Ice shelves, the floating extensions of ice sheets, play a critical role in determining the Antarctic Ice Sheet mass balance by modulating the discharge of grounded ice from Antarctica. The mass loss of Antarctica Ice Sheet with rapid thinning of the ice shelves and grounding line retreat (Rignot et al., 2014) is mainly attributed to basal melt rates through ocean heat transport to the ice shelves (Paolo et al., 2015). More than half of the net loss of ice for the Antarctic ice shelves could be accounted for by basal melting in the sub-ice shelf cavity (Depoorter et al., 2013; Rignot et al., 2013).

The basal melting of ice shelves occurs through various modes that are directly related to the water mass circulation into their cavities: inflows of cold or warm deep water (Mode 1 or 2), or warm surface water (Mode 3) (Jacobs et al., 1992). These circulation modes are closely linked to a variety of oceanographic thermal regimes in which ice shelves exist. Namely, ice shelves are characterized as ‘cold-water cavity’ or ‘warm-water cavity’, depending on whether the sub-ice shelf waters are dominated by cold dense shelf water (e.g., High Salinity Shelf Water; HSSW) or relatively warm Circumpolar Deep Water (CDW), respectively. The interactions of ice shelves with the various oceanic thermal environments, therefore, result in significant differences in the states of ice shelves.

Warm-cavity ice shelves, in which the inflow of CDW is the dominant process causing basal melting, can experience rapid melting in warm oceanic environments (Jacobs et al., 1996; Jacobs et al., 2011). Consequently, warm-water cavity ice shelves have received increasing attention recently by the community due to their vulnerability to rapid retreat under a warming climate (Dutrieux et al., 2014; Jenkins et al., 2018). On the other hand, basal melting in cold-water cavities is generally driven by HSSW which acts as a positive thermal forcing (Adusumilli et al., 2020) due to the freezing point’s dependence on pressure and salinity: the depression of the freezing point of seawater with increasing salinity and pressure in the form of a weak nonlinear function of salinity and a linear function of pressure (Millero, 1978). Polynyas are opened by katabatic winds which are persistent, cold, and strong offshore winds and have a major influence on air-sea fluxes (Jacobs and Giulivi, 1985; Sturman and Anderson, 1986). HSSW is formed through the coastal polynya processes including the sea–ice production and consequently brine rejection. The dense HSSW can effectively block the transport of warm deep water into the ice shelf cavities (Holland et al., 2020). Furthermore, it can access the grounding zones in the cold-water cavities with little modification along the seafloor, and in consequence, its intrusion triggers melting at the base of the deep ice shelf (MacAyeal, 1984; Jacobs et al., 1992). For the case of cold-water ice shelf cavities, variability of polynya activity, therefore, has a potential impact on the ice shelf stability in terms of the production of HSSW by atmospheric cooling and brine released during sea–ice formation in the coastal polynyas. Although warm cavity ice shelves (dominated by Mode 2 melting) have received widespread publicity, more surface area of Antarctica is covered by cold-water cavity ice shelves (Rignot et al., 2013) and a weak polynya state is more likely in the future climate, resulting in the reduction in local dense shelf water formation (Marsland et al., 2007; Aoki et al., 2022). It is thus necessary to resolve the mechanisms involved in the ocean-driven basal melting beneath cold-water cavity ice shelves to determine their stability.

As an underlying driver of basal melting of ice shelves, ocean heat transport significantly affects the ice shelf stability. The delivery of ocean heat into the ice shelf cavity is influenced by several factors (Gwyther et al., 2020). For instance, the bathymetry of the continental shelf can control access of warm off-shelf water masses and atmospheric interactions drive the modification of on-shelf waters. Ocean tides also interact with ice shelves in many ways: enhancement of the turbulent exchange of heat and freshwater through the ocean-ice boundary layer and contribution to ocean mixing by friction at the seabed. In addition to bathymetry in the continental shelf, cavity geometry (topographies of ice shelf and bedrock) has a significant impact on ocean circulation underneath the shelves, resulting in strongly steered ocean currents due to topography in the cavity. However, there is high uncertainty in the bottom topography and the geometry of the sub-ice shelf cavities because of difficulties in direct measurements. It is therefore essential to obtain a better description of the bathymetry beneath the ice shelf. These factors can have a significant impact on the magnitude and spatiotemporal distribution of basal melt rate of ice shelves.

