Case Study of Factors Affecting Tree Height Inversion Performance Using Sir-c/x Polarimetric Interferometric Sar Data of Hetian Test Site in China

In order to evaluate and study polarimetric SAR interferometry tree height inversion algorithm, three passes of SIR-C/X SAR L band full polarization SLC data were acquired for the same imaging area in the western region of Hetian district, Xinjiang Vygur Autonomous Region of PRC. China. The three images acquired in separate days of Oct. 7, Oct.8 and Oct. 9 of 1994 makes the study of the factors affecting performance of three-stage tree height inversion methodology possible. In addition to the SAR data, two scenes of Landsat TM/ETM+ data, one Land cover map and some sheets of 1:50000 topographic maps covering the test site were collected. Firstly, the INSAR preprocessing methods, such as sub-pixel accuracy co-registration, baseline estimation, flatten phase removal, effective wave number and incidence angel images generation, and the three-stage inversion method were outlined. Then data processing and tree height inversion results were analyzed. Some points established from physics theory and simulation POLInSAR data in some publications were confirmed by the primary results of the case study: It has been observed that the tree height image generated from full model inversion (FMI) is much noisy, and better inversion result was observed with vegetation bias removal (VBR) method; The maximum DEM difference (MDD) method can underestimate forest height. Some possible error sources that can lead to worse inversion result have been analyzed and the way to solve this problem has been suggested.