The case for a Framework for UnderStanding Ice-Ocean iNteractions (FUSION) in the Antarctic-Southern Ocean system

This contribution summarises the need for the development and implementation of a Framework for UnderStanding Ice-Ocean iNteractions (FUSION) to address significant and impactful unknowns in ice-ocean interactions in Antarctica. Guided by critical underpinning science questions, FUSION should comprise:

1. Definition of essential state variables for characterising ice-ocean interactions, technologies to observe the essential variables, parameters to ensure the variables are adequately characterised and directions for reconciling measurements across observational platforms and technologies;

2. An approach to integrate models and other technologies in the design of observational systems and programs, including novel methods in data assimilation into models, that operate across spatial and temporal scales and that can be used to inform measurement parameters and priority locations for observations;

3. An approach to reconcile observations and models to improve understanding of the processes underlying ice-ocean interactions and feedbacks that includes a standard protocol for how observations can be used to validate models, methods to integrate observations into models and parameterisations of fine-scale processes captured in high-resolution models to large-scale general circulation or Earth Systems Models; and

4. Acknowledgement of issues relating to data availability and to equity, diversity and inclusion, with an approach to ensuring best practices in these areas.

We propose that the implementation of FUSION be staged as follows:

• Develop an initial framework targeting one essential variable—ocean-driven ice shelf melt—through a series of dedicated workshops and a broad-scale community consultation process;

• Seek formal endorsement of the framework for ocean-driven ice shelf melt through international organisations and initiatives, including the Scientific Committee on Antarctic Research (SCAR), the Southern Ocean Observing System, the Scientific Committee on Oceanic Research, the International Union of Geodesy and Geophysics and its associations, namely the International Association of the Physical Sciences of the Ocean and the International Association of Cryospheric Sciences, the World Climate Research Program and its working groups, including the Southern Ocean Freshwater Input from Antarctica initiative and other relevant Model Intercomparison Projects. Endorsement and advertising via these organisations and initiatives will ensure maximum uptake within the community;

• Launch, test and refine the framework through fieldwork and synthesis conducted as part of the Antarctica International Science and Infrastructure for Synchronous Observation (Antarctica InSync) initiative 2027–2030 as part of the United Nations Decade of Ocean Science for Sustainable Development; and

• Develop FUSION to build on the framework for ocean-driven ice shelf melt, which will require a broader-scale multidisciplinary effort, coordinating with the international community to define the relevant essential variable and implement a strategy for Points 2–4 above.

Antarctic Ice Sheet mass loss is accelerating in response to ice-ocean interactions (Rignot et al., 2019; Adusumilli et al., 2020; Otosaka et al., 2023). Ocean-driven ice shelf melt from Antarctica impacts Earth and climate systems, human populations and ecosystems by raising the global mean sea level (Kulp and Strauss, 2019; Cooley et al., 2023) and altering heat and salt budgets of the ocean (Purkey and Johnson, 2012; Jacobs et al., 2022), with implications for the global overturning circulation (Purkey and Johnson, 2012; Gunn et al., 2023; Li et al., 2023) and sea ice coverage (Bronselaer et al., 2018; Purich and England, 2023). Freshwater fluxes from the Antarctic continent also alter nutrient flux and ocean biogeochemistry (Cape et al., 2019), with associated feedbacks on primary productivity and marine ecosystems (Arrigo et al., 2015; Cavan et al., 2019). Ice sheet models, including from the Ice Sheet Model Intercomparison Project (Seroussi et al., 2020), generally project increasing ocean-driven melt over the 21st century as the climate warms. However, estimates of ice mass loss vary widely, depending on the emission scenarios and ice physics represented in ice sheet models, and are associated with high uncertainty (Seroussi et al., 2020; Seroussi et al., 2023). Indeed, the Intergovernmental Panel on Climate Change Sixth Assessment Report (Fox-Kemper et al., 2021) could not rule out 5 m of sea level rise by 2150, due to deep uncertainties associated with processes of rapid ice sheet mass loss from Antarctica, including those triggered by ice-ocean interactions. Ice sheet modelling analyses suggest that tipping point thresholds which could lead to the disintegration of the West Antarctic Ice Sheet (WAIS) have not yet been crossed (Hill et al., 2023; Reese et al., 2023). However, a recent ocean modelling study suggests that even under the most extreme emission reduction scenarios ice shelf melt rates are likely to triple by the end of the century compared with historical rates, suggesting that exceedance of tipping point thresholds for WAIS could be imminent (Naughten et al., 2023).

Numerical modelling of the Antarctic Ice Sheet has advanced considerably over recent decades, and a number of development activities are underway to better capture ocean-driven ice melt, as well as the impacts of changes in ocean-driven melt on Earth and climate systems. The marine ice sheet and ocean model intercomparison projects (Asay-Davis et al., 2016) have made tremendous advances in understanding differences between model representations and parameterisations of ice shelf melt. Several international groups have—or are moving towards—coupling dynamic ice sheet models to ocean models or Earth System Models, which will enable more accurate representation and quantification of freshwater fluxes (Favier et al., 2019; Muntjewerf et al., 2021; Naughten et al., 2021; Smith et al., 2021). Initiatives such as the Southern Ocean Freshwater Input from Antarctica Initiative (Swart et al., 2023) will explicitly focus model experiments on the impacts of freshwater fluxes on regional and global scales in state-of-the-art coupled climate models. These experiments aim to produce a consistent and standardised modelling approach to understanding the impacts of Antarctic freshwater discharge on the climate system (Chen et al., 2023). Finally, significant improvements have been made in integrating biogeochemistry within the ocean model components of the Coupled Model Intercomparison Project suite of models (Séférian et al., 2020), facilitating the assessment of biases in the physical processes represented by these models (Fu et al., 2022).

Despite these advances in modelling ice-ocean interactions and their impacts, the processes underlying ocean-driven melt remain poorly understood. Uncertainties persist in: (i) the melt regimes of different Antarctic ice shelf-ocean systems due to a paucity of direct observations (e.g., Rosevear et al., 2022a; Rosevear et al., 2022b); (ii) understanding the influence of subglacial freshwater discharge (Wei et al., 2020a; Nakayama et al., 2021a; Gwyther et al., 2023) and different ocean water masses, both within and outside the ice shelf cavity (Lewis and Perkin, 1986; Jacobs et al., 1992; Jacobs et al., 1996; Jenkins et al., 2010; Thompson et al., 2018); (iii) feedbacks between the ice sheet, ocean and solid Earth near the grounding line (the interface between the grounded ice sheet and floating ice shelves; see summary in Rignot, 2023) and (iv) ice dynamic processes (Thomas 1973a; 1973b; Fürst et al., 2016; Reese et al., 2018b) that impact iceberg calving (Liu et al., 2015; Greene et al., 2022). Parameterisation of fine-scale melt processes within the ice-ocean boundary in coarser-resolution ocean and ice sheet models, without loss of fidelity to the magnitude of, and variability in, melt is an additional complication. Standardised protocols for observing key glaciological and oceanographic variables, and an approach for reconciling observational and modelling outputs related to ice-ocean interactions, do not exist for the Antarctic-Southern Ocean system.

