Simulated melt rates for the Totten and Dalton ice shelves

The Totten Glacier is rapidly losing mass. It has been suggested that this mass loss is driven by changes in oceanic forcing; however, the details of the ice–ocean interaction are unknown. Here we present results from an ice shelf–ocean model of the region that includes the Totten, Dalton and Moscow University ice shelves, based on the Regional Oceanic Modeling System for the period 1992–2007. Simulated area-averaged basal melt rates (net basal mass loss) for the Totten and Dalton ice shelves are 9.1 m ice yr −1 (44.5 Gt ice yr −1) and 10.1 m ice yr −1 (46.6 Gt ice yr −1), respectively. The melting of the ice shelves varies strongly on seasonal and interannual timescales. Basal melting (mass loss) from the Totten ice shelf spans a range of 5.7 m ice yr −1 (28 Gt ice yr−1) on interannual timescales and 3.4 m ice yr −1 (17 Gt ice yr−1) on seasonal timescales. This study links basal melt of the Totten and Dalton ice shelves to warm water intrusions across the continental shelf break and atmosphere–ocean heat exchange. Totten ice shelf melting is high when the nearby Dalton polynya interannual strength is below average, and vice versa. Melting of the Dalton ice shelf is primarily controlled by the strength of warm water intrusions across the Dalton rise and into the ice shelf cavity. During periods of strong westward coastal current flow, Dalton melt water flows directly under the Totten ice shelf further reducing melting. This is the first such modelling study of this region to provide a valuable framework for directing future observational and modelling efforts.

The Antarctic ice sheet contains enough ice to raise sea level by over 58 m (Fretwell et al., 2013). Observations of global mean sea level rise over the period 1993-2007, indicate an average rate of 2.61 ± 0.55 mm yr −1 , with an Antarctic contribution of 0.43 ± 0.2 mm yr −1 (Church et al., 2011). Projections of sea level rise have a large uncertainty 10 in the contribution of Antarctica, due to lack of observations of ice discharge rates and surface mass balance (Gregory et al., 2012). The ice sheet flows towards the ocean through both broad slow ice sheet flow and many fast flowing glaciers. Ice shelves form in coastal regions where the ice thickness of the ice sheet or glacier is insufficient to maintain contact with the bedrock and the 15 ice begins to float on the ocean. An ocean-filled cavity, insulated from the atmosphere, forms beneath. The sub-ice shelf seabed can reach depths of over 2800 m below mean sea level (e.g. the deepest part of the Amery Ice Shelf cavity; Galton-Fenzi et al., 2008). The freezing point of seawater decreases with increasing pressure. Therefore, at the back of deep ice shelf cavities, the decreased freezing point of seawater provides 20 a large potential for thermal driving of melting.
Increased melting leads to the localised thinning of an ice shelf, and potentially to the acceleration of the glacier behind it, as the buttressing effect of the shelf is decreased (Dupont and Alley, 2005). An acceleration in flow rate of a glacier causes a mass budget imbalance in the surrounding ice sheet due to the longer response time Introduction sheet susceptible to oceanic changes (Dupont and Alley, 2005). The ice shelf-ocean interaction is therefore an important control on the discharge of grounded ice into the oceans and subsequently on sea level. Recent satellite observations have shown rapid and accelerated thinning of glaciers along the coastal margins of Antarctica (Pritchard et al., 2009). For example, Pine Island Glacier thinned at rates up to 6 m yr −1 while 5 the Totten Glacier displayed thinning rates of 1.7 m yr −1 , three times the rate previously reported (Rignot, 2006). The high thinning and rapid retreat rates of marine terminating glaciers suggests a common oceanic driving, for example through increased basal melting (Pritchard et al., 2012). Glaciologically derived estimates of area-averaged basal melt rate for the Totten ice 10 shelf are 20 ± 9 m yr −1 (Rignot, 2002); 26 ± 8 m yr −1 (Rignot and Jacobs, 2002); and 10.5 ± 0.7 m yr −1 . These estimates are derived from satellite Interferometric Synthetic Aperture Radar (InSAR) observations of ice surface velocity and the first two are calculated as the difference in flux of ice from the Totten Glacier across the grounding line and across a flux gate 10-30 km downstream, divided by the area 15 of enclosed ice shelf. These measurements should thus be treated as an average melt rate for the grounding line region. The latter estimate was also derived from InSAR observations, but was calculated using the volume flux divergence for the whole ice shelf area. The last estimate is equivalent to a mass loss rate of 63.2 ± 4 Gt yr −1 . The InSAR estimate for the Moscow University ice shelf show basal melt 20 of 4.7±0.8 m yr −1 and a mass loss rate of 27.4±4 Gt yr −1 . It should be noted that this work defines the Dalton ice shelf as being separate to the Moscow University ice shelf. We define the Dalton ice shelf as the ice shelf separated from the ocean by the grounded peninsula which runs north-east from 67.2 • S, 118.5 • E. This is done since the grounded peninsula is likely to protect the Dalton ice shelf from oceanic 25 influences, and will change the mass loss characteristics of the Dalton as compared to the Moscow University ice shelf. These estimates, show the Totten and Dalton basal melting and mass loss to be exceptionally high compared to other ice shelves in East Antarctica   (Hellmer et al., 10 2012;Jacobs et al., 2011); regional upper-ocean hydrography (Hattermann et al., 2013); dynamic eddy-scale activity (Moffat et al., 2009); a link with sea ice conditions (Holland et al., 2010); polynya activity (Cougnon et al., 2013); wind-driven currents (Dinniman and Klinck, 2004); and changes in the Antarctic Circumpolar Current (Gille, 2008;Böning et al., 2008).

