Font Size: a A A

The Envisat Asar Applied Research, Invasive Plants, Leaf Area Index Inversion

Posted on:2011-01-19Degree:MasterType:Thesis
Country:ChinaCandidate:T XiaoFull Text:PDF
GTID:2190360302992081Subject:Physical geography
Abstract/Summary:PDF Full Text Request
In this paper, taking the Dongtan wetland of Chongming, Shanghai, China as study area, the principle and methodology to modeling leaf area index (LAI) inversion were particularly focused on for invasive plants such as Spartina Alterniflora and local plants such as Phragmites Australis by using ENVISAT dual-polarization advanced synthetic aperture radar (ASAR) imagery.Two kinds of inversion models were built at first in the study that one is empirical statistical model and another is back propagation (BP) algorithm based artificial neural network model, in which the mathematical relationship between the backscattering coefficient of ASAR image and LAI of invasive plant was embedded. The model parameters were then calculated by using the backscattering coefficient value generated from ASAR image and LAI value of plant achieved from field measurement. To obtain the parameters of empirical statistical model, the curve fitting method was employed.60 percent of the measured LAI values were used in the neural network training process based on BP algorithm while the other measured LAI values were used to verify the effectiveness of neural network model. Experimental cases were conducted comprehensively as followed process in this paper that the ENVISAT ASAR imagery the year 2005 and the measured LAI value of invasive plant were applied. In the experimental cases, the inversion models were compared by assessing accuracy of LAI inversion results with the modes of HH and VV polarizations via three metrics that mean error, root mean squared error and correlation coefficient. According to the optimal inversion model achieved through the comparison, the LAI values were finally predicted for research plants of the years 2004 and 2005 in the entire study area.Based on the comparison analysis and accuracy assessment of LAI inversion for invasive plants in the experimental cases as adopting the backscattering coefficient image without being filter-processed or the one with being processed by speckle noise filter in the modes of HH and VV polarizations, some important conclusions can be reached as following that:(1) With the same mode of polarization, Model III in the empirical statistical models that y=ax3+bx2+cx+d perform better in LAI inversion than Model I that y=exp(ax+b) and Model II that y=ax2+bx+c while the BP neural network models demonstrate the best performance in LAI inversion. (2) With the same mode of polarization, the LAI inversions by adopting the filter-processed backscattering coefficient image are significantly superior to the ones by adopting the backscattering coefficient image without being filter-processed.(3) By adopting the backscattering coefficient image without being filter-processed, the effectiveness of the LAI inversions via Modelâ…¢in the mode of HH polarization takes no difference with the one in the mode of VV polarization. However, by adopting the backscattering coefficient image without being filter-processed, the effectiveness of the LAI inversions via BP neural network models in the mode of HH polarization is greatly different to the one in the mode of VV polarization, where the effectiveness in the mode of HH polarization is better than the one in the mode of VV polarization.(4) By adopting the filter-processed backscattering coefficient image, the LAI inversions via Modelâ…¢in the mode of HH polarization performs as well as the one in the mode of VV polarization. However, by adopting the filter-processed backscattering coefficient image, the effectiveness of the LAI inversions via BP neural network models in the mode of HH polarization makes a bit difference to the one in the mode of VV polarization, where the effectiveness of the LAI inversions to Spartina Alterniflora in the mode of HH polarization is better than the one in the mode of VV polarization, while the effectiveness of the LAI inversions to Phragmites Australis in the mode of VV polarization is better than the one in the mode of HH polarization.The LAI inversion results were revealed in the paper that, by adopting the filter-processed backscattering coefficient image, the correlation coefficient of inversion LAI for Spartina Alterniflora via the optimal BP neural network model is 0.7087 in the mode of HH polarization while the one is 0.7113 in the mode of VV polarization, and the correlation coefficient of inversion LAI for Phragmites Australis via the optimal BP neural network model is 0.7147 in the mode of HH polarization while the one is 0.709 in the mode of VV polarization. The study in this paper demonstrates that the quantitative inversion for LAI of plant by using SAR multi-polarization data is reliable and feasible.
Keywords/Search Tags:Invasive plant, Leaf area index (LAI), ENVISAT ASAR, Backscattering coefficient, Remote sensing inversion, Dongtan of Chongming
PDF Full Text Request
Related items