Exceptionally Warm and Prolonged Flow of Warm Deep Water Toward the Filchner‐Ronne Ice Shelf in 2017

The Filchner‐Ronne Ice Shelf, fringing the southern Weddell Sea, is Antarctica's second largest ice shelf. At present, basal melt rates are low due to active dense water formation; however, model projections suggest a drastic increase in the future due to enhanced inflow of open‐ocean warm water. Mooring observations from 2014 to 2016 along the eastern flank of the Filchner Trough (76°S) revealed a distinct seasonal cycle with inflow if Warm Deep Water during summer and autumn. Here we present extended time series showing an exceptionally warm and long inflow in 2017, with maximum temperatures exceeding 0.5°C. Warm temperatures persisted throughout winter, associated with a fresh anomaly, which lead to a change in stratification over the shelf, favoring an earlier inflow in the following summer. We suggest that the fresh anomaly developed upstream after anomalous summer sea ice melting and contributed to a shoaling of the shelf break thermocline.


Introduction
The ice shelves fringing the Antarctic continent are central to the stability of the Antarctic Ice Sheet, whose mass balance is largely controlled by the interaction of the ice shelves with the ocean (Fürst et al., 2016). Enhanced basal melt rates and ice shelf thinning have been observed widely during the last decade, in particular in the Bellingshausen and Amundsen Seas (e.g., Paolo et al., 2015;Rignot et al., 2013). These changes have mostly been attributed to an increased transport of warm middepth Circumpolar Deep Water (CDW) onto the continental shelf (Nakayama et al., 2013;Pritchard et al., 2012). The exchange of water masses across the continental shelf break is largely controlled by dynamics and characteristics of the near-circumpolar Antarctic Slope Front (ASF; Jacobs, 1991) and the associated Antarctic Slope Current (ASC; see Thompson et al., 2018, for a review). The ASF and cross-slope exchange respond to local wind forcing (Gill, 1973;Graham et al., 2013;Stewart & Thompson, 2015)-mainly the along-coast component of the Antarctic easterlies-but also to buoyancy forcing, which is dominated by the seasonal sea ice cycle next to local input of glacial melt water (Hattermann, 2018;Núñez-Riboni & Fahrbach, 2009;Nøst et al., 2011). Furthermore, earlier studies suggest that complex processes such as dense water overflows, tides, and baroclinic instabilities can significantly influence the cross-slope and cross-shelf transport (Nøst et al., 2011;Stewart & Thompson, 2015;Stewart et al., 2018;St-Laurent et al., 2013).
The southeastern Weddell Sea, home of the large Filchner-Ronne Ice Shelf (FRIS), is of great climatic importance as dense water production on the 400 km broad continental shelf contributes significantly to the formation of Weddell Sea Deep and Bottom Waters (Nicholls et al., 2009). Both water masses are precursors of the globally relevant Antarctic Bottom Water (Foldvik, 2004), which ventilates the abyssal ocean (Reid & Lynn, 1971) and supplies the lower limb of the global overturning circulation (Sloyan & Rintoul, 2001). The Filchner Trough (FT, Figure 1) connects the deep ocean with the FRIS cavity, serving as outflow pathway of the potentially supercooled Ice Shelf Water (ISW; Darelius et al., 2014;Foldvik & Gammelsrød, 1988). ISW is formed by the interaction of High Salinity Shelf Water with the ice shelf base, resulting in temperatures below the surface freezing point (θ<−1.9°C). The trough also provides a potential inflow pathway for This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

