Researches On Multibaseline PolInSAR Forest Height Inversion Considering Vertical Structure Heterogeneity

As the largest terrestrial ecosystem on the earth,the forest plays a vital role in maintaining the earth’s ecological balance.Especially in the carbon cycle,forests effectively regulate the migration of carbon between animals,plants,and the environment through the photosynthesis and respiration of plants.Among the complex and numerous forest parameters,forest height information is closely related to ecological attributes such as forest growth and decline dynamics,aboveground biomass,and stock volume.Therefore,high-precision and large-scale forest height monitoring and inversion are very important for China to achieve the “double carbon” goals of carbon peaking and carbon neutrality.Polarimetric interferometric synthetic aperture radar(PolInSAR)technology has interferometry’s sensitivity to the spatial distribution and height of scatterers,and polarimetry’s sensitivity to the direction and shape of scatterers,so it can distinguish scatterers with different scattering mechanisms in the same resolution cell.Through decades of development,PolInSAR technology has been proved to be a powerful tool for forest height inversion.Compared with single baseline PolInSAR technology,multibaseline PolInSAR technology can provide more abundant observation information for forest height inversion,which can more effectively distinguish different scatterers in the same resolution cell and more accurately correspond to complex forest scenes under the real situation.However,there are still two key problems in multibaseline PolInSAR forest height inversion: 1)At the level of the inversion model,the traditional Random Volume over Ground(RVo G)model does not take into account the forest vertical structure heterogeneity,while empirical models such as Gaussian Vertical Backscatter model and Li DAR waveform-assisted models can describe the vertical structure heterogeneity,but they have the defects of subjective simplification of vertical structure and imaging mechanism mismatch between Li DAR and SAR,respectively.2)At the level of the inversion method,the a priori information between inversion model parameters is ignored,which makes it difficult for the traditional optimal baseline selection method to evaluate the matching relationship between the forest height and the interferometric geometry,and makes it fail for the traditional multibaseline joint inversion method to reasonably weight the multibaseline observations.As a result,neither of them can fully and effectively use the multibaseline observation information to obtain reliable forest height inversion results.According to the above key problems,this paper mainly carries out the following three parts of the research to form a multibaseline PolInSAR forest height inversion system,including model,method,and scheme.Research innovations and research contributions are as follows:(1)The coherent scattering model considering forest vertical structure heterogeneity is proposed.The proposed model accurately describes the scattering process of the SAR signal penetrating the forest area,which improves the applicability and inversion performance of the PolInSAR forest height inversion model.The traditional forest coherent scattering model cannot accurately portray the heterogeneous forest vertical structure in the SAR field of view,which reduces the performance of the model-based PolInSAR forest height inversion.Therefore,this paper introduces another multibaseline SAR technology,called synthetic aperture radar tomography(SAR tomography),which has the three-dimensional imaging capability and can fully reflect the scattering process in the forest area under the SAR field of view and reconstruct the forest vertical structure.The forest vertical structure information obtained by SAR tomography is introduced into the construction of the forest coherent scattering model,which provides a more rational forest height inversion model for multibaseline PolInSAR technology.The proposed model is verified by the airborne P-band SAR data of test site Mondah of the Afri SAR 2016 campaign.The experimental results show that,under the conditions of single-baseline and multibaseline data,the accuracy of forest height inversion based on the new model achieves 5.12 m and 3.15 m,respectively.Compared with the forest height inversion based on the traditional RVo G model,the accuracy improves by 48.34% and 56.01%,respectively.(2)The multibaseline PolInSAR forest height inversion based on coherent scattering model sensitivity is proposed.The proposed method reveals the nature of the influence of multibaseline interferometric geometry on forest height inversion,which provides a more reasonable quantitative expression of the inversion contribution of multibaseline observations.Taking the RVo G model as an example,based on the in-depth study of the a priori intrinsic connection between the coherent scattering model parameters,the matching relationship between forest height and multibaseline interferometric geometry is corrected by the coherent scattering model sensitivity.In terms of optimal baseline selection,the coherent scattering model sensitivity is used as a new quality evaluation index to select the optimal interferometric geometry in essence.In terms of the multibaseline joint inversion,the stochastic model and weight function are established according to the coherent scattering model sensitivity.The influence of unsuitable baseline observations on the inversion is eliminated through reweighted iteration.The proposed method is verified by the airborne P-band multibaseline SAR data of test site Mondah of the Afri SAR 2016 campaign.The experimental results show that the inversion accuracy of the proposed optimal baseline selection method is 5.10 m,and the inversion accuracy of the proposed multibaseline joint inversion method is 4.10 m.Compared with the traditional method,the new method effectively improves the accuracy and robustness of the multibaseline PolInSAR forest inversion.(3)The small sample SAR tomography-aided multibaseline PolInSAR forest height inversion scheme is proposed.The proposed scheme organically combines the new model and new method to establish the correlation between large-scale multibaseline PolInSAR observations and forest vertical structure by using small-scale SAR tomography,which realizes large-scale,fine-grained forest height inversion and provides a new idea for forest height mapping using SAR platform.For the future SAR forest monitoring missions carried out by China and abroad,this paper suggests conducting synergic observation of multibaseline SAR tomography and multibaseline PolInSAR.Due to the demanding data requirements of SAR tomography,tomographic imaging is only carried out in a small-scale forest area to provide forest vertical structure samples,which are used to construct the coherent scattering model considering forest vertical structure heterogeneity by machine learning approaches.The model is applied to the multibaseline PolInSAR iterative reweighted method for forest height joint inversion based on coherent scattering model sensitivity,finally obtaining the large-scale and high-precision forest height inversion information.The proposed scheme is verified by the airborne P-band multibaseline SAR data of test site Mondah of the Afri SAR 2016 campaign.The experimental results show that the proposed forest height inversion scheme takes full advantage of the two multibaseline SAR technologies.It has versatility.Only the small-scale SAR tomography and the large-scale three-baseline PolInSAR observation configuration are used for inversion.The inversion effect is equivalent to that of the large-scale SAR tomography and the large-scale ten-baseline PolInSAR observation configuration.The forest height inversion accuracy achieves3.20 m.