Modelling non-analogue elements of Pliocene North Atlantic warming

The strong warming seen in records of mid-Pliocene sea surface temperature (SST) in the North Atlantic has proved difficult to reproduce in climate model simulations. The results of the Pliocene Model Intercomparison Project (PlioMIP) Experiment 2 fail to produce a single simulation with North Atlantic sea surface temperatures (SSTs) as high as those indicated by the PRISM3 (Pliocene Research, Interpretation and Synoptic Mapping) data set. Direct comparisons between the data and models are hampered by differing techniques used in palaeoenvironmental reconstruction and physical climate simulations. However, even if current simulations are not directly comparable to the reconstructions of the North Atlantic, something must have forced these particularly high temperatures for at least parts of the mid-Pliocene warm period. The boundary condition changes defined in the PlioMIP Experiment 2 protocol are limited to CO₂, ice sheets, vegetation, land area change due to sea level rise and orography. Apart from small orographic changes imposed outside of the ice sheet regions, the rest of these factors would be expected to change under future anthropogenic climate change. As such the IPCC (Intergovernmental Panel on Climate Change) has referred to the mid-Pliocene as 'an accessible example of a world that is similar in many respects to ... the late 21st century'. However, there are a number of different palaeogeographic changes documented in published literature that are not incorporated into the PRISM3 palaeoenvironmental reconstruction used as model boundary conditions, particularly in the North Atlantic region. Although some of these would be expected under future climate change, e.g. a reduction in North Atlantic icebergs, many would not. Changes in the intensity of Icelandic mantle plume upwelling have resulted in changes in the sill depth of the Greenland-Scotland ridge over at least the last 40 million years. Pleistocene glacial erosion has created new ocean areas in the North Atlantic, including the Barents Sea. North American and European rivers have changed course, due to the newly created marginal seas and glacial rerouting, potentially affecting the salinity balance in the North Atlantic. We present simulations using the Hadley Centre coupled atmosphere-ocean HadCM3 model to estimate the impact of these changes on Atlantic Meridional Overturning Circulation (AMOC) and SST in the North Atlantic. By applying palaeogeographic changes to the standard PlioMIP Experiment 2 simulation individually and as a whole, we show that these can produce North Atlantic SST changes of a similar magnitude to data-model discrepancies. Palaeoclimate model simulations can only reproduce global and regional climate accurately if the boundary conditions given to the model are sufficient to capture all the significant changes in climate processes and dynamics. Incorporation of other important boundary condition changes and proper quantification of the model uncertainties due to unknown boundary conditions could explain existing data-model mismatches in the Pliocene North Atlantic.