Despite the significance of physical interactions between oceans and ice shelves, the processes involved in ocean-driven melting in the cavity are still not fully understood due to the lack of observations in ice shelf cavities, highlighting the need for comprehensive and high-resolution modeling studies on complex ocean–ice shelf systems. Here we select Nansen Ice Shelf (NIS), East Antarctica as a case study to explore a variety of the oceanographic processes involved in the basal melting through a regional coupled sea-ice/ocean model that includes physical interactions between the ocean and the ice shelf. The NIS is a cold-water cavity ice shelf and it recently suffered a calving event (Dziak et al., 2019). It is located beside the Terra Nova Bay polynya (TNBP) of the Western Ross Sea region in East Antarctica, where is one of the most intense sea–ice producing polynyas in Antarctica. Strong katabatic winds during the austral winter and a barrier effect of the Drygalski Ice Tongue (DIT) on sea–ice advected from the south act to create an area of low-ice concentration along the northern DIT and the western shore of TNB, forming a strong and active polynya characterized by increased sea–ice production (Sansiviero et al., 2017; Yoon et al., 2020). The sea–ice production and the associated brine rejection drives HSSW formation in the TNBP, in which 33% of the HSSW in the Ross Sea is produced (Fusco et al., 2009; Stevens et al., 2017; Jendersie et al., 2018). A coastal region of the NIS is considered to be the primary location of HSSW formation in TNBP (Yoon et al., 2020; Kim et al., 2022), which could facilitate the oceanic heat exchange with the adjacent ocean cavity, i.e., beneath the NIS. The NIS is therefore interesting and a unique testbed for examining how ocean-driven melting is regulated by dynamic and thermodynamic processes in sub-ice shelf cavities. Moreover, there are a few available ship-based hydrographic data measured in TNBP during austral summer, albeit limited, to validate the water mass circulation and properties simulated by the ocean model (Yoon et al., 2020). Particularly, airborne gravity surveys were conducted to infer the bathymetry in this region (Lee et al., 2019), which enable to use realistic geometry of the sub-ice shelf cavity for numerical simulations.

This study provides new insight into the characteristics of the basal melt rate of cold-cavity ice shelf. We firstly focus on spatiotemporal variability in basal melting driven by oceanic processes in the cavity beneath the NIS by analyzing the ocean modeling results. Next, we evaluate the impact of cavity geometry and the role of the tide on ice shelf basal melting by conducting sensitivity experiments.

Historical and recent observations have reported an active HSSW formation in the TNBP, showing a distinct seasonal variability with the maximum salinity in August–October (Buffoni et al., 2002; Rusciano et al., 2013; Yoon et al., 2020). The model successfully reproduces the densest shelf water that occupies the deeper part of TNBP ( > 800   m ), although the salinity values are slightly higher ( ∼ 0.05 ) than those observed ( Figures S4 , S5 ). The simulated HSSW has a remarkable seasonal variation in salinity with its maximum in September–October, which results from the active sea–ice production in TNBP. Sea-ice area in the TNB starts to increase in February/March and reaches its maximum extent in May. The maximum sea-ice extent is maintained throughout winter and then begins to decrease in October. Accordingly, the HSSW salinity at deep layers starts to increase gradually in April/May and has a maximum value in October, which is associated with winter polynya activity ( Figures S6 , S7 ). This suggests that the active sea–ice production during austral winter leads to HSSW formation through a brine supply from the surface.

The seasonal variability in HSSW salinity in the TNB is evident from the spatial patterns of salinity ( Figure 2 ). In February (austral summer), ice melting at the surface reduces the surface salinity to less than 34.5. The freshwater released from ice melting reduces the surface density (increasing buoyancy) and consequently strengthens the stratification. High-salinity water dominates in the lower column of the entire TNB, but more saline water with salinities higher than 34.8 is confined from the deepest part of Drygalski Basin (DB). In contrast, the salinity at all depths significantly increases to above 34.8 in September (austral winter), showing that the water columns of TNB become homogenous. This result demonstrates that HSSW formation through brine rejection drives deep convection. An interesting feature is that the HSSW that formed near the TNBP is evenly distributed in the NIS cavity, spreading along the seafloor to the grounding line ( Figure 2E ). The model results provide strong support for the dense shelf water flowing into the ice shelf cavity and accessing the grounding line.

