Informations et ressources scientifiques
sur le développement des zones arides et semi-arides

Accueil du site → Doctorat → Chine → 2017 → Remote Sensing Information Extraction and Inversion of Vegetation in Dryland Areas

China Forestry Science Academe (2017)

Remote Sensing Information Extraction and Inversion of Vegetation in Dryland Areas

叶静芸

Titre : Remote Sensing Information Extraction and Inversion of Vegetation in Dryland Areas

Auteur : 叶静芸;

Grade : Doctoral Dissertation 2017

Université : China Forestry Science Academe

Résumé
In dryland areas,vegetation information,such as Aboveground Biomass(AGB)and Fractional Vegetation Cover(FVC),is an important indicator of vegetation growth status evaluation and desertification monitoring.Remote sensing technology provides multi-band and multi-temporal data sources for vegetation information extraction.However,because desert vegetation distributed in most parts of the dryland areas is very sparse,so that the sensor sensitivity to detect the spectral information of vegetation is reduced,therefore the universal remote sensing model loses universality when extracting desert vegetation information in dryland areas.At same time,due to sensor resolution limits,there are many kinds of surface information in the medium and low resolution remote sensing image pixels and the mixed-pixel problem leads to the difficulty in extracting sparse desert vegetation information,as a result,the FVC and AGB of sparse desert vegetation has great uncertainty.Aiming at this issue,this thesis established a set of remote sensing method for extraction and inversion of desert vegetation.Using this method,in the semi-arid,arid and hyper-arid areas of China,the vegetation information extraction and inversion were carried out based on field investigation data and multisource remote sensing data.The main results and conclusions are as follows :(1)Methods of extraction and inversion of desert vegetation remote sensing information is divided into the following four steps : 1)to estimate the AGB in field plots by using allometric equations ;2)to extract remote sensing parameters highly correlated with sparse vegetation from the high,medium and low resolution remote sensing data ;3)to timely and spatially match the field plot and remote sensing data,and to establish the vegetation AGB and FVC models in different typical study areas and dryland areas by using linear and nonlinear regression models ;4)to select the optimal regression model to estimate the vegetation AGB and FVC in the study areas,and to correct the vegetation AGB and FVC estimation data derived from medium and low resolution data by the estimation data derived from high resolution data.(2)In the Mu Us Sandy Land,the desert-oasis ecotone on the northeastern edge of the Ulan Buh Desert and the Kumtag Desert,remote sensing feature parameters,such as vegetation index and texture index,were extracted from high-resolution data and the AGB regression models were established.based on field plot data.The results show that the R2 of the AGB regression model based on high resolution data is above 0.8,and the ratio of constructed and predicted data sets is close to 1.The estimation result is very well.In the study area of the Altun piedmont gobi desert in the Kumtag Desert research area,the segmentation method was used to divide field plots into different size sub-plots,which can effectively improve the estimation accuracy of the AGB model of sparse desert vegetation(R2 ??? = 0.98,RMSErel ???????????? = 10.98%).(3)In the Mu Us Sandy Land,Qinghai Gonghe Basin,the desert-oasis ecotone on the northeastern edge of the Ulan Buh Desert,Minqin and the Kumtag Desert,the regression models were established by using medium resolution data and field plot data.The results showed that for the optimal regression model of desert vegetation in semi-arid areas,the R2 was about 0.8 and the RMSErel was between 28-50%.For the optimal regression model of desert vegetation in arid areas,the R2 was about 0.5,and RMSErel was between 51-65%.For the optimal regression model of desert vegetation in hyper-arid areas,the R2 was about 0.4 and RMSErel was about 70%.(4)In the Mu Us Sandy Land and the Altun piedmont Gobi Desert in the Kumtag Desert research area,medium resolution data were corrected by using high resolution data as the conversion medium between the field data and medium resolution data.This method significantly improved the estimation accuracy of the vegetation AGB,and was more effective in the hyper-arid area.(5)The AGB of vegetation in dryland areas were estimated,by using the AGB data in the Mu Us Sandy Land and Gonghe Basin as the conversion medium of semi-arid area,using the AGB data of Minqin as the conversion medium of arid area,and using the AGB data of Altun piedmont gobi desert as the conversion medium of hyper-arid area.It significantly improved the estimation accuracy of the vegetation AGB in dryland areas.The temporal and spatial dynamics of the vegetation AGB and FVC from 2000 to 2016 were analyzed

Mots clés : Dryland areas; AGB; FVC; desert vegetation; sparse vegetation;

Présentation (CNKI)

Page publiée le 9 janvier 2018