10.1029/2020GL088119
Key Points: • Three 4-year-long mooring time series show anomalously warm and prolonged inflow along the eastern flank of the Filchner Trough in 2017 • Warm water persists on the shelf throughout winter and is associated with a fresh anomaly that leads to changes in the shelf density structure • We hypothesize that the fresh anomaly originated from anomalous summer sea ice melt upstream and caused shoaling of shelf break thermocline Supporting Information: • Supporting information S1 • Figure S1 • Figure S2 • Figure S3 Correspondence to: S. Ryan 2017). The seasonality has been linked to weaker along-coast wind stress during summer (all seasons are austral seasons), driving a shoaling of the ASF. Mooring data from the eastern shelf from 2013-2016 show a cessation of the warm inflow around June, followed by temperatures persisting around the surface freezing point or below throughout winter and spring (Ryan et al., 2017). In 2013, MWDW was observed in the vicinity of the Filchner Ice Shelf (FIS) front at the end of summer, as a results of a strengthened wind-driven coastal current (Darelius et al., 2016). While basal melt rates have been stable under FRIS to date (Paolo et al., 2015;Rignot et al., 2019), numerical model simulations predict that in a future warming climate MWDW may irreversibly enter the continental shelf via the FT and ultimately fill the FRIS cavity causing enhanced basal melting (Hellmer et al., 2012;Timmermann & Hellmer, 2013). FRIS is by volume the largest Antarctic ice shelf and buttresses several ice streams draining the Antarctic Ice Sheet (east and west) and therefore can potentially contribute several meters to sea level rise. Additionally, increased freshwater input from enhanced melting would ultimately affect the local dense water formation as shelf waters would become more buoyant leading to a shutdown of Ice Shelf Water formation and a flushing of the ice shelf cavity with MWDW (Hellmer et al., 2017). Thus, continuous efforts are being undertaken to monitor the warm inflow and understand key processes in the region.
This paper presents three extended mooring time series from the eastern flank of the FT that were installed from 2014 to 2018. The data reveal a particularly warm and prolonged inflow in 2017, compared to previous years. Hydrographic properties and possible forcing mechanisms of this event are investigated in this study.
Wind velocity at 10 m above sea level was obtained from the ERA-Interim reanalysis product with a horizontal resolution of 0.75° (Dee et al., 2011) and used to derive the surface wind stress following the method by Andreas et al. (2010), which implements changes to the drag coefficient depending on the sea ice concentration (see supporting information for more details on calculation). For consistency in the horizontal grid ERA sea ice concentrations are used for the calculation. For Box ups (green in Figure 1), the coordinate system is rotated to 245°to obtain the along-coast component. An extended time series (2002-2017) of polynya ice production rates derived for winter (April-September) is used (Paul et al., 2015b), which were computed from daily Moderate-Resolution Imaging Spectroradiometer (MODIS) median thin-ice thickness composites. Daily SSMI-SSMIS sea ice concentration data were obtained as a 5-day median-filtered and gap-filled product from the Integrated Climate Data Center in Hamburg (ICDC), Germany.

Prolonged and Exceptionally Warm Inflow in 2017
The hydrography along the eastern flank of the FT typically shows a distinct seasonal cycle with MWDW inflow during summer/autumn (−1.5 ≤ θ ≤ −1°C), followed by a decrease to near-freezing temperatures in winter, driven by a combination of seasonal suppression of the shelf break thermocline and local surface freezing and the development of deep mixed layers (Ryan et al., 2017; Figure 2). The seasonality is generally a robust feature in the 4-year record from 2013-2016. In 2017, an unprecedented strong and long warm inflow was observed across the whole mooring array (Figures 2a-2c) with maximum temperatures exceeding those previously observed by up to ∼0.5°C (i.e., σ > −1°C). The maximum temperatures are also associated with higher salinity (and density, Figures 2d-2f), which suggests the inflow to be composed of denser WDW, commonly found at greater depths over the continental slope. Alternatively, there has been less mixing with the overlaying WW during the cross-slope exchange.
In stark contrast to the previously observed seasonal cycle, temperatures at M 30.5W and M 31W remain above freezing until October, around −1.7°C and −1.5°C at ∼330 and ∼430 m, respectively. Thus, warmest water is found along the bottom. Cold and saline spikes between July and October near the bottom at M 31W (Figures 2b and 2h) suggest lateral/vertical movement of the ISW/MWDW interface past the mooring. The absence of these spikes at M 30.5W (Figures 2a and 2g) suggest that the interface remains to the east of it. Temperatures above freezing (∼−1.7°C), are also found over the slope at M 31.5W during winter (Figures 2c and 2i), however, the temperature is generally lower than at M 31W and M 30.5W , suggesting that the core of prolonged warm inflow is confined to the shallower eastern shelf. Observations at ∼77°S-384 m (Figure 2c, yellow line) show maximum temperatures above −1.5°C in autumn, which is close to the temperatures observed at this latitude during the strong warm inflow observed in 2013 (see M77 NORTH Darelius et al., 2016). In agreement with the observations at 76°S, temperatures remain above freezing between July and October. The LoTUS buoy was located at approximately 384 m depth; mooring observations in 2013 indicate that the main warm core follows the 400-500 m isobath and might therefore not have been fully captured by the LoTUS buoy. These results suggests that MWDW may have reached the FIS front in 2017, similarly to 2013 (Darelius et al., 2016).

