We used the Massachusetts Institute of Technology general circulation model (MITgcm) in its non-hydrostatic configuration (Marshall et al., 1997a, b) to model submarine ice melting and circulation in an idealised fjord 50 km in length and 5 km in width. In most simulations, the domain is uniformly 500 m deep. However, in some simulations, we include a sill which limits the overlying water depth to 100 m (uniform across the entire width of the fjord, and approximately 5 km wide in the along-fjord direction, with a Gaussian profile), centred 10 km from the open boundary (Fig. 1a). Model resolution is uniformly 500 m horizontally and 10 m vertically. The fjord sides are closed boundaries, while at the open ocean boundary we impose a 5 km sponge layer, in which conditions are relaxed towards those imposed at the boundary (e.g. Cowton et al., 2016; Sciascia et al., 2013; Slater et al., 2015).

The glacier-end of the domain is closed and consists of a virtual ice wall 5 km wide and 500 m high. In simulations incorporating subglacial discharge, this is input at a rate of 500 m3 s-1, a value typical of many of Greenland's tidewater glaciers (Mankoff et al., 2020a), at the centre of the base of the ice wall (Fig. 1a). The velocity of the subglacial discharge-driven plume (e.g. Fig. 1g) and the melting of the ice wall were calculated using the IcePlume package (Cowton et al., 2015). In common with several previous studies (Kimura et al., 2014; Slater et al., 2015; Xu et al., 2013), we implement a free slip condition on the fjord walls and ice front and do not simulate the effects of sea ice, atmospheric forcing or tides.

We use idealised representations of temperature and salinity profiles commonly observed at the mouth of Greenland's southeastern fjords during late summer as initial and open boundary conditions (Sutherland et al., 2014). In our standard set-up, this idealised profile is a cubic interpolation between 6 ∘C and 31 psu at the fjord surface, 0 ∘C and 34 psu at 100 m depth, 2 ∘C at 200 m and 3.5 ∘C at 500 m depth, where salinity is greatest at 35 psu (Fig. 1b–d). In this way, the upper several tens of metres represent waters that are seasonally warmed by solar insolation, whilst the relatively cold intermediate layer, centred 100 m below the fjord surface, represents the PW layer, which is underlain by warmer, more saline water representing the AW layer. Henceforth, we refer to this set of boundary conditions as BCstandard. In separate simulations, we use temperature minima at 100 m of -1 ∘C (PWcool) and 1 ∘C (PWwarm) and temperature maxima at 500 m of 2.5 ∘C (AWcool) and 4.5 ∘C (AWwarm) (Fig. 1e and f). Changing the temperature of the AW and PW layers causes corresponding changes in the vertical temperature gradient (Fig. 1e and f), the effects of which are discussed in Sect. 3.2. Initial and open boundary salinity are kept constant between simulations, but density changes between simulations are negligible. Boundary conditions were kept constant throughout each simulation. We focus on late-summer ocean conditions because of the greater availability of observations at that time to both force the model and with which to make comparisons.

Submarine iceberg melting is simulated using the IceBerg package within MITgcm (Davison et al., 2020), with an ice temperature of -10 ∘C (Inall et al., 2014; Luthi et al., 2002; Sciascia et al., 2013; Sutherland and Straneo, 2012). This package uses the velocity-dependent 3-equation melt rate parameterisation (Hellmer and Olbers, 1989; Holland and Jenkins, 1999; Xu et al., 2012). We chose to use this melt rate parameterisation rather than existing iceberg melt parameterisations (e.g. Bigg et al., 1997), because it enables us to resolve the vertical pattern of submarine melting of individual icebergs. The temperature and salinity fluxes associated with melting of individual iceberg faces within a grid cell are calculated based on local temperature, salinity and face-normal velocity. Face-normal current speed is calculated assuming that icebergs drift with the average current velocity along their draught (although we note that the iceberg locations are kept constant through each simulation). Melt-driven plumes are not simulated directly; instead, their effect on melt rates is parameterised by applying a minimum face-normal current speed of 0.06 m s-1 to each iceberg face. This minimum current speed is based on line plume modelling (Davison et al., 2020). The package does not include the effect of waves or mechanical iceberg break-up; therefore, melt rates calculated here are conservative. We use standard parameter values (Cowton et al., 2015; Davison et al., 2020; Jackson et al., 2020) for the drag coefficient (0.0025), and thermal and salt turbulent transfer coefficients (0.022 and 0.00062, respectively). The icebergs are rectangular in plan-view and have flat, vertical sides. All icebergs have length (l) to width ratios of 1.62:1 (Dowdeswell et al., 1992), and iceberg keel depth (d) is related to iceberg length through d=2.91l071 (Barker et al., 2004).

