Assessment of Spectral, Misregistration, and Spatial Uncertainties Inherent in the Cross-Calibration Study
Assessment of Spectral, Misregistration, and Spatial Uncertainties Inherent in the Cross-Calibration Study
复制标题
DOI：
10.1109/tgrs.2012.2228008
复制
发表时间：
2013-03-01
影响因子：
8.2
通讯作者：
Shrestha, Alok K.
中科院分区：
文献类型：
--
作者：
Chander, Gyanesh;Helder, Dennis L.;Shrestha, Alok K.
卫星传感器的交叉校准可以对不同地球观测系统获得的测量结果进行定量比较。交叉校准研究通常使用来自多个星载传感器的同时或近同时观测，通过回归分析建立逐带关系。本文的研究重点是评估交叉校准过程中固有的不确定性，包括不同光谱响应、光谱分辨率、光谱滤波器移位、几何错配和空间分辨率的贡献。利用环境卫星大气制图扫描成像吸收光谱仪和EO-1 Hyperion的高光谱数据，以及Landsat 7 Enhanced Thematic Mapper (TM) Plus和Terra中分辨率成像光谱辐射计传感器的相对光谱响应（RSRs）进行光谱不确定度研究。利用来自Landsat 5 TM的5种代表性土地覆盖类型（荒漠、牧场、草地、落叶林和针叶林）的数据进行几何错配和空间分辨率研究。利比亚4站点的光谱分辨率不确定度在0.25%以内，光谱滤波位移在2.5%以内，几何错配在0.35%以内，空间分辨率效应在0.1%以内。本文提出的1 -sigma不确定性是不相关的，因此，不确定性可以正交求和。此外，还建立了总体的总不确定度。总的来说，结果表明光谱不确定度比本文提出的其他不确定度更占优势。因此，需要对传感器RSR差异的影响进行量化和补偿，以避免交叉校准结果存在较大的不确定性。
Cross-calibration of satellite sensors permits the quantitative comparison of measurements obtained from different Earth Observing (EO) systems. Cross-calibration studies usually use simultaneous or near-simultaneous observations from several spaceborne sensors to develop band-by-band relationships through regression analysis. The investigation described in this paper focuses on evaluation of the uncertainties inherent in the cross-calibration process, including contributions due to different spectral responses, spectral resolution, spectral filter shift, geometric misregistrations, and spatial resolutions. The hyperspectral data from the Environmental Satellite SCanning Imaging Absorption SpectroMeter for Atmospheric CartograpHY and the EO-1 Hyperion, along with the relative spectral responses (RSRs) from the Landsat 7 Enhanced Thematic Mapper (TM) Plus and the Terra Moderate Resolution Imaging Spectroradiometer sensors, were used for the spectral uncertainty study. The data from Landsat 5 TM over five representative land cover types (desert, rangeland, grassland, deciduous forest, and coniferous forest) were used for the geometric misregistrations and spatial-resolution study. The spectral resolution uncertainty was found to be within 0.25%, spectral filter shift within 2.5%, geometric misregistrations within 0.35%, and spatial-resolution effects within 0.1% for the Libya 4 site. The one-sigma uncertainties presented in this paper are uncorrelated, and therefore, the uncertainties can be summed orthogonally. Furthermore, an overall total uncertainty was developed. In general, the results suggested that the spectral uncertainty is more dominant compared to other uncertainties presented in this paper. Therefore, the effect of the sensor RSR differences needs to be quantified and compensated to avoid large uncertainties in cross-calibration results.