| The middle section of Longquan Mountain is regarded as the“East Ecological Barrier”and the key area for planning and development in Chengdu.The parameters of vegetation biochemical information is estimated by using quantitative remote sensing technology,the practical application and applicability of different inversion methods are analyzed,it is necessary to improve the accuracy of the inversion model to monitor the growth and health of the vegetation in this area.Therefore,the chlorophyll of the middle section of Longquan Mountain was taken as research object in this paper,vegetation samples in the field were collected,spectral reflectance curve and leaf chlorophyll content of samples were obtained by the portable spectroradiometer SVC-HR1024i and SPAD-502 portable chlorophyll meter.Based on the measured spectral reflectance values of vegetation,measured chlorophyll content of leaves and Sentinel-2A multi-spectral remote sensing data,the reflectance spectra of vegetation in the study area were analyzed and the estimation models were constructed by data analysis software such as ENVI,EXCEL,MATLAB and SPSS22.0.The model accuracy of the vegetation index model,the statistical analysis model and the physical model were compared in the experiment,the best inversion estimation model for the inversion of the chlorophyll content of the vegetation in the Longquan Mountains was finally determined.This study provides a theoretical basis for remote sensing monitoring of vegetation physical and chemical parameters in the middle section of Longquan Mountain,and a technical basis for monitoring vegetation health status and achieving sustainable development in the region.The research results obtained mainly include the following aspects:?1?Based on the terrain correction of Sentinel-2A remote sensing image during image preprocessing,it is found that the atmospheric correction can only obtain the reflectivity ?TOA of the top by using the traditional FLAASH module atmospheric correction.The influence of factors such as the underlying surface should be eliminated.It is also necessary to use the Sen2Cor plug-in for atmospheric correction to obtain the eflectivity?BOA at the bottom of the atmosphere.The Sentinel-2A remote sensing image is preprocessed by atmospheric correction,topographic correction,and pixel decomposition to obtain a higher precision canopy spectrum,which provides data support for subsequent inversion analysis.?2?Through the analysis of the generated spectral simulation curve,it is found that the change of spectral reflectance in the range of 400nm-800nm is mainly caused by the difference of chlorophyll content in the case of considering the influence of vegetation physical and chemical parameters only at the leaf scale and canopy scale.The lower the chlorophyll content,the greater the difference.When the chlorophyll content is 10?g/cm2,the relative error between the spectral reflectance and the spectral reflectance of healthy vegetation reaches 73.78%.Secondly,the error of effect of leaf area index can reach 31.89%when the leaf area index is only 0.75.The last affected index is the leaf water content parameter,no matter how the water content parameter changes within the reference range,the error caused by it is about1.1%,which can be ignored.?3?At the leaf scale,six leaf-scale vegetation indices with correlation coefficient greater than 0.7 were selected.The vegetation index model with the highest coefficient of determination is mND705,R2 reaches 0.8323 and the standard error is9.91?g/cm2.Based on the statistical analysis method of leaf spectral characteristic variables,the cubic model constructed with green peak area is of the highest accuracy,R2 reaches 0.7778 and the standard error is 10.08?g/cm2.The BP neural network model with 6 vegetation indices as independent variables and measured chlorophyll content as the dependent variable has a coefficient of determination R2 of 0.7547 and a standard error of 10.33?g/cm2.The inversion result obtained by the simulated annealing algorithm?SA?with the measured chlorophyll content simulated solid annealing process has a coefficient of determination of 0.7912 and a standard error of10.13?g/cm2.With the help of the comparison of precision analysis,it can be seen that at the leaf scale,the inversion result of the vegetation index model is slightly higher than the inversion result of the other model in accuracy.When chlorophyll content in the physical model is lower than 60?g/cm2,and the prediction value obtained by the physical model inversion is generally lower than the actual value.When the chlorophyll content is higher than 60?g/cm2,the predicted value is closer to the actual value,it tends to be saturated.?4?At the canopy scale,vegetation indexes were constructed based on Sentinel-2A image data.The highest coefficient of determination R2 is the CIrededge?red edge vegetation index?model,which reaches 0.7214,and the standard error of the index is 56.21?g/cm2,which further demonstrates that the red edge band can accurately monitor the vegetation health status and changes in canopy chlorophyll content.The vegetation index based on Sentinel-2A data was used as the independent variable to construct the canopy-scale multiple linear regression model and BP neural network model.The multiple linear regression model has a coefficient of determination R2 of 0.7405 and a standard error of 56.74?g/cm2.The coefficient of determination of the inversion result of the neural network model is 0.7834,and the standard error is 48.27?g/cm2.The inversion model with the canopy chlorophyll content input simulated annealing algorithm?SA?has a coefficient of determination of0.8045 and a standard error of 47.91?g/cm2.Therefore,it can be seen from the accuracy comparison that the inversion accuracy of the simulated annealing algorithm is higher than the BP neural network algorithm,and the inversion results of the physical model are better than the inversion results of the vegetation index model.?5?The canopy spectral data was simulated by the scale conversion concept combined with the measured blade spectral data.The canopy-scale vegetation indexes were constructed based on the simulated canopy spectral data.The highest coefficient of determination R2 is the OSAVI model,which is 0.68,and the standard error is 59.16?g/cm2.This is because the two bands 800 nm and 670 nm calculated by OSAVI reduce the influence of soil background radiation.It can be seen from the comparative analysis that the accuracy of the vegetation index model based on Sentinel-2A data on the canopy scale is generally higher than that based on the canopy simulation spectrum.The vegetation index model based on simulated canopy spectral data is more susceptible to the vegetation's own biochemical parameters,underlying surface,water,soil background radiation and other factors.It can be seen that using Sentinel-2A image data to monitor the chlorophyll content of vegetation on a large scale is more advantageous,and the canopy spectroscopy data is more conducive to the inversion of the chlorophyll content of the small regional canopy. |
PDF Full Text Request