Dominant role of early winter Barents–Kara sea ice extent anomalies in subsequent atmospheric circulation changes in CMIP6 models

Abstract

The mid-latitude climate responses to Arctic sea ice loss remain unclear, partly because the atmospheric responses depend sensitively to the location of sea ice anomalies. Evaluating the role of regional Arctic sea ice extent anomalies is therefore essential to appreciate the extent of atmospheric responses. We investigated these responses but also the atmospheric precursors to regional Arctic sea ice extent anomalies in long pre-industrial control simulations from 36 CMIP6 climate models. This study examines changes in various atmospheric variables at different lead and lag times by performing a composite analysis between years of low and high sea ice extents in different Arctic seas. Stronger and more statistically significant relationships are found when the atmosphere leads to changes in sea ice than in the reverse direction, suggesting that in the CMIP6 models the atmosphere drives the sea ice rather than the opposite. The atmospheric circulation is found to be relatively insensitive to regional sea ice anomalies except in the Barents–Kara (BAKA) Seas, where the negative anomalies are followed by a robust negative winter North Atlantic Oscillation (NAO)-like pattern. Consistent with the so-called stratospheric pathway, a weakening of the stratospheric polar vortex (SPV) is simulated 1 month prior to the NAO change and can be partially explained by the sea ice anomaly in BAKA in preceding months. The magnitude of this weakening in models depends on other factors, such as the Siberian snow cover, the El Niño–Southern oscillation, and the quasi-biennial oscillation. Moreover, the SPV and the NAO index responses scale approximately linearly with the magnitude of the BAKA sea ice anomaly. These results highlight the inter-model consistency of the role played by the BAKA sea ice extent anomaly for the atmospheric responses under pre-industrial conditions. In upcoming work, the sea ice loss in this region should therefore primarily be considered, but under future conditions.

Appendix 1

See Figs. 12, 13, 14, 15 and 16.

Fig. 12

Multi-model zonal mean averaged over 60–75$^{\circ }$N in the meridional wave heat flux ([v*T*]) between years of December low sea ice extent years and years of December high sea ice extent from previous June (Lag − 6) to next June (Lag + 6) for the BAKA pattern (a). Atm stands for atmopshere and SI stands for sea ice. Multi-model mean difference between years of low and high December sea ice extent of [v*T*] at 10 hPa in November (Lag − 1, left), December (Lag 0, center) and January (Lag + 1, right) for the BAKA pattern. The contours show the multi-model mean climatological mean with intervals of 20 m K/s

Fig. 13

Multi-model zonal-mean difference in the zonal wind between years of low December sea ice extent and years of high December sea ice extent during the following winter (JFM, Lag + 1 to + 3) for the BAKA pattern (a), the OKH pattern (b), the CHUBER pattern (c), the LAB pattern (d) and the GR pattern (e)

Fig. 14

Multi-model zonal mean at 60$^{\circ }$ N in the zonal wind between years of December low sea ice extent years and years of December high sea ice extent from previous June (Lag − 6) to next June (Lag + 6) for the BAKA pattern (a), the OKH pattern (b), the CHUBER pattern (c), the LAB pattern (d) and the GR pattern (e). Atm stands for atmopshere and SI stands for sea ice

Fig. 15

Multi-model mean difference in sea level pressure between the January low years ($<-1\sigma$) and the January high years ($>1 \sigma$) in zonal wind in the stratosphere at 60$^{\circ }$N in October (Lag − 3) (a), in November (Lag − 2) (b), in December (Lag − 1) (c) and in January (Lag 0) (d)

Fig. 16

Multi-model mean difference in sea ice concentration between the January low years ($<-1 \sigma$) and the January high years ($>1 \sigma$) in zonal wind in the stratosphere at 60$^{\circ }$ N in July (Lag − 6) (a), and in February (Lag + 2) (b)