Summaries from the open Joint Commission on Ice-Ocean Interactions workshop in October 2022 (McCormack et al., 2022a) suggested the need for:

1. Coordinated international collaboration to collect the observational data needed to characterise and describe ocean-driven melt, including long-term monitoring to constrain both the mean state of and variability within the ice-ocean system;

2. Development, comparison and sensitivity testing of models of ocean-driven melt, to transfer knowledge between modelling efforts on different scales and constrain uncertainties in predictions of ice sheet mass loss; and

3. Development of methods to reconcile data streams that cross a broad range of temporal and spatial scales and to integrate these data effectively into models.

More recent research has motivated the need for a standardised approach to collecting observational estimates of ocean-driven ice melt across oceanographic, satellite remote sensing and geophysical instruments and platforms, and for reconciling multiple data streams that cross spatial and temporal scales (Cook et al., 2022). Coincident developments in modelling capabilities, including the adjoint method in ocean and ice sheet models (Heimbach and Losch, 2012; Goldberg and Heimbach, 2013; Morlighem et al., 2013; Goldberg et al., 2020; Nakayama et al., 2021c) and community-based forward modelling efforts (e.g., Seroussi et al., 2023), present opportunities for the use of models to understand ocean-driven melt and its sensitivity, which will aid in prioritising locations for data collection.

Similar considerations in Greenland ice-ocean interactions motivated the development of the Greenland Ice Sheet-Ocean Observing System (Straneo et al., 2019). Ice loss from the Greenland Ice Sheet has implications for sea level, ocean circulation and ecosystems, and like in Antarctica, understanding of the processes underlying ice-ocean interactions is hindered by a scarcity of observations. After building community consensus through workshops and summary documents, Straneo et al. (2019) proposed a framework for obtaining the necessary observations that included a number of key principles, namely (1) observations should be sustained over time and collected at a handful of key sites that span a range of ice-ocean geometries and ocean conditions; (2) observations at each site should include a chosen set of essential glaciological, oceanic and atmospheric variables and (3) standardised protocols for processing and sharing the data would be necessary to ensure best use of observations. Straneo et al. (2019) also stressed the importance of international collaboration and engagement with existing networks for the successful delivery of such a framework.

Here, we espouse the need for a framework providing guidance on combining observations and modelling to systematically improve understanding of ice-ocean interactions in the Antarctic and Southern Ocean: a Framework for UnderStanding Ice-Ocean iNteractions (FUSION). The development of FUSION is timely, given upcoming internationally coordinated and synchronous field campaigns, including Antarctica InSync which aims to assess connections between ice, ocean, climate environment and life around Antarctica, as well as activities proposed as part of the International Polar Year in 2032. In this Practice Bridge article, we describe key research priorities that should underpin FUSION: processes at the ice-ocean interface, ocean dynamics, grounding line dynamics, ice flow dynamics and modelling ice-ocean interactions. We outline the components of FUSION, including the essential variables needed to characterise ice-ocean interactions, developments required in stand-alone and coupled ocean and ice sheet models and approaches to reconciling observations and modelling. We conclude by highlighting barriers and opportunities to a successful framework implementation.

The thermodynamics and mechanics of the ice shelf-ocean boundary layer (ISOBL; Figure 1A), an ocean layer of metres to tens of metres in thickness adjacent to the ice shelf base, are at the heart of ice-ocean interactions. The ISOBL can be broadly described as a frictional boundary layer formed due to the presence of a mean ocean current flowing adjacent to a rough ice shelf base. It can be separated into three distinct layers (Soulsby, 1983): the viscous sublayer approximately 1 cm thick from the ice shelf base, the logarithmic (or surface) layer, which accounts for approximately 30% of the overall ISOBL thickness (Pope, 2000), and the outer mixed layer.

Melt and refreeze in the ISOBL are impacted by heat and salt exchanges between the ISOBL and the surrounding ocean, which depend on heat availability (Holland et al., 2008), the geometry of the ice shelf base and the ocean energetic conditions. For example, for the Larsen C and Filchner-Ronne ice shelves, which are both characterised as cold, energetic regimes, the current shear is a key control on melting (Jenkins et al., 2010; Davis and Nicholls, 2019). In warmer regimes, such as the Pine Island Glacier ice shelf cavity, buoyant meltwater dominates, with impacts on oceanic stratification and the processes of convection and diffusive-convection (Stanton et al., 2013; Begeman et al., 2018; Stewart, 2018).

Processes regulating melt and refreeze within the ISOBL occur on scales too fine to be captured by most ocean general circulation models. As a result, melt and refreeze in ocean general circulation models are typically represented by three equations (Holland and Jenkins, 1999) that capture: (1) the dependence of the local freezing point of seawater on pressure and salinity, (2) the balance between latent heat required for melt (or generated by refreezing) and the difference between the heat loss into the ice and the heat supply from the ocean and (3) the balance between the freshwater flux from melt and the flux of salt to the ice shelf.

Few measurements of the ISOBL exist, which limits understanding of its structure and processes for most Antarctic ice shelf cavities (Jenkins, 2021), and there are examples where the three equations used by most models do not adequately capture the relevant processes operating in Antarctic ice shelf cavities (Kimura et al., 2015; Davis et al., 2023). The physical properties of ice and water—density, specific heat capacity and thermal diffusivities—that enter the three-equation melt parameterisation are generally well constrained. However, a number of other terms are poorly constrained, including the turbulent transfer coefficients and the processes that underlie them (including turbulent mixing, stratification, the structure of the ISOBL and ocean currents; Jenkins et al., 2010; Malyarenko et al., 2020). For example, the ice shelf basal roughness is a critical control on turbulent heat flux via its influence on the friction velocity (Gwyther et al., 2015; Malyarenko et al., 2020), but is sparsely—if at all—sampled for the vast majority of Antarctic ice shelves. Little is also known about the melt regimes of most ice shelves, whether controlled by shear, convection, or diffusive-convection, and how these vary in time (Rosevear et al., 2022b). Addressing these uncertainties will require high-resolution observations across regimes and sustained monitoring of representative Antarctic ice shelf cavities.

