Comparative Studies On Leaf Area Index Retrieval Of Road Vegetation Based On GF-1 And ETM+ Images

The Leaf Area Index(LAI)is an important parameter to describe the vegetation structure.It is directly related to many ecological processes of vegetation such as transpiration,photosynthesis and respiration.It can describe the change of the number of vegetation leaves,the change of the canopy structure,the vitality of the plant,and the environmental conditions.It can also be used as the basic parameter to construct many biological and physical models in the ecosystem.The rapid development of remote sensing technology has made it possible to obtain a large area of leaf area index.With China Remote Sensing career advancement,a large number of independent researches and development of satellite have launched.Among a new generation of high-resolution satellites,GF-1 stands out.In order to investigate the adaptability of GF-1 images in the LAI monitoring,this paper compared the same period of ETM+ satellite images in three aspects of sensor spectral response characteristics,the accuracy of inversion model and LAI space consistency,with Li-Tan Highway in Hunan Province selected as the research object and the field measured LAI sampling data,The main research work and results are as follows:(1)This paper compared the image spectral response function and the original image band.Comparing the spectral response capabilities of GF-1 images with Landsat ETM+images,it is found that GF-1 and ETM+ have similar spectral response in the visible light bands(blue,green,red)and near-infrared.Through linear regression analysis of the band reflectance of GF-1 images and Landsat ETM+ images by SPSS 22.0,it is found that there is a clear correlation between the band reflectance,although there are some differences.It is showed that GF-1 images can be used to monitor the growth of vegetation crops and used for inversion of LAI.(2)By constructing an empirical regression inversion model with single-band and six vegetation indices including RVI,DVI,NDVI,TVI,SAVI,and MSAVI,this paper compared the LAI inversion results of GF-1 images with that of ETM+ images.Through the correlation analysis between modeling factors and measured LAI with the help of SPSS22.0,it is found that the correlation between single-wavelength and measured LAI was relatively low,and the correlation between the vegetation indexes and the measured LAI was significant.With the modeling determination coefficient R2 into consideration,the best LAI inversion model was selected.The optimal model for LAI inversion of GF-1 images is the exponential function model of TVI.R2 is 0.639.The optimal model of ETM+ images is the polynomial function model of NDVI,R2 is 0.600.The accuracy of LAI inversion of GF-1 image is slightly higher than that of ETM+ image.The leaf area index values retrieved from two kinds of images showed a consistent linear relationship with the measured leaf area index.(3)Except for the empirical model,the Extreme Learning Machine is used.The inversion results showed that the LAI predicted by the ELM is in good agreement with the measured LAI.The root mean squared error of the predicted LAI of the GF-1 images is 0.495,and the accuracy EA is 86.37%.The RMSE of the predicted of the ETM+ images is 0.509,and the EA is 85.54%.The precision of the predicted LAI based on ELM was significantly higher than that of empirical model,which showed that the machine learning method has a predictive advantage over the empirical model.(4)This paper compared the spatial distribution of the predicted LAI in the region of the Li-Tan Expressway,and it is found that the LAI predicted by the two images have the same distribution.The 30m resolution of ETM+ image is a comprehensive summary of surface information at the pixel scale.There are more mixed pixels,so that the distribution of predicted LAI in the two polar regions is shifted to the middle interval.The 16m resolution of GF-1 remote sensing data is the description of the detailed information of the ground surface,and the probability of occurrence of mixed pixels is significantly reduced,and the inversion results are consistent with the field survey.Therefore,higher pixel resolution is more advantageous in the LAI inversion of road vegetation.