Hyperspectral Estimation Models Of Biochemical Composition For Dicranopteris Dichotoma Leaves From Different Ecological Restoration Stages In The Eroded Red Soil Areas Of Southern China

Dicranopteris dichotoma which is an important zonal herbaceous plant for ecological restoration and reconstruction in the red eroded area of southern China can control soil erosion effectively.By using hyperspectral remote sensing technology,the growth and health status of dicranopteris dichotoma can be monitored rapidly and timely without any damage.Through field sampling and laboratory analysis,the hyperspectral reflectance data and the corresponding chlorophyll and nitrogen contents of dicranopteris dichotoma leaf were collected from six different ecological restoration stages which include the age of 6,12,18,36 years and two control plots in Zhuxi small watershed of Changting County,Fujian Province.This study analyzed the morphological characteristics and hyperspectral curve properties of leaf,transformed the original spectral into first derivative,and determined the sensitive wavebands to create univariate,multivariate stepwise regression and BP neural network models for the content of biochemical components.The best estimating models for chlorophyll contents and nitrogen contents of Dicranopteris dichotoma leaf were determined by accurate analysis and evaluation test.The objective of this research is to provide effective technical support for ecological restoration by building hyperspectral estimation models of biochemical components with a rapid and non-destructive method for monitoring vegetation growth.Through the study it can be concluded as follows:?1?Contrasted leaf morphology,chlorophyll and nitrogen contents of dicranopteris dichotoma leaf at different ecological restoration stages,with the ecosystem restoration gradually,the specific leaf area increased,the leaf dry matter content decreased,the content of chlorophyll and nitrogen continued to increase and became the highest in the control plot of FSL where is the best ecological restoration stage.?2?The original spectral data of dicranopteris dichotoma leaf from different ecological restoration stages showed the similar spectral features.The highest slope was in the red edge,the reflectance of short wave near-infrared in the FSL control plot where is at late ecological restoration stage was stay the highest.Comparing different slope position at the same ecological restoration stage,the original spectrum reflectance below was higher than reflectance upside;the original spectrum reflectance of gully was higher than the ridge.?3?The sensitive wavelength of original spectral for dicranopteris dichotoma leaf from different ecological restoration stages were 407nm,603nm,and the optimal wavebands of first derivative were 463nm,554nm,674nm,739nm.The correlation between chlorophyll content and the original spectral showed that the correlation in red edge was high,and the correlation between chlorophyll content and first derivative showed that the correlation was extremely significant level from 350nm to700nm in the visible light.?4?In univariate regression estimating models of chlorophyll content of dicranopteris dichotoma leaf,the linear model y=0.096x-66.543 established by?r index and the quadratic model y=4.693-1.506x+0.139x2created by RVI_603/40703/407 index got the highest model accuracy,and the R2of estimation model reached 0.741.For the multivariate regression models,the modely?28?-52.997?10?.0103?₁-.0029?₂?10?.0525?₃basing on hyperspectral indices of?r,?y and SDr,with the highest test results and the highest verification accuracy,was the best estimating model.?5?Of all the estimation models of nitrogen content of dicranopteris dichotoma leaf,the multivariate regression estimating model Y=13.45+25.562NPCI+0.292DCNI established by NPCI,DCNI index was better than others in four multivariate regression models,with the highest R2,the lowest RMSE and RE%?R2=0.597,RMSE=1.304,RE=15.0%?,and so it was the best estimating model.In four BP neural network models of nitrogen content,the estimating model established by NPCI,SIPI,mNDVI₇₀₅,Vog index was more appropriate for estimating nitrogen content,with the highest R2of estimated value and measured value,the minimum RMSE and RE value?R2=0.916,RMSE=1.06,RE=7.52%?.