspectral lidar in estimating plant biochemical and structural parameters at three levels,including 2-D leaf-level,plant-level,and organ-level.The main contents and findings are as follows:（1）This study assessed the capability of HSL to monitor the chlorophyll and nitrogen concentrations at the 2-D leaf level.Based on three datasets,namely the PROSPECT-5 synthetic dataset,the ANGERS public dataset,and HSL measurements,a sensitivity analysis was conducted to analyze the reflectance property of each hyperspectral lidar wavelength.Additionally,the best spectral index was selected from the 20 selected ratio and normalized spectral indices.The results indicated that the CI_{red edge}index was most compatible with hyperspectral lidar wavelengths and had the best performance in chlorophyll estimation.Compared with a single spectral index,the PLSR model had a stronger capability in extracting spectral information of hyperspectral lidar returned signals,thus improving the retrieval of biochemical parameters.（2）This study explored the capability of HSL in monitoring plant photosynthetic traits(V_cmax and J).Twenty wavelengths（ranging from 523 to 833 nm）with a higher signal-to-noise ratio were utilized to retrieve V_cmax and J by using two estimating methods（the reflectance-based method and the trait-based method）.The results showed that both the reflectance-based method and the trait-based method are highly applicable to hyperspectral lidar wavelengths,with the reflectance-based method having a better performance.The novel HSL system has great potential in the high-throughput measurement of plant photosynthesis.（3）This study explored the ability of HSL in estimating the properties of individual plants in 3-D space,achieving the simultaneous extraction of plant structural and biochemical traits.Ratio spectral indices and the Lambertian-Beckmann model were applied for eliminating the incidence angle effects of the hyperspectral point cloud,and the 3-D distribution of biochemical parameters was subsequently characterized for maize and Kniphofia uvaria plants based on the relationship between reflectance and biochemical properties.Thus,the variation of biochemical traits under different fertilizer conditions and growth stages can be further analyzed in 3-D space.The results showed that HSL has the ability in detecting the reflectance variation at any 3-D position,and in analyzing the heterogeneity characteristics of plant properties（e.g,chlorophyll,nitrogen,and photosynthetic traits）.The system can be further utilized for understanding the distribution mechanism of plant properties and optimizing radiative transfer models.（4）This study tested the feasibility of HSL in extracting plant phenotypes at the organ level.Based on the segmented maize leaf and stem point clouds,spectral indices and structural traits were utilized for retrieving nitrogen concentration and biomass,respectively.The W portioning,dynamics of the N concentration,and the variation in N with W accumulation under differed N conditions and growth stages were further analyzed.The results indicated that HSL has the ability in extracting nitrogen and biomass with high accuracy at the organ level,thus it can be applied to explore the N distribution mechanism of plants.