Title
On the errors incurred by solar radiation calculations in Earth System models
Authors
Michael Prather and Juno Hsu (junoh@uci.edu)
Abstract
Sunlight drives the Earth’s weather, climate, chemistry, and biosphere. Recent efforts to improve solar heating codes in climate models focused on more accurate treatment of the absorption spectrum or fractional clouds. A mostly forgotten assumption in climate models is that of a flat Earth atmosphere. Spherical atmospheres intercept 2.5 W⋅m−2 more sunlight and heat the climate by an additional 1.5 W⋅m−2 globally. Such a systematic shift, being comparable to the radiative forcing change from preindustrial to present, is likely to produce a discernible climate shift that would alter a model’s skill in simulating current climate. Regional heating errors, particularly at high latitudes, are several times larger. Unlike flat atmospheres, constituents in a spherical atmosphere, such as clouds and aerosols, alter the total amount of energy received by the Earth. To calculate the net cooling of aerosols in a spherical framework, one must count the increases in both incident and reflected sunlight, thus reducing the aerosol effect by 10 to 14% relative to using just the increase in reflected. Simple fixes to the current flat Earth climate models can correct much of this oversight, although some inconsistencies will remain.
In addition to the errors caused by ignoring the spherical geometry, different classes of errors from various approximations used in radiative transfer code of the climate models (CLIRAD, RRTMG, LLNL) are also quantified. This includes the use of broad wavelength bins to integrate over spectral features; multiple-scattering approximations that alter the scattering phase function for clouds, aerosols, and gases; treatment of fractional cloud cover including cloud overlap; and the approximation of ocean surface albedo by a constant. These approximations, coded as options in the Solar-J code can be further tested in E3SM to investigate how they impact long climate simulations.