Representative wavelengths absorption parameterization applied to satellite channels and spectral bands

Abstract

Accurate modeling of wavelength-integrated radiative quantities, e.g. integrated over a spectral band or an instrument channel response function, requires computations for a large number of wavelengths if the radiation is affected by gas absorption which typically comprises a complex line structure. In order to increase computational speed of modeling radiation in the Earth׳s atmosphere, we parameterized wavelength-integrals as weighted means over representative wavelengths. We parameterized spectral bands of different widths (1cm−1, 5cm−1, and 15cm−1) in the solar and thermal spectral range, as well as a number of instrument channels on the ADEOS, ALOS, EarthCARE, Envisat, ERS, Landsat, MSG, PARASOL, Proba, Sentinel, Seosat, and SPOT satellites. A root mean square relative deviation lower than 1% from a “training data set” was selected as the accuracy threshold for the parameterization of each band and channel. The training data set included high spectral resolution calculations of radiances at the top of atmosphere for a set of highly variable atmospheric states including clouds and aerosols. The gas absorption was calculated from the HITRAN 2004 spectroscopic data set and state-of-the-art continuum models using the ARTS radiative transfer model. Three representative wavelengths were required on average to fulfill the accuracy threshold. We implemented the parameterized spectral bands and satellite channels in the uvspec radiative transfer model which is part of the libRadtran software package. The parameterization data files, including the representative wavelengths and weights as well as lookup tables of absorption cross sections of various gases, are provided at the libRadtran webpage.In the paper we describe the parameterization approach and its application. We validate the approach by comparing modeling results of parameterized bands and channels with results from high spectral resolution calculations for atmospheric states that were not part of the training data set. Irradiances are not only compared at the top of atmosphere but also at the surface for which this parameterization approach was not optimized. It is found that the parameterized bands and channels provide a good compromise between computation time requirements and uncertainty for typical radiative transfer problems. In particular for satellite radiometer simulations the computation time requirement and the parameterization uncertainty is low. Band-integrated irradiances at any level as well as heating and cooling rates below 20km can also be modeled with low uncertainty.