A plausible emergence of new convection sites in the Arctic Ocean in a warming climate


               <jats:p />


Deep-convection sites might appear in the Arctic Ocean
Atlantic Meridional Overturning Circulation (AMOC) transports enormous amounts of heat from low to high latitudes and thus, it is vital for the climate.It is widely predicted to decline under global warming [1], which is attributed to sea-ice melt in Greenland and the Arctic [2], that can stabilize the stratification and prevent deep convection in the Labrador Sea [3].
Deep convection overturns the upper limb of the AMOC and thereby modulates the AMOC strength.For the past 3 kyr, paleoclimate proxies have shown that the AMOC variability was related to deep convection in the Labrador Sea (e.g.[4].).Climate models have also shown that the Labrador Sea convection drives the multi-decadal AMOC variability [5].Furthermore, deep convection has remained anomalously weak over the past 150 years [3].
However, observed changes were found in the locations of deep convection where the AMOC is mainly modulated.The zonal mooring line crossing the subarctic Atlantic, as the first AMOC observation in this region, has observed significantly stronger AMOC east of Greenland relative to the Labrador Sea since 2014 [6], further confirmed by a 60 year reconstruction [7].The amounts of overturning in the Nordic, Iceland, Irminger Sea account for 39%, 48% and 13% of the total overturning in the eastern subarctic Atlantic, respectively [8].This phenomena implies that deep convection in the eastern subarctic Atlantic is more dominant at modulating the AMOC, different than the former cognition of dominance in the Labrador Sea.
There is growing observational evidence that finds enhanced ocean convection near the sea ice edge in the Nordic Sea, following sea ice retreat [9,10] (figure 1).We question, with the rapid decline of Arctic sea ice under global warming, is it possible that the dominant deep-convection site occurs in the Arctic Ocean?
The northern end of the AMOC has been suggested to be in the Arctic Ocean [11].Climate modeling under future warming scenarios found that the convection and subduction sites would appear in the Arctic Ocean following sea ice retreat [12][13][14], and the AMOC emerges and strengthens north of the Greenland-Scotland Ridge (GSR) [13] (figure 1).Deep convection first declines in the Labrador Sea and strengthens in the Nordic Sea, and then declines in the Nordic Sea and eventually strengthens near the sea ice edge of the Nansen Basin [13].Besides, the strengthening of this part of the AMOC partially counteracts the weakening of the AMOC at 26 • N [13].

The potential mechanism that drives the extension of the AMOC towards the Arctic Ocean
In the paleoclimate, with a transition from stadial to interstadial, the dominant convection site moves northward across the GSR following sea ice retreat, and both the convection and AMOC strengthens [15].Considering the similarity to the climate predictions mentioned above, the relevant mechanisms for the paleoclimate could shed light on understanding the future AMOC.
Via prescribing deglacial global warming in an ocean model, a mechanism for the AMOC increase is proposed as follows [16].Despite the freshwater input from the melting glaciers and sea ice, with sea ice retreat, the increasing area of open ocean facilitates air-sea interaction that induces deep convection.The enhancement of convection causes a decrease of subsurface temperature that intensifies the northward salt transport from the low latitudes.The additional salt at the convection site would erode the halocline and thus break the salinity-driven stratification.This enables the vertical temperature inversion to cause a rapid strengthening of deep convection and thus, the AMOC.
This mechanism has implications for the future AMOC response to the Arctic sea ice retreat.Sea ice loss activates heat loss from the ocean surface, and the increase of the first-year-ice region (change from multi-year-ice region) causes more brine rejection during sea ice formation in winter.Both processes could enhance the convection.The inflow of Atlantic waters into the Arctic Ocean increases during the late Holocene, with an acceleration to the recent maximum of the past 2 kyr [17].The shallow mixed layer depth and strong halocline in the Arctic Ocean was capable of preventing the exchange between the cold, fresh surface water and the warm, salty subsurface Atlantic water.However, as sea ice decline induces stronger vertical mixing, observations evidenced unprecedented upward intrusion of Atlantic water (called 'atlantification' of the Arctic Ocean), because of weakened stratification in the mixed layer and erosion of halocline [18].This process has the potential of fundamentally breaking the salinity-driven stratification in the Arctic Ocean.It could be the starting point of deep convection in the Arctic Ocean, as suggested by climate modeling [12].

Discussion and implication
To summarize, in addition to the currently predicted slowdown of AMOC, we emphasize the possibility of an AMOC extension and enhancement towards the Arctic Ocean under global warming.In other words, the anthropogenic warming may induce an AMOC strengthening on the regional scale, accompanied with the large-scale AMOC slowdown.Regarding the effect of Arctic sea ice decline on the AMOC, we highlight that the induced freshening prevents deep convection in the north Atlantic and slows down the general AMOC, but more open-ocean areas may activate air-sea interaction and thus, the convection and AMOC in the Arctic Ocean.
Once the AMOC extends to the Arctic Ocean, the concept of 'Atlantification' in the Arctic Ocean would not only be limited as the upward intrusion of Atlantic water, but further include the resemblance of hydrography and circulation structure to the North Atlantic.Bringing huge amounts of heat to the Arctic, this additional response of the AMOC to global warming could have severe impacts on the Arctic environment and ecosystem.Deep convection acts as a heat and carbon pump and facilitates the exchange of oxygen [14].Its occurrence in the Arctic Ocean could thus cause local transitions in the ecosystem.Furthermore, the AMOC extension could cause the migration of marine species in the Atlantic towards the Arctic Ocean.These ecology impacts could affect fisheries and the livelihoods of people in the Arctic who still depend on marine resources.
Given the concern of a tipping point when the AMOC overturns in the Arctic Ocean, the mechanism behind this possibility still needs to be clarified by climate prediction.A conceptual model of AMOC extension to the Arctic Ocean has been proposed [19].Although the current generation of climate models show large uncertainty in simulating deep convection in the Arctic Ocean [20], the possibility of occurrence still cannot be ruled out.With the future development of climate models, reducing biases in simulating the Arctic will give us a clearer picture of whether the AMOC would extend to the Arctic Ocean.

Figure 1 .
Figure 1.Schematic showing the potential emergence of convection and AMOC in the Arctic Ocean.The circles and crosses denote convection sites.The arrows denote the AMOC, with red and blue representing its warm upper limb and cold lower limb.The dashed circle and arrows denote the possibility in the future Arctic Ocean, respectively.LAB: Labrador Sea; IRM: Irminger Sea; GIN: Nordic Sea.