Water mass properties and circulation within ice shelf cavities play a pivotal role in influencing the thermodynamics and mechanics in the ISOBL and the location and magnitude of ice shelf melt. Melt can be categorised into three modes (Figure 1B–D) based on the water mass that serves as the primary heat source (Jacobs et al., 1992). In the first mode (Figure 1B) melting is primarily induced by water masses at the surface freezing-point temperature, which can serve as a heat source for ice shelf melting at depth due to the pressure dependence of the freezing point of water and adiabatic compression of the water mass. Mode 1 melting is dominant in the Weddell Sea and Ross Sea sectors (Jacobs et al., 1992) and below the Amery Ice Shelf (Herraiz-Borreguero et al., 2013), where Dense Shelf Water (DSW) is formed during sea ice production in coastal polynyas (e.g., Foldvik et al., 2004; Tamura et al., 2012; Silvano et al., 2023), but the mechanism contributes to the thermal driving wherever seawater interacts with glacier ice at greater depths below the sea level (Smith et al., 2020b). In the second mode (Figure 1C) melting is induced by Circumpolar Deep Water (CDW) transported polewards across the shelf break and onto the continental shelf (e.g., Jacobs et al., 1996; Dutrieux et al., 2014). Mixing of CDW on the continental shelf leads to modified CDW (mCDW), which has a higher water temperature (1–2°C) than DSW, leading to much higher ice shelf melt rates. The mCDW is widespread in the Amundsen and Bellingshausen Seas and some locations along the East Antarctic coast, including the Sabrina and Knox coasts (Jacobs et al., 1996; Dutrieux et al., 2014; Rintoul et al., 2016; Thompson et al., 2020; Ribeiro et al., 2021; Schulze Chretien et al., 2021), while more intermittent inflows are observed in the Atlantic sector of the Southern Ocean (Nøst et al., 2011; Darelius et al., 2016; Hirano et al., 2020; Lauber et al., 2023). The third mode (Figure 1D) is the ice shelf melting caused by Antarctic Surface Water (ASW), which is produced during summer months as surface waters warm. ASW contributes predominantly to melting close to the ice shelf front (Malyarenko et al., 2019; Aoki et al., 2022), but remnants of this water mass have been observed to impact ice-ocean interactions several hundred metres below the surface further inside ice shelf cavities (Hattermann et al., 2012; Lauber et al., 2024). Ocean dynamics driving these modes of ice shelf melting are closely linked to the varying coastal regimes around the Antarctic continent (Thompson et al., 2018), which together with other factors that lead to changes in the modes of melting are critically under-observed and still poorly represented in most large-scale models.

Another water mass significant for ice shelf melt is subglacial freshwater discharge at glacier grounding lines (Figure 1E). This water mass can drive buoyant plumes that rise along the sloping ice shelf base (Jenkins, 2011) and enhance turbulent mixing and hence basal melting (Hewitt, 2020), contributing to the evolution of ice shelf basal channels (Drews et al., 2017; Hofstede et al., 2021). Ocean modelling simulations show that melt from freshwater discharge is generally concentrated near the grounding line (Nakayama et al., 2021a; Gwyther et al., 2023). However, freshwater plumes may affect the general cavity circulation (Washam et al., 2023) and travel over 100 km from the source location (Gwyther et al., 2023), influencing ocean stratification and reducing sea ice production on the continental shelf (Goldberg et al., 2023). In other cases, the enhanced stratification induced by subglacial discharge suppresses mixing within the ISOBL (Davis et al., 2023) through processes that usually are not represented in the most commonly used global ocean model parameterizations. Similar to observations of the ISOBL, there are few direct measurements of subglacial freshwater in Antarctica. Satellite optical imagery, which can be used to infer the presence of ice shelf basal channels, suggests that subglacial outflow is widespread in Antarctica (Le Brocq et al., 2013). Other studies have demonstrated a link between the location of channels, subglacial discharge and high ice shelf melt rates (Washam et al., 2019; Wei et al., 2020a; Dow et al., 2022). Although a number of subglacial hydrology modelling studies estimate discharge volume and rate (Dow et al., 2020; Dow, 2023; Pelle et al., 2023; Pelle et al., 2024), there are no direct observations quantifying discharge, which remains a key limitation in estimating its impact on melt.

Projections of how the Southern Ocean will change in the future and impact ice shelf melt remain highly uncertain. Biases and inter-model discrepancies in state-of-the-art climate models outpace projected changes in the Antarctic coastal regime (Barthel et al., 2020; Jourdain et al., 2020), and a key challenge remains in understanding how the driving forces behind the three main water masses that contribute to Antarctic ice loss will change. In the Amundsen Sea sector of West Antarctica, where the largest melt rates are observed, variations in the thermocline depth—the interface between cold Winter Water and warm mCDW—control ice shelf melt rates (Dutrieux et al., 2014; Webber et al., 2019). One leading hypothesis suggests that the thermocline depth is controlled by off-shelf winds (Steig et al., 2012; Dutrieux et al., 2014), but the exact mechanisms remain uncertain. In the Ross Sea, downstream of the Amundsen Sea, a long-term freshening has been observed, which is likely caused by increased ice-shelf meltwater from the upstream Amundsen Sea (Jacobs et al., 2002; Jacobs et al., 2011; Nakayama et al., 2014; Jacobs et al., 2022). These changes have the potential to alter the magnitude of ice shelf melting and Antarctic Bottom Water formation (Nakayama et al., 2020). A similar situation is found in the Weddell Sea sector, where changes in synoptic winds affect sea ice production that reflects on melting and dense water formation (Hattermann et al., 2021; Janout et al., 2021), and a potential tipping point was identified in relation to self-sustained enhanced access of CDW onto the continental shelf (Hellmer et al., 2012; Hellmer et al., 2017; Hazel and Stewart, 2020; Naughten et al., 2021). By contrast, large parts of the East Antarctic coast are currently protected from CDW by the pronounced Antarctic Slope Front (Thompson et al., 2018), but little is known about the future stability of this system (Barthel et al., 2020) and recent studies suggest that warmer water is closing in on the continent (Herraiz-Borreguero and Naveira Garabato, 2022; Lauber et al., 2023; Gao et al., 2024). Furthermore, the bathymetry of the continental shelf plays a key control in determining mCDW access to ice shelf cavities, yet remains sparsely sampled. Accurately representing these processes requires high-resolution global or circum-Antarctic Ocean models with ice shelf capabilities that achieve good model-data agreement. Such capability improvements in models will enable critical knowledge gaps to be addressed, such as an uncertainty assessment of the impact of the Amundsen Sea glacial melt on downstream hydrography. While ongoing studies provide qualitative insights, a quantitative understanding of how the Southern Ocean might evolve over decades to centuries, with a thorough assessment of uncertainties, is urgently needed.

Key to constraining uncertainties in future projections are sustained ocean observations, from the ice shelf cavity scale across the continental shelf to the Southern Ocean, and the development of effective techniques for integrating observations and numerical modelling (Section 4.2; e.g., Mazloff et al., 2010; Forget et al., 2015; Zhang et al., 2018). Sustained summer-time observations over a few decades exist in the Weddell Sea, Ross Sea, Amundsen Sea and a few regions along the East Antarctic coast (Jacobs et al., 2002; Janout et al., 2021; Lauber et al., 2023), although very few winter-time measurements exist in these regions. However, observations are generally lacking around the Antarctic continental shelf, including in regions of East Antarctica that have been highlighted as particularly vulnerable to rapid retreat under warming climate conditions (Golledge et al., 2015; Garbe et al., 2020). Fine-scale bathymetric and hydrographic data over the continental shelf and inside ice shelf cavities are needed to better estimate ocean circulation and heat supply to ice shelf cavities. Finally, approaches to effectively synthesise observational data into models to achieve better model-data agreement are needed. Data assimilation techniques employing Green’s functions or adjoint methods applied to Antarctic continental shelf regions have been shown to improve the representation of the ocean and sea ice state, Winter Water masses and intrusions of mCDW (Nakayama et al., 2017; Nakayama et al., 2021c). Nevertheless, ocean processes driving ice-ocean interactions are challenging to simulate even when advanced assimilation techniques are applied.

