Past and future of the Arctic sea ice in HighResMIP

Arctic sea-ice area and volume have dramatically decreased since the beginning of the satellite era. This alarming rate of ice decline raises a key scientific question: how soon will the Arctic meet the first “ice-free” summer? Coupled climate models are the primary tools to provide projections of future sea ice conditions. Increasing the horizontal resolution of general circulation models is a widely recognized way to improve the representation of the complex processes at high latitudes, and to obtain trustworthy predictions of ice variability. Here, we investigate the past and future changes of sea ice cover at hemispheric and regional scales using model outputs from the High Resolution Model Intercomparison Project (HighResMIP, Haarsma et al. 2016) of the sixth phase of the Coupled Model Intercomparison Project (CMIP6). 

The main objective is to investigate the impact of ocean/atmosphere model resolution on the representation of Arctic sea ice area (SIA) and volume (SIV)  and their seasonal, interannual variability, and trends in the recent past. Model results over the period 1950–2014 are compared to a set of observational datasets. 

All models project substantial sea ice shrinking: from 1950 to 2050 the Arctic loses nearly 92% of SIV from 1950 to 2050. The individual models simulate the first summer ice-free Arctic as early as 2019 and as late as 2050. The ensemble mean of the three best performing models suggests the event to happen by 2044. Along with the overall reduction of sea ice cover, there are changes in the structure of sea ice cover: the marginal ice zone (MIZ) dominates the ice cover by the mid-XXI which implies the shift to a new sea ice regime closest to the Antarctic conditions. The MIZ-dominated Arctic might suggest to adapt and modify model physics parameterizations and sea ice rheology.

Our analysis does not present a strong relationship between ocean/atmosphere spatial resolution and sea ice cover representation: the impact of horizontal resolution rather depends on the model used and the examined variables. However, the refinement of the ocean mesh has a more prominent effect compared to the atmospheric one, mainly due to a more realistic representation of the sea ice edges as a result of better simulated ocean currents and heat transports in the Northern Atlantic Ocean.