热红外甲烷廓线物理反演中稳定性机理分析和模型研究

项目名称： 热红外甲烷廓线物理反演中稳定性机理分析和模型研究

项目编号： No.41201353

项目类型： 青年科学基金项目

立项/批准年度： 2013

项目学科： 地理学

项目作者： 张莹

作者单位： 中国科学院遥感与数字地球研究所

项目金额： 24万元

中文摘要： 甲烷是一种非常重要的温室气体，星载热红外高光谱传感器能够快速、经济、可重复地获取宏观尺度上甲烷分布信息，为研究全球气候变化提供了很好的支持。经典的大气甲烷廓线物理反演模型由于平滑误差、输入温湿廓线误差和观测误差等因素的影响导致其迭代速度慢甚至不收敛，得不到稳定的解。通过Levenberg-Marquardt（LM）模型可以对经典的物理反演模型进行修正，增加解的稳定性，此时LM模型中的两个参数成为快速、稳定求解的关键。本研究通过定量分析经典的物理反演模型求解不稳定的形成机理，然后据此建立模型求解LM模型中的参数，并且选择合适的反演通道，以提高甲烷廓线物理反演模型的收敛速度、反演精度和稳定性。课题以中国为研究区域，基于Atmospheric Infrared Sounder (AIRS) 卫星观测资料进行反演试验，并且利用地基观测数据、全球大气成分本底站数据、飞机观测资料进行结果验证和误差分析

中文关键词： 热红外遥感；温室气体；大气层析遥感；高光谱遥感；

英文摘要： Methane is one of the most abundant greenhouse gases in the troposphere. Recent missions of hyper-spectral infrared spaceborne observations provide a promising way to understand the relative strengths of the methane sources and sinks due to large spatial and temporal coverage. The classic optimal estimation (OE) model is currently adopted as the core algorithm in the numerical weather analysis. However, in the methane physical retrieval, the error from model smoothing, the error of fitted temperature and humidity and temperature profile, the measurement noise will slow down the convergent rate even lead to non-convergence. Levenberg-Marquardt (LM) model adds a correction term in the OE model so that it can interpolate between the Gauss-Newton algorithm and the method of gradient descent to stabilize the OE solution. And the correction term, expressed by the product of a positive definite matrix and a Lagrange factor, is the key issue in the LM model. In this study, a quantitative error analysis of classic OE model in the methane retrieval is performed to study the mechanism of instability. And thus build the model to solve the parameters in LM model to improve the convergent rate,accuracy and stability. After selecting the fitting spectral range, the retrieval experiment is applied to Atmospheric Infrared Sounde

英文关键词： Infrared Remote Sensing；Sensing;Greenhouse Gases；atmospheric profile sounding；Hyper-spectral remote sensing；