PYDOME: python library for radiative transfer computations

Radiative transfer model simulations are an important component of atmospheric correction and trace gas retrieval algorithms. During the last few years at the German Aerospace Centre (DLR) we developed a software package PYDOME for the simulation of satellite-based measurements of reflected and scattered solar radiation in the ultraviolet and visible spectral ranges. The tool is based on a one-dimensional scalar radiative transfer model, which incorporates the exact solver, namely, the discrete ordinate method with matrix exponential (DOME) (Doicu&Trautmann, 2009, Efremenko et al. 2017), as well as several approximate solvers (the single scattering solver, the two-stream solver and others). The model is implemented for coupled atmosphere-ocean systems. It takes into account the change in the refraction index across the interface between two media and the wind-induced roughness of the water surface described by the Cox-Munk model. In addition, PYDOME can also treat a specular reflection from the surface (a “mirror” case). PYDOME is equipped with tools for modeling the atmospheric state. The interfaces to OPAC database (Hess et al, 1998) of aerosol models and molecular spectroscopic database HITRAN are provided. The computations of optical properties of the atmosphere are implemented via object-oriented programming paradigm. The atmospheric state can be constructed in an interactive mode by using the corresponding methods. For instance, the class Map contains a table of scattering, extinction and absorption coefficients together with expansion coefficients of phase functions for each atmospheric level. Molecular scattering, molecular absorption, aerosol and cloud models provide maps, which are then mixed into a new map corresponding to all atmospheric constituents. The optical parameters are derived from the resulting map. Besides the radiance field, the PYDOME tool provides the additional quantities, such as the path radiance, the spherical albedo, the direct transmittance, the diffuse transmittance, as well as the irradiance at the given altitude. These parameters can be stored in the look-up tables (either in ASCII or hdf5 formats) and then used in the atmospheric correction algorithms. PYDOME is implemented in python3 and relies on the NumPy library. The code is vectorized by means of the matrix algebra. Thus, most of the time consuming for-loops are eliminated. Consequently, the performance of PYDOME is comparable to the models implemented in FORTRAN or C. For the retrieval of atmospheric properties from remote sensing measurements by using the Gauss-Newton algorithm, PYDOME can be used for computing the Jacobian matrix comprising the partial derivatives of the radiances with respect to atmospheric parameters. The code is differentiated automatically by overloading NumPy functions. In the talk, we summarize the implementation issues and present comparisons of PYDOME results in several spectral ranges with those of the radiative transfer codes used worldwide by the scientific community (DISORT, LIDORT, LibRadTran, MODTRAN). References Doicu A., Trautmann T. Discrete ordinate method with matrix exponential for a pseudo-spherical atmosphere: Scalar case. Journal of quantitative spectroscopy and radiative transfer. 110(1), 146, 2009 Efremenko D.S., Molina Garcia V., Gimeno Garcia S., Doicu A. A review of the matrix-exponential formalism in radiative transfer. Journal of quantitative spectroscopy and radiative transfer. 196, 17, 2017. Hess M., Koepke P., Schult I. Optical Properties of Aerosols and Clouds: The Software Package OPAC. Bulletin of the American Meteorological Society. 76(5), 831, 1998