Study On Atmospheric Correction Methods For MERIS Data Over Taihu Lake

Due to the influence of atmospheric absorption and scattering,only about 10% of the global solar radiation received by the satellite sensors can reflect the real water-leaving radiance.Therefore,atmospheric correction is particularly important for the study of quantitative remote sensing of water color,and how to effectively remove atmospheric interference directly affects the accuracy of the subsequent inversion of water color parameters.In this paper,Atmospheric correction methods of hperspectral remote sensing images(MERIS)for lake Taihu case-2 water are researched,based on the further study of the existing atmospheric correction principles and methods.Based on the 6S atmospheric correction model and the actual aerosol data in lake Taihu,the variation characteristics of the aerosol optical thickness are analyzed and the effect of standard aerosol model on atmospheric correction is discussed.Through the user-defined aerosol interface in the 6S model,the training dataset is established,and an atmospheric correction method based on neural network technology is built.Then,atmospheric correction for hperspectral remote sensing images can be achieved.The main research contents and results were summarized as follows:(1)According to the continuous aerosol observation data in lake Taihu from 2006 to 2010 provided by AERONET web site,the variation characteristics of the aerosol optical thickness are analysed to provide prior knowledge for determining aerosol optical thickness in subsequent method studies.To validate the impacts of aerosol model on Case-2 water atmospheric correction of lake Taihu,the standard aerosol model and the measured aerosol parameters are used for MERIS image atmospheric correction by input into the 6S atmospheric radiative transfer model.Experimental results show that the wrong assumption of aerosol model will cause large errors in the atmospheric correction of remote sensing images.(2)Through 6S atmospheric correction model,a 13-band training dataset of MERIS images is simulated can be obtained with different geometrical parameters,different aerosol models and different aerosol optical thicknesses,and the BP neural network to generate the atmospheric correction inversion model.The water-leaving reflectance can be extracted with the input of the geometric information and the satellite radiance of 13-band of MERIS images after preprocessing,which doesn’t need the synchronous aerosol parameter.(3)Evaluated the reasonableness of the atmospheric correction method in this paper by compare with 6S atmospheric correction model and C2 R Case-2 water atmospheric correction algorithm that come with BEAM using the measured data collected in lake Taihu.The results show that in the 6S model,the “continental Model” aerosol model is used for atmospheric correction,and the correction result is seriously higher than the measured value.The average relative error and standard deviation are very large.Although the C2 R algorithm has the best stability,the average relative error were up to 70% or more.That means this method doesn’t suitable in lake Taihu.The method proposed in this paper better than the other two methods,and the water-leaving reflectance in numerical and morphology are close to the measured.