基于Landsat8 TIRS数据的海表温度反演算法对比  

Comparison of Sea Surface Temperature Inversion Algorithms Based on Landsat8 TIRS Data

作　　者：朱博 陈正华 陆永强 黄荣永 ZHU Bo;CHEN Zhenghua;LU Yongqiang;HUANG Rongyong(School of Resources,Environment and Materials,Guangxi University,Nanning 530004,China;School of Marine Sciences,Guangxi University,Nanning 530004,China)

机构地区：[1]广西大学资源环境与材料学院,广西南宁530004 [2]广西大学海洋学院,广西南宁530004

出　　处：《海洋技术学报》2023年第3期9-19,共11页Journal of Ocean Technology

基　　金：广西自然科学基金资助项目(2020GXNSFAA297245);广西南海珊瑚礁研究重点实验室自主基金资助项目(GXLSCRSCS2021102)。

摘　　要：海表温度(Sea Surface Temperature,SST)是研究气候变化的重要参数,具有重要研究意义。为了选出适用于近海海域的最优温度反演算法,本文基于Landsat8卫星遥感数据,以北部湾海域为研究区,对比分析了包括辐射方程传输法(Radiative Transfer Model,RTM)、单窗算法(Mono-window model,MW)、单通道算法(Single-channel model,SC)、线性劈窗算法(Linear Split-window Algorithm,SW_(1))和非线性劈窗算法(Non-linear Split-window Algorithm,SW_(2))在内的海表温度反演算法的反演精度并进行了敏感性分析。同时本文利用劈窗协方差-方差比值法(Split Window Covariance-variance Ratio,SWCVR)来反演大气水汽含量数据,减少了温度反演过程中对外部数据的依赖,研究结果表明:基于Landsat8 TIRS(Thermal Infrared Sensor)数据的SWCVR法进行大气水汽含量反演的效果较好,误差约在0.5 g/cm2;与实测海温数据相比SW_(2)与SC算法精度较高,误差约为0.6 K;RTM与SW_(1)算法次之,误差约为1.6 K与1.9 K;MW算法精度较低,误差约为2.5 K;与AVHRR(Advanced Very High Resolution Radiometer)SST产品进行相比两种劈窗算法的精度较高,误差约为1 K和1.3 K,SC算法精度较劈窗算法略低,误差约为1.4 K左右,RTM与MW算法精度较低,误差约为2 K与3 K;SW_(2)算法对参数的敏感性最低,其次是SC算法、SW_(1)算法与MW算法,RTM算法的敏感性最高。Sea surface temperature is an important parameter in the study of climate change and has great research significance.In order to select the suitable temperature inversion algorithm for the study of offshore waters,this paper compares and analyzes the inversion of sea surface temperature inversion algorithms including the Radiative Transfer Model(RTM),the Mono-window Model(MW),the Single Channel model(SC),the Linear Split-window Model(SW_(1))and the Non-linear Split-window Model(SW_(2))based on Landsat8 satellite remote sensing data,using the Beibu Gulf waters as the study area.The sensitivity analysis was also performed.The Split-window Covariance-covariance Ratio method(SWCVR)is also used to invert the atmospheric water vapour content data,reducing the dependence on external data in the temperature inversion process.The results show that:1The SWCVR method based on Landsat8 Thermal Infrared Sensor(TIRS)data for atmospheric water vapor content inversion is better,with an error of about 0.5 g/cm2;the accuracy of the SW_(2) and SC algorithms is higher compared with the measured SST data,with an error of about 0.6 K;the RTM and SW_(1) algorithms are second,with an error of about 1.6 K and 1.9 K;the MW algorithm is less accurate,with an error of about 2.5 K;the accuracy of the two Split-window Algorithms is higher compared with the Advanced Very High Resolution Radiometer(AVHRR)SST product,with an error of about 1 K and 1.3 K;the accuracy of SC algorithm is slightly lower than that of Split-window algorithm,with errors of about 1.4 K,and the RTM and MW algorithms are less accurate,with errors of about 2 K and 3 K;The SW_(2) algorithm has the lowest sensitivity to parameters followed by the SC algorithm,SW_(1) algorithm and MW algorithm,and the RTM algorithm has the highest sensitivity.

关 键 词：海表温度 Landsat8 反演算法 大气水汽含量 

分 类 号：TP701[自动化与计算机技术—检测技术与自动化装置]