Estimation Of Aboveground Biomass Of Vegetation In Dryland Areas Based On Ground And Multi-source Remote Sensing Data

China’s dryland area is large and widely distributed,with an arid climate and poor soil.The dryland area is mainly distributed with desert and grassland vegetation,which has a simple structure and is low and sparse.Vegetation coverage and Aboveground Biomass(AGB)are important indicators of vegetation growth evaluation and desertification monitoring in dryland areas.Remote sensing technology has become the primary means of vegetation information extraction and inversion.However,optical remote sensing vegetation information extraction technology is mostly applied to forest and grassland vegetation,and the research on desert vegetation is relatively weak.In recent years,the study of sparse vegetation has attracted much attention,but there are few studies on remote sensing estimation of AGB of sparse vegetation,lack of systematic research,and there are still many deficiencies in estimation methods.In this thesis,the sparse vegetation in semi-arid,arid and hyper-arid areas is taken as the research object.According to the characteristics of natural geography and spatial distribution of vegetation in dryland areas,Mu Us Sandy Land,Gonghe Basin,Ulanbuh Desert and its northeast desert oasis transition zone,Minqin area,Kumtag Desert and its surrounding areas are selected as typical research areas,based on a large number of long-term investigated plot data and multi-source remote sensing data with different spatial resolution,the remote sensing estimation method of vegetation,especially sparse vegetation AGB in dryland area was systematically studied.This study will provide data and technical support for the estimation of AGB and carbon storage of vegetation and desertification monitoring in dryland areas.The main results and conclusions are as follows:(1)The remote sensing estimation results of aboveground biomass of vegetation in drylnad areas:The average aboveground biomass of Mu Us Sandy Land was 231.16 g/m~2(RMSErel=54.28%)and Gonghe basin was 177.97 g/m~2(RMSErel=67.38%);The average aboveground biomass of desert vegetation was 90.73 g/m~2(RMSErel=24.15%)and 138.14 g/m~2(RMSErel=104.08%).The average aboveground biomass of Minqin study area was 127.9 g/m~2(RMSErel=154.26%);The average aboveground biomass of desert vegetation was 118.28 g/m~2(RMSErel=71.51%)in front of the north foot of Altun mountain,103.89 g/m~2(RMSErel=68.23%)in Aqike Valley and 136.72 g/m~2(RMSErel=133.23%)in Dunhuang West Lake wetland.Based on the above ground sample data and MODIS data,the average AGB in the semi-arid,arid and hyper-arid areas of northern China was 517.29 g/m~2(RMSErel=47.93%),297.17 g/m~2(RMSErel=72.77%)and 39.99 g/m~2(RMSErel=84.93%)respectively.(2)Different spatial resolution remote sensing data have different applicability for estimating the aboveground biomass of sparse vegetation in the dryland areas.High spatial resolution satellite remote sensing data can accurately describe the spatial and temporal distribution characteristics of vegetation in arid areas.The RMSErel mean values of vegetation aboveground biomass in semi-arid areas,arid areas and hyper-arid areas are 24.37%,16.62%and 41.72%,respectively,which is more suitable for high-precision remote sensing estimation of sparse vegetation aboveground biomass in the dryland areas.The RMSErel mean values of medium spatial resolution remote sensing data in semi-arid,arid and hyper-arid regions were62.36%,72.77%and 84.93%,respectively,which were more suitable for estimating AGB in arid and semi-arid regions with higher vegetation coverage.The RMSErel mean values of low spatial resolution remote sensing data in semi-arid,arid and hyper-arid area were 47.93%,72.77%and84.93%,respectively.The performance was better in semi-arid areas with high vegetation coverage,but in arid and hyper-arid regions,the performance was poor.(3)Applicability of the above ground biomass estimation model in the dryland areas:the accuracy of using NDVI and RVI to estimate the above ground biomass of sparse vegetation in arid area is significantly higher than that of the improved soil adjusted vegetation index(MSAVI),When using medium spatial resolution remote sensing data,linear regression model is the best.(4)The influencing factors of sparse vegetation AGB estimation are as follows:1)the size of sparse vegetation AGB and vegetation coverage are the internal factors affecting AGB estimation.The results show that from semi-arid area to arid area and then to hyper-arid area,using the exact spatial resolution remote sensing data and the same estimation method,the error of AGB remote sensing estimation results is higher and higher,and the accuracy is gradually reduced.From the semi-arid area to the arid area and then to the hyper-arid area,with the climate becoming dryer,the AGB per unit area of vegetation decreased significantly,and the vegetation coverage showed the same change trend as the AGB.2)The spatial resolution of remote sensing images is an important external factor affecting the estimation of the sparse vegetation AGB.The excellent correlation between AGB and NDVI is the basis of AGB remote sensing estimation.The results showed that with the decrease of AGB,the NDVI value of the corresponding pixel gradually narrowed.That is the sensitivity of NDVI value to AGB gradually decreased,which made it more difficult to extract AGB.Besides,there is a good correlation between NDVI of high spatial resolution remote sensing images and AGB in different types of dryland areas.Still,the correlation between NDVI of medium and low spatial resolution remote sensing images and AGB is weak,especially in the hyper-arid area.(5)The applicability of using higher spatial resolution remote sensing data to correct low spatial resolution remote sensing estimation results:1)Using high spatial resolution remote sensing estimation to correct medium spatial resolution remote sensing estimation results in the Mu Us Sandy Land and the Gobi area at the northern foot of Altun Mountain,respectively.The RMSErel of the result decreased from 62.36%to 54.28%;the estimated RMSErel of the Gobi area decreased from 114.34%to 83.16%.It can be seen that when extracting sparse vegetation AGB information,using high spatial resolution remote sensing data estimation results to correct medium spatial resolution remote sensing data can effectively improve the estimation accuracy of sparse vegetation AGB using medium spatial resolution remote sensing data.2)When estimating the AGB of sparse vegetation in the dryland areas,the results of estimating the AGB of vegetation with low spatial resolution remote sensing data were compared by using high and medium spatial resolution remote sensing data,which could not effectively improve the accuracy of estimating the AGB of vegetation correction.Therefore,the AGB correction method is not suitable for the correction of the estimation results of low spatial resolution remote sensing data using high and medium spatial resolution remote sensing data.