The decomposition of cloud–aerosol forcing in the UK Earth System Model (UKESM1)

Climate variability in the North Atlantic influences processes such as hurricane activity and droughts. Global model simulations have identified aerosol–cloud interactions (ACIs) as an important driver of sea surface temperature variability via surface aerosol forcing. However, ACIs are a major cause of uncertainty in climate forcing; therefore, caution is needed in interpreting the results from coarse-resolution, highly parameterized global models. Here, we separate and quantify the components of the surface shortwave effective radiative forcing (ERF) due to aerosol in the atmosphere-only version of the UK Earth System Model (UKESM1) and evaluate the cloud properties and their radiative effects against observations. We focus on a northern region of the North Atlantic (NA) where stratocumulus clouds dominate (denoted the northern NA region) and a southern region where trade cumulus and broken stratocumulus dominate (southern NA region). Aerosol forcing was diagnosed using a pair of simulations in which the meteorology is approximately fixed via nudging to analysis; one simulation has pre-industrial (PI) and one has present-day (PD) aerosol emissions. This model does not include aerosol effects within the convective parameterization (but aerosol does affect the clouds associated with detrainment) and so it should be noted that the representation of aerosol forcing for convection is incomplete. Contributions to the surface ERF from changes in cloud fraction (fc), in-cloud liquid water path (LWPic) and droplet number concentration (Nd) were quantified. Over the northern NA region, increases in Nd and LWPic dominate the forcing. This is likely because the already-high fc there reduces the chances of further large increases in fc and allows cloud brightening to act over a larger region. Over the southern NA region, increases in fc dominate due to the suppression of rain by the additional aerosols. Aerosol-driven increases in macrophysical cloud properties (LWPic and fc) will rely on the response of the boundary layer parameterization, along with input from the cloud microphysics scheme, which are highly uncertain processes. Model grid boxes with low-altitude clouds present in both the PI and PD dominate the forcing in both regions. In the northern NA, the brightening of completely overcast low cloud scenes (100 % cloud cover, likely stratocumulus) contributes the most, whereas in the southern NA the creation of clouds with fc of around 20 % from clear skies in the PI was the largest single contributor, suggesting that trade cumulus clouds are created in response to increases in aerosol. The creation of near-overcast clouds was also important there. The correct spatial pattern, coverage and properties of clouds are important for determining the magnitude of aerosol forcing, so we also assess the realism of the modelled PD clouds against satellite observations. We find that the model reproduces the spatial pattern of all the observed cloud variables well but that there are biases. The shortwave top-of-the-atmosphere (SWTOA) flux is overestimated by 5.8 % in the northern NA region and 1.7 % in the southern NA, which we attribute mainly to positive biases in low-altitude fc. Nd is too low by −20.6 % in the northern NA and too high by 21.5 % in the southern NA but does not contribute greatly to the main SWTOA biases. Cloudy-sky liquid water path mainly shows biases north of Scandinavia that reach between 50 % and 100 % and dominate the SWTOA bias in that region. The large contribution to aerosol forcing in the UKESM1 model from highly uncertain macrophysical adjustments suggests that further targeted observations are needed to assess rain formation processes, how they depend on aerosols and the model response to precipitation in order to reduce uncertainty in climate projections.