Abstract. Smoke aerosols (BBA) from African fires may strongly impact Earth’s radiation budget in the southeast Atlantic (SEA), but the sign and magnitude of the overall radiative effect (RE) remains uncertain. Aerosol-climate models are needed to separately quantify direct, indirect, and semi-direct REs. Here we evaluate improved simulations with the UK Met Office’s Unified Model and with them explore how REs depend on the method used to match observed meteorology (nudging or running forecasts reinitialized at different frequencies). REs are calculated as differences in radiative fluxes between simulations with and without smoke emissions, and with and without aerosol absorption. All model setups agree on net warming for the SEA dominated by the direct effect. Simulated smoke, clouds, and the direct effect agree better with observations than previous studies using the same model, though biases in aerosol extinction and liquid water path remain. Changes in cloud droplet number concentration due to BBA self-lofting influence how cleanly we can separate cloud effects into semi-direct and indirect effects. Total RE, which remains unaffected, ranges from +3.0 to +7.9 W m⁻². The 4.9 W m⁻² spread arises mainly from simulated semi-direct effects. Forecasts three days long or less probably do not allow time for plausible differences in boundary layer properties due to semi-direct effects to accumulate. Free running simulations with and without smoke accumulate differences in meteorology that are likely spurious ‘butterfly effects’. We recommend future research quantifying BBA REs over weeks to months use meteorological forcing techniques that allow aerosol absorption to affect the boundary layer.