One of the primary objectives of the New Zealand Earth System Model (NZESM) is to reduce shortwave radiation biases over the Southern Ocean, which are related to deficiencies in representation of clouds and aerosols in this region. This is a subject of active research with multiple hypotheses being tested including cloud microphysics, cloud–aerosol interaction, horizontal homogeneity and differences in the frequency of cloud regimes related to different weather systems being examined. Comparison with observations is necessary for the identification and resolution of these deficiencies. Unfortunately, observations in the Southern Ocean are scarce, with satellites providing the most extensive spatial and temporal coverage, especially instruments such as MODIS and ISCCP and active instruments such as radar and lidar (laser lidar) on the CloudSat and CALIPSO satellites. However, these instruments lack the capability to observe low-level cloud when there is a higher-level overlapping cloud. We present a multi-year dataset of shipborne and ground-based ceilometer, radar and aerosol observations in the Southern Ocean, which allows for cloud to be seen ‘from below’. In particular, we discuss the use the Cloud Feedback Model Intercomparison Project (CFMIP) Observation Simulator Package (COSP) to compare the ceilometer measurements with NZESM simulations. The COSP simulator currently does not support ground-based lidars with a matching wavelength, but the ACTSIM lidar simulator in COSP requires only a few modifications to support the analysis of ceilometer data. Using an instrument simulator such as COSP allows us to account for the limited view of the ceilometer and signal attenuation in the atmosphere. We apply the newly developed ceilometer simulator to the NZESM atmospheric state output in the regions of the Southern Ocean where shipborne or ground-based observations are available, and compare the resulting backscatter distribution and algorithmically derived products such as cloud base between the model and observations.