Refinement of the geophysical fluid dynamics laboratory solar benchmark computations and an improved parameterization for climate models

[1] A recent intercomparison study of solar radiative transfer models has revealed a notable difference (5%) in the total spectrum column absorptance, for a specified clear-sky atmospheric profile, between two principal line-by-line benchmark results (namely, the Geophysical Fluid Dynamics Laboratory (GFDL) and the Atmospheric and Environmental Research, Inc. models). We resolve this discrepancy by performing a series of "benchmark" computations which show that the water vapor continuum formulation, spectral line information, and spectral distribution of the solar irradiance at the top of the atmosphere are key factors. Accounting for these considerations reduces the difference between the two benchmarks to less than 1%. The analysis establishes a high level of confidence in the use of benchmark calculations for developing and testing solar radiation parameterizations in weather and climate models. The magnitude of the change in absorption in the newer GFDL benchmark computations, associated with the use of a more recent spectral line catalog and inclusion of the water vapor continuum, has also necessitated revising the solar parameterization used in the operational GFDL general circulation model (GCM). When compared with the newer reference computation, the older parameterization shows an underestimate of the clear-sky heating rate throughout the atmosphere, with the error in the atmospheric solar absorbed flux being about 20 W m^-2 for a midlatitude summer atmosphere and overhead Sun. In contrast, the new parameterization improves the representation of the solar absorption and reduces the bias to about 5 W m^-2. Another important feature of the new parameterization is a nearly 50% reduction in the number of pseudomonochromatic columnar calculations compared to the older formulation, with only relatively small increases in the biases in absorption for cloudy layers. This yields a reduction of about 10% in the GCM computational time. The effect of the new parameterization on the simulated temperature in the new operational GFDL climate GCM is also examined. There is an increased solar heating; this yields temperature increases exceeding 1 K in the lower stratosphere.