Observed vulnerability of Filchner-Ronne Ice Shelf to wind-driven inflow of warm deep water

The average rate of melting at the base of the large Filchner-Ronne Ice Shelf in the southern Weddell Sea is currently low, but projected to increase dramatically within the next century. In a model study, melt rates increase as changing ice conditions cause a redirection of a coastal current, bringing warm water of open ocean origin through the Filchner Depression and into the Filchner Ice Shelf cavity. Here we present observations from near Filchner Ice Shelf and from the Filchner Depression, which show that pulses of warm water already arrive as far south as the ice front. This southward heat transport follows the eastern flank of the Filchner Depression and is found to be directly linked to the strength of a wind-driven coastal current. Our observations emphasize the potential sensitivity of Filchner-Ronne Ice Shelf melt rates to changes in wind forcing.

A ntarctic ice shelves are thinning at an accelerating rate 1 , leading to reduced buttressing of the ice sheet 2 with a consequential increase in glacial flow and sea-level rise. The ice shelf thinning is caused by enhanced basal melt 3 and is therefore linked to increased heat transport from the open ocean towards the ocean cavity beneath the ice shelves. In the areas west of the Antarctic Peninsula where the ice shelves are observed to be rapidly thinning 1,3 , warm water of open ocean origin has direct access to the ice-shelf cavities [3][4][5] . In the southern Weddell Sea, however, where the large Filchner-Ronne Ice Shelf (FRIS) resides ( Fig. 1), the water on the wide continental shelf separating the ice-shelf cavity from the deep ocean is cold, and the majority of the water entering the ice-shelf cavity is at the surface freezing point 6 . Basal melt rates in this region are therefore currently low 7 , although recent modelling studies 8,9 suggest that changing ice conditions during the next century will affect the transfer of momentum from the atmosphere to the ocean and redirect the coastal current in a way that will allow the Warm Deep Water (WDW) to access the Filchner Ice Shelf (FIS) cavity through the Filchner Depression (FD, Fig. 1). The melt rates would then increase dramatically-from 0.2 to 4 m per year 8 .
WDW is the Weddell Sea variant of the relatively warm Circumpolar Deep Water that enters the Weddell Sea in the east 10 and is carried westward in the southern limb of the Weddell gyre. The core of the WDW is typically 0.6-0.9°C, cooling towards the west 10,11 , and found at B300 m depth in the interior of the Weddell basin, below a layer that is chilled by the Antarctic atmosphere. Easterly winds along the Antarctic continent cause converging Ekman transport and down-welling along the coast 12 , depressing the WDW and the thermocline to depths below the shelf break. Weaker summer winds allow for a shoaling of the thermocline and a seasonal inflow of WDW along the eastern flank of the FD 13 . FD is a deep trough that cuts across the eastern part of the broad continental shelf in the southern Weddell Sea, and extends beneath FIS to the south (Fig. 1). The depression can act as a direct conduit for warm water southward, similar to the troughs across the continental shelves in the Amundsen and Bellingshausen Sea 14,15 . The FD is filled with dense Ice Shelf Water (ISW), that is, water colder than the surface freezing point, which emerges from the ocean cavity beneath FIS 6 . The ISW extends from the sea floor up to above the depth of the FD sill (600 m), confining the warmer, summer inflow to shallower depths. Hydrographic sections from the region in the austral summer typically show a 100 m thick warm layer along the 400 m isobath 6,16 ; in January 2013 the warm layer was observed as far south as 76S (ref. 17).
The observations presented here show that the warm water eventually reached as far south as the Filchner ice front and that the southward transport of warm water is linked to the strength of a wind-driven coastal current, thus emphasizing the potential sensitivity of FRIS melt rates to changes in wind forcing.

