Retrieval Of Vegetation Canopy Fuel Moisture Content From Time Series Multi-polarimetric SAR Data

Wildfires destroy natural environment and species diversity,release greenhouse gases,affect global climate,and threaten human lives and property security.Therefore,assessment and early warning of wildfire risk is vitally significance.Vegetation canopy fuel moisture content(CFMC),defined as the proportion of water content over dry mass,is the focus for wildfire risk assessment and early warning.Currently,the development of optical remote sensing provides a unique way to accomplish this objective.However,optical signal is highly susceptible by weather conditiond due to the weak penetration,leading to it's difficult to accurately monitor the spatio-temporal variation of vegetation CFMC.Instead,microwave remote sensing provides the new opportunity for vegetation CFMC retrieval because of microwave signal high sensitivity to vegetation moisture,all-time work capacity,and unaffected by illumination and weather condition.This thesis,taking the microwave scattering model and polarization decomposition method as the theoretical basis,in-orbit active microwave remote sensing data and in-situ measured CFMC data as the data basis,fully explores the feasibility of CFMC retrieval from microwave remote sensing data.Main accomplished work can be summarized as the following three parts:(1)Based on the quad-polarimetric Radarsat-2 data,the feasibility of polarimetric decomposition parameter for grassland CFMC estimate was evaluated.And the multiple linear regression model was built using the stepwise regression method to map the spatial CFMC distribution.The experimental result indicates that the polarimetric decomposition parameter can be used for grassland CFMC estimate.R and RMSE between measured and estimated grassland CFMC is 0.658 and 30.319%,respectively.(2)Based on bare soil scattering Dubois model,vegetation scattering ratio model and Topp soil dielectric model,an semi-empirical surface scattering model for surface vegetation parameter retrieving from active microwave remote sensing data was bulit.Then utilizing five Ruoergai prairie measured vegetation parameters,the effect of built model was evaluated.The experimental result shows that whether the built model is befit for a certain vegetation parameter retrieval mainly decided by the parameter itself.For Ruoergai prairie grassland CFMC,the built model is unsatisfactory.The best performance of the built model to retrieve vegetation parameters is as follows:vegetation wet weight(R=0.723,RMSE=0.448 kg/m~2),vegetation dry weight(R=0.674,RMSE=0.129 kg/m~2),CFMC(R=0.394,RMSE=49.546%),Leaf Area Index(R=0.761,RMSE=0.863 m~2/m~2),Normalized Difference Vegetation Index(R=0.598,RMSE=0.032).(3)Based on bare soil scattering Line model and vegetation scattering Water Cloud Model,an semi-empirical scattering model for retrieving CFMC from active microwave remote sensing data was established.Compared with the built model described in(2),it contains only three model parameters(VH and VV backscattering coefficients,CFMC)and therefore can be applied to the region without soil moisture measurement or retrieved product.The performance of this model for forest CFMC retrieval was evaluated using the public American national fuel moisture data.The result shows this model can be effectively applied to quantitatively retrieve time series forest CFMC(based on the three-fold cross-validation strategy,RMSE between measured and retrieved CFMC is 19.53%,12.64%,and 15.45%,respectively),which demonstrates that the feasibility of active microwave remote sensing for vegetation canopy CFMC retrieval.