For this study we apply the RFMIP Aerosol-IRF protocol to both GFDL’s AM4 and NCAR’s CESM 1.2.2. Our results demonstrate errors in fluxes as high as 30%. We find very different spatial patterns of error between the two models. This can partially be attributed to the different aerosol properties of the two GCMs. However, we also found choices made in the radiative transfer solution to have a significant impact. The usage of a two stream solver in both models resulted in an underestimate of atmospheric absorption and attenuation of flux at the surface. Perhaps more surprisingly, we found that the width of the spectral bands also played an important role. At visible and UV wavelengths CESM, which uses the RRTMG radiation scheme, has far fewer bands than the GFDL radiation scheme (5 vs. 13). We show that these broader bands of RRTMG result in an overestimate of absorption by aerosol. By interpolating the CESM aerosol properties to the narrower GFDL band structure it was possible to almost completely remove this error. This implies that a finer band structure in the UV and visible can greatly improve the representation of aerosol radiative transfer in GCMs.
The two stream and band width errors have a notable geographic variation and exhibit a dependence on both the aerosol specie type and background environment. Therefore, highlighting that errors in aerosol radiative transfer for a particular GCM are uniquely tied to its aerosol distribution and spectral properties.