Computation of Solar Radiative Fluxes by 1D and 3D Methods Using Cloudy Atmospheres Inferred from A-train Satellite Data

This study used realistic representations of cloudy atmospheres to assess errors in solar flux estimates associated with 1D radiative transfer models. A scene construction algorithm, developed for the EarthCARE mission, was applied to CloudSat, CALIPSO and MODIS satellite data thus producing 3D cloudy atmospheres measuring 61 km wide by 14,000 km long at 1 km grid-spacing. Broadband solar fluxes and radiances were then computed by a Monte Carlo photon transfer model run in both full 3D and 1D independent column approximation modes. Results were averaged into 1,303 (50 km)2 domains. For domains with total cloud fractions Ac < 0.7 top-of-atmosphere (TOA) albedos tend to be largest for 3D transfer with differences increasing with solar zenith angle. Differences are largest for Ac > 0.7 and characterized by small bias yet large random errors. Regardless of Ac, differences between 3D and 1D transfer rarely exceed ±30 W m−2 for net TOA and surface fluxes and ±10 W m−2 for atmospheric absorption. Horizontal fluxes through domain sides depend on Ac with ∼20% of cases exceeding ±30 W m−2; the largest values occur for Ac > 0.7. Conversely, heating rate differences rarely exceed ±20%. As a cursory test of TOA radiative closure, fluxes produced by the 3D model were averaged up to (20 km)2 and compared to values measured by CERES. While relatively little attention was paid to optical properties of ice crystals and surfaces, and aerosols were neglected entirely, ∼30% of the differences between 3D model estimates and measurements fall within ±10 W m−2; this is the target agreement set for EarthCARE. This, coupled with the aforementioned comparison between 3D and 1D transfer, leads to the recommendation that EarthCARE employ a 3D transport model when attempting TOA radiative closure.