Estimation Of Fractional Vegetation Coverage With Radar Vegetation Index

Fractional Vegetation Coverage(FVC)is a direct and quantitative indicator to reflect the basic information of vegetation,and an important indicator affecting soil erosion.FVC information is also important for management of soil erosion.Remote sensing technology can quickly monitor vegetation on large areas and is effective for quantitative estimation of FVC.In southern China,cloudy and rainy weather is a major factor restricting the application of optical remote sensing technology.So radar remote sensing is very suitable for applications in southern China with its all-time and all-weather imaging capability.This thesis aims to develop method of FVC estimation with radar data to solve data acquisition problem in southern China.In this work,we used Radarsat-2 fully polarimetric data to estimate FVC of the vegetation in Changting county of Fujian Province,which is famous for soil erosion.In our developed method,we combined radar vegetation indices(RVI)and dimidiate pixel model to estimate FVC.First,four RVIs were calculated from both the polarimetric information and intensity information of Radarsat-2 data.Then,the pixels of each RVI image were modelled with dimidiate pixel model to estimate FVC.The estimated FVC images were resampled to five pixel sizes for studying scale effect on estimation.Finally,FVC calculated from high resolution image,instead of field data,was used to validate the estimated FVC.The main contribution of this thesis is the proposal of a method to estimate FVC by combining dimidiate pixel model with RVI.In detail,the following conclusions were drawn:1)Polarimetric information is superior to intensity information for estimating FVC.The results based on RVIvan and RVIFreemanRVIFreeman are better than that based on RVIr and RVI,at the same scale.2)At different scales,estimating FVC with different RVI gets different changing rules with scale changing.The accuracy of estimation based on polarimetric information improves with increasing scale.The best estimation is based on RVIFreemanbasedonRVIFreeman at the 100m scale(R2=0.718).The accuracy of estimation based on intensity information doesn't have apparent scale rules,and its optimal estimation results are not at the largest scale.3)At different scales,estimating FVC with different RVI gets different errors.The error of estimation based on RVIVan and RVIFreeman reduces with scale increasing.When it reaches the largest scale 100m,The absolue error is 0.2 or less,floating around 0.The errors of estimation based on RVIr and RVI? are greater than those based on RVIVan and RVIFreeman as a whole.They don't have scale rules that estimation based on RVIr gets the least error at scale 50m,and estimation based on RVIr gets the least error at scale 75m.4)Advices about the scale of estimation of FVC were given.In practice,it is necessary to make a choice between high accuracy and high resolution,choosing the right scale to meet the accuracy requirements of the study.The best result of this study is based on RVIFeeman at scale 100m.when the research scale is undemanding,we can choose RVIFreeman at this scale.When the scales are demanding,RVIVan,at the scale of 12.5m is a good choice.According to different requirements in the actual study,reseachers must balance the accuracy and resolution,and select the appropriate scale and RVI to achieve research purposes.Our results show that a relatively accurate estimation of FVC can be made using fully polarimetric radar data and our method.When the optical data is unavailable,radar remote sensing is a good alternative.Our method combining dimidiate pixel model with a radar image to estimate FVC not only resolves the data access issue,but also finds a new way for the quantitative extraction of FVC.Meanwhile,the estimation can provide decision support information for soil erosion control work.