Fresh Anomaly
After peak temperatures are reached in late autumn, decreasing temperatures are generally associated with a continuous freshening (Figures 2g-2i), due to increased entrainment of upper-layer WW as the shelf break thermocline deepens (Årthun et al., 2012). In 2014-2016, salinity on the eastern shelf (M 30.5W and M 31W ) reached a minimum in June/July (∼34.4) in concert with the cessation of the warm inflow and followed by a gradual salinity increase during winter. In 2017, however, the increase in salinity is absent and continuous freshening is observed instead over the eastern shelf. The salinity minimum is reached around October (S ≤ 34.4) and coincides with the termination of the warm signal on the eastern shelf (31°W, 30.5°W). The freshest water is found at middepth (upper sensors) on the eastern shelf, particularly at M 30.5W , suggesting advection by the coastal current as a potential source of the freshening.
In Temperature-Salinity (TS) space, MWDW lays on the mixing line between WDW and WW (gray line in Figures 2d and 2e). The MWDW observed on the eastern shelf during winter in 2017 indicates mixing with a slightly fresher end-member than in previous years (Figures 2d and 2e). These properties resemble Eastern Shelf Water (ESW), which is generally found upstream of the FT on the narrow continental shelf, where Ekman-driven downwelling transports surface water, further freshened by input of glacial meltwater, to greater depth (Fahrbach et al., 1994). Thus ESW typically has lower salinities compared to WW, while both have temperatures close to the freezing point (34.28 ≤ S ≤ 34.40; Carmack, 1974).
In 2014-2016, the ISW layer (cold, saline) extended onto the eastern shelf along the bottom from about October onward, with ISW being observed at M 31W almost continuously every autumn, while reaching as far as M 30.5W sporadically (Ryan et al., 2017;Figures 2a, 2b, 2g, and 2h). This is not observed in 2017. The combination of the freshening signal and the absence of salinification during winter, leads to significantly fresher conditions on the eastern shelf at the beginning of the following summer (December 2017) compared to previous years at this time. Hence, 2017 was not just characterized by a warmer and prolonged warm inflow but also a pronounced fresh anomaly. In the next section, the resulting changes in the vertical density structure and implications are investigated further.

Modified Shelf Density Structure
A Hovmoeller diagram of temperature at M 31.5W from 2016 to 2018 (Figure 3a) shows that generally the upper interface of ISW (−1.9°C isotherm), lays above 380 m during summer and deepens in the presence of the warm intrusions, extending down to a depth of approximately 500 m in late autumn. In 2016, temperatures between ∼380to 500-m depth decrease in late autumn and reach the surface freezing point in August, indicated by a sharp jump and subsequent vanishing of the −1.9°C isotherm. At the same time, ISW appears in the bottom layer at M 31W (not shown) in agreement with the seasonal cycle described by Ryan et al. (2017).
In 2017, significantly warmer temperatures ( > −1.2°C) are present ( Figure 3a) and extend down to 500 m in May/June. In contrast to 2016 (and likely to previous years), the ISW interface shoals only gradually while warmer water remains present at middepth throughout winter. In January 2018, the ISW layer ultimately extends to above the upper sensor, that is, ∼380 m. CTD profiles in the FT (depth > 500 m) from February 2018 show that the upper 400 m of the water column remain about 0.1 kg m −3 lighter (∼0.5-1 psu fresher) compared to 2013 and 2016 ( Figure 3b). This indicates that the observed fresh water anomaly during winter 2017 altered the density structure over the FT significantly. This is also visible in the comparison of CTD sections along 76°S from different years (2013,2014,2016,2018; see Figure S1).
MWDW entering the continental shelf in summer typically covers a density range between 27.7 and 27.8 kg m −3 (Figures 2d-2f). The bottom layer over the eastern shelf and at middepth (∼400 m) in the FT (at 76°S) had densities above 28.0 kg m −3 in summers (January/February) of 2013, 2014, and 2016 (Figures 3c and S1), thus "blocking" the lighter MWDW at these depth levels. In 2018 however, the hydrographic section at 76°S suggests that the modified density structure in the upper 400 m allows for an earlier warm water inflow at greater depth, in particular over the FT ( Figure S1) but also over the eastern shelf where in previous years MWDW did not appear at the lower sensor of M 31W until the second half of March (Figure 2 Ryan et al., 2017). This could potentially lead to larger southward heat transport in 2018, however no data extending beyond February 2018 is available yet.

