The relationship between intermittent coherence and precision of ISBAS InSAR ground motion velocities: ERS-1/2 case studies in the UK
The relationship between intermittent coherence and precision of ISBAS InSAR ground motion velocities: ERS-1/2 case studies in the UK
复制标题
DOI：
10.1016/j.rse.2017.05.016
复制
发表时间：
2017-12-01
影响因子：
13.5
通讯作者：
Sowter, Andrew
中科院分区：
文献类型：
--
作者：
Cigna, Francesca;Sowter, Andrew
利用差分干涉合成孔径雷达（干涉合成孔径雷达）和小基线子集（SBAS）方法对C波段卫星雷达图像进行非城市和半植被地区的地面运动信息反演具有挑战性。通过利用6个堆栈的中分辨率ERS-1/2 SAR图像在1992年和2000年之间在英国的四个感兴趣的地区，本文演示了最近开发的处理方法间歇SBAS（ISBAS）的性能。这种方法建立在传统的低分辨率SEAS方法的基础上，通过放松选择图像像素的方法来处理和考虑非城市目标的间歇性，能够将运动结果的覆盖范围扩展到整个土地覆盖类型，即使是那些通常不利于干涉合成孔径雷达的土地覆盖类型。平均而言，新的ISBAS实现提供了比SBAS多4到26倍的覆盖范围，地面运动解决方案的空间覆盖范围从SBAS的4-12%土地像素增加到ISBAS的39-99%。尽管仅依赖于小基线干涉图网络的时间子集，但间歇性相干像素显示的速度标准误差平均为0.8-1.4毫米/年，因此保持了亚毫米至毫米的精度。间歇性的相干性和估计的地面运动速度的标准误差之间的经验关系计算的六个数据集的每一个，并确认误差控制的独立观察用于每个图像像素提取ISBAS解决方案的数量。特别地，间歇相干像素的速度标准误差ε（vel）与所使用的最佳相干干涉图的数量n（i）的平方根成反比，并且可以对六个数据集的平均值建模为ε（vel）= 11 / root n（i）mm/year。所建立的经验关系还允许对n（i）的ISBAS阈值做出知情决定。这是通过在速度估计中设置最大可接受误差n（MAX），然后计算相应的最小值n（i）以接受将保证期望精度的间歇相干像素来实现的。在“大SAR数据”及其衍生的“大干涉合成孔径雷达数据”的当今时代，我们讨论了使用数千甚至数百万地面变形时间序列的巨大数据集的前景-如使用ISBAS产生的，特别关注大数据的准确性和需要在“大干涉合成孔径雷达数据”周期的质量评估检查点。(C)2017由Elsevier Inc.出版
Retrieving ground motion information for non-urban and semi-vegetated areas using differential Interferometric Synthetic Aperture Radar (InSAR) and Small Baseline Subset (SBAS) approaches with C-band satellite radar imagery is challenging due to temporal decorrelation. By exploiting six stacks of medium resolution ERS-1/2 SAR images acquired between 1992 and 2000 over four regions of interest in the UK, this paper demonstrates the performance of the recently developed processing method Intermittent SBAS (ISBAS). This approach builds upon the conventional low-resolution SEAS method and, by relaxing the approach to selecting image pixels to process and accounting for the intermittent nature of non-urban targets, is capable to extend the coverage of motion results across the full range of land cover types, even those typically unfavourable for InSAR. On average, the new ISBAS implementation provides 4 to 26 times more coverage than SBAS for the processed regions, with the spatial coverage of ground motion solutions increasing from only 4-12% land pixels with SBAS, to 39-99% with ISBAS. Despite relying only on temporal subsets of the networks of small baseline interferograms, intermittently coherent pixels show velocity standard errors of 0.8-1.4 mm/year on average, hence retain sub-millimetre to millimetre precision. The empirical relationships between intermittent coherence and standard errors in the estimated ground motion velocity are computed for each of the six datasets, and confirm that errors are controlled by the number of independent observations used for each image pixel to extract the ISBAS solution. In particular, velocity standard errors epsilon(vel) for the intermittently coherent pixels are inversely proportional to the square root of the number of best coherence interferograms used, n(i), and can be modelled as epsilon(vel) = 11 / root n(i) mm/year on average for the six datasets. The established empirical relationship also allows informed decisions on the ISBAS threshold for n(i) to be made. This is achieved by setting the maximum acceptable error in the velocity estimate epsilon(MAX), and then computing the corresponding minimum n(i) to accept an intermittently coherent pixel that will guarantee the desired precision. In the present era of 'big SAR data' and their derived 'big InSAR data', we discuss perspectives on the use of huge datasets of thousands or even millions of ground deformation time series - such as those produced using ISBAS, with a particular focus on the veracity of big data and the need for a quality assessment check-point in the 'big InSAR data' cycle. (C) 2017 Published by Elsevier Inc.