The role of ocean and atmospheric dynamics in the marine-based collapse of the last Eurasian Ice Sheet

Rising sea levels pose a global threat, motivating efforts to understand the stability, dynamics and meltwater fluxes of ice sheets. A key knowledge gap is the non-linear interplay among oceanic and atmospheric circulation, mass balance, subglacial conditions, and calving dynamics that controls marine sectors of Greenland- and Antarctic-like ice sheets, yet direct observational timescales are short. Former Northern Hemisphere ice sheets provide longer-term context, but marine sectors remain less constrained than terrestrial parts. This study reconstructs the last deglaciation of the marine sectors of the Eurasian Ice Sheet (EIS) and its disintegration into the British-Irish (BIIS), Fennoscandian (FIS) and Kara-Barents Sea-Svalbard (KBSIS) components. Leveraging extensive new geomorphic mapping, high-resolution seismic data, and a comprehensive, recalibrated radiocarbon dataset, the authors produce 1 ka time-slice reconstructions from 20 to 14 ka (extended to 25–10 ka using published terrestrial and western shelf reconstructions). These reconstructions are coupled with a steady-state ice sheet model to estimate evolving ice volumes and meltwater fluxes, and with eastern Nordic Seas summer SST proxy records, to evaluate the relative roles of atmospheric SMB forcing versus oceanic controls on the retreat of marine sectors.

- Data compilation and mapping: The study compiles and updates radiocarbon-dated deglacial chronology and glacial geomorphology across the North Sea, Mid Norwegian margin, Svalbard shelves and Barents/Kara Seas, drawing on Olex bathymetry and other seabed imagery, shallow acoustics, and published datasets. Focused landforms include ice marginal features (terminal moraines, grounding-zone wedges) and glacial lineations. Mapping for the Mid Norwegian margin includes an extensive new effort; other regions largely synthesize published work. - Geochronology: Only marine radiocarbon dates (molluscs, foraminifera) from marine sites are used to time the marine deglaciation. All dates are recalibrated with Marine20 using OxCal. Given spatiotemporal variability and correlation uncertainties, ΔR is conservatively set to 0. Dates of limited relevance or problematic context were excluded; stratigraphic context and potential for reworking were evaluated. - Ice-sheet reconstruction and modelling: Time-slice margins (25–10 ka) are converted to 3D steady-state ice geometries using ICESHEET v1.0, assuming perfectly plastic rheology. Inputs include ice margins, basal topography (GEBCO2020), and spatially parameterized basal shear stress guided by topography and sediment cover (PaleoMIST 1.0), with sensitivity experiments varying shear stress ±20% to bracket volume uncertainty. Glacial isostatic adjustment is computed with SELEN v4 using an Earth rheology based on VM5i; shoreline migration and grounding line adjustments are included. The evolving bed is iterated once with GIA (net effect ~−5% on volume). Global sea-level partitioning (marine vs terrestrial sectors) follows Waelbroeck et al. Volume-to-sea-level conversions assume constant ocean area. - Ocean temperature time series: Summer SSTs are reconstructed from planktonic foraminiferal assemblages in two dated piston cores under Atlantic inflow pathways: LINK17 (North Sea margin) and JM03-373PC (western Svalbard margin). Transfer functions use C2 and WAPLS (1 component) with the 100-µm modern dataset; RMSEP ~1.93 °C. Chronologies are radiocarbon-based, calibrated consistently with the deglacial dataset. SST estimates emphasize relative trends and warmest summer conditions, acknowledging reduced sensitivity below ~4 °C and unrepresented seasonal/sea-ice complexities.

