Evaluating radiation parameterizations under quadrupled carbon dioxide

The opacity of the gaseous component of the atmosphere varies by orders of magnitude over very small spectral intervals and further depends on the distribution of temperature, pressure, and the concentrations of all radiatively-active gases. Atmospheric models of all kinds employ parameterizations, such as those based on k-distributions or exponential sum fits, to represent this enormous complexity with a relatively small set of spectral quadrature points. These parameterizations are trained against reference calculations encompassing some set of atmospheric conditions. That range may be wide or narrow, depending on the parameterization. Parameterizations show a surprisingly large range of absolute accuracies under present-day conditions and their performance degrades in atmospheric conditions far from the training set.

The accuracy of radiation parameterizations is not well-known under one widely-used perturbation: concentrations of carbon dioxide quadrupled from preindustrial concentration, which has been used to provide strong forcing to determine the climate sensitivity of coupled atmosphere-ocean models.

We examine the behavior of the radiation parameterizations of many climate models participating in the CMIP ensemble under 4xCO2 conditions. The parameterizations are compared to reference line-by-line models thought to be absolutely accurate to within 1 W/m2. We examine four diverse base atmospheres from the present-day drawn from the Continual Intercomparison of Radiation Codes for which the accuracy of each parameterization is already known. Fluxes are recomputed for each base case assuming only that CO2 concentrations have quadrupled. We investigate the accuracy of the parameterizations' forcing estimate relative to the reference model examining, among other things, the degree to which accuracy in present-day flux indicates accuracy in forcing from increased CO2, and the dependence of forcing accuracy on temperature and humidity structure.