Grounding lines are key indicators of the overall stability of ice streams, as they advance and retreat in response to climate conditions (Joughin et al., 2012; Rignot et al., 2014). The location of the grounding line is estimated based on the limit of the inland tidal flexure of ice shelves from repeated altimetry or interferometric synthetic-aperture radar measurements (Figure 1E; Friedl et al., 2020). This transition between the grounded and floating ice is at the interface between the ice, ocean and solid Earth, and it is influenced strongly by tides, ocean variability on multiple timescales and subglacial water discharge at the grounding line, making this region challenging to analyse and understand.

Recent observational and modelling efforts have enabled greater insights into the processes operating in the grounding zones relevant for ice-ocean interactions. Improved remote-sensing coverage now facilitates analysis of short-time changes in grounding zone positions, ice velocities and thicknesses, and a better understanding of the complex patterns of changes (Milillo et al., 2019). This enables insight into the drivers of short-term changes, including tidal variations and their role in influencing long-term grounding line stability (Freer et al., 2023). Observations have enabled quantification of melt rates as high as 60 m yr⁻¹ within the grounding zones of Antarctic and Greenland glaciers (Milillo et al., 2019; Ciraci et al., 2023). The largest melt rates occur due to rapid, pressurised seawater intrusions via freshwater wedges (Wilson et al., 2020) that are capable of producing vigorous basal melting. Intrusions typically occur several kilometres into the grounding zones (Warburton et al., 2020), but could occur over larger areas in response to layered seawater intrusions into grounded ice regions (Robel et al., 2022).

Despite advances in understanding the properties and processes important for ice-ocean interactions that occur within the grounding zone, these properties and processes are notoriously difficult to observe and model. The assumption of hydrostatic equilibrium employed in remote-sensing estimates of ice shelf melt is not valid in the first few kilometres downstream of the grounding line, limiting the applicability of this method within regions of generally highest melt rates (Rignot et al., 2013; Adusumilli et al., 2020). Accurate estimates of ice sheet thickness at the grounding line are critical for estimating discharge and hence for determining ice shelf thickness change due to melt, and yet the thickness of less than one-third of the Antarctic grounding line has been surveyed at sufficiently high resolution (SCAR RINGS Action Group, 2022). Ocean models are limited in their ability to resolve these areas and therefore to accurately estimate melt rates, as the thin water column in the grounding zone needs to be captured with extremely high horizontal and vertical resolution (Nakayama et al., 2019). Ice flow models also require extremely high resolution—both in model mesh size and in the underlying topography data (Gagliardini et al., 2010)—as well as appropriate numerical treatment of melt rate in the vicinity of the grounding line (Arthern and Williams, 2017; Seroussi and Morlighem, 2018) to accurately capture grounding line migration. The convergence of these limitations in both observations and models impacts our understanding of processes happening close to the grounding line and therefore the reliability of projections of ice mass loss contribution to sea level rise (Seroussi et al., 2019).

Discharge of grounded ice into the ocean accounts for over 90% of the total ice mass loss from Antarctica (Gardner et al., 2018). Discharge has been increasing in recent decades (Smith et al., 2020a; Diener et al., 2021; Otosaka et al., 2023), predominantly driven by ocean-induced ice shelf melt and calving that induce changes in ice shelf buttressing (Greene et al., 2022). Of considerable concern in West Antarctica and the subglacial basins of East Antarctica is the potential for a rapid increase in discharge linked to ice sheet instabilities, as may have occurred during past periods of rapid retreat (Jones et al., 2022; Stokes et al., 2022). Understanding the ice flow dynamics underlying these processes of retreat, and the timescales over which they may occur, is critical to reducing uncertainty in projections of sea level rise.

The ice flow dynamics underlying discharge (Figure 1F) are highly sensitive to changes in ice shelf buttressing, via the resistive (or membrane) stresses transmitted by the ice shelf to the upstream grounded ice sheet (Hindmarsh, 2006). Melt at the ice-ocean interface can lead to a reduction in buttressing through ice shelf thinning, with subsequent increases in ice flow and discharge (Gudmundsson et al., 2019). The degree to which ice shelf thinning or calving results in a reduction in buttressing depends on the grounding line position, ice shelf geometry and the influence of pinning points or significant topography (Matsuoka et al., 2015; Fürst et al., 2016; Reese et al., 2018b). However, cavity geometries—and hence pinning points—are poorly constrained for most of Antarctica’s ice shelves (Morlighem et al., 2020), which can impact the fidelity of the ice sheet response to ice-ocean interactions simulated by ice sheet models (Favier et al., 2016).

Ice-ocean interactions that induce changes in buttressing can impact ice flow dynamics over a broad range of spatial and temporal scales. Highly localised ice shelf thinning can impact ice flow acceleration hundreds of kilometres upstream (Reese et al., 2018b). However, ice dynamics tend to be most sensitive to ice shelf thinning close to the grounding line, and monitoring these regions is therefore critical (Section 3.3). In addition to the instantaneous response on the distribution of stresses, perturbations near the grounding line can lead to a longer-term diffusive response of the ice sheet, as it readjusts to its evolving geometry in a positive feedback between surface slope and driving stress that propagates upstream over time (Payne et al., 2004; Parizek et al., 2013). Critical to simulating the ice sheet response to ice-ocean interactions are improved representation of the processes of ice flow by sliding (Weertman, 1957; Budd et al., 1979; Schoof, 2005) and deformation (Glen, 1952; 1953; 1955) in models, constraints on the relative contributions of these processes to flow (McCormack et al., 2022c; Rathmann and Lilien, 2022) and how the flow processes vary depending on properties at the ice-bed interface zone, including bed topography and roughness (Castleman et al., 2022; Law et al., 2023), basal thermal regime (Dawson et al., 2022; McCormack et al., 2022b; Reading et al., 2022) and subglacial hydrology (Dow, 2023).

The main contributing factor to the uncertainty in sea level rise projections is whether Antarctica, and particularly West Antarctica, will undergo dynamic instability and the likely timescales of the ice sheet response (Fox-Kemper et al., 2021). The marine ice sheet instability theory posits that grounding line retreat on retrograde bed slopes can be self-sustaining once triggered (e.g., through increased ocean-driven melt), even if the forcing returns to previous levels (Weertman, 1974; Schoof, 2007). Recent modelling studies have focused on the conditions and timescales under which a tipping point threshold into irreversible retreat may be crossed in West Antarctica (Garbe et al., 2020; Rosier et al., 2021; Hill et al., 2023; Reese et al., 2023), showing that current grounding line retreat is not yet irreversible, but could become so under present-day climate forcing. This finding highlights the importance of monitoring properties and processes operating in grounding zones, as discussed above (Section 3.3). Ocean variability can also greatly impact the timing of the onset of grounding line retreat (Hoffman et al., 2019; McCormack et al., 2021) and sea level rise, due to marine ice sheet instability (Robel et al., 2019), and remains poorly constrained. The marine ice cliff instability is another proposed instability process, whereby ice shelf breakup leads to the exposure of unstable ice cliffs, the collapse of which can lead to self-sustaining and rapid retreat (DeConto and Pollard, 2016; Bassis et al., 2021). The incorporation of the marine ice cliff instability process into ice sheet models can lead to even greater uncertainties in high-end scenarios of sea level rise, particularly beyond 2100 (DeConto et al., 2021; Fox-Kemper et al., 2021). While surface melt and hydrofracture are generally understood to be the main drivers of ice shelf collapse (Munneke et al., 2014), as evidenced by the disintegration of a number of ice shelves along the Antarctic Peninsula (Cook and Vaughan, 2010), ocean-driven basal melt may pre-weaken ice shelves, increasing their vulnerability to atmospheric-driven collapse (Banwell and Macayeal, 2015; Robel and Banwell, 2019). However, marine ice cliff instability has not been observed directly and remains poorly understood (Edwards et al., 2019), as are the dynamics underlying ice shelf hydrofracture and calving more broadly. Importantly, understanding the timescales on which ice shelf collapse may occur is critical (Clerc et al., 2019), which highlights the need for concerted monitoring efforts across Antarctic ice shelves.

