Forest Aboveground Biomass Estimation Using PolInSAR Data

As the largest carbon pool in the terrestrial biosphere,forests play an important role in both carbon sequestration and emission.In order to quantify the forest biomass and carbon storage at global scale,numerious remote sensing techniques have been developed,including optical,Light Detection And Ranging(LiDAR),passive microwave,and Synthetic Aperture Radar(SAR)remote sensing.SAR remote sensing is of particular interest for forest biomass mapping due to its relative weather independence,stronge penetration,and high spatial and temporal resolution.Based on SAR,the recently developed advanced techniques,e.g.Interferometric SAR(InSAR),Polarimetric SAR Interferometry(PolInSAR),and SAR Tomography(TomoSAR),showed unique potential on forest biomass estimation.Based on the two most potential techniques,i.e.PolInSAR and TomoSAR,this study mianly concentrates on the general issues associated with Pband SAR for forest height and biomass estimation,including the temporal decorrelation,residual ground scattering contribution,complex underlying surface,topographic influence,saturation problem,and so on.It is expected that the results could provide new insights on improving the accuracy and robustness of the PolInSAR-and TomoSARbased forest height and biomass estimation.Four kinds of work were conducted:(1)Exploring the impacts of spatial baseline on Canopy Height Model(CHM)and Digital Terrain Model(DTM)retrievals using P-band PolInSAR data.Because of nonvolume decorrelation and other unavoidable errors,the robustness of retrieval heights is sensitive to the spatial baseline length,which relates forest parameters to measured coherence.It is important to find an optimal spatial baseline length to improve the retrieval accuracy.Within the context of the random volume over ground(RVoG)model and the three-stage inversion method,we aimed to quantify the influence of spatial baseline on the CHM and DTM inversions at P-band,which are distinct from the inversions at higher frequency due to the non-negligible ground contributions.This information assists in optimal baseline selection and the development of robust inversion schemes.(2)Improving forest height retrieval by reducing the ambiguity of volume-only coherence using multi-baseline PolInSAR data.Non-volume decorrelation(e.g.temporal decorrelation)united with the unknown ground contribution will bring a 2-D ambiguity to volume-only coherence,making the inversion underdetermined even when multiple baselines are available.In the context of RVoG model and three-stage algorithm,this study theoretically presented the varied response of different baselines to both nonvolume decorrelation and ground contribution,and then proposed a new multi-baseline inversion method to reduce the 2-D ambiguity.The proposed method includes two steps,calculating the common overlapped ambiguity from different baselines and fixing the extinction coefficient,to more accurately retrieve the volume-only coherence and forest height.It makes no assumptions on non-volume decorrelation and ground contribution,and is relatively robust among different baseline combinations,which could make the method to be robust in the real applications.(3)Biomass estimation in dense tropical forest using multiple information from single-baseline P-band PolInSAR data.SAR derived information can be correlated to forest biomass.However,the relationship is sensitive to other disturbances,which will reduce the biomass estimation accuracy and robustness.This study developed an improved Aboveground Biomass(AGB)estimation approach by integrating multiple information derived from single-baseline PolInSAR data.The approach involves three kinds of steps to improve the biomass estimation.Firstly,decomposing different scattering machanisms to remove the ground scattering component based on RVoG model.Secondly,employing the PolInSAR-derived DTM to compensating the topographic influence on backscatter and forest height inversion.Thirdly,linearly combining PolInSAR-derived forest height and volume backscatter into biomass estimation models.Backscatter is satuated for high biomass regions,and forest height shows limited correlation with high AGB due to the varying forest basal area.It is expected that combining backscatter and forest height can compensate their respective weaknesses and produce more accurate AGB estimates.(4)Potential of texture from SAR tomographic images for forest aboveground biomass estimation.SAR texture has been demonstrated with potential to improve the forest biomass estimation in addition to backscatter.Texture is generally quantified based on the 2D variance between pixel and its neighboring pixels.However,the strong penetration of SAR signals makes each pixel contains the contributions from all the scatterers inside forest canopy.Consequently,the traditional texture derived from SAR images will be affected by the forest vertical heterogeneity.In order to separate and explore the influence of forest vertical heterogeneity on texture-based biomass estimation,we introduced TomoSAR technique into the traditional texture analysis.The GLCM textures were calculated for layered backscatter at different height.The stepwise regression method was used to construct the biomass estimation models that were crossvalidated between different regions.