AbstractThe cloud vertical distribution and overlap of four large-scale models operating at different horizontal and vertical resolutions have been assessed using radar and lidar observations from the Baltex Bridge Campaign of CLIWA-NET. The models range from the global European Centre for Medium range Weather Forecast (ECMWF) model, to the Regional Atmospheric Climate Model (RACMO) and the Rossby Centre Atmospheric (RCA) regional climate model, to the non-hydrostatic meso-scale Lokal Model (LM). Different time averaging periods for the radar data were used to estimate the uncertainty of the point-to-space transformations of the observations. Relative to the observations, all models underestimated the height of the lowest cloud base. Clouds occurred more frequently in the models but with smaller cloud fractions below 7 km. The findings confirm previous cloud radar studies which found that models overestimate cloud fractions above 7 km. Radar-observed clouds were often thinner than the model vertical resolutions, which can have serious implications on model cloud overlap and radiation fluxes. The radar-derived cloud overlap matrix, which takes into account the overlap of all vertical layers, was found to be close to maximum-random overlap. Using random overlap for vertically continuous clouds with vertical gradients in cloud fraction larger than 40–50% per kilometre gave the best fit to the data. The gradient approach could be improved by making it resolution- and cloud system-dependent. Previous cloud radar overlap studies have considered the overlap of two cloud layers as a function of maximum and random overlap. Here, it was found that the two-layer overlap could be modelled by a mixture of maximum and minimum overlap.