A number of ice and ocean models, as well as coupled ice-ocean models, have been developed to advance our understanding of ice-ocean interactions across various spatial and temporal scales. This section considers recent advances in models of the ocean, models of the ice sheet and finally coupled models.

Stand-alone large-eddy simulations directly simulate ice-ocean interactions in the ISOBL over scales of a few millimetres to metres (Gayen et al., 2016; McConnochie and Kerr, 2018; Mondal et al., 2019; Vreugdenhil and Taylor, 2019; Rosevear et al., 2021; Na et al., 2022). These process-oriented models show promise in addressing some of the uncertainties associated with the ISOBL and its role in regulating melting. For example, while the three-equation parameterisation is built on the assumption of a steady-state melt regime, large-eddy simulations provide evidence that steady-state is not always the case: tidal variations may lead to switching between shear-controlled and diffusive-convection-controlled melt, and under diffusive-convective-controlled melt, growth of thermal and saline diffusive sublayers close to the ice shelf base can prevent the development of a steady-state regime (Rosevear et al., 2021). Though applying such high-resolution simulations at geophysical scales is impractical, findings from large-eddy simulations may guide the development of parameterisations to represent processes that are sub-grid scale in coarser-resolution ocean general circulation models. However, to date, there are very few in-situ observations to validate or constrain these simulations, as discussed above (Section 3.1).

High-resolution ocean models (approximately 2 km or finer; Graham et al., 2016) have been developed to understand how ocean circulation and ocean heat pathways evolve over the continental shelf and within ice shelf cavities and how they impact ice-ocean interactions (Dinniman et al., 2011; St-Laurent et al., 2013; Hattermann et al., 2014; Nakayama et al., 2019, Nakayama et al., 2021b; Goldberg and Holland, 2022). Such high resolution is critical over the continental shelf and near ice shelf cavities to resolve complex ocean circulation in the presence of rough ice and complex ocean bathymetry, and a number of circumpolar ocean models have been developed at these scales (e.g., Richter et al., 2022; Gallmeier et al., 2023). Despite ongoing work required to improve the representation of small-scale physics in these models, good model-data agreement to capture water mass hydrography, ocean circulation and short-term variability has been achieved (Nakayama et al., 2019). Further efforts are required to couple such high-resolution ocean models with atmosphere models, for example, to capture processes in polynya regions.

The inclusion of ice shelf cavities into global ocean models—fundamental in capturing ice-ocean interactions—is a relatively recent development; very few models have evolving cavity geometries. For global ocean models that do include evolving cavity geometries, there is a trade-off between model resolutions and model integration periods. For some models with longer model integrations, grid resolution over the continental shelf regions and ice shelf cavities is not sufficiently high (typically >1°) to conduct in-depth studies of ice-ocean interactions. Although possible to employ unstructured meshes to refine the resolution over the continental shelf and in ice shelf cavities (Timmermann et al., 2012), the computational cost of such modelling experiments can be prohibitive due to the large domain size and requirements for sufficient resolution. Good model-data agreement is also less achievable for coarser-resolution ocean models as: (1) global models require long model integration to achieve steady states and (2) appropriate representation of important processes and properties is required globally, or over a larger domain, which requires the adjustment of parameter values to achieve a reasonable global fit, rather than being able to model such processes and properties explicitly, as is feasible in regional-scale models.

Turning to ice sheet models, melting at the ice-ocean interface is an external forcing imposed directly using observed estimates of melt, output from ocean models, or calculations from melt parameterisations. The review of Asay-Davis et al. (2017) summarises a number of different parameterisations and coupled models that have been developed and used.

Melt parameterisations have increased in complexity significantly over recent years, from simple depth-dependent parameterisations (Favier et al., 2014; Seroussi et al., 2014) to complex parameterisations that include the role of plumes (Lazeroms et al., 2018), overturning circulation (Reese et al., 2018a) and subglacial freshwater discharge (Pelle et al., 2023). Such parameterisations are calibrated and evaluated based on remote-sensing estimates of melt rates (Schodlok et al., 2016; Jourdain et al., 2020). Comparison of these results with ocean simulations at high resolution or coupled ice-ocean simulations highlights their potential in capturing the overall melt simulated, as well as limitations in accurately reproducing complex spatial patterns of melt; the high melt rates observed near the grounding line are typically underestimated by all parameterisations (Favier et al., 2019; Burgard et al., 2022). These limitations have a significant impact on simulated grounding line evolution and the ice sheet response to ocean-induced melt, as both the overall magnitude of melt and its spatial distribution impact the response of ice streams (Durand et al., 2009; Seroussi et al., 2020).

Ice fronts and rifts play a critical role in ocean heat intrusion into ice shelf cavities and enhanced melting at the ice shelf base. Including kilometre-wide rifts in the geometric description of ice shelf cavities changes mixing processes in the cavity and increases ocean heat intrusion and melt in these cavities (Poinelli et al., 2023a). As these rifts develop, they impact the extent of ice shelves through ice front retreat. The incursion of ocean heat into the cavities is also altered by such ice front retreats and in turn impacts the patterns of melt rate under ice shelf cavities (Poinelli et al., 2023b). Many ice shelves around Antarctica display such large-scale rift systems and the impact of explicitly modelling them could be significant (Walker et al., 2013). Nevertheless, while ice flow models are starting to include calving laws at large scale and high resolution (Choi et al., 2018; Choi et al., 2021), calving laws have been designed mostly for Greenland glaciers, currently remain poorly validated by observations and may not capture the relevant processes underlying calving for each ice shelf cavity system.

Significant progress has been made in coupling ice sheet and ocean models to better capture these processes and understand feedbacks at the ice-ocean interface and beyond. These efforts include both idealised and realistic ice-ocean systems, both of which have highlighted the limitations of parameterisations, which tend to underestimate ocean-induced melt rates and ice mass loss (Goldberg et al., 2012; De Rydt and Gudmundsson, 2016; Seroussi et al., 2017; Pelle et al., 2021). Efforts are also underway to couple ice sheet models into large-scale general circulation models (Muntjewerf et al., 2021; Smith et al., 2021; Comeau et al., 2022), for example, as used in the Coupled Model Intercomparison Project. For example, along with incorporating models for both the Greenland and Antarctic Ice Sheets, the UK Earth System Model also incorporates schemes to capture iceberg discharge as a source of freshwater and extends the ocean model to capture ice shelf cavities (Smith et al., 2021). However, these climate models are run at relatively coarse resolution, and ice shelf melt parameterisations may still be required to capture spatial patterns of melt rates accurately, especially for small ice shelves.