- Spatiotemporal deglaciation: The EIS marine sectors retreated asynchronously. The Norwegian Channel Ice Stream (NCIS) underwent rapid retreat, deglaciating to Skagerrak/Swedish west coast by ~17 ka. The BIIS and FIS separated at ~18.7 ka along a line east of Shetland and west of the Norwegian Channel; associated drainage of an ice-dammed lake south of Dogger Bank occurred at ~18.7 ka. - Northern sectors: Along the western Barents Sea margin, initial retreat from the shelf edge began in Storfjorden Trough at ~20 ka, with broader retreat ~1 ka later. The Bjørnøya Trough recorded episodic retreat with a dated readvance at ~17 ka and multiple grounding-zone wedges. The FIS and KBSIS separated in the Barents Sea at ~15 ka via a coastal corridor; by 14 ka most of the KBSIS had collapsed, with residual ice on high Arctic archipelagos. Eastern KBSIS deglaciation was later than DATED-1 estimates: main Saint Anna/adjacent troughs remained ice covered until ~16 ka; Novaya Zemlya began partial deglaciation ~15 ka and the northern island was not ice-free until ~14 ka. - Ice-sheet volumes and sea-level equivalent (SLE): Peak EIS volume occurred at 19 ka: 7.9 (7.1–8.6) ×10^6 km³, equivalent to 20.04 (17.98–21.90) m SLE. From 20–14 ka, the EIS lost 12.57 (11.24–13.78) m SLE. Peak eustatic SLR contribution of ~3.95 m occurred at ~15 ka, partitioned roughly equally between Fennoscandian and Barents Sea sources. Of the marine-based component at this peak, 87% (~1.75 m) derived from KBSIS collapse. - Marine margin exposure: The marine-terminating margin length increased from ~38% of the total (5055 km) at 20 ka to ~47% (~7700 km) by 16–17 ka, highlighting enhanced vulnerability of marine sectors. Marine grounding persisted longer at higher latitudes. - Forcing mechanisms: Southern marine sectors (North Sea/NCIS) deglaciated early (19–17 ka) largely under atmosphere-driven SMB forcing and ice-dynamics feedbacks along retrograde beds, rather than subsurface ocean warming (surface waters cooled during AMOC slowdown). In contrast, northern marine sectors (Barents Sea/BeIS) were strongly influenced by gradual subsurface warming of eastern Nordic Seas waters, promoting submarine melt and grounding-line retreat despite cold atmospheres; decreasing precipitation after the LGM further reduced SMB. - Ocean–ice feedbacks and AMOC: The main EIS meltwater pulse (16–14 ka) coincided with KBSIS collapse and HS1 stratification (warm deep/cold surface waters). Proxy SSTs suggest a brief AMOC strengthening near 15 ka; modelling implies EIS-sourced meltwater alone likely had limited, short-lived impacts on North Atlantic overturning relative to external (Laurentide, Greenland) inputs.

The reconstruction clarifies how marine-based sectors of the EIS responded nonlinearly and asynchronously to regional forcing. Early North Sea/NCIS retreat, occurring while near-surface North Atlantic/Nordic Seas were cooling, implicates atmospheric-driven SMB increases in melt, compounded by bed geometry and subglacial hydrology (e.g., widespread tunnel valleys). Conversely, Barents Sea retreat was delayed and closely tied to progressive subsurface warming and grounding-line instability along deep troughs (e.g., BeIS), with reduced precipitation after the LGM further lowering SMB. The modelled ice-volume evolution tracks summer insolation increases but diverges between 16–14 ka due to rapid KBSIS collapse, underscoring the dominant role of marine sectors in abrupt ice loss. Separation of ice-sheet saddles in the south (~19–18 ka) and north (~16–15 ka) marks tipping points that accelerated deglaciation and meltwater fluxes. Despite substantial meltwater export, proxy and modelling results suggest limited long-lived suppression of AMOC directly attributable to EIS inputs, highlighting the influence of broader North Atlantic forcing. These findings address the central question by distinguishing where atmospheric SMB versus ocean thermal forcing dominated retreat, providing analogues for contemporary marine-based ice-sheet sensitivity and feedbacks.

Marine sectors governed the abrupt, nonlinear collapse of the last EIS. The North Sea deglaciation occurred relatively early and was primarily driven by atmospheric SMB and dynamic feedbacks, while delayed Barents Sea retreat was driven by subsurface ocean warming and reduced precipitation. Peak EIS volume (~20 m SLE at 19 ka) gave way to rapid losses culminating ~15 ka, when KBSIS collapse dominated marine-based contributions to sea level. The work refines regional deglaciation chronologies (e.g., later deglaciation in eastern Barents/Novaya Zemlya than previously inferred) and quantifies evolving marine margin exposure. These reconstructions provide a benchmark for fully coupled ice–ocean–atmosphere models that capture ice–climate feedbacks. Future work should prioritize improved chronological control in data-poor marine regions (eastern Barents/Kara Seas), better constraints on spatial–temporal marine reservoir effects (ΔR), and transient, physics-based modelling that assimilates multi-proxy oceanographic and glaciological data.

- Spatial data gaps and uneven coverage: Fewer dates and mapped landforms in the eastern Barents and Kara Seas and in parts of the southeastern North Sea increase regional uncertainty. - Chronological uncertainties: Marine radiocarbon dating is affected by variable ΔR; a conservative ΔR=0 approach was adopted but may bias absolute ages. Some regions (e.g., Novaya Zemlya) have sparse or conflicting age control. - Proxy and method constraints: SST reconstructions based on planktonic foraminifera have reduced sensitivity below ~4 °C, do not resolve seasonal/sea-ice complexities, and rely on transfer functions with RMSEP ~1.93 °C. - Modelling simplifications: Steady-state, perfectly plastic ice reconstructions approximate transient ice dynamics; basal shear stress parameterization and limited GIA iterations introduce volume uncertainties (partly bracketed by ±20% stress perturbations). Ocean/atmosphere forcings are not explicitly coupled. - Exclusion of OSL and some datasets: OSL dates from the southern North Sea were not incorporated due to large uncertainties, potentially omitting constraints where radiocarbon control is sparse.