15
Here we examine the temporal variability and driving mechanisms controlling the basal melt rates of the Totten and Dalton ice shelves using a numerical model. The Totten and Dalton ice shelves are prime candidates for investigation, due to their rapid thinning rates, high estimated basal melt and the limited existing research. Section 2 presents different hypotheses for drivers of basal melting. Section 3 provides an Introduction  . Ice shelf thinning by increased basal melt suggests increased oceanic heat available for melting at the ice shelf cavity . Increased heat supply to ice shelf cavities could result from intrusions of warm water across the continental shelf or changes in atmospheric forcing, such as buoyancy fluxes and wind stress. This paper investigates the mechanisms that influence the oceanic heat supply 5 that drives basal melting.

Circumpolar Deep Water intrusions
Increased heat supply for ice shelf melting has been suggested to be due to periodic incursions of modified Circumpolar Deep Water (MCDW) onto the continental shelf which can then flow along isopycnals and be delivered to the ice shelf base (Jacobs et al., 1996). Observations near and beneath several Antarctic ice shelves show the signature of MCDW. In the case of PIIS, which is showing high melt rates, water measurements taken from an autonomous underwater vehicle show MCDW mixing with cold melt water. This indicates that MCDW is responsible for the high melt rates of PIIS Jacobs et al., 2011).

15
In the Totten Glacier region, Conductivity, Temperature and Depth (CTD) measurements have indicated the presence of MCDW on the south side of the continental shelf break, ∼ 200 km from the Totten ice shelf (Williams et al., 2011). However, no observations yet exist to determine the temporal and spatial variability or causal mechanism of the MCDW incursion. Consequently, MCDW can not be definitely attributed as the 20 water source driving melting of the Totten ice shelf.

Interaction with bathymetric features
Observations suggest that intrusions of warm water onto the continental shelf is highly dependent on bathymetry features (e.g. Martinson and McKee, 2012 bathymetry features such as the shelf break (as opposed to small scale sills and ridges) can also lead to net onshore intrusion. Inertia of the along-slope water flow can carry the water up the slope and onto the shelf, depending on the curvature and orientation of the continental shelf (Dinniman and Klinck, 2004;Klinck and Dinniman, 2010). In the case of Pine Island Glacier, modelling has shown warm waters being channelled 5 through irregular bathymetric features and submarine troughs (Thoma et al., 2008). The bathymetry in front of the Totten ice shelf is suggestive of possible pathways of MCDW intrusion onto the continental shelf at ∼ 120 • E.