Finally, community-led modelling initiatives have made significant contributions to improving the representation of ice-ocean interactions in models. These include Model Intercomparison Projects, which broadly aim to understand how and why models perform differently and to make recommendations for improvement of the representation of physical and numerical schemes. For example, the Marine Ice Sheet Model Intercomparison Project (MISMIP) and its follow-on efforts MISMIP3D and MISMIP+ have facilitated the assessment of grounding line migration schemes and their implementation, first for flow line models and then for three-dimensional cases (Pattyn et al., 2012; Pattyn et al., 2013). While early efforts highlighted the impact of model resolution and sub-grid grounding line scheme as a main driver of the results, the latest results from MISMIP+ (Cornford et al., 2015) demonstrated that model accuracy is now impacted mostly by the choice of basal friction law rather than numerical schemes. Similar to the MISMIP efforts, the Ice-Shelf Ocean Model Intercomparison Project was established to investigate ice-ocean processes in ice shelf cavities. Here, melt rates were shown to vary strongly as a function of vertical resolution and treatment of temperature, salinity and freshwater flux in the boundary layer (Losch, 2008; Mathiot et al., 2017). Initiatives exist for understanding the impact of single processes related to ice-ocean interactions; these include the Calving Model Intercomparison Project (Jordan and Pattyn, 2023) and Antarctic BUttressing Model Intercomparison Project (Sun et al., 2020) and a coupled ice-ocean model intercomparison project is also underway (Asay-Davis et al., 2016). By providing a framework both for how data can be integrated into models effectively and how models can be used to inform data collection, FUSION will be distinct from these Model Intercomparison Projects; however, FUSION should nevertheless draw from these initiatives to establish a baseline of model capabilities in representing ice-ocean interactions and highlight avenues for further improvements.

To advance understanding of ice-ocean interactions, the research community requires improved observational data including: (i) sustained observations to separate long-term trends from short-term variability; (ii) filling the large data gaps around Antarctica (e.g., measurements of bathymetry on the continental shelf) and (iii) data targeting under-observed processes (e.g., melt at deep grounding lines). Recent international collaborative projects have made progress on some of these issues. For example, autonomous underwater vehicle explorations in the cavities under the Kamb Ice Stream and Thwaites Glacier have produced unprecedented fine-scale observations of ocean conditions in difficult-to-reach locations at the grounding line and within basal crevasses (Lawrence et al., 2023; Schmidt et al., 2023; Washam et al., 2023). Meanwhile, the International Thwaites Glacier Collaboration has made the first simultaneous measurements of basal melt rate and ice shelf cavity ocean properties (Davis et al., 2023). FUSION should clearly define the essential variables needed to understand ice-ocean interactions and the required data properties to meet the needs of data users, ensuring that these targeted studies produce data that can be scaled up to tackle continental-scale problems, with long-term impact.

The World Meteorological Organisation’s Global Climate Observing System defines a set of ‘essential climate variables’ intended to provide the observational data needed to understand and predict changes to the global climate (World Meteorological Organisation, 2022). Many of the defined variables are relevant to ice-ocean interactions (e.g., ice shelf thickness and grounding line position, ocean temperature and salinity). However, the existing inventory also misses variables that are key to understanding and predicting ice-ocean interactions, such as ice shelf basal melt rate, subglacial water discharge and continental shelf bathymetry. Also important is to define the necessary parameters to make an observational dataset functional to end users, such as the required frequency, duration and precision of observation. Essential variables were defined for the Greenland Ice Sheet-Ocean Observing System encompassing the ocean, ice sheet and atmosphere systems, as well as the observational parameters required to measure them (Straneo et al., 2019). FUSION, focusing on Antarctica, could build on the Greenland Ice Sheet-Ocean Observing System recommendations, adapting them to suit the unique environmental setting of the Antarctic-Southern Ocean system. A consistent framework for observing ice-ocean observations would allow field programs with diverse field sites and discipline backgrounds to make observations that are functional for the broadest possible range of potential end users. A clearly defined set of parameters for observations also gives the necessary specifications to inform future technological developments.

The development of FUSION is also an opportunity to reflect on data that already exist and prioritise the gaps that need to be addressed. These data gaps may be geographical: for example, the substantial gaps in bathymetric data on the Antarctic continental shelf which significantly hamper understanding of oceanic processes crossing the shelf (Dorschel et al., 2022; McMahon et al., 2023). There may be gaps in the type of data collection needed to understand physical processes: for example, direct observations of melt at the grounding line (Rignot, 2023), concurrent observations of melt rates and ocean variables in the ice shelf cavity to test melt rate parameterisations (Rosevear et al., 2022a), or observations in ice shelf cavities with a broader range of dominant melt mechanisms (Rosevear et al., 2021). There may also be gaps in existing technologies, such as the need for improved geolocation of floating instruments, deployed under ice, which spend long periods of time unable to reach the ocean surface to relocate via GPS and to transmit data.

The required observations are likely to be diverse in location, requiring expensive and challenging logistics to reach field sites far from permanent Antarctic research stations, as well as bulky and expensive equipment and highly specialised field teams. International collaborations can help to provide the scale of support needed for this type of cutting-edge science. One notable example is the SCAR RINGS Action Group, which clearly defines an essential gap in Antarctic bed topography data, allowing the international research community to target a broad range of field projects more effectively and address the problem collectively (SCAR RINGS Action Group, 2022). A key aim of this group is to provide a protocol for collecting airborne- and ground-based measurements of the Antarctic Ice Sheet, thereby ensuring that data collected by different instruments and platforms are consistent and comparable. Similarly, FUSION seeks to use community consensus to help the international research community to coordinate and prioritise observational data collection, but even further it seeks to develop guidance for integrating the use of models into the framework, including for example in experimental design.

The ice-ocean scientific community has seen great improvements in previous years in its ability to model the physics of the ice-ocean interface at multiple scales and with dynamic boundaries (Snow et al., 2017; Jordan et al., 2018; Gladstone et al., 2021; Middleton et al., 2021; de Diego et al., 2023; Scott et al., 2023; and references above). For example, ocean and ice sheet models provide critical insight into the processes driving ice-ocean interactions and their sensitivities where observational data are often limited. Furthermore, models are able to provide a more complete picture of the ice-ocean system and its connectivity with the broader ocean and ice sheet, enabling investigation of both internal and external drivers of change across a range of spatial and temporal scales. At the same time, the volume and diversity of observations of ice-ocean interactions has grown immensely with the onset of new satellite sensors and new technologies (Gourmelen et al., 2017; Adusumilli et al., 2020; Davis et al., 2023; Schmidt et al., 2023). However, our ability to leverage these observations within models has not kept pace, due to the nature of the observations, lack of constraints in our inherent errors and limitations in our understanding of how to best assimilate observations in models. FUSION can provide essential guidance on how to (i) reconcile multiple observational data streams; (ii) reconcile the representation of processes in fine-scale models with those of larger-scale stand-alone or coupled ocean, ice sheet and general circulation models (e.g., via sub-grid scale parameterisations) and (iii) assimilate data into models.

