Atlantic water intrusion triggers rapid retreat and regime change at previously stable Greenland glacier

The study addresses why some previously stable, shallow and well-grounded Greenland tidewater glaciers can undergo sudden, rapid retreat and dynamic change. It focuses on ocean–glacier interactions at the ice–ocean interface, where three primary mechanisms can force terminus change: (1) submarine melting driven by the advection of warm, deep Atlantic Water that undercuts calving fronts; (2) enhanced terminus melt from near-ice circulation and plumes fueled by subglacial discharge sourced from surface melt; and (3) changes in calving rates due to the formation and breakup of rigid ice mélange that modulates backstress. Despite similar external forcing, glaciers often respond heterogeneously due to fjord and glacier geometry, subglacial hydrology, and ocean circulation, complicating prediction. Southeast Greenland experienced distinct episodes of retreat attributed either to AW intrusion (early 2000s) or to high surface melt and mélange loss (2016). The purpose of this study is to characterize recent changes at K.I.V Steenstrups Nordre Bræ (Steenstrup), quantify retreat, thinning, and acceleration, and determine the dominant forcing mechanisms and evolving sensitivities that triggered its destabilization beginning in 2018.

Prior work identifies multiple controls on Greenland tidewater glacier behavior: AW-driven submarine melt initiating calving undercutting; subglacial discharge–driven plume melt linking atmospheric forcing to ocean–ice heat transfer; and mélange-related backstress changes modulating calving. Studies show spatial heterogeneity in responses, attributed to fjord depth and sills, glacier geometry (including retrograde beds), hydrology, and shelf current structure. Deep termini (e.g., Jakobshavn, Kangerlussuaq) have been linked to mélange rigidity changes and/or AW intrusion, while some shallow, well-grounded glaciers are thought protected by bathymetric sills. In Southeast Greenland, early-2000s retreats were associated with AW reaching fjords via the Irminger Current, whereas a synchronized 2016 retreat has been linked to atmospheric forcing via elevated surface melt and/or mélange loss without a concurrent AW anomaly. These contrasting findings underscore the need to disentangle the dominant drivers at less-studied outlets such as Steenstrup.

Calving front positions: Terminus positions along the centerline were manually digitized from Landsat 4–8, ASTER, and Sentinel-1 imagery from 1985–2021, extending a prior dataset.

Glacier velocity: Changes in speed were analyzed using ITS_LIVE data. Annual mosaics (1985–2015) and all available scene-pair velocities (2016–2021) with >1% coverage were used. Time series sampling used 1×1 km boxes at points 6–23 km from the 2016 front; median speeds and reported errors were used when coverage exceeded 70%. For 2019–2021, custom annual mosaics were computed from scene-pair speeds and errors, producing weighted means and standard errors; differences with 2016 were tested for significance using two-tailed unpaired t-tests.

Mélange presence: Sentinel-1A/B IW SAR image pairs (6-day intervals) were processed (GrIMP-style speckle/feature tracking) to map the extent of rigid mélange in 2020–2021. The method detects coherent, rigid mélange; non-rigid mélange is not mapped. Mélange length along an extended centerflowline was compared to contemporaneous terminus positions to quantify potential backstress.

Ice discharge: Monthly discharge (1985–2021) was computed across an upstream flux gate perpendicular to flow, sampled every 250 m across-gate. Surface velocities from optical (Landsat/ASTER MIMC, SETSM) and SAR (TerraSAR-X/TanDEM-X, GrIMP Sentinel-1) were combined, smoothed, and gap-filled via a Kalman filter using median seasonal variability. Depth-averaged velocities were assumed equal to surface velocities. Ice thickness was derived by differencing DEM surface elevation (AeroDEM through ArcticDEM) and BedMachine v4 bathymetry. Discharge equals the sum over bins of thickness × velocity × ice density (910 kg m−3). Uncertainty was estimated via a 1000-member Monte Carlo, yielding typical uncertainties up to ~20% and as high as ~50% in earlier records due to velocity and bed uncertainties.

Topographic analysis: 2-m ArcticDEM strips (2016–2021) were referenced to mean sea level (EIGEN-C64 geoid) and coregistered to 2016 DEMs. Flotation potential was assessed from BedMachine v4 bed topography by comparing ice surface heights to theoretical flotation thickness, indicating regions at or near flotation.

Ocean reanalysis and validation: Monthly ocean potential temperatures (1991–2020) from CMEMS Arctic Ocean Physics Reanalysis were converted to thermal forcing using salinity- and pressure-dependent freezing point calculations. Anomalies were computed over the continental shelf sampling zone CEI. CMEMS fields were evaluated against available OMG CTDs within CEI (2016–2020), showing a mean reanalysis–CTD difference of −0.03 °C, acknowledging limited spatial/temporal CTD coverage and known reanalysis limitations on shallow shelves.

Bathymetry: Regional bathymetry from IBCAO v4.1 and local fjord bathymetry from OMG MBES were used to define sill depths and fjord geometry.

Undercutting melt parameterization: Monthly mean undercutting rate across the submerged calving face was parameterized as a function of depth-averaged thermal forcing in the lower 60% of the water column, subglacial discharge (basin-integrated RACMO2.3p2 surface runoff plus constant basal melt), and water depth at the front (h = 320 m from OMG MBES). The glacier front width was set to 4000 m; basin delineation followed Mankoff et al. Constants followed Rignot et al. (A = 3×10−4, α = 0.39, B = 0.15, β = 1.18). Nominal uncertainty of modeled undercutting was 26%.

