SMAC: a simplified method for the atmospheric correction of satellite measurements in the solar spectrum

Abstract This paper describes a computationally fast and accurate technique for the atmospheric correction of satellite measurements in the solar spectrum. The main advantage of the method is that it is several hundred times faster than more detailed radiative transfer models like 5S and that it does not require precalculated look-up tables. The method is especially useful for correcting the huge amounts of data acquired by large-field-of-view high-repetitivity sensors, like the ones on board polar orbiting and geostationary meteorological satellites. The technique is based on a set of equations with coefficients which depend on the spectral band of the sensor. Semi-empirical formulations are used to describe the different interactions (absorption, scattering, etc.) of solar radiation with atmospheric constituents during its traverse through the atmosphere. Sensor specific coefficients of each equation are determined using a best fit technique against the computations of the 5S code (Simulation of Satellite Signal in the Solar Spectrum, Tanré et al. 1990). Other radiative transfer models could be used. Once coefficients for a specific spectral band are determined, the inputs of the model are vertically integrated gaseous contents, aerosol optical depth at 550 nm, geometric conditions and reflectance at the top of the atmosphere (TOA). TOA reflectances were calculated using our method and then compared to the TOA reflectances calculated by 5S for a wide range of gaseous and aerosol contents, illumination and observation conditions for various sensor spectral bands. In the case of NOAA-9 AVHRR visible data the maximum relative error is 2·35 per cent (i.e. 0·01 for a reflectance value of 0·4) and the corresponding rmse is 0·0018. For NOAA-9 AVHRR near-infrared, Meteosat-1 visible, Landsat-5 TM band 1 and Landsat-5 TM band 4 the maximum relative errors are 3·11, 4·0, 1·65 and 2·37per cent respectively. The corresponding values of the rmse are 0·0022, 0·0015, 0·0017 and 0·0012. The method can be used both in the direct and in the inverse mode, i.e., to compute TOA reflectance knowing the surface reflectance (e.g., for fast sensitivity studies), or conversely to retrieve surface reflectance from the TOA reflectance. It can easily be implemented in operational data preprocessing computer code, since only band specific coefficients need to be updated when new sensors are flown, while the routines remain the same.