Ocean-driven ice shelf melt rate is a key example of an essential variable for which estimates can vary widely depending on the underlying method used to derive it. Methods for estimating melt can be classified broadly as oceanographic (e.g., Jenkins et al., 2018), geophysical (e.g., Nicholls et al., 2015), or satellite-based (Adusumilli et al., 2020). Geophysical estimates, for example using autonomous phase-sensitive radio echo sounders, give very accurate measurements with high temporal frequency, but are limited to point locations. By contrast, satellite altimetry-based methods provide broad spatial coverage and provide long-term temporal averages, but are limited by assumptions regarding firn compaction and hydrostatic floatation. Both methods suffer in locations where ice shelves are highly crevassed. Oceanographic estimates, encompassing ship-, autonomous vehicle-, or instrument-based methods to generate estimates over a range of temporal and spatial scales, may suffer from similar limitations as geophysical or satellite methods or may be limited to collection in summer months. The range of spatial and temporal scales over which each of these different methods apply can present a challenge in reconciling these different data streams and making meaningful comparisons of melt (e.g., see Figure 2). Cook et al. (2022) propose a framework that meets this challenge, ensuring comparability of the different estimates, quantification of errors in the methodologies and improved accuracy and reliability of melt estimates. This framework could provide an appropriate basis for a broader framework on ice-ocean interactions (i.e., FUSION) that would facilitate model validation and the accuracy of simulations that employ data assimilation.

A critical gap in modelling ice-ocean interactions is in reconciling the representation of key processes across scales. For example, ice shelf basal roughness impacts ice-ocean interactions at sub-metre scales (Watkins et al., 2021; Section 3.1). However, the relatively coarse resolution of regional and global models cannot resolve such scales, and parameterisations to capture the impact of fine-scale roughness at resolutions appropriate for modelling do not yet exist. There is the need to develop such parameterisations through reconciling fine-scale process-based models (e.g., large-eddy simulations) with large-scale models of ice-ocean interactions. Incorporating this aspect in FUSION will be essential to ensuring that any improvements made in fine-scale modelling and data collection have direct translation into improved modelling.

Observations at the ice-ocean interface give at best incomplete or indirect measurements of properties that are essential to characterise or understand. Assimilation of data in models (Figure 3) seeks to address this issue by providing a model state, consistent with observations, and an estimate for ‘unknown fields’, such as ice damage and small-scale roughness, that together can be used to interrogate the processes underlying ice-ocean interactions. A powerful form of assimilation, variational data assimilation, has been developed over several decades for weather forecasting (Rabier et al., 2000) and more recently applied to large-scale ocean (Wunsch et al., 2009; Verdy and Mazloff, 2017) and ice sheet (Morlighem et al., 2010; Goldberg et al., 2015) models. Variational data assimilation involves constraining large datasets subject to model physics, and to make this tractable the mathematical adjoint (or gradient) of the model must be found. In most cases this effort requires automatic differentiation—the same technology that allows training of deep neural networks (Baydin et al., 2018). A drawback of variational data assimilation is that automatic differentiation is difficult to implement in mature models, meaning that adjoint-based assimilation and state estimation generally does not benefit from community-wide experimentation and intercomparison, as is the case in the ice-ocean modelling community (Asay-Davis et al., 2016; Cornford et al., 2020). In operational ocean data assimilation, simpler methods are therefore used, such as optimal interpolation or filtering methods (Brasseur et al., 2005), but such methods can often introduce ‘jumps’ in ocean properties, which can confound any rigorous analysis (such as heat budgets) carried out with the assimilation products and limit use as initial conditions for long-term forecasts (Wunsch et al., 2009).

Another drawback is that quantifying how observational (including instrumental and methodological) and model-based errors propagate to errors in the state estimate is challenging, and even more challenging is to determine how to minimise uncertainty in observations. These are considerations that should be examined in FUSION. In spite of these drawbacks, the adjoint method demonstrates considerable utility, both in quantifying the model state and in determining the sensitivity of any model variable to another, thus presenting as a powerful tool in determining where more or improved observations would be most impactful (Heimbach and Losch, 2012; Goldberg et al., 2020; Nakayama et al., 2021c).

Efforts are needed to bring model-data synthesis in ice-ocean models to the level of sophistication that exists in other geoscientific modelling areas, with more advanced frameworks and diverse data streams. FUSION should include model-data synthesis as a key component to provide the basis for these developments, drawing on lessons learnt from past community initiatives in these areas (e.g., European Centre for Medium-Range Weather Forecasts support of 4DVar; NASA support of a global ocean state estimate).

The overarching aims of an Antarctic FUSION are to define the developments needed to characterise the state and variability of the Antarctic-Southern Ocean system and the processes that underlie this system, through integration and reconciliation of observations and models, particularly targeting the knowledge gaps highlighted in Section 3, and to facilitate improved accuracy in projections of future changes in this system. FUSION should define systematic approaches to the following elements (discussed in Section 4):

1. Measuring ocean, ice sheet and atmosphere properties that are necessary and sufficient to establish a baseline of the state and variability of the Antarctic-Southern Ocean system and the processes underlying ice-ocean interactions. This approach requires identification of the essential variables, technologies to observe the essential variables, parameters to ensure that the variables are adequately characterised (i.e., sampling frequency and locations) and directions for reconciling measurements across observational platforms and technologies to ensure spatial and temporal consistency in datasets acquired. Data collected in accordance with FUSION should encompass both sustained monitoring programs (i.e., programs of sustained long-term observations), to ensure that the state and variability of each system is adequately characterised, and campaign-style programs that are vital for process-based understanding. This element will draw on the successful approaches of the Framework for Ocean Observing (Lindstrom et al., 2012), which ‘guides our implementation of an integrated and sustained ocean observing system to deliver maximum impact for our user base and society’, and the Greenland Ice Sheet-Ocean Observing System (Straneo et al., 2019), which captures the fundamental components of an observational framework for the Greenland Ice Sheet.

2. Integrating models and other technologies in the design of observational systems and programs. Novel methods in data assimilation can quantify the sensitivities of any essential variable to changes in another variable, for example, changes in ocean-driven ice shelf melt rate as a function of thermocline depth or ice thickness and can be used to inform measurement parameters and priority locations for observations. The approach can draw on principles from the Southern-Ocean Observing System Design Capability Working Group, which uses models and other technologies to facilitate optimal experimental design, that is, minimising the amount of data and effort required to characterise a variable or system. Applications of the observing system design include providing observational targets for the Animal Borne Ocean Systems program (McMahon et al., 2021) and the placement of ocean moorings to measure heat fluxes in the Southern Ocean (Wei et al., 2020b).

3. Reconciling observations and models to improve understanding of the processes underlying ice-ocean interactions and feedbacks. This approach requires the development of a standard protocol for how observations should be used to validate models, methods to integrate observations into models (i.e., data assimilation) and parameterisations of fine-scale processes captured in high-resolution models to large-scale general circulation ocean and ice sheet models.

