Understanding the vertical distribution of aerosol helps to reduce the uncertainty in the aerosol life cycle and therefore in the estimation of the direct and indirect aerosol forcing. To improve our understanding, we use measurements from four deployments of the Atmospheric Tomography (ATom) field campaign (ATom1–4) which systematically sampled aerosol and trace gases over the Pacific and Atlantic oceans with near pole-to-pole coverage. We evaluate the UK Earth System Model (UKESM1) against ATom observations in terms of joint biases in the vertical profile of three variables related to new particle formation: total particle number concentration (NTotal), sulfur dioxide (SO2) mixing ratio and the condensation sink. The NTotal, SO2 and condensation sink are interdependent quantities and have a controlling influence on the vertical profile of each other; therefore, analysing them simultaneously helps to avoid getting the right answer for the wrong reasons. The simulated condensation sink in the baseline model is within a factor of 2 of observations, but the NTotal and SO2 show much larger biases mainly in the tropics and high latitudes. We performed a series of model sensitivity tests to identify atmospheric processes that have the strongest influence on overall model performance. The perturbations take the form of global scaling factors or improvements to the representation of atmospheric processes in the model, for example by adding a new boundary layer nucleation scheme. In the boundary layer (below 1 km altitude) and lower troposphere (1–4 km), inclusion of a boundary layer nucleation scheme (Metzger et al., 2010) is critical to obtaining better agreement with observations. However, in the mid (4–8 km) and upper troposphere (> 8 km), sub-3 nm particle growth, pH of cloud droplets, dimethyl sulfide (DMS) emissions, upper-tropospheric nucleation rate, SO2 gas-scavenging rate and cloud erosion rate play a more dominant role. We find that perturbations to boundary layer nucleation, sub-3 nm growth, cloud droplet pH and DMS emissions reduce the boundary layer and upper tropospheric model bias simultaneously. In a combined simulation with all four perturbations, the SO2 and condensation sink profiles are in much better agreement with observations, but the NTotal profile still shows large deviations, which suggests a possible structural issue with how nucleation or gas/particle transport or aerosol scavenging is handled in the model. These perturbations are well-motivated in that they improve the physical basis of the model and are suitable for implementation in future versions of UKESM.