Shoaled Thermocline
CTD profiles at the shelf break east of the FT (Figures 4a and 4b) show that in 2014 the thermocline is suppressed below the shelf break depth (∼600 m) close to the Brunt Ice Shelf (yellow), while in 2017, profiles at similar positions, suggest a shallower thermocline (Δ∼50-100 m, orange). Profiles in both years were obtained in the first half of February (8-10 February). In addition, warm temperatures of up to −0.5°C are observed along the bottom further south within a small trough cutting across the continental shelf between the FT and Brunt Ice Shelf (blue). In contrast, nearby profiles from 2014 (purple) show temperatures below −1.5°C, suggesting weaker or no warm inflow via this trough. CTD casts only provide snapshots in time, however, the combination of the shallower thermocline and the presence of warmer water on the shelf further south, indicate that a stronger heat flux onto the continental shelf is already occurring east of the FT in 2017, due to a shoaled thermocline.
To investigate upstream versus local wind forcing, we define two near-shore boxes (see Figure 1). The along-slope wind stress at the end of 2016/beginning of 2017 in both boxes does not show a strong negative anomaly, which would support a relaxation of the ASF, compared to climatological values (Figure 4c). There is even a general tendency to above-average easterly wind stress between November and February, thus, neither local nor remote wind forcing is favoring a shoaling of the thermocline during summer (Figure 4c). Below-average wind stress is found in both boxes during March and April, potentially leading to a relaxation of the thermocline. A 3-year time series from a shelf break mooring located at the mouth of the FT between 2007 and 2009 (S4E, Figure 1; Årthun et al., 2012) indicates that the thermocline generally reaches its shallowest position toward the end of March. Afterward it deepens rather rapidly within 1 month below the shelf break depth of around 450 m (Ryan et al., 2017), "shutting off" the WDW supply. Interannual variability at S4E shows that weak wind stress in autumn (April-June) 2007 is associated with a longer and shallower presence of WDW at the shelf break (Årthun et al., 2012; Figure S3). Thus, the autumn wind stress anomalies in 2017 potentially contributed to a shallower thermocline depth and a prolonged inflow, however, this remains speculative as no shelf break observation are available yet during this time. While the along-coast wind forcing in autumn 2017 was favorable for a prolonged onshore flow and southward transport of MWDW, the high temperature of the inflow and the suggested thermocline shoaling upstream of the FT cannot be explained by the wind stress.

Thermocline Preconditioning due to Sea Ice Melt
It has been suggested that, in addition to wind forcing, sea ice and the associated seasonal freshwater input during summer melting have an impact on thermocline dynamics by counterbalancing the wind-driven Ekman downwelling (Daae et al., 2017;Hattermann, 2018;Stewart & Thompson, 2016;Zhou et al., 2014).
A middepth fresh anomaly is observed at all moorings during the second half of 2017 (Figures 2g-2i), resulting in modifications of the vertical density structure on the shelf. This winter freshening is in contrast to the steadily increasing salinity at freezing point temperatures in the previous years (Figures 2g-2i), which had been attributed to a deepening of the mixed layer associated with brine rejection from winter sea ice production (Ryan et al., 2017). Estimates of 2017 (April-September) polynya ice production off Brunt Ice Shelf were not significantly below the multiyear average (Figure 4e). Reduced local sea ice formation and associated reduced salt input during winter are thus likely not the source of the observed fresh anomaly. We hypothesize that this anomaly has been advected into the region from further upstream via the coastal current. This is supported by the difference in the MWDW mixing line suggesting mixing with ESW instead of WW.
Exceptionally low sea ice conditions were observed in summer 2016/2017 in Box ups and Box loc (Figure 4d). Maps of sea ice concentration ( Figure S4) show early melting at the Greenwich Meridian in November and continuous melting along the coast with almost no sea ice cover in the southeastern Weddell Sea in February. The large-scale melting is likely to induce a freshwater anomaly along the coast due to Ekman transport of the fresh surface layer toward the coast. This is consistent with findings by Graham et al. (2013), who suggest that during summer a fresh anomaly develops near the Fimbul Ice Shelf (0-5°W) which then propagates southwestward along the coast. Assuming a coastal current velocity of 0.1-0.3 m s −1 (Graham et al., 2013) and 0.03 m s −1 for the Filchner inflow (Darelius et al., 2016;Ryan et al., 2017) the advective time scale from Greenwhich Meridian to the mooring location at 76°S ranges from 4-7 months. This is consistent with our observations which show the onset of the fresh anomaly in June/July and maximum in October. Furthermore, large-scale melting occurred along the whole coast upstream, likely causing a freshening of the entire continental shelf along the coast of Dronning Maud Land. Increased baroclinic instabilities along the ASF, due to larger horizontal density gradients, can drive a shoaling of the thermocline (Hattermann, 2018;Nøst et al., 2011) upstream of the FT, consistent with our shelf break observations.

