Atlantic Water warming increases melt below Northeast Greenland's last floating ice tongue

Greenland has only three glaciers with floating ice tongues, of which Nioghalvfjerdsfjorden Glacier (79 North Glacier, 79NG) is the largest. While its overall extent remained relatively stable, substantial basal thinning has been detected over the last decades. Two potential drivers can enhance basal melt: warming Atlantic Intermediate Water (AIW) entering the sub-ice cavity via the Northeast Greenland shelf circulation, and increased subglacial discharge driven by atmospheric warming and enhanced surface runoff. Observations alone have been insufficient to distinguish their relative importance. This study addresses the central research question: What controls interannual to decadal variability and long-term trends in basal melt beneath 79NG—the ocean (AIW temperature/transport) or the atmosphere (subglacial discharge)? The work is motivated by implications for glacier stability, buttressing, upstream ice discharge, and global sea level rise.

Prior studies and observations indicate that warm, salty Atlantic-origin waters are routed to the Northeast Greenland shelf via Fram Strait and Norske Trough, providing oceanic heat to the 79NG cavity. Mooring and CTD measurements have documented AIW presence and hydraulically controlled inflow constrained by a sill at the calving front, as well as undercutting of grounded ice. Satellite-derived estimates for 2011–2015 suggested negative mass balance with basal melt exceeding grounding-line ice discharge. However, sustained in situ observations at 79NG began only in 2016, limiting multi-decadal inference. Existing modeling for Greenland glacier cavities (including Petermann and Ryder) has largely used idealized or 2D setups, lacking a realistic, fully coupled 3D representation of shelf-to-cavity pathways and cavity circulation. Theoretical and empirical work suggests an above-linear dependence of basal melt on ocean thermal forcing and a square-root dependence on subglacial discharge for ice shelves/tongues, whereas tidewater glacier regimes can show different scalings (linear with thermal forcing, cubic with discharge).

The authors use the global, unstructured-mesh Finite-volume Sea ice-Ocean Model (FESOM2.1) with an ice-shelf component resolving the 79NG and Zachariæ Isstrøm (ZI) cavities. The mesh is refined to 700 m in the 79NG cavity (sufficient to resolve the local Rossby radius), ~2.5–4 km on the Northeast Greenland shelf and Arctic, and coarser elsewhere. The simulation spans 1970–2021 (after a 1960–1969 spin-up). Vertical discretization uses 86 z-levels (5 m in the top 100 m; up to 25 m layer thickness to 900 m depth, the cavity maximum). Bathymetry and ice shelf geometry are from RTopo-2.0.4, with fixed ice-shelf geometry through time. Atmospheric forcing uses JRA55-do v1.4 at 3-hourly resolution. River runoff is from JRA55-do except around Greenland, where solid ice discharge and liquid water discharge (runoff) are from Mankoff et al.; runoff is routed to coasts at the surface except within the 79NG cavity, where subglacial discharge is injected along the grounding line at ~500 m depth and spread over a 10 km radius (line-plume representation). Sea surface salinity is weakly restored to PHC climatology. Ice-ocean interaction applies the three-equation boundary-layer formulation with velocity-dependent heat/salt exchange; drag coefficient Cd = 1.25×10^-3 (sensitivity tests with larger Cd increased absolute melt but not variability). Eddy effects use GM and isoneutral diffusion where the mesh is coarse; GM is disabled for mesh finer than 30 km. The control experiment (REF) covers 1960–2021 with a 3-minute time step on HPC resources. Sensitivity experiments include: (1) CLIM (2000–2021) replacing interannual subglacial discharge with a climatological seasonal cycle to isolate ocean-temperature effects; (2) discharge-scaling experiments (2010–2014) with 0%, 25%, 50%, 75%, 100%, 125%, and 150% of reference subglacial discharge. Statistical analysis detrends time series before computing correlations and assesses significance at the 95% level. Hydrographic observations in southern Norske Trough are compiled into decadal means for evaluation. Pathways and lags are identified using pointwise lagged correlation between detrended AIW temperatures and basal melt.