In Sect. 3.1, we consider a range of iceberg concentrations, maximum keel depths and size-frequency distributions, whilst using only the BCstandard boundary conditions. In all set-ups, iceberg concentration is uniform across the fjord and decreases linearly from a maximum adjacent to the virtual ice wall to a minimum 10 km from the open boundary. In Sect. 3.1, iceberg concentration (defined as the percentage of the fjord surface in plan-view occupied by icebergs), is 80 % adjacent to the ice wall and decreases to 5 % in our c1 experiment, and is reduced to 75 %, 50 % and 25 % of these values in our c0.75, c0.5, and c0.25 experiments, respectively. Regardless of concentration, we used a maximum iceberg keel depth of 300 m and the size-frequency distribution of the icebergs is described using a power law with an exponent of -2, which is similar to that observed in Sermilik Fjord (Sulak et al., 2017). In separate simulations, we assign maximum iceberg keel depths of 50, 150, 250, 350 and 450 m, whilst maintaining the c1 concentration and the -2 power law exponent. We then vary the size-frequency distribution power law exponent from -1.6 to -2.1 in increments of 0.1 (covering the range observed to date in Greenland's fjords (Rezvanbehbahani et al., 2020; Sulak et al., 2017)), whilst retaining the c1 concentration and the 300 m maximum keel depth. In Sect. 3.1 we show the results from simulations both with and without subglacial discharge, to demonstrate the effect of icebergs in isolation and in combination with subglacial discharge.

In Sect. 3.2 onwards we consider five realistic combinations of iceberg concentration, maximum iceberg keel depth and power law exponent, in order to approximate the range of iceberg geometries and distributions found in Greenland's fjords in summer (Fig. 2). In these set-ups, iceberg concentration decreases linearly in the along-fjord direction away from the glacier between specified maximum and minimum values (Table 1) and icebergs are distributed randomly in the across-fjord direction (Fig. 2). These iceberg set-ups range from those representing a fjord hosting few and small icebergs, such as Kangerlussuup Sermia Fjord (Sulak et al., 2017) (scenario 1), to those representing an iceberg-congested fjord, such as Sermilik Fjord (scenario 5) (Fig. 2; Table 1). In all simulations shown in this section (Sect. 3.2) 500 m3 s-1 subglacial discharge is injected into the fjord as described in Sect. 2.1.

The effect of iceberg melt on glacier-adjacent water properties depends on iceberg geometry, iceberg concentration and iceberg size-frequency distribution (Fig. 3) as well as on the presence or absence of subglacial discharge. In the absence of subglacial discharge, icebergs modify glacier-adjacent water properties (here defined as the average properties of the water within 2 km of the ice wall; Fig. 1a) in two main ways.

Firstly, they cause substantial (6–7.5 ∘C) cooling in the upper ∼60 m of the water column, relative to the initial conditions (Fig. 3a–c). The amount of cooling in this near-surface layer depends somewhat on iceberg concentration, with steady-state water temperature varying between ∼-1.5 and ∼0 ∘C over the range of iceberg concentrations considered but is otherwise relatively insensitive to changing iceberg geometry and distribution (Fig. 3a–c). Secondly, warming of up to ∼1 ∘C occurs below ∼80 m because iceberg melting causes localised freshening at depth. The resulting iceberg melt-modified water (i.e. the mixture of iceberg freshwater and ambient water at depth) is less dense than the surrounding water and rises buoyantly towards the fjord surface. The vertical extent and magnitude of the resulting warming generally increase with maximum iceberg keel depth (Fig. 3b), because icebergs with deeper keels cause upwelling of deeper AW (which in this case is also warmer (Fig. 1b)). This warming effect does not extend to the fjord surface, because the stronger stratification near the surface limits upwelling and because iceberg-ocean contact areas are much greater near the surface, so that cooling due to localised iceberg melting dominates. When subglacial discharge is included, the effect of iceberg melt on glacier-adjacent water properties at depth (below 60 m) is similar to that in simulations without subglacial discharge but glacier-adjacent water temperatures in the upper ∼60 m of the water column display a greater range and the cooling of the near-surface waters is considerably reduced (Fig. 3d–f). This is because the subglacial discharge causes strong upwelling of AW towards the fjord surface and increases rates of fjord-shelf exchange, which counters some of the iceberg-induced cooling of near-surface waters.

