首页 > , Vol. , Issue () : -
森林高度是反映森林资源数量和质量的重要参数，极化干涉合成孔径雷达(Polarimetric Synthetic Aperture Radar Interferometry，PolInSAR)技术在森林高度反演中极具潜力。由于森林散射特征受波长影响明显，由此引起的散射机理差异使得基于PolInSAR技术反演的森林高度结果具有很大的不确定性。为了定量化该不确定性的影响，本文以模拟森林场景为例，对PolInSAR技术森林高度反演中常用的4种方法——极化相位中心高度估测法、复相干相位中心差分法、复相干幅度反演法和相干幅度、相位联合反演法在常用的4个微波波段——P、L、C和X中的森林高度估测结果进行了分析，明确了匀质森林场景中，算法选择、波段选择引起的森林高度估测结果的不确定性。研究结果表明：在森林场景基本一致的情况下，估测算法的选择直接影响森林高度估测结果精度，其中基于干涉幅度信息的复相干幅度反演法在4个波段的估测结果中精度均最高，但各估测点的估测结果离散度及不确定度较大，P、L、C和X四个波段不确定度分别为0.198、0.134、0.293以及0.278；波长对4类估测方法估测结果的影响差异明显。复相干幅度反演法的反演结果几乎不受波长的影响，而相干幅度、相位联合反演法受波长影响明显，在长波长反演结果中精度较高，在短波长反演结果中精度显著降低。此外，以传统的HV极化代表冠层散射相位中心，HH-VV极化代表地表散射相位中心，采用复相干相位中心差分法会严重低估森林高度。不确定度具有波长和算法选择依赖性，短波长中复相干相位中心差分法估测结果不确定度最低，长波长中极化相位中心高度估测法不确定度最低，而复相干幅度反演法估测结果则在多个波段中的不确定度均最高。
Forest height is an important indicator for the quality and quantity of forest resource. Polarimetric interferometric synthetic aperture radar (PolInSAR) technology has been demonstrated and validated as a potential way for forest height inversion and mapping in the last years. Indeed, air and spaceborne PolInSAR data has been applied in a variety of temperate, boreal, and tropical forest. However, since the effects of different microwave frequencies on forest scattering mechanisms and the different theory base of each estimation algorithm, uncertainties are obvious in forest height estimation and mapping using PolInSAR data. In order to clarify the uncertainties resulted from different microwave frequencies and selected algorithms in the procedure of forest height inversion, this paper discusses critical effects of the selected 4 inversion algorithms and 4 typical microwave length, using a simulated forest scene, on the performance of forest height estimation in a comprehensive way. The 4 inversion algorithms include Polarimetric phase center height estimation method (PPC), complex coherence phase center differencing algorithm (CCPCD), Coherence amplitude inversion method (CAI) and Hybrid inversion method. The involved microwave length are P, L, C and X band. The results demonstrate that the effects of the wavelength and estimation algorithm are obvious on the performance of forest height estimation using PolInSAR data. First, the selected estimation algorithm directly affects the accuracy of forest height estimation results when the microwave wavelength is same. The estimated results from CAI and Hybrid inversion methods agree well with the average forest height in the simulate forest scene at the 4 selected microwave bands, but values of the latter are slightly lower than the former. In addition, the estimated result of the former algorithm is more discrete with the maximum standard deviation value of 7.64 m, while the maximum standard deviation of the latter is 5.02 m. The uncertainty ratios of P、L、C and X are 0.918, 0.134, 0.293 and 0.278, respectively. The results reveal great underestimation of CCPCD method, especially HV channel phase was selected as the canopy scattering phase center and HH-VV phase was selected as the surface scattering phase center to retrieve the forest height. Second, the wavelength effects depend on the selected estimation algorithms. It shows no obvious effect on CAI method. However, it shows great effects on the performance of Hybrid inversion method. The estimation results from Hybrid inversion method show a better performance at long wavelength (P- and L- band), but a worse performance at short wave length (C- and X- band). HV PPC method show great underestimation on forest height at P-band, but it performs well in forest height estimation in other bands, notwithstanding a slightly underestimation. The uncertainties of estimation results depend on wavelength and algorithm selections. Short wavelength with CCPCD method and long wavelength with PPC method show better performance and the lowest uncertainties on forest height estimation. While CAI method shows the highest uncertainties in forest height estimation at P、C and X band. The conclusions demonstrated in the manuscript are obtained through a study of simulation data, where the forest scene is homogeneous and ideal, since the real forest has strong heterogeneity due to the influence of environment, forest type, structure, etc., it is necessary to further consider the effects on the selected estimation algorithms and wavelength resulted from the different forest scattering mechanisms affected by forest type, structure and so on. We will consider the related effects in the future research.