Eddies
Eddies may also transport and mix heat across the shelf break. Observations in the 10 western Antarctic Peninsula have characterised eddies as transporting Circumpolar Deep Water onto the continental shelf (see Moffat et al., 2009). The Antarctic Circumpolar Current is recognised as an important generator of eddies, which carry heat poleward (Rintoul et al., 2001). Closer to the continental shelf, interaction of the westwards flowing coastal current with bathymetric features, such as ridges along the shelf break 15 may lead to eddy generation. Numerical models have shown topographic Rossby wave breaking is an important factor for generating eddies, which carry a significant amount of heat across the continental shelf (St-Laurent et al., 2013). Direct modelling of eddy generation and evolution is limited by model grid resolution and bathymetry detail. Eddy-permitting flow can be simulated with horizontal grid 20 resolutions of approximately 2-3 km (Klinck and Dinniman, 2010), however, a horizontal grid resolution of 1 km is required in order to resolve individual eddies (St-Laurent et al., 2013). The modelling work presented here (with a horizontal grid resolution of 2.5-3.5 km) is eddy-permitting, where the mean flow and associated interactions are resolved. Introduction

Exchange with the atmosphere
In analysing the source of increased heat supply to the Totten ice shelf region, we must also consider the effect of the atmospheric forcing on the ocean. Atmospheric forcing consists of transfers of momentum (surface wind stress) and buoyancy (heat and/or salt fluxes), such as the forcing of dense water formation in polynya regions.

5
In some cases, local recurring conditions favour wind stress patterns that drive currents over the continental shelf break and on to shelf. For example, modelling of PIIS showed a strong correlation between flow of CDW onto the continental shelf and local synoptic wind stress conditions (Thoma et al., 2008).
The polynya located to the east of the Totten ice shelf produces dense High Salinity Shelf Water (HSSW). HSSW flow may interact in different ways with the local environment, and until observations are taken, modelling presents the only method to understand the implications of the polynya for the oceanography of the region. Kämpf (2005) suggested that dense water cascading down submarine canyons and off the shelf break could induce upwelling of deeper water onto the shelf, which was con-15 firmed in numerical simulations and laboratory experiments (Kämpf, 2007). Alternatively, HSSW produced by polynya activity could be guided by the westwards flowing coastal current, directly towards the mouth of the Totten ice shelf cavity, potentially cooling sub-ice shelf waters and quenching melting. Ice concentration observations taken by the Special Sensor Microwave Imager are 20 used to infer heat and salt flux out of the ocean (Tamura et al., 2008). The heat flux out of the Dalton ice tongue polynya using the method of Tamura et al. (2008) is shown in Fig. 2a, where a positive value indicates loss of heat to the atmosphere and formation of HSSW, and is considered as a proxy for polynya activity. A peak in polynya activity coincides with the Austral winter, due to strong wind over the region (see Fig. 2a, green line) and low air temperatures. Long-term polynya activity can be analysed by considering the polynya heat flux anomaly -calculated as the difference of the heat flux from the 1992-2007 mean, OSD 10,2013 Simulated melt rates for the Totten and Dalton ice shelves smoothed by a 12 month moving average filter, as shown in Fig. 2b. The heat flux anomaly for the Dalton ice tongue polynya region is derived from the buoyancy flux data (see Tamura et al., 2008) used to force the open ocean surface of the model. A positive heat flux anomaly indicates above average heat loss from the ocean and stronger than average polynya activity and thus, the resulting signal is a proxy for interannual polynya 5 strength.

Description of region
The Totten ice shelf (centred at 67 Antarctica (See Fig. 1). The Sabrina coast is characterised by marine terminating glaciers, ice shelves (such as the Totten, Moscow University and Dalton ice shelves), large tracts of year-round ice adjoining the land (fast ice), extensive seasonal sea ice growth and a region of strong sea ice formation (the Dalton ice tongue polynya). The continental shelf seas are approximately 1000 m deep, with the shelf break occurring between 65 • S to 65.5 • S in this region. The Antarctic Circumpolar Current is located in the deep abyssal waters to the north of the continental shelf break (Orsi et al., 1995).