In reality, changes in iceberg concentration, keel depth and size-frequency distribution do not occur in isolation and there are characteristic relationships between those iceberg descriptors (Sulak et al., 2017). Fjords hosting large glaciers, such as Sermilik Fjord and Helheim Glacier in east Greenland, tend to contain both high iceberg concentrations and large, deeply draughted icebergs, whilst those with lower iceberg concentrations, such as Kangerlussuup Sermia Fjord, also tend to contain smaller icebergs. To better represent the range of iceberg conditions found in Greenland's fjords, we consider five iceberg scenarios (Fig. 2; Table 1), ranging from a fjord with low iceberg concentration, shallow iceberg keels and fairly uniform iceberg sizes (iceberg scenario 1), to a fjord with high iceberg concentration, deep iceberg keels and a large range of iceberg sizes (iceberg scenario 5). For each of these scenarios, we examine steady-state glacier-adjacent water temperatures for a range of ocean boundary conditions, with and without a shallow (100 m) sill. We therefore consider three different PW and AW temperatures in turn (Fig. 1e and f), and examine the resulting glacier-adjacent water properties for each of the five iceberg scenarios. To isolate the effect of iceberg melting from other processes, we compare each of the above simulations to identical simulations without icebergs.

Our simulations suggest that several changes to glacier-adjacent water properties can occur due to submarine iceberg melting. In almost all simulations, we simulate pronounced (>2 ∘C) cooling in the upper several tens of metres of the water column. Deeper in the water column (between ∼80 m and the maximum iceberg keel depth), both iceberg-induced cooling and warming can occur (e.g. Figs. 4 and 6), depending on the balance between cooling due local iceberg melting and warming due to melt-driven upwelling.

The balance between these processes depends on the iceberg contact area at depth available for local melting (and therefore cooling) and on the temperature of the upwelling water. When vertical temperature gradients are steep (e.g. with AWwarm; Fig. 6), icebergs can cause warming between their maximum keel depth and the surface layer. This is particularly apparent in the PW layer, where the temperature difference between an upwelled parcel of water and that at the parcel's new neutral buoyancy depth in the PW layer is greatest, and where iceberg melt rates (and therefore melt-driven cooling) are generally smaller because of the low water temperatures. In contrast, when vertical temperature gradients are shallower (e.g. with AWcool), cooling due to localised melting dominates (blue lines in Figs. 6d and e and 7c). These effects tend to reduce vertical temperature variations of glacier-adjacent waters compared both to simulations without icebergs and compared to conditions at the fjord mouth.

Detailed near-glacier hydrographic observations against which to make comparisons are sparse, but those that do exist provide some useful insight into the applicability of our model results to Greenland's fjords. The pronounced surface and near-surface cooling (relative to conditions at the mouth) that we simulate is a common feature in Greenland's fjords. For example, a transect of conductivity, temperature, depth (CTD) casts along Sermilik Fjord revealed cooling of approximately 4 ∘C in the upper ∼50 m (Straneo et al., 2011, 2012), which was also reproduced in a detailed modelling study of Sermilik Fjord that included icebergs (Davison et al., 2020). Similar along-fjord near-surface cooling has also been observed in other iceberg-congested fjords, such as Illulissat Isfjord (Beaird et al., 2017; Gladish et al., 2015) and Upernavik Isfjord (Fenty et al., 2016), both in West Greenland. In Illulissat Isfjord, the cold surface layer usually extends along-fjord to a shallow sill at the fjord mouth, where icebergs frequently become grounded (Gladish et al., 2015).

Iceberg-induced changes to water properties below ∼ 80 m are harder to identify in hydrographic observations, most likely because they also contain the signature of glacial plumes resulting from subglacial discharge, or other external forcings. Our modelling suggests that if vertical temperature gradients are shallow, then icebergs can cause cooling over large depth ranges (e.g. Fig. 7c). As one example, hydrographic observations in Kangerdlugssuaq Fjord showed relatively uniform near-glacier temperatures with substantial cooling in both the upper 100 m and between depths of 300 and 400 m, relative to a transect acquired at the fjord mouth (Straneo et al., 2012), consistent with the modelling results presented here. Iceberg melt-induced warming of parts of the water column is even harder to identify in published hydrographic observations because of the difficulty in distinguishing it from relatively warm subglacial discharge-driven plume outflow.