Results
Warm water at the Filchner Ice Shelf front. Two oceanographic moorings, M NORTH and M SOUTH , ran from January 2013-2014 on the eastern flank of the FD at 77°S (500 m depth) and 78°S (700 m depth), and recorded the arrival of a pulse of warm water (Fig. 2a,c). The warm water is a mixture of WDW and surface water, and is colder than the WDW found off shelf. The warm pulse reaches M NORTH in mid-January, when it was observed at 350 m depth above a 100 m thick layer of ISW. The first, sporadic arrival of warm water at the FIS front (M SOUTH ) occurs at the beginning of March with a more persistent and deeper presence from the end of the month. On 8th April, the warm layer suddenly deepens to 450 m, followed by a second deepening to 500 m 1 week later. The properties of the warm inflow slowly erode, and from May onwards the temperature at the mooring location is again at or below the surface freezing point. Warm water is present at M NORTH until the beginning of June.
All available temperature and salinity profiles from the front of FIS are shown in Fig. 3, indicating that no WDW has previously been observed at the ice front. In this area, observations from ships are made only during summer (from January to early March), likely too early in the season for warm water to have arrived at the ice front. From February through September 2011, however, a set of oceanographic profiles was collected by a diligent Weddell Seal in close proximity to M SOUTH (ref. 18). The profiles show that during 2011 the warm inflow, while present at the sill 13 , does not reach the front of FIS. Warm water higher in the water column was observed near the ice front in March, 2011 (ref. 18) (and also previously 16 ) but this is relatively fresh surface water heated by solar radiation during summer, and it is too shallow to access the ice-shelf cavity.
The wind-driven coastal current. The warm inflow through FD that leads to the dramatic increase in basal melt rates in the predictions in ref. 8 is caused by a redirection of the coastal current flowing westward along the southern Weddell Sea coast (Fig. 1). Upstream of the FD, the coastal current is merged with the Antarctic Slope Front current and has a strong barotropic component 19 that is set up by the prevailing easterly winds ( Supplementary Fig. 1a) and the Ekman transport that converges towards the coast. When the continental shelf widens at 27°W the coastal current bifurcates: the inner branch follows the coast 6 and the outer branch continues along the continental shelf break 20 .
The currents observed at M NORTH suggest that the current bifurcates a second time: when reaching the divergent isobaths of the FD, part of the current is diverted southward along the eastern flank of the Depression, transporting the warm water present at the shelf break during summer towards the FIS front. The observed southward flow at M NORTH is highly variable and strongly affected by the upstream along-coast wind observed, for example, at the nearby Halley research station ( (r ¼ À 0.47, po0.01) between wind and current agrees with the typical lag of 0.5-1.0 days observed in other regions where directly forced wind-driven slope currents are found 21,22 .
The effect of strong easterly winds is thus twofold and depends on the time scale considered: on longer (monthly) time scales it depresses the thermocline above the continental slope, shutting off the inflow of warm water towards the continental shelf. On shorter (daily) time scales, it strengthens the coastal current and enhances the southward transport of warm water available on the shelf. The baroclinic response, that is, the depression of the pycnocline, thus depends on the mean wind, while the response of the barotropic current reflects the day to day variability in atmospheric forcing. A combination of generally weak easterly winds, leading to a relaxation of the pycnocline, interrupted by short and intense wind events, setting up strong barotropic currents, would allow warm water to reach the FIS front.  (Fig. 4b, Supplementary Fig. 3 and Supplementary Note 1): there were three episodes of near gale winds at Halley during January-February 2013 with none during the same period in 2011. The weak winds in 2013 potentially allowed for a shallow thermocline and a large warm inflow during early summer that then was advected southward by the wind-driven currents, reaching the FIS front some 350 km to the south about 3 months later. The preconditioning above the slope and the shoaling of the thermocline during spring and early summer is likely the most important factor, allowing for the barotropic current and eddy exchanges 23 to transport warm water onto the continental shelf. The inflow dynamics and the thermocline depth are influenced by several factors including the presence of a variable fresh surface layer due to summer sea ice melt 24 , sea ice concentration 8 , shelf salinity 9,25 and interannual variability of the circulation and properties of ISW within the FD. Sea ice conditions in the region differed greatly between the 2 years ( Fig. 4c and Supplementary  Fig. 4) and, contrary to modelling results which show increased inflow for low shelf salinity 9 , the salinity on the shelf was higher by B0.05 in 2013 compared with 2011 (Fig. 4d,e). The relative importance of these factors and the wind stress on the interannual variability of the warm inflow are left as a challenge for the emerging community of high-resolution regional models to disentangle.
Heat transport and warm water access to the cavity. An upper bound of the southward heat transport is obtained using the mean velocity and the maximum temperature and layer thickness observed at M NORTH . Assuming that the inflow width extends from M NORTH to the coast (100 km), the upper bound heat flux is 1.6 TW during the period of inflow (or 0.7 TW when averaged over a year). This is a substantial, geophysically significant flow of heat, enough to melt 70 km 3 or 65 Gton of ice annually, or about half of the FRIS basal mass budget 26 .
The draft at the ice front is 400-450 m (ref. 27) while the warm water is observed down to 500 m at M south . If some of it were to enter the ice-shelf cavity, it would be able to make contact with B20% of the FIS base 27 . The mean current at M SOUTH (500 m depth) is however directed northward, away from the ice front, both during the period when warm water is observed at the location and when temperatures are below freezing. The southward flow of warm water presumably lies above shallower isobaths to the east of M SOUTH .
The data presented here do not reveal the extent to which the warm water penetrates into the ice-shelf cavity. Relatively strong across ice front tidal currents observed at M SOUTH (0.15 m s À 1 ), suggest that the warm water present at the ice front can reach several km into the cavity, enhancing basal melt in the frontal region 28 . Independent evidence suggests that the melt rates in the frontal region can be substantial 29 ; basal melt in the outermost 100 km of FRIS accounts for 40% of the total melt below the ice shelf 26 . The warming and shoaling of the warm water core observed since the 1980s (ref. 30) have likely affected the frequency and the heat content of the warm water pulses reaching the FIS. Highresolution time series of ice-shelf thickness show a thinning on the eastern side of the FIS front between 1995 and 2012 (refs 1,7), while the entirety of the FRIS gained mass during the same period 1 .
The remarkable changes projected to occur in the FRIS cavity during the next century-a 2°C increase of the water temperature within the ice-shelf cavity and a 20-fold increase in basal melt rates 8 -will have consequences not only for the tributary ice streams and sea-level rise, but for also for the hydrography in a region known to produce a large fraction of Antarctic Bottom Water 31 . We show that the mechanism responsible for the dramatic changes in ref. 8-a redirection of the coastal current-is realistic although important small scale physical processes are not correctly resolved in their coarse model. The wind and ice conditions in 2013 were unusual but not extreme, and it is likely that warm water has occasionally reached the FIS Wind. Time series of wind stress and variance of the wind were calculated using the observed wind at Halley Station (75°35 0 S, 26°34 0 W) and the boxplots (Fig. 4a,b) were constructed using monthly mean values from the period 1978 to 2014. The effect varying ice concentration upstream was incorporated in the value of the drag coefficient C D following ref. 34 and the mean ice concentration in 'Coast' (Fig. 3c).
The variance was calculated in 4 week-long sliding windows from the high-pass filtered (3 days) time series. Wind and wind stress were calculated along the major wind axis, shown in Fig. 4, which is rotated 165°counterclockwise from east. To show that the winds at Halley are representative of a broader region upstream, we have correlated the wind observations from Halley with the along slope wind component from ERA Interim 35 during the period 1979-2014 ( Supplementary Fig. 2c,d). The correlation is high (r40.8) for a large portion of the coastal area upstream of FD.