基于变分原理的星载遥感器红外通道在轨光谱定标方法研究

The sensor spectral performance determines the outputs from scenes, as well as the calibration targets that are used to calculate the calibration coefficients in the radiometric calibration. It is significant for radiometric calibration, and high quality radiometric radiance requires accurate spectral and radiometric calibrations. However, the spectral response function (SRF) used in the radiometric calibration may not always accurately characterize the actual in-orbit instrument spectral performance, due to the possible inaccurate spectral calibration before satellite launch, and/or possible SRF change caused by the harsh space environment. Recent researches indicate that the real SRF may present a large shift of 10cm-1 from the laboratory measurement, thus lead to an approximate 2-3K bias in the calibrated radiance, and about 30% relative errors in the quantitative applications. As there's no spectral measurement assembly onboard for thermal infrared channels, on-orbit spectral response monitoring and correcting are diffcult. On requirements of quantitative remote sensing, an on-orbit spectral calibration method is brought out for thermal infrared channels. And it is developed to retrieve on-orbit spectral response functions using co-located high spectral resolution observation data based on one dimensional variational theory (1DVR), which can help to correction the spectral errors both in spectral drift and deformation. To make sure the feasibility of the numerical model and the reliability of the retrieval results, more details are carefully considered in this project, i.e., observation samples acquiring and radiance simulation, 1DVR numerical model construction and its solving method, as well as error control and so on. The implementation of this project can solve the in-orbit thermal infrared spectral calibration problem of broadband sensor, and help to improve the quality of satellite data and promote its quantitative application ability. It not only has high scientific research significance, and more practical value.

光谱响应特性是卫星资料应用中一个非常重要的参数，决定了遥感器能够探测到的特征辐射。研究表明，卫星发射后遥感器光谱响应可能发生显著变化，造成观测偏差可达2-3K,而定量应用误差可达30%,严重影响卫星数据质量及其定量应用效果。由于缺乏星上光谱测量装置，红外通道光谱响应无法实现在轨测量。项目从定量遥感的需求出发，基于一维变分原理，针对通道式遥感器研究其红外通道在轨光谱定标方法，发展基于高光谱同步观测资料的光谱响应函数的数值求解算法，实现对光谱漂移和形变误差的同时订正，并以风云卫星载荷为例实现算法应用和验证。通过对样本获取、模拟计算、解算模型构建与方程求解以及误差控制等几个方面的深入研究，确保数值求解算法的可行性以及解算结果的可靠性。本项目的实施可解决通道式遥感器无法实现在轨光谱定标的难题，有助于提高卫星数据质量以及定量产品精度，不仅具有较高科学研究意义，更具广泛业务推广价值。

光谱响应特性是卫星资料应用中一个非常重要的参数，决定了遥感器能够探测到的特征辐射。研究表明，卫星发射后遥感器光谱响应可能发生显著变化，造成观测偏差可达2-3K,而定量应用误差可达30%,严重影响卫星数据质量及其定量应用效果。由于缺乏星上光谱测量装置，红外通道光谱响应无法实现在轨测量。本项目从宽通道遥感器红外波段高精度定量遥感的需求出发，研究了基于一维变分原理的光谱响应函数在轨反演算法，实现了对光谱响应的在轨解析和光谱订正，并以风云卫星载荷为例实现算法应用和验证。.项目实施过程中，.1)发展了一套联合传统星星交叉和近重合观测的红外高光谱互定标新方法，解决传统互定标方法匹配样本种类单一且对光谱响应不敏感问题，大大降低互定标方法的不确定性。以此为基础形成行业标准，应用到风云卫星上，实现对多个光学载荷观测偏差的长时间评估；.2)发展了一套基于一维变分理论的光谱响应反演算法，研究构建了包含观测误差项和背景场约束的反演模型，评估了背景场和误差贡献项对反演结果的影响，解决数值求解对观测样本依赖性大、方程不适定最优解不稳定等系列问题，实现基于L-BFGS方法的快速求解。创新性的实现了光谱响应的在轨解析，解决了通道式遥感器无法实现在轨红外光谱定标的难题；.3)针对普遍存在的由于实验室大气吸收而产生发射前的光谱测量误差，发展了光谱响应污染误差分离和在轨订正算法。从机理上分析了发射前光谱定标误差的来源和特征，提出了对污染误差的直接修正方法，并在风云载荷上得到应用。.本项目研究成果有助于提升在轨定标精度，改进卫星资料数据质量，提高定量应用性能，对气象卫星资料在数值天气预报模式以及全球天气、气候和气候变化等综合领域中的定量应用都具有促进意义。该成果不仅具有较高科学研究意义，更具广泛业务推广价值。