TanDEM-X Pol-InSAR Performance for Forest Height Estimation
TanDEM-X Pol-InSAR Performance for Forest Height Estimation
复制标题
DOI：
10.1109/tgrs.2013.2296533
复制
发表时间：
2014-10-01
影响因子：
8.2
通讯作者：
Papathanassiou, Konstantinos P.
中科院分区：
文献类型：
--
作者：
Kugler, Florian;Schulze, Daniel;Papathanassiou, Konstantinos P.
Tandem-X和TerraSAR-X平台共同构成了空间中第一台星载单程极化干涉仪。这使得首次能够在没有时间去相关干扰的情况下获取星载极化合成孔径雷达干涉测量(POL-InSAR)数据。本文件旨在评估此类数据在森林应用方面的潜力。为此，研究了在北部、温带和热带地区采集的单极和双极数据，以表征干涉相干性测量的X波段穿透和极化多样性。针对单极和双极情况，提出并实施了POL-InSAR森林高度反演方案，并与所有站点的LIDAR参考测量进行了交叉验证。单极反演依赖于外部地面数字地形模型(DTM)，对相关系数r(2)在0.80至0.98之间的所有站点都表现良好。双极反演不需要外部数字地面模型，而是取决于整个林层的能见度。因此，其表现随森林结构和季节的不同而不同：夏季收购北方试验地(r(2)=0.86)和冬季收购温带试验地(r(2)=0.77)表现最好。对于热带试验点，只能建立弱相关(r(2)=类似于0.50)。
TanDEM-X and TerraSAR-X platforms form together the first spaceborne single-pass polarimetric interferometer in space. This allows, for the first time, the acquisition of spaceborne polarimetric synthetic aperture radar interferometry (Pol-InSAR) data without the disturbing effect of temporal decorrelation. This paper aims to assess the potential of such data for forest applications. For this, single- and dual-pol data acquired over a boreal, a temperate, and a tropical site were investigated to characterize X-band penetration and polarization diversity of the interferometric coherence measurements. Pol-InSAR forest height inversion schemes have been proposed and implemented for the single-and dual-pol cases and cross validated against LIDAR reference measurements for all sites. The single-pol inversion relies on an external ground digital terrain model (DTM) and performed well for all sites with correlation coefficients r(2) between 0.80 and 0.98. The dual-pol inversion does not require an external DTM but depends on the visibility of the whole forest layer. Accordingly, its performance varied with forest structure and season: The best performance was achieved for the summer acquisition of the boreal test site (r(2) = 0.86) and for the winter acquisition of the temperate test site (r(2) = 0.77). For the tropical test site, only a weak correlation (r(2) = similar to 0.50) could be established.