To further compare our modelling results to observations, we examined CTD casts acquired as part of the Oceans Melting Greenland (OMG) project (https://omg.jpl.nasa.gov/, last access: 14 April 2020; data available at: https://omg.jpl.nasa.gov/portal/data/OMGEV-AXCTD, last access: 14 April 2020). In keeping with our simulation design, we selected pairs of CTD casts acquired less than a week apart, one near or outside the fjord mouth and the other as close as possible to the tidewater glacier at the head of the fjord. These profiles (Fig. 8) show many of the characteristics that we have simulated here. Specifically, the profiles show that near-surface water temperatures are substantially colder adjacent to tidewater glaciers compared to those observed outside each fjord, and the observed temperature differences between the mouth and near-glacier region are comparable to those simulated here. In all but two of the surveyed fjords (Illulissat Isfjord and Timmiarmiut Fjord, shown in Fig. 8e and f), the profiles also show warming at intermediate depths (∼50–200 m) relative to the waters outside the fjord, consistent with our simulations using icebergs scenarios 3–5, particularly using our AWwarm boundary conditions (Fig. 6c–e). These observations do not allow us to quantify the relative contributions to intermediate depth warming between plume outflow and iceberg melt-induced upwelling. However, we note that the vertical pattern and magnitudes of intermediate depth warming are similar to those simulated here. In addition, the intermediate depth warming occurs over a large depth range, which is not easily explained by plume outflow and is consistent with our simulations. Some of the profiles also show notable cooling at depth (e.g. Illulissat Isfjord, Fig. 8e), which we are only able to reproduce in simulations including a shallow sill (e.g. the red line in Fig. 4j). Our simulations may underestimate cooling at depth because power law size-frequency distributions underestimate the number of very large icebergs (Sulak et al., 2017) and because the parameter values used in our melt calculation may underestimate submarine melt rates (Jackson et al., 2020).

In our simulations, we have generally considered a glacier-fjord system in which the glacier face and subglacial discharge interact with the entire water column, and with icebergs affecting a range of depths between the surface and their keels, which is a coarse representation of many fjords in Greenland. In many other fjords in Greenland, glacier grounding lines are shallower, such that the calving front and subglacial discharge interact predominately with the surface and PW layers. Although our simulations do not encompass this geometry, they still provide some insights into the potential effect of icebergs on near-glacier conditions in these fjords. With this geometry, subglacial discharge is injected directly into the PW layer. Therefore, plume outflow is relatively cool and we would expect, based the simulations presented here, that iceberg-driven cooling of the surface layer to be significant (resembling Fig. 3a–c). In addition, icebergs calved from such shallow glaciers would not be able to cause upwelling of warm AW (as in our scenarios 1 and 2), and so we would not expect any iceberg melt-driven warming of the PW layer. Overall, based on the insights gained from our simulations we expect that the effect of iceberg melt on near-glacier water properties in shallow fjords therefore largely manifests as a cooling in the upper several tens of metres of the water column, thereby reducing vertical variations in water column temperature. Such patterns have been observed in fjords hosting glaciers with relatively shallow (∼250 m) grounding lines resting in the PW layer (e.g. Mortensen et al., 2020).

If iceberg-induced changes to glacier-adjacent water properties significantly affect the magnitude and/or the vertical pattern of glacier submarine melting, then icebergs may play an important role in modifying glacier response to ocean forcing. To assess the effect of icebergs on glacier submarine melting, we first consider how iceberg melt impacts subglacial discharge-driven plume dynamics and then assess how the simulated temperature changes could affect melt rates across the parts of glacier fronts that are not directly affected by subglacial discharge-driven plumes.

To examine the effect of icebergs on subglacial discharge plume-driven glacier submarine melting, we evaluated plume properties for a single set of ocean boundary conditions (BCstandard; Fig. 1b–d) using each of the five iceberg scenarios. We find that submarine iceberg melting has negligible influence on plume vertical velocity and only a modest influence on plume temperature, meaning that plume-induced glacier submarine melt rates appear relatively insensitive to the changes in temperature and salinity induced by changes in iceberg geometry, concentration and size-frequency distribution (Fig. 9).

