GF-1卫星PMS(GF-1 PMS)数据具有高空间分辨率、短重访周期的特点,可以在地表类型识别、参数提取中发挥重要作用。但由于缺少2.1μm附近的短波红外波段,使得气溶胶反演时地表反射率的精确确定非常困难,从而导致其高精度大气校正难以开展,限制了该数据的应用。本文提出了一种地表反射率数据支持的气溶胶反演方法,用于GF-1PMS数据的大气校正。其基本思想是:使用现有的地表反射率数据集为GF-1PMS数据提供地表反射率,用于确定GF-1PMS图像中浓密植被像元(DDV)的分布,基于确定的浓密植被像元反演气溶胶光学厚度(AOD),并用于大气校正。这里使用的地表反射率数据集为合成的无云MODIS地表反射率产品,对GF-1PMS数据做了空间尺度的转换。为降低两类数据配准误差对地表反射率确定的影响,提出了使用区域NDVI分布百分比匹配的方法,回避了像元的直接匹配,为GF-1PMS数据提供DDV的空间分布。为验证该方法的有效性,利用北京、太湖两个AERONET站点观测的气溶胶光学厚度对气溶胶反演结果进行精度验证,结果表明,气溶胶反演算法精度较高,稳定性较强。AOD反演结果应用于北京和敦煌地区的GF-1PMS数据大气校正,获得的地表反射率与地面实测的地表反射率的误差低于0.015,且大气校正后影像对比度明显提高。

GF-1 PMS data have an important role in land cover detection and quantitative information extraction because of its high spatial resolution and short revisit period. Land surface reflectance calculated by using atmospheric correction is a key step in the application of GF-1 PMS data in land surface parameter detection. However, atmospheric correction is relatively limited because of the lack of a shortwave infrared near 2.1μm, which is necessary for Aerosol Optical Depth(AOD) retrieval that requires correction. This study aims at developing a method to retrieve the AOD and obtain high-precision land surface reflectance from atmospheric correction.
A new algorithm for AOD retrieval from GF-1 PMS data was developed by introducing a prior surface reflectance data set, and the retrieved AOD was applied in atmospheric correction. In this method, Moderate-Resolution Imaging Spectro-radiometer surface reflectance outputs(MOD09) were used to obtain the land parameters required in AOD retrieval with Dense Dark Vegetation(DDV) method, with which the AOD was obtained from GF-1 PMS data. The land parameter obtained from the surface reflectance data set was the Normalized Difference Vegetation Index(NDVI), which was used to determine the pixel distribution of DDV. GF-1 PMS data were processed to match the selected MOD09 data in spatial resolution. The percentage matching strategy of regional NDVI distribution was adopted to reduce the effect of registration errors produced from pixel-by-pixel matching of the two types of data.
Aerosol Robotic Network ground measurements AOD in the Beijing and Taihu stations have been used to evaluate the accuracy of the derived AOD from GF-1 PMS. Results show that AOD retrieval with this method can obtain a high precision. The accuracy of land surface reflectance obtained from atmospheric correction of GF-1 PMS has been estimated by using the surface reflectance of typical types measured in the Beijing and Dunhuang areas. The uncertainty of land surface reflectance for different land covers is less than 0.015.
This work has improved the application level of GF-1 satellite, particularly its quantitative application, by developing a method to obtain the AOD and land surface reflectance with high precision and provide an example for the processing of the same type of satellite data. The method can also be used to achieve AOD inversion and atmospheric correction of other terrestrial observation satellite data.