To better quantitatively describe the HSSW distribution, a tracer experiment is computed by inserting a passive tracer into the ocean model domain. A passive tracer is released at the surface layer where the surface salinity is higher than 34.8 due to active sea–ice production in TNBP to determine the contribution of HSSW formed by brine rejection at the surface. In other words, the passive tracer represents the dense shelf water formed by the polynya’s sea–ice production, which is not released during austral summer when the surface salinity is less than 34.8 due to sea–ice melting. As expected, no tracer is observed at the surface layer (10 m) in February, but its concentration gradually increases from the surface to the bottom layer ( Figures 3A–C ). The tracer representing the HSSW is distributed over the DB (more than 80% of the concentration at 900 m), indicating that the HSSW formed during the previous winter is consistently distributed on the bottom layer throughout the year. Unlike the austral summer, a high concentration of the tracer is distributed along the northern margin of the DIT and the western shore of TNB at the surface in September, similar to that of an L-shaped polynya in the TNB ( Figures 3D , S7D ). Because the density of water increases with salinity and surface water freezing is accompanied by brine rejection, the thus-formed near-surface water tends to sink. Therefore, the tracer concentration of more than 90% is dominant in the lower layers of the water column in the TNB, consistent with the salinity pattern ( Figures 2 , 3 ). These tracer distributions strongly support the HSSW formation through brine rejection from winter polynya sea–ice formation and the resultant deep convection revealed by the salinity distribution. In addition, the tracer enters the NIS cavity and flows along the seafloor to the grounding line, showing more than 90% of the tracer concentration that gradually decreased from the ice front to the grounding line ( Figures 3 , 4 ). These modeling results with the passive tracer experiment demonstrate the seasonal variability of HSSW formation in the TNB.

Tides are known to affect sea–ice production and basal melting under ice shelves (e.g., Jourdain et al., 2019; Hausmann et al., 2020). In addition, the spatial distribution of basal melting is very sensitive to model geometry as well as tides (Mueller et al., 2012). Because the ocean underneath the ice shelves is completely isolated from the atmosphere by thick ice, tides instead of winds become a major source of oceanic kinetic energy for conversion to mixing in the ice shelf cavities. The tidal currents in the cavity interact with the topographies of the ice base and bedrock, inducing turbulent mixing. The tide-induced turbulent mixing within the ice shelf cavity affects basal melting by changing the heat and salt exchange through the ice shelf-ocean boundary layer. In this section, impacts of tides and cavity geometry on the NIS basal melting are estimated by conducting two additional experiments; one is without tides, and the other one uses original RTopo2 ( Figure 7 ). A comparison of the sensitivity experiments with the control simulation exhibits a significant difference during summer but a small one during winter ( Figure S9 ). Because the water is completely mixed from surface to bottom and the very cold-salty water occupies the whole water column during most of the cold season. Thus, we focus on the impact of tides and realistic cavity geometry on the basal melting beneath the NIS during the warm season when thermohaline stratification occurs in the TNB.

Tides increase the basal melting of the shallow-draft outer portions of the ice shelf in February (austral summer), particularly those shallower than 200 m, making up a substantial increase of 30% relative to the result without tides ( Figure 8A ). This is similar to the modeling result for the Ross Ice Shelf reported by Arzeno et al. (2014). Tidal currents in high latitude oceans tend to be quite barotropic because of the weak stratification (MacAyeal, 1984; Robertson et al., 1998). Here the spatial variability of monthly barotropic currents is characterized by using the time- and depth-averaged tidal current speed ( Figure 9 ). Comparing the results with and without tides demonstrates that tides drive strong barotropic currents in the cavity and a robust thermal exchange at the ocean-ice interface. The strong barotropic current speed appears on the shallow ice draft region, particularly along the sloping ice shelf base ( Figure 9B ), which induces turbulent mixing by interaction with the ice shelf topography. This facilitates the heat exchange between the ocean and ice shelf ( Figure 9E ), resulting in substantial basal melting. The increased basal melting in areas of shallow ice draft in the upper part of the cavity, particularly near the ice front, can be also linked to an enhanced heat absorption within the TNBP, which can be induced by reduced sea–ice production due to tidal forcing. Koentopp et al. (2005) investigated the effects of tides on the seasonal fluctuations of the sea-ice cover using a sea-ice model and showed a decrease in sea-ice concentration when including tidal forcing into the model. They found that during summer enhanced absorption of radiation and additional entrainment of warmer waters into the mixed layer by tidal current are responsible for this reduction. During austral summer when significant open water exists, including the coastal TNBP, the tidal effects lead to smaller minimum sea-ice extent than those without tides and thus increased solar absorption in the surface layer, driving rapid surface heating in TNBP ( Figure 8B ). Consequently, the simulation with tides exhibits enhanced solar heating of the surface water and the subsequent intrusion of this relatively warm water into the upper part of the NIS cavity where the ice draft is lower than 400 m ( Figure 8C ). The experiment without tides further suggests that the inflow of solar-heated surface waters from the adjacent TNBP is important in driving NIS basal melting during summer.

