Uncertainty of SW Cloud Radiative Effect in Atmospheric Models Due to the Parameterization of Liquid Cloud Optical Properties

Clouds are largely responsible for the spread of climate models predictions. Here we focus on the uncertainties in cloud shortwave radiative effect due to the parameterization of liquid cloud single scattering properties (SSPs) from liquid water content (LWC) and droplet number concentration (N), named parameterization of cloud optical properties. Uncertainties arise from not accounting for the droplet size distribution (DSD)—which affects the estimation of the effective radius (reff) and modulates the reff‐dependency of the SSPs—and from averaging SSPs over wide spectral bands. To assess these uncertainties a series of reff‐dependent SSPs parameterizations corresponding to various DSDs and spectral averaging methods are derived and implemented in a radiative code. Combined with the DSD‐dependent estimation of reff they are used to compute the bulk radiative properties (reflectance, transmittance, absorptance) of various clouds (defined in terms of LWC and N), including a homogeneous cloud, more realistic case studies, and outputs of a climate model. The results show that the cloud radiative forcing can vary up to 20% depending on the assumed DSD. Likewise, differences up to 20% are obtained for heating rates. The estimation of reff is the main source of uncertainty, while the SSPs parameterization contributes to around 20% of the total uncertainty. Spectral averaging is less an issue, except for atmospheric absorption. Overall, global shortwave cloud radiative effect can vary by 6 W m−2 depending on the assumed DSD shape, which is about 13% of the best observational estimate. Plain Language Summary Climate predictions differ a lot from one model to another, and the difficulty to simulate how clouds will behave in a warmer world is largely responsible for that. The radiative effect of clouds depends on the size of the individual droplets forming a cloud, a quantity that is not explicitly represented in climate models. In this study we investigate how not accounting for the detailed droplet size distribution affects the capability of climate models to reliably predict the radiative effect of clouds. By assuming a variety of droplet size distributions in a set of simulations, we observe that apparently similar clouds in climate models can have very different radiative impacts depending on the assumed distribution. This is primarily attributed to the estimation of the effective radius of cloud droplets, a key quantity that drives cloud radiative properties. D