15
The Antarctic Circumpolar Current is composed of several water masses, with Antarctic Surface Water overlying relatively warm and salty Circumpolar Deep Water (CDW). Poleward mixing of CDW with cold and dense Antarctic waters produces "modified" CDW or MCDW (Orsi et al., 1995), at the southern boundary of the Antarctic Circumpolar Current. 20 The embayed position of the Totten ice shelf, located on the eastern flank of Law Dome and with a broad continental shelf, makes it relatively isolated from the eastwards flowing Antarctic Circumpolar Current. The two main oceanic currents near to the coastal ice shelves flow westwards. The Antarctic Coastal Current (ACoC) flows within 50 km of the coast and is associated with the east wind drift. The Antarctic Slope 25 Current (ASC) is topographically controlled and is an important influence on transport across the continental shelf break (Jacobs, 1991). 10,2013 Simulated melt rates for the Totten and Dalton ice shelves A large polynya forms west of the Dalton ice shelf tongue (Massom et al., 1998). Polynyas form when strong gravity-driven katabatic winds flow down the Antarctic ice sheet and turn westwards under Coriolis acceleration, removing freshly formed sea ice. Heat is then lost from the uncovered ocean, leading to more sea ice formation and brine rejection as a latent-heat polynya. These are important as they produce large 5 quantities of HSSW, which contribute to bottom water formation (Rintoul et al., 2001), an important driver of the global climate system (Jacobs, 2004). The easterly katabatic winds are an important controlling factor in polynya strength (Massom et al., 1998). As the ACoC is also controlled by the strength of the easterly winds, a correlation between the strength of both the polynya and the ACoC is to be expected. 10

Model details
A modified version of Regional Oceanic Modeling System (ROMS; Shchepetkin and McWilliams, 2005), a 3-D primitive equation finite difference model, is used to simulate ice shelf-ocean interaction. ROMS was modified to include realistic tidal forcing, thermodynamic interaction between ocean and the ice shelf (see 15 2012; Dinniman et al., 2007) and frazil thermodynamics (see Galton-Fenzi et al., 2012). A three-equation formulation parameterises ice-ocean thermodynamics, following Holland and Jenkins (1999).
The model domain (part of which is shown in Fig. 1 ness, T C , is set to a nominal depth, T C = 300 m, representative of the typical thickness of an ice shelf cavity along the cavity centreline (Galton-Fenzi et al., 2008) while along the grounding line, T C = 0. In the grounding line region, the centreline depth is not set. The choice of T C along the centreline is justifiable, as uncertainty in this depth should only affect the spatial distribution of melt/freeze regions  and not open ocean circulation and heat transport into the cavity. Furthermore, the agreement between modelled basal melt rates and glaciological estimates (see Sect. 5.1) attests to the scale of the ocean cavity being of the correct order. Between the centreline and grounding line, T C is linearly interpolated. Fast ice is included from the maps of Fraser et al. (2012) with a 5 m draft (Massom et al., 2001). Seasonal sea ice formation is parameterised by the heat and salt flux algorithms of Tamura et al. (2008). Currents, heat and salt fluxes on the lateral boundaries are relaxed to monthly values from the ECCO2 reanalysis product (Menemenlis et al., 2008). The surface is forced with wind stress from COREv2 (Large and Yeager, 2009) and the buoyancy fluxes of Tamura et al. (2008).

15
The model was spun up by 32 yr of repeated 1992 forcing, prior to observed forcing for the period 1992-2007, following Cougnon et al. (2013).

Computed ice shelf melt rate
Melting and freezing at the ice-ocean interface is calculated for the ice in the domain. 20 Area-averaged melt rates for the Totten and Dalton ice shelves are shown in Fig. 2c, while the spatial melt rate distributions are shown in Fig. 3. These are at the same times as the corresponding circulation patterns in Fig. 4. In Fig. 2c Totten and Dalton ice shelves, which are used for calculating the area-averaged melt rates and mass loss, are defined as 5402 km 2 and 5113 km 2 , respectively. In the grounding line region, we expect the melt rate calculated using this method to be among the highest rate found under the Totten ice shelf. The model area-averaged melt rate in the Totten grounding line region ranges from 16 m ice yr −1 to 45 m ice yr

Ocean circulation features
Simulated model currents are shown in Fig. 4a-d. We find four different atmospheric and ocean circulation regimes, and subfigures in Fig. 4  The ASC is also reproduced, flowing westwards along the continental shelf break (see Fig. 4b). The current is channelled by bathymetric features, such as where the shelf break curves north-west at 116 • E, 64.5 • S and approximately flows between the 2000 m and 2500 m isobaths. Currents for the ASC can be as high as 20 cm s −1 , but have a mean flow of approximately 5 cm s −1 . These model simulated currents agree 5 reasonably with sparse nearby observations (see Bindoff et al., 2000;Williams et al., 2011). Modelling studies of circum-Antarctic ASC transport predict similar velocities . The bathymetry of the continental shelf break acts to guide flow onto the shelf. The ASC interacts with several barriers as it flows westwards through the model domain. 10 The eastern edge of the rise located at 122 • E, 66 • S, henceforth named "Dalton rise" can act to guide water southwards towards the mouth of the Dalton ice shelf. West of the Dalton rise, a large basin with an average depth of 750 m exists, henceforth named the "Totten basin". Interaction of the westwards flowing ASC with the the western edge of the Dalton rise generates a southwards flowing jet clockwise around the Totten basin.