• Prior to 2018, Steenstrup’s terminus was stable for decades: the 2015 front was ~200 m from the 1985 mean position, with low seasonal variability (SD ~155 m). Discharge in 2016 was 3.34 Gt a−1 (82nd percentile of GrIS outlets).
• From 2018 to 2021, the terminus retreated ~7.1 km, creating a new ~6 km-long fjord. Retreat segments: ~3.2 km in 2018 (mid-May to Jan), ~3.0 km in 2019, ~2.1 km in 2020; partial advance in early 2021 followed by 2.9 km retreat by December 2021.
• Flow acceleration: Front speeds increased by >270% to peaks of 16.8 m d−1 (Aug 2020) and ~15.0 m d−1 (Oct 2021). Statistically significant speedups propagated inland up to ~40 km (e.g., 24% increase from 1.7 to 2.1 m d−1 at 40 km). The main trunk accelerated while two distributaries decelerated.
• Discharge doubled to 6.37 Gt a−1 by 2021 (93rd percentile), from 3.34 Gt a−1 in 2016.
• Surface lowering and dynamic thinning intensified post-2018: −10 to −20 m a−1 within ~6 km of the 2016 front in 2016–2018; up to ~50 m a−1 thinning 8–10 km upglacier in 2018–2021; cumulative losses >200 m (2016–2020). Thinning and retreat along a retrograde bed left ~1 km of the tongue at or near flotation by 2021.
• Forcing attribution: The major 2018 retreat coincided with an exceptional AW thermal anomaly in CMEMS reanalysis, reaching within ~100 m of the surface. Anomalies peaked at +3.3 °C (380 m depth) and +2.4 °C (186 m, above the −180 m sill), enabling AW to overtop the proglacial sill. Nearby OMG CTD (site 144, ~40 km from terminus) showed >4 °C water up to 130 m depth in Aug 2018. In contrast, 2018 did not show exceptional near-surface temperature or subglacial discharge anomalies.
• Modeled undercutting rates peaked at 1.05 m d−1 in July 2018 (record high) and 0.86 m d−1 in July 2019, driven primarily by thermal forcing; the 2016 peak was 0.69 m d−1. Historic temporary retreats (e.g., 2003, 2010, 2016) align with smaller undercutting anomalies from combined forcing.
• Mélange regime change: Prior to 2018, mélange was minimal and localized. By winter 2020/21, a rigid, fast-flowing mélange formed across the new fjord, correlating with significant winter advance and enhanced seasonal variability consistent with mélange buttressing.
• Stabilization in 2021 likely reflects reaching a secondary sill, reduced driving stress from thinning, and mélange buttressing, but the glacier remains dynamically out of balance with quadrupled velocities and continued vulnerability to AW and mélange variability.

Findings show that Steenstrup, long considered a shallow, well-grounded glacier protected by a proglacial sill, underwent rapid retreat, thinning, and acceleration beginning in 2018 due to an unprecedented intrusion of warm Atlantic Water that overtopped the sill. The event produced record modeled undercutting and initiated dynamic thinning and flow acceleration that propagated tens of kilometers inland. Unlike other SE Greenland retreats in 2016 driven by high surface melt and/or mélange loss, Steenstrup showed limited sensitivity to atmospheric forcing prior to 2018; its geometry likely constrained subglacial discharge and limited plume-driven melt. Post-retreat, glacier geometry changed to increase sensitivity to mélange, leading to strong seasonal advance/retreat in 2020/21. The results challenge expectations that AW is least effective at shallow, grounded fronts and demonstrate that even sill-protected outlets can be vulnerable when exceptional AW anomalies occur. Stabilization in 2021 at a secondary sill may be temporary given ongoing dynamic imbalance, retrograde bed sections inland, and anticipated collapse of distributaries that could reduce mélange buttressing. The study thus refines understanding of tidewater glacier vulnerability, emphasizing the roles of episodic AW access, fjord/sill geometry, and evolving mélange regimes.

This study documents an unprecedented regime change at K.I.V Steenstrups Nordre Bræ: ~7 km retreat, up to ~50 m a−1 thinning near the front, a >270% speedup at the terminus, and a doubling of discharge between 2018 and 2021. The retreat was not triggered by elevated surface melt or mélange loss but by exceptional intrusion of warm Atlantic Water that overtopped a previously protective sill, producing record undercutting melt. Subsequent fjord formation and thinning enabled rigid mélange development and strong seasonality by 2021, indicating an evolving sensitivity to mélange backstress. The findings show that apparently stable, shallow, sill-protected glaciers can rapidly destabilize under AW forcing and contribute substantially to ice-sheet-wide discharge. Future work should prioritize in-fjord CTD observations to validate and refine ocean reanalysis in shallow shelf settings, improved bed and fjord bathymetry near termini, and targeted numerical modeling to test process interactions and produce near-term projections that can be evaluated over years.

• Ocean reanalysis uncertainties: CMEMS products have known limitations on shallow continental shelves (lack of eddy resolution, thermocline depth biases). Validation is limited by sparse OMG CTDs near the study site and variable CTD timing among years.
• Bathymetry/topography uncertainties: Poorly constrained bed topography and proglacial bathymetry near the terminus (BedMachine limitations; limited radar lines) introduce systematic uncertainties in thickness, flotation assessments, and discharge magnitudes.
• Discharge uncertainty: Monte Carlo analyses indicate uncertainties up to ~20% (and as high as ~50% in early records), primarily from velocity and bed errors; however, timing and relative changes are robust.
• Mélange mapping: The SAR-based approach detects only rigid mélange; non-rigid mélange presence and its effects are not captured.
• Generalizability: Conclusions about AW intrusion are supported by reanalysis and limited nearby CTDs; lack of in-fjord CTD profiles constrains definitive attribution and detailed fjord dynamics.