Although subglacial discharge-driven plume dynamics appear to be relatively insensitive to iceberg-induced modification of glacier-adjacent water properties, submarine melting distal to glacial plumes (“background melting” (e.g. Slater et al., 2018)) may be more directly affected. Qualitatively, the iceberg melt-induced changes to glacier-adjacent water properties presented above suggest that iceberg melt will affect background glacier melt rates in three key ways: (1) at and near the fjord surface cooling will reduce background melt rates, (2) in the PW layer background melting will usually increase due to upwelling of warmer AW, and (3) in the AW layer iceberg melt-induced changes in background melt rates are expected to be modest with slight increases in fjords with steep vertical temperature gradients and slight decreases in other fjords (assuming icebergs penetrate into the AW layer). These effects will be more pronounced in fjords with higher concentrations of larger (and thus deeper keeled) icebergs. In fjords where icebergs cause cooling near the surface and warming at depth, we expect icebergs will increase glacier undercutting through impacting submarine melt rates, which may in turn influence the rate and mechanism of calving (Benn et al., 2017; James et al., 2014; O'Leary and Christoffersen, 2013).

To explore these effects quantitatively, we calculate the percentage change in background melt rate of the glacier terminus due to iceberg-induced modification of glacier-adjacent water temperature (relative to simulations without icebergs). Modelling studies indicate that background melt rates scale linearly with ocean temperature (Sciascia et al., 2013; Slater et al., 2016; Xu et al., 2013); thus, changes in temperature, T, should cause proportional changes in background melting (Jackson et al., 2014). We choose to focus on relative changes in melt rate rather than absolute changes, because of poor constraints on important melt rate parameter values (Jackson et al., 2020). We calculate the relative change in submarine melt rate, SMR, following Jackson et al. (2014), as 1ΔSMR=Tib-Tf-Tnib-TfTnib-Tf100, where the subscripts “ib” and “nib” indicate simulations with “icebergs” and “no icebergs”, respectively, and Tf is the in-situ freezing point, given by 2Tf=λ1S+λ2+λ3z, where λ1-3 are constants representing the freezing point slope (-0.0573 ∘C psu-1), offset (0.0832 ∘C) and depth (0.000761 ∘C m-1), respectively (Cowton et al., 2015). S is the local salinity (horizontally averaged within 2 km of the terminus) and z is depth in the water column. It is worth noting that changes in melt rate calculated using this method assume that all changes in heat supply are accommodated by changes in submarine melt rates, and so this method provides an indication of the maximum relative changes in submarine melt rates expected due to changes in ambient ocean temperature.

Using this approach, we find that the impact on water properties resulting from iceberg melt substantially modifies background glacier submarine melt rates. Firstly, in the upper 50 m and using BCstandard, iceberg melt causes a 34.9 % reduction in melt rate on average. Even in iceberg scenario 1, iceberg melt causes a 29.5 % reduction in melt rate over this depth range. Secondly, between 100 and 200 m depth, iceberg melt causes a 13.5 % increase in melt rate on average when using BCstandard, but this increases to 59.2 % when using PWcool (for which warming of the PW layer due to upwelling is most pronounced). Changes in iceberg melt rates in the AW layer are minimal, with the most pronounced effect being a 5.4 % increase in the 200–400 m depth range using iceberg scenario 5 and PWwarm. When averaged through the entire water column, these effects largely compensate for each other, resulting in a net 3.1 % decrease in melt rates with BCstandard. Overall therefore, this analysis suggests that iceberg melt can influence the vertical pattern of glacier terminus background melting by decreasing melt rates at the surface and increasing them in the PW layer, with minimal changes in the AW layer.

