Height values of trees are an important indicator of the health and viabilityof forests. At present, it is the main biophysical parameter observable fromremote sensing images, in particular from Polarimetric Interferometric SAR(PolInSAR) data. It is important to have these values as accurately aspossible. The accuracy of estimated tree height obtained by PolInSAR isaffected by temporal decorrelation. Modeling this correlation is the focus ofthe current thesis.The first chapter explores modeling of the structure function. We used theFourier-Legendre series and combined it with the Gaussian motion functionfor modeling the vertical displacement of the scatterers. This improved theheight estimation accuracy using a single-baseline PolInSAR image pair. Theimprovement was higher when applied in P-band than in L-band. The reasonis the different interaction of the ground and vegetation layer and the lowerpenetration of L-band. The penetration depth becomes important if we areinterested in reconstructing the vertical profile of trees at a higher resolution.In this case, P-band should be used; this fortunately will be available insatellite sensors in near future. For L-band, the exponential function asassumed by the RVoG and RMoG model was equally good.The second chapter proposes the use of the Polarimetric Coherence Tomo-graphy (PCT) model to estimate height from multi-baseline SAR tomostackdata. In the past, temporal decorrelation was considered as a separate sourceof error that is independent of the canopy. It thus causes biased heightestimates. Merging of a Fourier-Legendre series from the PCT model witha temporal decorrelation function from the Random Motion over Ground(RMoG) model has been explored to solve this problem. Results showed animprovement of height estimation accuracy after applying this modification.The optimal number of terms of the Fourier-Legendre series varied for eachpixel. This can be used as an indicator of the complexity of the vegetationlayer as for multi-layer dense forests, more terms are required. This chaptershows that increasing the number of unknown parameters can be done viasegmenting the area into different height classes and selecting the optimumnumber of unknown parameters for each class.The third chapter focuses on obtaining the most accurate height mapsfrom PolInSAR. This is important by itself, whereas height also serves as themain biophysical parameter contributing to the estimation of biomass. Theeffect of mitigating temporal decorrelation was thus examined on biomassiii retrieval accuracy. This research developed new allometric equations for thispurpose and tested different strategies for regression. This was challenging dueto the lack of sufficient field data. The strategy to develop a new allometricequation based on height only is important. A parameter usually measuredduring fieldwork is H100, defined as the basal area weighted average of the100 highest trees in each plot,. This chapter showed that the relation betweenPolInSAR height and H100 is weak, because PolInSAR height estimates theaverage of heights inside the plots and does not simply coincide with H100.The fourth chapter discusses how to take temporal decorrelation intothe estimation of tree heights. It addresses the sensitivity of the proposedmodified model to the choice of complex coherence estimation method. Thebasic step of estimating height in any of the explained models is the selec-tion of homogeneous pixels. To do so, we distinguished polarimetric frompolarimetric-interferometric information. By addressing the pixel selectionwe could jointly take the phase and the magnitude values of the pixels intoaccount. We employed two adaptive methods to define statistically homo-geneous pixels. Height estimation accuracy increased after applying theadaptive methods. Since the proposed adaptive methods are computationallymore intensive, a trade-off between the desired accuracy and computation isrequired prior to selection of any method.To summarize, this dissertation improved the accuracy of tree heightestimation from airborne fully polarized InSAR data by carefully addressingtemporal decorrelation. This is potentially of use for future SAR satellitemissions.
|Qualification||Doctor of Philosophy|
|Award date||28 Feb 2019|
|Place of Publication||Enschede|
|Publication status||Published - 28 Feb 2019|