Long-term forecasting of the behaviour of CO2 injected in industrial quantities into sub-surface reservoirs is generally performed using models including very limited geological detail. This practise, although dictated by restrictions with respect to model resolution and CPU cost, partly rests on an unverified but widely held assumption that geological detail will not influence simulation outcomes on the spatial and temporal scales relevant for geological sequestration. The present study aims to partly assess the validity of this assumption by selecting a series of realistic geological features and investigate their impact when modelling CO2 sequestration.Injected CO2 is primarily retained in structural and stratigraphic traps at the top of the reservoir interval. We will therefore investigate how different top-surface morphologies will influence the CO2 storage capacity. To this end, a series of different top surfaces are created by combining different stratigraphic scenarios with different structural scenarios. The models are created stochastically to quantify uncertainty. Theoretical upper bounds on the volume available for structural trapping are established by a geometrical analysis. Estimates for actual trapping from a single point source are calculated in a simplified and efficient manner by a spill-point analysis and more accurately by a detailed flow simulation that assumes vertical equilibrium. The experiments show that the morphology of the top seal has a large impact on the storage capacity and migration patterns. Comparing the two simulation approaches, our results verify that structural trapping can be roughly estimated by spill-point analysis, while estimation of residual trapping requires detailed flow simulations. The migration rate may be retarded if the plume height is at the same scale as the top surface relief. Among the studied scenarios, we found that the offshore sand ridges model, which has the largest relief, has the largest total capacity for storage of CO2 by structural and residual trapping.For the fluid flow simulation method, we also investigate the effect of grid resolution. Upscaling by simple grid coarsening will smooth fine-scale morphology that retards the CO2 and hence generally overestimate the plume migration. The error in the upscaling is largest for the small structures that dominate the trapping mechanisms induced by faults and buried beach rides in a flooded marginal-marine setting. Buried offshore sand ridges, on the other hand, are characterised by relatively large structures and are hence less sensitive to grid resolution.