As well as affecting glacier-adjacent water temperatures, iceberg melt likely affects submarine melt rates in other ways not examined here. For example, the cooling and freshening of the surface and near-surface layers induced by iceberg melting may prevent or hinder plume surfacing (De Andrés et al., 2020), and may expedite sea ice formation after the melt season, promoting the development of an ice mélange. In addition, mechanical iceberg break-up, iceberg calving and iceberg rotation can cause vigorous mixing of fjord waters, which can temporarily increase glacier and iceberg submarine melt rates (Enderlin et al., 2018), and increase the iceberg-ocean contact area available for melting. Iceberg melt-induced invigoration of fjord circulation can increase oceanic heat flux towards tidewater glaciers (Davison et al., 2020), likely resulting in faster terminus submarine melting. Icebergs likely also exert a mechanical influence on the circulation and plume dynamics at the ice-ocean interface (Amundson et al., 2020) and may prevent plume surfacing (Xie et al., 2019).

Current state of the art projections of dynamic mass loss from the Greenland Ice Sheet (Goelzer et al., 2020) are forced by far-field ocean temperature profiles, provided by ocean modelling output that does not include fjord-scale processes (except for the obstruction of shelf-water intrusion by shallow sills) (Slater et al., 2019, 2020). The results presented here suggest that such an approach is broadly appropriate for fjords with maximum iceberg keel depths of less than 200 m and iceberg concentrations less than ∼20 % on average, where iceberg modification of glacier-adjacent water properties appears to be limited other than in the upper several tens of metres (Figs. 4 and 6). The majority of Greenland's fjords likely fall into this category (Mankoff et al., 2020b; Sulak et al., 2017). Even in such fjords, however, this approach would not capture the surface and near-surface cooling caused by iceberg melting. In order to capture this near surface cooling, one relatively simple modification to such an approach could be to reduce surface water temperature to close to the in-situ melting point during winter periods, and proportionally to the iceberg surface area at the fjord surface during summer periods.

In fjords hosting icebergs with keel depth greater than or equal to 200 m and with average concentrations of more than ∼20 % (i.e. our iceberg scenario 3 or higher), iceberg modification of glacier-adjacent water properties becomes increasingly more important. In such fjords that also exhibit relatively shallow sills, icebergs act to cool glacier-adjacent water throughout the water column, with the amount of cooling proportional to the draught and concentration of the icebergs, as well as to the temperature of the ambient water at the fjord mouth (Fig. 4). In such fjords that do not have shallow sills, the effect is more complicated with both iceberg melt-induced warming and cooling, depending on the vertical temperature gradient of the water column and iceberg concentration at depth. Overall, these changes to the water column temperature can cause non-negligible (up to several tens of percent) changes in terminus submarine melt rates across the large areas of the calving front that are not directly affected by plume-inducing subglacial discharge. The vertical pattern of changes to terminus submarine melt rates (reduced near the surface and increased at intermediate depths) induced by iceberg melting is expected to exacerbate undercutting of glacier termini, with potentially important impacts on calving rates (Benn et al., 2017; Ma and Bassis, 2019; O'Leary and Christoffersen, 2013; Todd and Christoffersen, 2014). Although fjords hosting icebergs this large and numerous are relatively few in number, it is these fjords (and the glaciers hosted by them) that contribute the most to dynamic mass loss from the Greenland Ice Sheet (Enderlin et al., 2014; Khan et al., 2020).

All of the results presented here were extracted from the final 10 d of simulations that were run to a quasi-steady state (i.e. the variable of interest had stabilised). In our domains without sills, steady state of temperature and salinity was generally reached after just 10–30 d. However, our simulations with sills could take as many as 1000 d to reach such a steady state because fjord-shelf exchange is reduced. For an equivalent steady state to be reached in reality, open ocean conditions, subglacial discharge and iceberg size and distribution would also have to remain quasi-stable for an equivalent time period. In reality, this is unlikely to occur (particularly in fjords with shallow sills) because subglacial discharge and coastal and open ocean conditions change on subseasonal to seasonal time scales (Moon et al., 2017; Mortensen et al., 2014; Noël et al., 2016; Sutherland et al., 2014; Sutherland and Pickart, 2008). In reality therefore, glacier-adjacent water properties in fjords with shallow sills are likely a complex amalgamation of temporally evolving source waters, modified by processes operating within the fjord. In addition, some variations in coastal conditions can be transmitted towards glaciers very rapidly. During winter, strong wind events on the east coast of Greenland drive fast shelf-forced flows (or intermediary currents) in glacial fjords, delivering coastal waters to tidewater glaciers over just a period of a few days, and potentially reducing the magnitude of iceberg-driven modification (Jackson et al., 2014, 2018). Such currents are strongest in winter, when hydrographic observations are sparse, so this remains speculative.