Conclusions
Two years of additional mooring observations from the southern Weddell Sea continental shelf (76°S) show an unprecedented long-lasting and warm inflow of MWDW along the eastern flank of the FT in 2017, reaching at least as far south as 77°S. At all moorings, maximum temperatures exceeded −1.0°C, and temperatures around −1.5°C persisted on the continental shelf throughout winter. In contrast, previous observations showed a return to near-freezing conditions during winter/spring. Shelf break CTD profiles from February 2017 indicate a shoaled thermocline, which likely lead to the onshore transport of deeper and thus warmer MWDW. Darelius et al. (2016) attribute a strong warm inflow in 2013 to persistent below-average alongcoast wind stress that relaxed the shelf break thermocline. These conditions were not observed in 2017. Instead, we propose that freshened shelf waters from upstream along the coast of Dronning Maud Land contributed to the observed shoaling of the thermocline, via increased baroclinic instabilities.
TS properties of the MWDW suggest mixing of WDW with fresher ESW, typically found upstream on the narrow eastern Weddell Sea (Graham et al., 2013), in 2017 instead of WW. In addition, salinities observed in the FT associated with the prolonged inflow, were significantly below the 2014-2016 average. We hypothesize that the low salinity anomaly is an advected rather than a locally forced signal, as estimates of sea ice production in the FT area in the previous winter only showed average values. The anomaly possibly originated from enhanced summer sea ice melting upstream along the coast of Dronning Maud Land, from where it was transported southward within the ASC and ultimately contributed to a shoaling of the thermocline (Daae et al., 2017;Hattermann, 2018;Stewart & Thompson, 2016;Zhou et al., 2014). However, due to the lack of observations upstream, this remains speculative.
Our results suggest that the modified density structure on the continental shelf with lighter densities in the upper 400 m favors an earlier onset of onshore flow of MWDW at greater depth in 2018. This mechanism is different from the processes proposed by Hellmer et al. (2017), who suggested that less sea ice formation on the southern Weddell Sea continental shelf and associated decrease in ISW density, cause an irreversible flow of MWDW into the FT toward the end of this century. Our study shows that the projected upper ocean freshening, either by enhanced melting of sea ice or fringing ice shelves, along the coast of the eastern Weddell Sea might be an additional if not an earlier threat to the stability of FRIS and highlights a sensitivity to upstream processes, which are more susceptible to the changing climate. Therefore, in addition to continuing the hydrographic monitoring in the FT, upstream observations are urgently needed to constrain our proposed hypothesis and provide a better estimate on the vulnerability of the Antarctic Ice Sheet.

Data Availability Statement
The time series of sea ice production rates up to 2015 is available under https://doi.pangaea.de/10.1594/ PANGAEA.848612 (Paul et al., 2015a), and the extension will be provided upon request by Dr. Stephan Paul at the Ludwigs-Maximilian University Munich (stephan.paul@lmu.de). ERA-Interim data are available online (from https://apps.ecmwf.int/datasets/data/interim-full-daily/levtype=sfc/). Sea ice data can be downloaded online (at ftp://ftp-icdc.cen.uni-hamburg.de/asi_ssmi_iceconc/ant/