热红外遥感图像中有云像元的地表温度估算方法研究

Cloud is an essential obstacle for remote sensing (RS) application to agriculture and land surface temperature (LST) retrieval from RS data. Two kinds of cloud can be identified in RS images: occlusion and cover. Cloud occlusion is due to difference between sensing direction and solar radiation. In cloud occlusion, the ground was not visually seen by remote sensor but still was under sunshine. However, in cloud cover, the glound can not only be seen by the sensor but also blocked from sunshine. This is the general case in low spatial resolution images such as MODIS frequently used in agricultural monitoring. In this study, we intend to develop an applicable approach to esitmate LST for the cloud-covered pixels in MODIS images. Impact of cloud cover on LST change relies on the fact that cloud reduces or blocks the solar radiance reaching the ground. However the machenism behind the impact is very complex. LST of cloud-covered pixels changes not only with the radiance but also with the characteristics of ground surface and cloud cover itsself such as form, size,thickness, and period. In order to investigate the impact, we classifies the pixels into 3 typical surfaces: natural, urban and water. Natural surface can be seen as composed of vegetation and bare soil in different fractions. The same is for urban urface: vegetation and buiding. Thus we have 4 typical surfaces: vegetation, bare soil, building and water. Surface energy balance model will be used to simulate the change of LST with the radiance reaching the typical surfaces of vegetation, bare soil and building under various conditions. On the basis of this LST-radiance relationship, we are able to establish an approach to estimate the LST for the cloud-covered pixels in MODIS images. This can be done through the estimation of radiance for the pixels using the scatering models and the composition of the pixels from the 4 typical surfaces. Field observations and experiments will be done in the study to provide required data sets for the simulation and validation of the relationship. We intend to focus our study in the 3 important agricultural regions of China: North China, Middle Yangtze Basin and South China. The available 4 agricultural experiment stations (Beijing Shunyi, Shandong Dezhou, Jianxi Ji'an and Guangdong Jiangmen) locating in the 3 regions will be used for field observation and experiments.Such parameters as LST, atmospheric scattering radiance, global radiance, cloud condition, crop growing condition, air temperature and humidity, soil temperature and moisture, wind speed and sunshine hour will be measured in the stations. After proofing its applicability, we intend to program the approach for fast estimation of LST for entire China from MODIS data, which is required by agricultural monitoring.

云是农业遥感应用和热红外图像地表温度反演的重要障碍。热红外图像中的云现象可分为：遮挡和覆盖。云遮挡指云遮挡了遥感观测而没有阻挡太阳照射，常见于高分辨率图像中。云覆盖不仅是遮挡遥感观测而且也阻挡了太阳照射，常见于MODIS等数据中。本项目将以农业遥感常用的MODIS数据为例，重点研究云覆盖像元的地表温度估算方法。云覆盖对地表温度变化的影响，主要因为云减弱了到达地表面的大气辐射强度，但云覆盖像元的地表温度变化，还因地表类型、云覆盖特征(形状、大小、时间、厚度等)、地理纬度等不同而变得复杂。本研究首先根据地表热量平衡原理，模拟典型地表类型的地表温度随大气辐射强度的变化规律，然后根据像元尺度地表构成，通过估算云像元的大气辐射强度，建立云覆盖像元地表温度估计方法。在华北、长江中下游和华南地区野外试验站开展云覆盖、大气辐射和地表温度变化的观测验证，为农业遥感应用提供技术支撑。

云像元地表温度（LST）遥感反演，是热红外遥感研究的前沿学术难题。本项目的研究目标是，阐明云对LST变化的作用过程和影响机制，分析有云情况下不同类型地表(植被和裸地)的LST变化规律，建立云像元LST遥感反演方法。经过4年的努力，按照计划完成了预期研究目标。基于野外观测试验和遥感图像分析，阐明了云对LST的影响机理，分析了云对LST变化的影响程度，利用地表能量平衡模型模拟了不同云覆盖条件下LST变化过程，提出了云像元的LST遥感反演无窗算法，研发了云像元地表温度遥感反演系统，突破了云对LST度遥感反演的障碍。项目成果LST遥感反演无窗算法，具有重要的理论价值和现实意义，应用前景广阔。在研究中，还深入地研究了地表热辐射方向性问题，提出了LST图像像元分解方法，改进了Landsat8数据的LST遥感反演算法。目前共发表13篇论文(SCI 7篇)，专著1部，申请专利1个，培养毕业了3名博士研究生和1名硕士研究生。