15
This interaction, which is dependent on the meridional location of the ASC front, floods the Totten basin with warmer water, supplying heat that is available to drive the melting of the Totten ice shelf.
Observations to support these model results do not exist, as the region is covered by sea ice for much of the year. This makes long term estimation of sea surface currents 20 by satellite observation impossible, and ship-borne observations sparse and difficult to acquire.

Heat
In general, temperature at each layer of the model decreases polewards. The ASC is typically between 0 • C to 0.4 • C, though at several points during the 1992-2007 time pe- 25 riod, the temperature rises to approximately 0.5 • C. The water that makes up the ACoC (−1 • C to −0.5 • C) is cooler than the ASC, as a result of mixing with cold meltwater from glaciers and ice shelves, and heat lost to the atmosphere.

Discussion
Model results show variability in melt rate over the course of each year. However, the interannual Totten melt rate, the dashed black line in Fig. 2c, shows variation from high to low melt rate and vice versa, with a ∼ 2-3 yr period. The ∼ 2-3 yr modulation of melt rate signal is noticeable as an increasing annual average melt between 1992 to 5 1994, 1998 to 2002, 2004 to 2007 and decreasing annual average melt between 1994 to 1998 and between 2002 and 2004. This pattern of increasing and decreasing melt repeats over the 1992-2007 period. However, there is no overall trend between 1992 and 2007 Totten ice shelf melt discernible above the ∼ 2-3 yr melt variation. Thus, we focus analysis on understanding the processes governing simulated interannual 10 variability in melt rates. Satellite-derived estimates of ice shelf thinning indicate that for the period 2003-2008 thinning of the Totten ice shelf was the fourth highest in Antarctica (Pritchard et al., 2012). Over the period 2003-2007, the model area-averaged melt rate increases to its highest value, in agreement with an ice shelf that is rapidly thinning.

15
Oceanic heat is supplied to both ice shelves across the continental slope. However, variability in heat supply at the ice shelves is also dependent on the nearby Dalton ice tongue polynya. HSSW produced by the polynya can flow westwards, reducing supply of heat into the Totten ice shelf cavity. We proceed by examining model currents and atmosphere-ocean exchange together with variability in ice shelf melt rate.

Atmosphere-ocean exchange
Strong katabatic winds that drive polynya activity, are also partly responsible for driving the ACoC. Generally, as polynya activity increases from March to April, large quantities of HSSW are formed in the model. When the polynya is strongly active, the ACoC is also strong and the resulting westward flow directs most of the HSSW towards the Tot-Introduction into the Totten basin), access to the ice shelf cavity to drive high melt of the Totten ice shelf. Times when HSSW production and ACoC are strong leads to cold, dense HSSW entering the ice shelf cavity and decreased melting. However, the timing of the peak in polynya activity and thus HSSW production in June lags the minimum of Totten ice shelf 5 melt (compare Fig. 2c and a). Increasing polynya activity causes the rate of change of melt rate to become negative and so the melt rate passes through a local maximum. The melt rate continues to decrease until polynya activity decreases, at which point the rate of change of melt rate become positive. The melt rate then passes through a local minimum and increases. The lag between polynya maximum and Totten melt minimum 10 is created because the bulk volume of water below the ice shelf requires time to cool as HSSW is added and time to heat up as warm water is added. The Dalton ice shelf exhibits a similar temporal pattern of melt rate as the Totten ice shelf, suggesting that polynya activity is an influential factor on Dalton melt rate as well.
The heat flux anomaly for the Dalton ice tongue polynya region is shown in Fig. 2b.

15
Here, a positive signal indicates stronger than average heat loss to the atmosphere and stronger HSSW production. From 1993 to 1995, the heat flux anomaly is negative during the middle of the year, suggesting weaker peak polynya strength. Consequently, both ice shelves display strong melting. From 1996 to 1998, the heat flux anomaly is strongly positive, 20 leading to strong HSSW production and the Totten progressively decreasing in melt rate during this period. From mid-1999 to mid-2002, the heat flux anomaly is negative and consequently, HSSW production drops and oceanic heat supply increases to both ice shelves. Annual average melt is thus higher for this period than for the 1996-1998 period when the heat flux anomaly was strongly positive.