• Mean cavity circulation and melt: Simulations show a bottom-intensified AIW inflow through the deepest channel at the main calving front (69 mSv, exceeding mooring-based 46 ± 11 mSv), hydraulically constrained by a sill; gravity current speeds exceed 0.3 m s^-1 with cooling due to entrainment. AIW circulates cyclonically; the meltwater plume accelerates along the southern flank, reaching >0.3 m s^-1, and detaches near neutral buoyancy, exporting meltwater primarily through the main calving front at ~150 m depth (62 mSv, 82%; Dymphna Sound 14 mSv, 18%). Based on freshwater input (0.54 mSv) and outflow (~76 mSv), the plume contains ~1–2% meltwater. The long-term mean basal meltwater flux is 17.0 ± 7.9 km³ yr^-1, aligning with mooring heat-transport estimates (17.8 ± 5.2 km³ yr^-1) and exceeding satellite-based estimates (11.9 ± 1.6 km³ yr^-1). Maximum local basal melt rates reach ~85 m yr^-1 near the grounding line.
• Trends and variability: Annual mean basal melt exhibits strong interannual variability (σ ≈ 3.9 km³ yr^-1) and an increasing trend of 2.3 km³ yr^-1 decade^-1 over 1970–2021 (3.4 km³ yr^-1 decade^-1 over the last two decades). AIW maximum temperature at the calving front increases by 0.19 °C decade^-1 (σ ≈ 0.36 °C), consistent with upstream Fram Strait warming. Subglacial discharge more than doubled from 3.3 km³ yr^-1 (1970s) to 6.7 km³ yr^-1 (2010s), trend 0.9 km³ yr^-1 decade^-1 (σ ≈ 2 km³ yr^-1).
• Dominant control by ocean temperature: Basal melt correlates strongly with AIW temperature at the calving front (r = 0.85, 1970–2021, 99% significant). With climatological discharge (CLIM), the correlation strengthens (r = 0.89 for 2000–2021) compared to REF (r = 0.77). Lagged correlations trace AIW temperature anomalies and associated melt variability backward along the pathway: Norske Trough (≈1-year lag), shelf break (≈2 years), and eastern Fram Strait (≈3 years), indicating predictability from WSC monitoring.
• Sensitivity relationships: In the absence of interannual discharge variability, basal melt scales approximately quadratically with AIW temperature. Discharge-scaling experiments show a square-root dependence of melt on subglacial discharge; without summer discharge, melt is nearly seasonally constant. In the REF, interannual basal melt and subglacial discharge are weakly correlated (r = 0.22, not significant).
• Future implications: CMIP6 multimodel projections indicate increases in maximum water-column temperature at the Norske Trough entrance between 1981–2000 and 2081–2100 of about +1.8 °C (SSP1-2.6), +2.7 °C (SSP2-4.5), +3.6 °C (SSP3-7.0), and +4.7 °C (SSP5-8.5). Applying the quadratic sensitivity suggests potential basal melt rates of approximately 40, 63, 92, and 140 km³ yr^-1, respectively; even SSP1-2.6 implies ~161% increase relative to 1981–2000. A projected 3.3× increase in Greenland runoff (CMIP5 RCP8.5) would increase annual mean basal melt by ~18% via the square-root scaling, indicating ocean dominance persists.

The results directly address the central question by demonstrating that AIW temperature variability, rather than interannual subglacial discharge variability, primarily controls basal melt beneath 79NG on interannual to decadal timescales. The strong correlation between basal melt and AIW temperature at the calving front, and the coherent lagged pathway from Fram Strait (≈3-year lead), highlight the role of shelf-to-cavity oceanic heat transport. This finding implies predictability potential using existing Fram Strait WSC moorings and underscores the importance of large-scale ocean variability in modulating glacier–ocean interactions. Although subglacial discharge has increased over recent decades and intensifies seasonal melt via plume dynamics, its interannual variability contributes comparatively little to annual basal melt variability. The quantified quadratic (temperature) and square-root (discharge) sensitivities provide a mechanistic basis to translate projected ocean and runoff changes into basal melt projections. Under future warming, ocean-driven increases in AIW temperature likely continue to dominate basal melt, with implications for thinning, buttressing, and potential transition to a tidewater state under high-emission scenarios. The approach and findings are consistent with relationships observed for other Greenland and Antarctic ice shelves and suggest broader applicability across similar systems.

This study provides a dynamically consistent, global-to-cavity modeling framework resolving the 79NG cavity at 700 m that links Atlantic Water variability to basal melt beneath Greenland’s largest floating ice tongue over 1970–2021. It shows that interannual basal melt variability is dominated by AIW temperature changes, traceable from Fram Strait with multi-year lags, and quantifies sensitivities: quadratic dependence on ocean temperature and square-root dependence on subglacial discharge. Mean circulation, melt patterns, and fluxes are consistent with available observations. Applying these sensitivities to CMIP6 projections suggests substantial increases in basal melt across emissions scenarios, implying continued ocean dominance and potential destabilization under strong warming. Future research should incorporate time-evolving ice-shelf geometry and grounding lines, improved plume representations (including channelized outlets and axisymmetric plumes), tides, and coupled ice sheet–ocean feedbacks to refine projections and assess stability thresholds.

• Fixed ice-tongue geometry: The model does not include time-evolving ice-shelf thickness/grounding line; observed thinning over recent decades is not represented.
• Subglacial discharge representation: Implemented as a line plume uniformly along the grounding line rather than localized axisymmetric outlets; plume geometry affects entrainment and melt rates.
• Missing sub-ice channelization: Kilometer-scale basal channels that can focus melt and modify plume structure are unresolved, potentially underestimating localized melt.
• No tides: Tidal motions in the cavity are omitted; observations suggest they are weaker than mean AIW inflow but could modulate mixing locally.
• Warm temperature bias: A ~0.3 °C positive temperature bias traced to the Nordic Seas (insufficient eddy resolution and lateral heat flux representation) likely raises mean melt rates while having limited impact on variability and sensitivities.
• ZI geometry changes not represented: The neighboring glacier’s transition to a tidewater state in the 2010s is not captured in this configuration.