A fast and accurate parametric model of estimating thermal infrared atmospheric parameters: a case study for LANDSAT 8 TIRS sensor

Abstract

ABSTRACT The traditional methods used in atmospheric correction depend on the empirical relationships or the atmospheric radiative transfer model (RTM). However, both of them have some deficiencies. The empirical method often depends highly on the training data; it would be applicable under certain conditions. In contrast, the RTM-based method is time consuming and has to run the code each time, which is also not an appropriate choice for operational estimation of the atmospheric parameters. In this study, a fast and accurate parametric model of estimating atmospheric parameters for Landsat 8 thermal infrared sensor was proposed to balance the accuracy and efficiency. The layer optical thickness for water vapour, water vapour continuum and other gases are well parameterized in the proposed parametric model, where the band coefficients can be easily determined by a least square fitting over MODTRAN simulations. The results show that the root mean squared error (RMSE) of the total transmission is 0.004 and the RMSE of for both the atmospheric upward and downward radiances are about 0.0004. Therefore, the proposed model could be used without the help of any atmospheric RTM to achieve the same accuracy as RTM does, but with less computational demand; that is, this model would have a better application market.