Several currently available broadband parameterizations for longwave and shortwave radiation have been combined to produce a computationally fast radiation parameterization that is well suited for atmospheric circulation models. The main features of the parameterization are the ability to include overlapping partly cloudy layers in the longwave, the use of a delta-Eddington technique to treat clouds in the shortwave, and a computational structure that is amenable to vectorization on supercomputers. Selected results of off-line one-dimensional computations using the code have been compared with more rigorous methods as part of an international intercomparison program and found to be quite accurate. INTRODUCTIONThe parameterization of radiation for atmospheric circulation models is currently under extensive study. The impetus for these efforts is primarily twofold. Tests have demonstrated the need for retaining high accuracy in the parameterization of radiation [Fels and Kaplan, 1975;Ramanathan et al., 1983], but in order to achieve sufficient accuracy, the computation time for the numerical simulation typically increases substantially. This is particularly true for terrestrial radiation, since the interaction of each model layer with every layer may have to be considered, and the absorbing properties of the atmospheric constituents vary considerably with wavelength. There is therefore a need for a radiation scheme that is computationally fast yet accurate. The parameterization discussed here is primarily intended for use in atmospheric circulation models, including those used for climate studies and for global and regional numerical weather prediction; and it is particularly suitable for the study of cloudiness, because it accommodates variable cloud fractions and cloud properties.A recent review of radiation parameterizations for climate models [Stephens, 1984] Since the host model for which the parameterization was initially designed (the UCLA/Goddard general circulation model) has a cloud generation scheme that allows cloud formation in every tropospheric layer, an important design goal is that the algorithm not lose its vectorizability in the presence of arbitrary cloudiness. We stress this seemingly technical consideration, since historically many of the algorithms for computing radiative fluxes have been nearly unvectorizable.For longwave absorption, we have used the broadband transmission approach of Chou [-1984

No abstract

No abstract

No abstract

The cloud liquid water feedback in climate models consists of the increase (decrease) in optical depth of clouds resulting from higher (lower) liquid water contents that might accompany tropospheric warming (cooling). The change in cloud liquid water with temperature is shown to depend on the rate of change of the slope of the moist adiabat with respect to temperature, and it is a strong function of temperature. The value of this rate of change in the tropics is about half that in mid and high latitudes and is much less than the value obtained by assuming that liquid water scales with the saturation mixing ratio.