Research On Improving Gordon Atmospheric Correction Algorithm For Turbid Inland Waters

With the markedly complex composition of water constituents, Lake Taihu is a typical inland Case-II water. There are great differences in the optical properties between Lake Taihu and other coastal waters. Therefore, the traditional atmospheric correction algorithm for Case-II waters cannot be simply applied to this water. Taking Lake Taihu as the study area, this study was focused on how to improve the traditional Gordon atmospheric correction for HJ-1, ALOS multi-spectral data and MERIS hyper-spectral data. According to the analysis of the error of current Case-II water atmospheric correction methods and the characteristics of different image data, corresponding improved methods of atmospheric correction for this lake were proposed. The improved Gordon atmospheric correction algorithms were established for the hyper-spectral data and the MERIS data, based on the stable radiance of green band and the bands of the vapor and Oxygen absorption, respectively. Compared with the traditional Gordon algorithm, the improved algorithms could be more applicable to Lake Taihu with higher accuracy. Thus, this study could provide useful information for developing the atmospheric correction algorithm of other waters, and could provide the data basement for subsequent application such as the inversion of the water quality parameters using remote sensing images.The main conclusions of this study are as follows.First, the improved Gordon algorithm based on the green band is established for the multi-spectral data. Combining the idea of dark pixel atmospheric correction and formulates a clean-water pixel selection strategy, the input parameter of atmospheric correction parameter was calculated; then the water leaving radiance of clean water in green band was estimated according to the measured spectral data. From this way, it could reduce the errors or negative values caused by the assumptions that water leaving reflectance in short-wave infrared band is approximately0and the exponential extrapolation from the short-wave infrared bands.Second, the algorithm for correcting MERIS hyper-spectral data was proposed based on the Oxygen and water vapor absorption bands. MERIS L2p data was used to aid selecting relatively clean water pixels according to the band setting, and Level2product characteristics. On the other hand, using Oxygen and water vapor absorption bands to replace the traditional atmospheric correction bands could also reduce the errors caused by subjectively selection of clean water pixels and the absolute assumption of water leaving reflectance in short-wave infrared band is approximately0.Third, multiple aerosol scattering values in each band of five aerosol models were simulated using6S radiative transfer equation in a certain geometry, atmospheric model and visibility, by which, the input parameters of the atmospheric correction could be provided. Then, the determined exponential model of atmospheric correction parameter and the wavelength difference from6S simulation were applied to the study area to calculate the aerosol scattering.