The sensitivity result using cavity geometry estimated by original RTopo2 reveals an excessive basal melting just near the NIS frontal band ( Figure 10A ). Strong time-averaged barotropic currents and hence an enhanced turbulent exchange rate for heat are restricted in a narrow frontal band of NIS ( Figures 9C, F ), corresponding to the excessive melting region ( Figure 10A ). RTopo2 bathymetry shows seafloor hill-like topography at the entrance of the NIS cavity ( Figures 7E , 10C ). This topographic feature induces abrupt changes in water column thickness and strong tide-topography interaction at the ice shelf front ( Figure 9C ). Consequently, locally enhanced tide-induced vertical mixing occurs under the NIS calving front due to interaction with the unrealistic topography ( Figure 9F ), which is responsible for warmer/cooler water at/above the depth of the sill. Namely, the surface layer becomes cooler due to substantial basal melting around the NIS calving front while the water at the depth of the sill becomes warmer due to mixing through tide-topography interaction ( Figures 10B, C ).

Using a regional coupled sea-ice/ocean circulation model including the ocean–ice shelf dynamics/thermodynamics, a numerical simulation with tides and realistic cavity geometry is conducted on the NIS in the TNB, East Antarctica, to explore ocean-driven basal melting of a cold-water cavity ice shelf. The results highlight the seasonal spatiotemporal variability in melt rates from different modes of basal melting occurring at different depths of the NIS base. During austral summer, the basal melt rate near the grounding line is moderate (for ice draft deeper than 400 m), whereas that for shallow-draft outer regions of the NIS increases toward the ice shelf front, contributing 78% of the total NIS melting rate. The enhanced melting near the ice shelf front is dominated by Mode 3 processes. Solar-heated surface waters advect under the ice shelf and drive strong melting in shallower ice drafts along the ice shelf front, suggesting that melting at the shallow ice shelf base is the dominant source of the NIS basal mass loss during austral summer. Meanwhile, during austral winter, the basal melt rate near the ice front dramatically decreases, whereas basal melting near the grounding line increases much above summer values. The increased basal melt rates near the grounding line are closely linked to the active formation of HSSW in the TNBP during austral winter, which is characterized by Mode 1 melting processes. The intrusion of HSSW produced by active TNBP that flows into the cavity beneath NIS toward the deep grounding line explains 36% of the total melting rate of NIS in winter.

To identify the contribution of tides and realistic cavity geometry to basal melt rates of the NIS, two sensitivity experiments are performed using the same initial and boundary conditions as in the control simulation: one is without tides, and the other one uses original RTopo2. The additional experiments demonstrate that basal melting is very sensitive to tide and cavity geometry. Tidal effects increase the melting of NIS, particularly contributing as much as 30% to the shallow ice draft region along the sloping ice shelf base. Tides induce turbulent mixing by interacting with topographies of ice base and bedrock, stimulating the exchange of heat and salt between the ocean and ice shelf. Sea-ice production is also reduced due to tides, leading to surface heating in TNBP during summer. In addition, basal melt rates beneath the NIS responded directly to the cavity geometry. Spurious vertical mixing is locally induced and enhanced by interaction between tides and the unrealistic topography, resulting in excessive basal melting near the NIS frontal band where a sill is featured in original RTopo2. The sensitivity experiments have shown that tides and realistic cavity geometry bring a significant improvement in the estimation of basal melt rates through a numerical model.

Our results suggest that accurately modeling ocean-driven melting beneath ice shelves is crucial to predicting their temporal and spatial mass loss. Seasonal variability signatures for different ocean-driven melting modes also have implications for the regional basal mass balance around the ice shelves. As heat absorption and HSSW formation are driven by atmospheric processes in polynya, ice shelf melting is likely to vary with atmospheric conditions and is associated with ocean circulations on interannual and longer time scales. As sea–ice concentration decreases because of ocean warming, ice shelves change with thinning ice fronts and grounding line retreat by the ocean and so are more vulnerable to a warming climate system. The potential for climate-related ice shelf melting provides motivations for further studies, including observational and modeling efforts focused on ocean–ice shelf interactions and longer-term effects on basal melting beneath ice shelves.

The original contributions presented in the study are included in the article/ Supplementary Material , further inquiries can be directed to the corresponding author.

TK, EKJ, and J-HM contributed to the conception and design of the study. J-SH performed the numerical experiments and produced all Figures. TK, J-HM, and J-SH contributed to the data processing and analysis. S-KS and WSL contributed to scientific discussion. EKJ and WSL provided the study materials. The paper was written by TK and J-HM with input from all authors. All authors contributed to the article and approved the submitted version.