The mid-Pliocene warm period between 3.29–2.97 million years before present (Ma BP) (Haywood et al., 2010, 2011) has been suggested as an analogue for future warmer-than-present climate states (Jansen et al., 2007, p. 440). This time-slice is furthermore of relevance for the understanding of Cenozoic (since 66 Ma BP, Walker et al., 2013) climate evolution as it represents the last warm period before the establishment of a bipolar glaciated climate state that resulted from the onset of Northern Hemisphere glaciation (about 2.75 Ma BP, Maslin et al., 1998). The mid-Pliocene has been a major focus of recent climate modelling efforts. Given the principal imperfections of climate models, which represent the real climate system per definition with reduced complexity and approximation, it is of interest to evaluate the performance of climate models in the simulation of paleoclimate. The Pliocene Model Intercomparison Project (PlioMIP), established by Haywood et al. (2009a, 2010, 2011), is a major coordinated effort to identify strengths and weaknesses of climate models in the context of mid-Pliocene climate. This dissertation aims at research topics that are related to the PlioMIP. A first aim is to evaluate the mid-Pliocene climate simulated with the Community Earth System Models (COSMOS). Furthermore, the modelling methodology of the PlioMIP is reviewed and discussed with a focus on potential improvements with respect to future intercomparison studies and a more complete and accurate simulation of paleoclimate. An outcome of the model-intercomparison is a persistent deviation between mid-Pliocene climate as inferred from simulations and proxy reconstructions (e.g. Dowsett et al., 2013; Salzmann et al., 2013). One possible explanation for this mismatch are past Earth surface conditions that influenced the geologic record but have not been considered in idealized climate simulations. An assumed driver of climate evolution are modifications in ocean gateways. In the PlioMIP, a present day state of ocean gateways is assumed, although there is the indication of ocean gateway reconfigurations during this time-slice of the Cenozoic. Hence, in this dissertation ocean gateways are considered in sensitivity studies with a climate model of reduced complexity. The focus is on their impact on the state of the Atlantic Ocean circulation and characteristics. Another inference from the PlioMIP is the necessity of including dynamic ice sheet models into the simulation of the mid-Pliocene climate system, as prescribed ice sheet reconstructions potentially bias the simulated climate. In order to fully exploit the methodological capabilities of a coupled atmosphere-ocean / ice sheet model in the context of paleoclimate and rapid climate change, the COSMOS have to be improved. In particular, there is the need for a flexible discharge transport model that is able to adapt the continental discharge routes to varying orography and sea level height. In order to satisfy this demand, the COSMOS are extended in this dissertation with the novel Flexible Hydrological Discharge Model. Its properties and characteristics are described, and various examples of practical application in the framework of global climate models are presented.