【作者】 张璐；

【导师】 高炜； 施润和；

【作者基本信息】 华东师范大学 ， 地图学与地理信息系统， 2016， 硕士

【摘要】 随着工业化和城市化进程的加速,空气污染越来越严重,引起社会广泛关注。气溶胶光学厚度(Aerosol Optical Depth, AOD)是大气环境的重要评价指标,获取其空间分布信息对于大气环境监测、污染治理具有重要意义。卫星遥感可获得空间覆盖范围广、时间尺度上连续的信息,已成为气溶胶光学厚度监测的重要手段。近年来我国遥感卫星发展迅速,相继有多颗搭载了中高分辨率传感器的国产资源环境遥感卫星发射升空。研究如何利用这些国产遥感卫星数据进行AOD反演并得到精度较高的反演结果,具有重要的现实意义和探索价值。本文以北京市为研究区,运用深蓝算法针对HJ-1A/B/CCD、GF-1/WFV和CBERS-04/WFI等国产卫星遥感数据开展了AOD的反演研究。AOD反演前的预处理研究是本文工作重点之一。国产遥感卫星数据用于AOD反演时会遇到参数不全的问题。CBERS-04/WFI相机尚未提供大气层外波段平均太阳光谱辐照度数据(Mean solar exoatmospheric irradiances over band, ESUNb),而这是AOD反演时所需的重要参数。本文基于CBERS-04/WFI的波谱响应函数和9条常用太阳光谱曲线数据开展了CBERS-04/WFI相机ESUNb的计算,依据已有官方ESUNb值的中等空间分辨率传感器确定了适合于计算CBERS-04/WFI传感器ESUNb值的太阳光谱曲线,计算得到CBERS-04/WFI相机的ESUNb值。由于这3种遥感传感器均为宽幅面传感器,同一景影像的观测角度数据差距较大,计算AOD时不能将整幅影像的观测角度看作定值。GF-1/WFV数据尚未提供与遥感影像相对应的观测角度数据,因此根据GF-14台WFV相机成像的几何位置关系和影像的中心角度数据计算了逐像元的卫星观测角度数据。基于深蓝算法反演AOD的一个关键问题是构建基于国产卫星传感器蓝波段的地表反射率数据库。本文利用MODIS 8天合成的地表反射率产品MOD09A1来构建适合HJ-1A/B/CCD、GF-1/WFV和CBERS-04/WFI的地表反射率数据库。考虑到CBERS-04/WFI、GF-1/WFV、HJ-1A/B CCD1和MODIS传感器的光谱响应函数有一定的差异,需要进行波段转换修正。从ENVI标准波谱数据库提取了22种典型地物的地表反射率曲线,结合这几种传感器蓝波段波谱响应函数计算出各类地物在蓝波段的地表反射率,通过回归拟合建立了国产卫星传感器和MODIS蓝波段地表反射率的线性转换关系,并应用于MODIS地表反射率数据从而构建基于这3种国产卫星传感器蓝波段的地表反射率数据库。基于6S辐射传输模型循环迭代模拟,构建了针对CBERS-04/WFI、GF-1/WFV、HJ-1A/B CCD1这3种传感器的多维查找表,据此建立了观测角度数据与大气参数之间的关系。在此基础上,基于查找表运用深蓝算法进行逐像元计算,得到北京地区多个时相的AOD。基于AERONET观测数据对AOD遥感反演结果进行精度验证,结果表明,CBERS-04/WFI和HJ-1A/B CCD1的AOD反演结果与地基数据的相关性显著(P<0.01),相关系数分别为0.941和0.919；GF-1/WFV数据匹配后的验证数据少,反演结果与地基验证数据的平均绝对误差为0.07。此外,将AOD反演结果与MOD气溶胶产品MOD04进行了对比,表明反演得到的AOD与MODIS AOD之间有显著的相关性(P<0.01)。其中,基于HJ-1A/B CCD数据反演的AOD与MOD04产品的相关性最高(R=0.931),CBERS-04/WFI次之(R=0.863),GF-1/WFV数据的相关性相对较低,为0.590。验证结果表明,本文基于多种国产遥感数据的AOD反演结果较合理。更多还原

【Abstract】 With the rapid industrialization and urbanization, air pollution becomes an increasingly serious problem and has attracted special attention throughout the society. Aerosol optical depth (AOD) is a key indicator of the atmospheric environment. It has great significance to obtain the spatial distribution information of AOD for atmosphere monitoring and pollution controlling. Satellite remote sensing has been an improtant method of AOD monitoring, which can provide spatially continuous information over large areas and frequent time intervals. With the rapid development of China’s remote sensing satellite, several domestic satellites with medium or high resolution sensors were launched in recent years. It is of great significance to carry out research on retrieving AOD by these domestic satellite data. In this paper, HJ-1A/B/CCD, GF-1/WFV and CBERS-04/WFI were employed to retrieve AOD based on deep blue algorithm over Beijing.Data pre-processing is an important part of this paper, including pre-processing for remote sensing data and calculating necessary parameters. The parameters of domestic satellite data are not enough for AOD retrieval. The ESUNb (Mean solar exoatmospheric irradiances over band) for CBERS-04/WFI is not provided yet and need to be calculated before retrieving AOD. Extraterrestrial Solar Spectral Irradiance and Spectral Response Function (SRF) are necessary to calculate ESUNb. Nine released solar spectra and a number of medium spatial resolution sensors whose ESUNb had been officially released were chosen to determine the optimal solar spectra for medium resolution sensors. Then the ESUNb for CBERS-04/WFI were calculated. The data pre-processing for remote sensing data including radiometric calibration, re-projecting, cloud removing and computing viewing geometry et al. For CBERS-04/WFV and HJ-1A/B CCD1 data, the viewing zenith and viewing azimuth were provided officially. However, GF-1/WFV has not provide viewing geometry files. By analysing the geometry information of the four WFV sensors and the viewing angle for central pixels which can be obtain from the header files of the image, the pixel-by-pixel viewing angles were calculated.One of the key steps in retrieving AOD by deep blue algorithm is generating the land surface reflectance database for domestic sensors. There are some differences between the spectral responses of CBERS-04/WFV, GF-1/WFV, HJ-1A/B CCD1 and MODIS.22 spectral data were selected from ENVI standard spectral library to derive the linear relationships between domestic sensors and MODIS. Based on the linear equations, the MODIS land surface reflectance was converted to domestic sensors. Besides, the lookup Tables (LUT) were generated using Second Simulation of the Satellite Signal in the Solar Spectrum (6S) model for each sensors. Based on the LUTs and land surface reflectance database, the AODs were retrieved using deep blue algorithm.The accuracy of retrieved AOD was assessed by AERONET ground observed data. Result indicated that the AOD result of CBERS-04/WFI and HJ-1/CCD1 showed significant correlations with AERONET data (P<0.01), with correlation coefficients of 0.941 and 0.919 respectively. There was less validation samples of GF-1/WFV, which gave a MAE (mean absolute error) value of 0.07. In addition, the AOD results were also compared with MODIS AOD product (MOD04). Result suggested notable correlations between retrieved AOD of three domestic sensors and MODIS AOD (P<0.01). The highest correlation was observed between the retrieved AOD of HJ-1/CCD and MODIS AOD, with the correlation coefficient of 0.931. CBERS-04/WFI achieved the second higher correlation coefficient of 0.863. GF-1/WFV showed the lowest correlation coefficient of 0.590. The satisfactory validation results demonstrated the validity and reliability of the retrieved AOD from domestic satellite data in this paper.更多还原