Estimation Of Biochemical Parameters Of Plant Leaves Using Multi-Angular Reflectance Based On Spectral Angle Cosine Method

Leaf Chlorophyll Content(LCC)and leaf water content(expressed as Equivalent Water Thickness,EWT)are crucial biochemical parameters of plant leaves,which determine the physiological state and many biochemical processes of plants.Efficient and dynamic monitoring of these two biochemical parameters is of great significance.Remote sensing detection technology can obtain spectral information of plants and achieve rapid and non-destructive estimation of leaf biochemical parameters.However,the non-Lambertian property of the leaf determines that the leaf reflection information actually includes two parts:diffuse reflection from the inside of the leaf,and specular reflection that only occurs on the surface of the leaf.Due to the fact that specular reflection does not contain biochemical information and changes with incident-observation geometry and leaf surface properties,plant leaf spectra show additive spectral intensity differences at different observation angles.Multi-angular detection exists objectively in various remote sensing detection processes,so specular reflection is an important factor in reducing the estimation accuracy of leaf biochemical parameters.It is important to find an effective method to remove spectral intensity differences caused by specular reflection in multi-angular detection.Spectral Angle Cosine(SAC)is originally a method for classifying ground objects from remote sensing images,which can eliminate multiplicative spectral intensity differences in the target spectrum,and identify ground object types by measuring spectral shape similarity between the target spectrum and the reference spectrum.In the study of estimating leaf biochemical parameters,based on reflection spectral data detected in the vertical direction,the spectral shape similarity between the target spectrum and the reference spectrum can be compared using the SAC method to achieve the estimation of leaf biochemical content.However,under multi-angular spectral measurement conditions,the SAC method,which is better at eliminating the multiplicative spectral intensity differences of the target spectrum,cannot completely remove the additive spectral intensity differences displayed at different observation angles for the same leaf sample due to the influence of specular reflection.Therefore,specular reflections shown as additive variables in multi-angular reflection can be effectively removed through the SAC method by converting them into multiplicative variables after appropriate spectral preprocessing.This study selected the leaves of 12 plants species to obtain multi-angular hyperspectral reflectance data in the incident principal plane.Three types of leaf multi-angular spectra were used as the target spectra,including the Bidirectional Reflectance Factor(BRF)and two preprocessing spectra,namely,Continuum Removal(CR)and Band Depth(BD),and the normalized specific absorption coefficient(Nk)spectra of biochemical parameters were used as the reference spectra.Using the SAC method,three types of Spectral Angle Cosine Index(SACI),namely,multi-angular BRF-SACI,multi-angular CR-SACI,and multi-angular BD-SACI,were established,and their estimation abilities for LCC and EWT were tested.The results showed that the correlation between multi-angular BRF-SACI and LCC or EWT was poor(represented by the determination coefficient,R~2),with LCC:R~2=0.83,EWT:R~2=0.76;The correlation between multi-angular CR-SACI and LCC or EWT was poor,with LCC:R~2=0.86 and EWT:R~2=0.82.In contrast,multi-angular BD-SACI has the strongest correlation with LCC or EWT,with LCC:R~2=0.90 and EWT:R~2=0.94.Based on the principle of the strongest correlation,the Optimal Spectral Interval(OSI)for chlorophyll and water is defined.The two OSIs are LCC-OSI:658-748 nm,EWT-OSI:1872-2026 nm,both located in the absorption peak band of biochemical parameters.In these two intervals,the spectral shape of leaf multi-angular BD varies with changes in leaf chlorophyll and water content,and the spectral intensity shows multiplicative differences in the absorption band at different observation angles.The characteristics of spectral shape and spectral intensity of multi-angular BD within the optimal spectral intervals jointly explain the mechanism of estimating leaf biochemical parameters by multi-angular BD-SACI,that is,using the SAC method to measure the spectral shape similarity of multi-angular BD and Nk to show changes in leaf biochemical content;At the same time,the SAC method effectively removes specular reflections that shown as multiplicative variables in the absorption band of multi-angular BD.In addition,this study used five independent datasets to validate the ability of multi-angular BD-SACI to estimate leaf biochemical parameters.They are multi-angular validation dataset,three measured datasets,namely,LOPEX,ANGERS,and IFGG,and PROCOSINE modeled dataset.These datasets cover a variety of spectral measurement methods(multi-angular measurement,integrating sphere,leaf clip,and modeled data),measurement regions(different countries and climatic regions),and plant species.The validation results show that multi-angular BD-SACI can achieve high-precision estimation of LCC or EWT(represented by relative Root Mean Square Error,r RMSE).For multi-angular validation dataset,LCC:r RMSE=24.37%,EWT:r RMSE=21.20%;For the summary of three measured datasets,LCC:r RMSE=17.65%,EWT:r RMSE=29.71%;For the PROCOSINE modeled dataset,LCC:r RMSE=14.08%,EWT:r RMSE=21.88%.Multi-angular BD-SACI has universal applicability for different spectral measurement strategies and various plant species.Therefore,using the SAC method based on multi-angular BD provides a new idea for estimating chlorophyll and water content based on multi-angular spectral characteristics of leaves,and provides an effective tool for quantifying biochemical parameters of plant leaves.