Summer Arctic temperatures in PMIP4 Last Interglacial simulations and their link to Arctic sea ice

Investigating climate models responses and feedbacks under warmer climates is useful in building confidence in future climate projections. Here we focus on the summer sea ice in the Arctic during the Last Interglacial period (LIG), when the Arctic was warmer than the Pre-Industrial period (PI) by around 4.5 ± 1.7 K. Given that it is difficult to ascertain the state of Arctic sea ice from marine core proxies of sea ice state, we focus instead on summer surface air temperature (SSAT) in CMIP6-PMIP4 simulations and compare these with equivalent proxy data. All 12 models we have analysed show both warmer SSAT and a reduction in summer sea ice in the LIG compared to the PI, with an average warming of +3.6K and an average 52% decrease in minimum sea ice area.

 

We find that model-observation differences in LIG SSAT are linearly related to the percentage loss of summer Arctic sea ice. However this general finding does not fit the CNRM model result, which is an outlier. This simulation captures the observed pattern of SSAT, without being close to ice-free. However peculiarities in the CNRM set-up (forcing and sea ice model tuning) means it is unclear what can be drawn from this one result. CNRM aside, models tend to yield more accurate LIG SSAT changes when they are closer to an ice-free state in summer. The models which feature sea ice losses larger than the multi-model-mean sea ice loss, tend to have the smallest model-observation SSAT errors. The results of this study provides caveated support to the argument that Arctic could have been ice-free during LIG summers. That said, a careful examination of the SSAT dataset would also be value to the LIG community, given that these results are dependent on the LIG SSAT observational dataset.