顾及邻近像元效应的高空间分辨率卫星热红外地表温度反演方法研究

According to the previous literature, land surface temperature (LST) retrieval methods do not take the adjacency effect into account. However, with the development of high spatial resolution of thermal infrared sensor, the adjacency effect is becoming more significant and remarkable. Thus, it is necessary to consider how to tackle with the adjacency effect to improve the accuracy of LST retrieval methods, and this is a bottleneck problem. As to this problem, this study will develop a land surface temperature retrieval method based on correction of adjacency effect using the high spatial resolution and multi-spectral thermal infrared data from satellite GF-5. At first, with the help of thermal infrared radiative transfer theory, the remote sensing physics and mathematical formula derivation, the point spread function (PSF) is modified and its application spectral range is expanded from visible/near infrared (VNIR) region to thermal infrared (TIR) region. Secondly, according to the observation geometric relationship between the sensor, target pixel and the adjacent pixels, the adjacent effect is estimated based on the extended PSF module. Thirdly, a high spatial resolution surface temperature inversion method considering the effect of adjacent pixel is established by using the medium and low spatial resolution surface temperature inversion algorithm. At last, according to the spectral response function of GF-5, an inversion algorithm of land surface temperature is developed based on the land surface temperature retrieval method of the third step, then the land surface temperature products from GF-5 is validated through the field observation experiments. This study will provide a new land surface temperature retrieval approach for high spatial resolution satellite GF-5, and will provide land surface temperature products with high accuracy from GF-5 for some relating to application areas.

现有地表温度反演算法均未考虑邻近像元效应的影响。但随着热红外传感器空间分辨率的提高，邻近像元效应越来越显著，如何消除邻近像元效应的影响，进一步提高地表温度的反演精度是目前亟需解决的一个瓶颈问题。本研究拟利用我国自主发射的GF-5卫星数据所同时具备的高空间分辨率和多光谱特性，开展考虑邻近像元效应影响的地表温度反演方法研究。首先，根据遥感物理原理和数理公式推导，将可见光-近红外波段的点扩散函数(PSF)模型拓展至热红外波段；其次，利用拓展的PSF模型，根据传感器、目标和邻近像元所构成的几何关系，估算邻近像元效应的影响；再次，基于热辐射传输理论，借鉴中低空间分辨率地表温度反演算法的发展思路，建立考虑邻近像元效应影响的高空间分辨率地表温度反演方法；最后，发展适合GF-5的地表温度反演算法并验证反演结果。本研究可为即将发射的GF-5卫星提供高精度地表温度反演算法，为相关领域提供GF-5地表温度产品。

地表温度（LST）是陆地表层系统过程中一个重要的特征物理量，是地-气之间能量交换与相互作用的关键参数，准确获取高精度的高空间分辨率卫星遥感温度产品有助于理解地气之间的能量交换过程。在前期城市热环境研究的基础上，本项目针对目前高空间分辨率热红外遥感反演温度模型精度较低的问题，并考虑到目前还没有一个适合GF-5 卫星的地表温度反演模型，为满足高空间分辨率地表温度反演模型精度达到1K的科学需求，本研究提出了考虑邻近像元效应的地表温度反演模型。首先，高空间分辨率的城市下垫面精细分类和光谱不确定性处理，为提取城市下垫面几何信息和准确构建目标像元、邻近像元、传感器三者之间的关系奠定了基础。其次，基于辐射传输理论和物理观测几何，定量模拟的邻近像元效应的变化及其与各影响因子的关系。最后，将邻近像元效应定量变化模型与传统地表温度反演模型进行耦合，构建高分热红外遥感地表温度反演模型，进而获取高精度的地表温度产品。.在完成本项目预定的研究目标基础上，项目团队成员继续深入思考了高光谱高空间分辨率地表温度反演与农田干旱监测的问题。针对目前基于温度和比辐射率进行农田干旱监测中存在的问题，本研究提出了基于高光谱高空间分辨率热红外遥感数据反演农田中相关作物的温度和比辐射率，构建温度和比辐射率与农田土壤水之间的定量变化关系模型，从而直接监测农田旱情，同时设计了相关农田干旱热红外遥感监测实验，基于HyperCAM热红外成像光谱仪开展了基于温度/比辐射率分离的相关农作物干旱监测研究。.项目实施以来，通过一系列技术攻关，完成了城市复杂下垫面的精细分类和结构信息提取研究，完成了邻近像元效应的定量变化研究并构建了高精度的地表温度反演模型，并基于实验室高光谱热红外成像光谱仪，额外开展了高光谱高空间分辨率地表温度与比辐射率分离方法研究并将其应用于农田干旱监测。依托本项目的研究，第一项目标注SCI论文4篇，已申请专利1项。