Shelf exchange processes and heat supply for melting
The primary source of heat for melting is supplied across the continental shelf. 10 Bathymetry is important for controlling intrusion of warm water by guiding the slope current. Model results show that regional topographic features (the Dalton rise and Totten basin) act to guide water onto the continental shelf. These features are regions of intermittent warm inflow. The mechanism by which the flow of the ASC can be diverted by bathymetry is probably different for each topographic feature.

15
Geostrophic flows tend to conserve potential vorticity f /H, where f is the Coriolis parameter (which varies with latitude) and H represents the water depth, leading to the tendency of geostrophic flows primarily following isobaths. If a change in bathymetry is encountered, resulting flow will tend to divert to a different latitude to counter the change in depth. In the model results, we see this by the poleward current generated 20 as the ASC interacts with the edge of the Totten basin. The narrow current formation is supported by idealised modelling studies, which suggest a significant onshore flow generated by troughs, in agreement with previous dynamical understanding ( St-Laurent et al., 2013).
Melting of the Dalton ice shelf is strengthened by occasional flow of warm water into the basin of Paulding Bay, east of the Dalton rise. Model currents suggest counterclockwise circulation forming in Paulding Bay when the ASC impacts with the north-

Tables Figures
Back Close

Full Screen / Esc
Printer-friendly Version

Interactive Discussion
Discussion Paper | Discussion Paper | Discussion Paper | Discussion Paper | eastern tip of the Dalton rise. The resulting circulation is onto the Dalton rise and into the ice shelf cavity (see Fig. 4b).
The difference between the simulated Totten and Dalton melt rates is shown in Fig. 2d. A positive value indicates the Totten melting is lower than Dalton while a negative value indicates Dalton melting is weaker than Totten. The signal is smoothed by 5 a 6 month moving average and ±1 standard deviation is shaded. A difference greater in magnitude than 1 standard deviation indicates periods when the melt rates of the Totten and Dalton ice shelves are significantly uncorrelated. Since polynya activity drives the melt rate signal on a seasonal timescale, significantly different melt rates indicate times when the melt rate of at least one of the ice shelves is not purely driven by 10 polynya activity.
Late 1996, late 2000, late 2003 and late 2007 are times when there is strong supply of heat to the Dalton ice shelf, but not to the Totten ice shelf (See Fig. 2d). During these times, oceanic currents supply heat over the Dalton rise only to the Dalton ice shelf, but polynya activity and formation of cold, dense HSSW is strong enough to limit Totten 15 melting (see Fig. 4b). Circulation also shows cold Dalton melt water leaving the cavity and flowing westwards to the Totten ice shelf cavity.
Late 1998, early 1999 and late 2007 are times when the Totten ice shelf is strongly melting but the Dalton ice shelf is not. Polynya activity at these times is weaker than average, leading to lower rates of HSSW production. Model circulation at these times 20 show weak flow of HSSW to the Totten ice shelf (see Fig. 4c). Warm water that intrudes into the Totten basin can consequently enter the ice shelf cavity and drive increased melting. Weaker westwards ACoC flow results in HSSW flowing into the nearby Dalton ice shelf cavity, decreasing Dalton melting. 25 The melt rates displayed by both the Totten and Dalton ice shelves are a complex combination of a seasonal signal with occasional periods of very high and very low melt. Comparing melt rate and polynya activity, as shown in Fig. 2c  in general, periods of lower melt are correlated with periods of stronger polynya activity. However, there were several short periods when ocean circulation features caused the melt rate of the Dalton and Totten ice shelves to be out of phase. This suggests that a combination of atmosphere-ocean exchange and shelf break flow mechanisms is required to fully explain the melt rate of the Dalton and Totten ice shelves. 5 We propose a set of four main states of circulation and polynya activity, that drive melt of the Dalton and Totten ice shelves. These four states describe the configuration of on-shelf flow and polynya activity, and explain the seasonal and interannual variation of the melt rate of the Totten and Dalton ice shelves. These states are described as follows: both ice shelves. This state is illustrated in Fig. 5a and shown in model currents in Fig. 4a and melt rate anomaly distribution in Fig. 3a.