As a first step, we will target the development of a framework for ocean-driven ice shelf melt, through an International Association for the Physical Sciences of the Ocean Best Practice Study Group, led by the Joint Commission on Ice-Ocean Interactions. This framework will present a standardised, best practice approach to observing, measuring and modelling ice shelf melt, and include methodologies to reconcile melt estimates across instruments and platforms, guided by the elements outlined in Section 5.1. We aim to publish the framework by late 2025, in time to provide the basis for fieldwork testing associated with Antarctica InSync, as below. Our proposed implementation approach is as follows:

• Develop an initial framework targeting one essential variable—ocean-driven ice shelf melt—through a series of dedicated workshops and a broad-scale community consultation process.

• Seek formal endorsement of the framework for ocean-driven ice shelf melt through international organisations and initiatives, including the SCAR, the Southern Ocean Observing System, the Scientific Committee on Oceanic Research, the International Union of Geodesy and Geophysics and its associations, namely the International Association of the Physical Sciences of the Ocean and the International Association of Cryospheric Sciences, the World Climate Research Program and its working groups, including the Southern Ocean Freshwater Input from Antarctica initiative and other relevant Model Intercomparison Projects. Endorsement and advertising via these organisations and initiatives will ensure maximum visibility within the community and broad-scale adoption.

• We will leverage existing or planned (but funded) missions to launch, test and refine the framework through fieldwork and synthesis. For example, we plan to coordinate with existing and planned fieldwork opportunities through the SCAR, the Antarctica InSync initiative 2027–2030 as part of the United Nations Decade of Ocean Science for Sustainable Development, the Network for the Collection of Knowledge on Melt of Antarctic Ice Shelves (a key initiative coordinating estimates of melt derived from autonomous phase-sensitive radio echo sounders) and other individual or group programs. For example, we propose to set up a working group associated with Antarctica InSync that engages relevant stakeholders, including groups that have funded and/or planned fieldwork to deploy instruments that can be used to monitor/estimate melt. We will also establish connections and coordinate with model- or data-based programs, including relevant satellite missions and World Climate Research Program Model Intercomparison Projects or working groups, where appropriate.

Development of the broader FUSION will build on the framework for ocean-driven ice shelf melt, but will require a larger-scale, multidisciplinary effort, coordinating with the international community to define the relevant essential variables and implement a strategy, in keeping with the underpinning science questions (e.g., Section 3).

Progress in understanding ice-ocean interactions, and especially introducing data synthesis into ice-ocean models, requires advances in our observations of melt processes (Section 4). There are substantial challenges in gathering the data required, due to the remote location of field sites, the challenge of maintaining equipment over extended periods, and the heavy logistical costs of deploying instrumentation under an ice shelf (Section 4.1). These challenges have been exacerbated by a recent slowdown in field access as many nations have cut back field programs (Mervis, 2023) on top of the delays and cancellations caused by COVID-19 (Liggett and Herbert, 2021).

There are also significant inequities in access to the field support required to undertake these observations. The expense of Antarctic logistics means that collecting observational data is closed to many nations (Robel et al., 2024), including those most heavily affected by the consequences of ice sheet mass loss (e.g., Pacific Island nations). Similarly, the high cost and long timeline of Antarctic field logistics presents a barrier to early-career researchers, who are typically on short-term contracts and require rapid outputs for career progression (Levine et al., 2020). This barrier is on top of the well-documented barriers faced by researchers with diverse identities (including race, ethnicity, gender identity, sexuality, socioeconomic status, language and ability) in Antarctic science (Griffiths et al., 2021). These barriers are exacerbated in the field, where discrimination, harassment and bullying are an ongoing problem (Nash et al., 2019; Nash, 2021; Seag et al., 2023). A recent report from the SCAR found that female and early-career researchers were impacted more adversely by COVID-19 than their counterparts, experiencing adverse effects in stress levels, financial security and career progression (Liggett and Herbert, 2021).

Gathering the coordinated datasets required to advance the field will require a sustained funding commitment from national Antarctic programs, and the broad geographical range of observations will mean that international coordination is key to success. Given the challenges outlined above, plans to coordinate field observations need to consider and implement strategies for developing and retaining a diverse, skilled workforce, which can draw on a range of grassroots efforts working to promote diversity in the polar science community (Griffiths et al., 2021). There is no single solution to these issues, but the International Thwaites Glacier Collaboration (ITGC) provides a model for building a more positive and inclusive culture in large field programs (Karplus et al., 2022). The ITGC has published useful resources outlining expected community values and behaviours (ITGC IDEA Committee, 2022), as well as recommending best practices for field and ship-based work (ITGC IDEA Committee, 2021).

FUSION will necessarily rely on extensive community consultation, working with international groups and initiatives, such as the SCAR, the Southern-Ocean Observing System and the Scientific Committee on Ocean Research, to ensure broad agreement amongst ice-ocean researchers on the approach developed and widespread adoption of FUSION. Such engagement will be necessary in ensuring that the recommendations in FUSION around field survey protocols and design are adopted by programs that coordinate fieldwork. Programs like Antarctica InSync and the International Polar Year in 2032, through their widespread endorsement by multiple national Antarctic programs, provide opportunities to share and optimise logistics and to test initial framework design.

Finally, the dissemination of data in a timely fashion and in accordance with FAIR (findable, accessible, interoperable, reuseable) data principles is key to underpinning progress on all fronts. Sharing data and knowledge helps the field to advance more rapidly, fosters interdisciplinary research and increases transparency and access to scientific knowledge for stakeholders outside the research community. The ‘democratisation of data’ can help to improve equity, by making high quality field observations open to researchers who are unable to undertake fieldwork themselves (Levine et al., 2020). FUSION should incorporate clearly defined data streams to ensure ice-ocean research is more accessible to a broader range of scientists as well as other stakeholders.

In a time of accelerating ice mass loss from Antarctica, observing, modelling and understanding ice-ocean interactions is critical. The development and implementation of an Antarctic framework for understanding these interactions—FUSION—will offer a systematic approach to addressing key knowledge gaps in the Antarctic-Southern Ocean system and provide an opportunity to accelerate improvements in projections of sea level rise and climate and ecosystem change into the future. Key to success in this initiative is coordinated international collaboration, underpinned by a commitment to equity, diversity and inclusion and sustained funding by national Antarctic programs.

All data included in this study have been published previously and are available publicly via their corresponding references.

The authors wish to thank Dr Luciano Pezzi and Dr Jody Deming for editing the paper, and Dr Marquez and three anonymous reviewers for their suggestions that greatly improved the manuscript. They also thank Stacey McCormack from Visual Knowledge for the adaptation of Figures 1 and 3.

FSM was supported under an Australian Research Council (ARC) Discovery Early Career Research Award (DECRA; DE210101433) and the ARC Special Research Initiative (SRI) Securing Antarctica’s Environmental Future (SR200100005). SC received grant funding from the Australian Government as part of the Antarctic Science Collaboration Initiative program. HS was supported by grants from NASA’s Cryospheric Science Program (#80NSSC21K1939 and #80NSSC22K0383). DNG was supported by NERC grants NE/T001607/1 and NE/S006796/1. The Joint Commission on Ice-Ocean Interactions is a joint commission of the International Association of the Physical Sciences of the Oceans and the International Association of Cryospheric Sciences and received funding from these parent organisations to host the online workshop in October 2022.

The authors have declared that no competing interests exist.

FSM led the conception and design. SC developed Figure 2. All authors contributed to writing and editing the manuscript and approved the submitted draft for publication.