Melt rate forcing by a combination of polynya activity and circulation
-State B: Polynya activity is strong, but now the ASC interacts with the bathymetry offshore of the Dalton rise, which gives rise to a warm current over the Dalton rise, directly to the mouth of the Dalton ice shelf. The increased melting leads 20 to a strong meltwater outflow westwards towards the Totten ice shelf, inhibiting melting of the Totten. The Totten ice shelf shows very low melt rates, while the Dalton ice shelf displays strong melting, as illustrated in Fig. 5b and shown in model currents in Fig. 4b and melt rate anomaly distribution in Fig. 3b. little warm inflow, the lowered HSSW production is enough to inhibit melting. This state is shown in Fig. 5c and shown in model currents in Fig. 4c and in melt rate anomaly distribution in Fig. 3c.
-State D: Polynya activity and ACoC strength is much weaker than in States A and B. As a result, HSSW production is lowered and transport of dense, cold water 5 to both the Dalton and Totten ice shelves decreases. The ASC interacts with the Dalton rise, leading to a warm currents flooding the Totten and the Dalton ice shelves. Both ice shelves thus display high melting, but the weak ACoC means that Dalton melt water (which now flow eastwards) does not limit Totten melting as in State B. This state is illustrated in Fig. 5d and shown in model currents in The interactions described by these four states, represented in Fig. 6, summarise the melt rate controls of the Dalton and Totten ice shelves. This interaction chart shows the strong negative impact of weak polynya activity and ACoC strength on heat supply to both ice shelves. Warm water intrusions that flow over the shelf break can interact with 15 the shelf bathymetry and supply heat to either ice shelf, which will result in enhanced basal melting. If heat is supplied to the Dalton ice shelf, increased melting may lead to a decrease in heat supply for the Totten ice shelf (via outflow of cold Dalton melt water).
These interactions provide a hypothesis describing likely causes of melting of these under-researched ice shelves. A similar mechanism has been proposed linking basal 20 melt of the Mertz Glacier Tongue and a nearby polynya (Cougnon et al., 2013). This is also the first evidence of the interaction between two ice shelves, where the melt rate of one ice shelf affects the melt rate of the other. This indicates that the melt rate and glaciology of the Totten ice shelf cannot be studied in isolation from the Dalton ice shelf. Introduction

Conclusions
Remote sensing observations show thinning of the Totten Glacier and suggest strong basal melting of the Totten ice shelf, driven by changing ocean conditions as the most likely explanation. It is generally believed that enhanced exchange of heat across the shelf break, through a series of possible processes, is the main mechanism causing 5 increased basal melting. This study shows that, along with exchange of heat across the shelf break, atmosphere-ocean interaction processes on the continental shelf can modify oceanic heat supply to ice shelves. Dense water formation in the Dalton ice tongue polynya strongly modulates the seasonality of melting for both ice shelves. Melting of the Totten ice shelf is increased when 10 the interannual strength of the Dalton polynya is below average, and vice versa. The Dalton ice shelf melt rate is primarily controlled by intrusions of warm water across the Dalton Rise and into the ice shelf cavity. Simulated area-averaged melt rates for the Totten ice shelf agree well with recent glaciological estimates, and suggest the Dalton ice shelf, like the Totten, as being a region of high basal melting.

15
Furthermore, this study suggests important evidence for the interaction between two ice shelves, where the melting of one directly influences the melting of the other. Lastly, strong interannual variability in heat supply to the ice shelves and subsequent melt rates suggest that observations will need to be long to correctly determine attributable mechanisms. 20 This is the first such modelling study of this region, and will provide valuable information for directing future observations. OSD OSD 10,2013 Simulated melt rates for the Totten and Dalton ice shelves   OSD 10,2013 Simulated melt rates for the Totten and Dalton ice shelves  6. The interaction between the Totten and Dalton ice shelves, the Dalton ice tongue polynya and on shelf flow is summarised. A negative interaction is shown by , while a positive interaction is shown by ⊕. Strong polynya and ACoC flow has a negative impact on warm water supply to both ice shelves, while warm water flow into the Totten basin and over the Dalton rise has a positive impact on basal melting of the Totten and Dalton ice shelves, respectively. Increased basal melting of the Dalton ice shelf can have a strong negative affect on the Totten ice shelf. The dependence of this last interaction on strong ACoC flow is shown by the dashed line.