Redrawing the Iceland–Scotland Overflow Water pathways in the North Atlantic

The study investigates how Iceland–Scotland Overflow Water (ISOW), a key component of the lower limb of the Atlantic Meridional Overturning Circulation (AMOC), spreads from its entry point into the North Atlantic. Historically, ISOW has been depicted as a counter-clockwise deep boundary current around the subpolar gyre, turning north along the western flank of the Reykjanes Ridge after passing the Charlie-Gibbs Fracture Zone (CGFZ). However, accumulating evidence suggests more complex pathways. Understanding these routes is crucial for interpreting the AMOC’s geographic structure and the transport of high-latitude climate signals (e.g., anthropogenic carbon) into the deep ocean. The research questions addressed are: What are the dominant ISOW pathways downstream of the CGFZ, and how does the North Atlantic Current (NAC) and its deep-reaching eddies/meanders influence the partitioning among these pathways?

Past observations and models have revealed complexity beyond a single boundary-following ISOW path. East of the Reykjanes Ridge, ISOW’s shallower component can transit westward through deep gaps before the latitude of the CGFZ (e.g., Lankhorst & Zenk 2006; Xu et al. 2010; Zou et al. 2017; Petit et al. 2018). Some ISOW bypasses the CGFZ westward route and continues south along the eastern flank of the Mid-Atlantic Ridge (MAR) into the Western European Basin (Fleischmann et al. 2001; van Aken 2000; Xu et al. 2018). West of the CGFZ, hydrography shows a higher-salinity tongue extending W–WNW suggesting alternative interior spreading (McCartney 1992; Stramma et al. 2004; Sarafanov et al. 2012; Daniault et al. 2016). Modeling highlighted a deep mean W–WNW current from CGFZ toward southern Greenland carrying much ISOW (Xu et al. 2010). Deep-Argo evidence includes one float directly reaching the Flemish Cap from the CGFZ (Racapé et al. 2019). Within the CGFZ, moorings and LADCP show ISOW transport reversals linked to the position of the eastward-flowing NAC (Bower & Furey 2017; Schott et al. 1999), but downstream NAC impacts remained unclear. These studies motivated a re-examination of ISOW pathways using new Lagrangian data and eddy-resolving models.

Observations: Twenty-one neutrally buoyant RAFOS floats, ballasted for the ISOW layer (depths 1800–2800 m; densities ≥27.80 kg m⁻³; salinity >34.94; in situ T ~2.4–3.4 °C), from the OSNAP program were analyzed. Nine floats were released in the northern valley of the CGFZ on June 28, 2014; twelve were deployed along the eastern flank of the Reykjanes Ridge and drifted through/near the CGFZ during 2014–2018. Float lifetimes ranged 78–730 days (mean 498 days). Underwater tracking used moored sound sources (2014–2018), with daily records of acoustic travel time, pressure, and temperature; trajectories were reconstructed upon surfacing (nominally 2-year missions). Mooring-derived mean velocity and salinity sections across CGFZ (2010–2012; from Bower & Furey 2017) provided context. Surface absolute dynamic topography (ADT) from AVISO at 10-day intervals was used to assess NAC eddy/meander interactions with float segments (10-day pieces). Modeling: The eddy-resolving FLAME model (1/12°) covering 18°S–70°N with 45 vertical levels was used (1990–2004; 3-day output) forced by NCEP/NCAR reanalysis anomalies after climatological spin-up. ISOW at CGFZ (35.3°W, 52–53°N) was defined using density and salinity: model threshold σθ ≥27.82 kg m⁻³; salinity threshold 34.95 (≤1995) and 34.96 (≥1996). Monthly releases of 3593 synthetic floats in the northern valley of the CGFZ were integrated forward for 2 years using 3-D velocity fields; for extended analyses, trajectories were integrated to 10 years (with recycled velocity fields to extend beyond 2004). Probability distributions were computed on 0.25°×0.25° grids by counting float passes and normalizing. Mean age fields and first-occurrence statistics at 45°N (subpolar–subtropical boundary proxy) were calculated along longitude bins (1°). Eddy kinetic energy (EKE) west of the CGFZ (box [38°W–34°W, 51°N–54°N]) was computed at 10 m and 2500 m; surface and deep EKE time series correlations were assessed. Floats were classified as launched during high-EKE periods if 12-month mean EKE > long-term mean +1σ (0.022 m² s⁻²) and low-EKE if < mean −1σ (0.015 m² s⁻²); 509 and 752 floats met high- and low-EKE criteria, respectively. Pathway quantification categorized simulated floats into five groups based on spreading and surface end locations: west–northwestward interior (western subpolar gyre, west of Reykjanes Ridge, north of 52°N); southward along western MAR (final south of 52°N, west of MAR); eastward into eastern subpolar gyre; local recirculation near CGFZ ([38°W–32°W, 52°N–54°N]); and continuous deep boundary current following the western flank of the Reykjanes Ridge between the 2000–3000 m isobaths. Uncertainties were estimated by repeatedly sampling 21-float subsets (10,000 trials) to mirror the observed sample size. HYCOM eddy-resolving simulations were also consulted for qualitative comparison (supporting material), showing consistent large-scale spreading patterns.

• Observed RAFOS floats (n=21) downstream of the CGFZ split primarily between two unexpected pathways: eight floats exhibited quasi-unidirectional west–northwestward motion reaching the central Labrador and Irminger Seas (three reached the central Labrador Sea); eight flowed southward along the western flank of the Mid-Atlantic Ridge (two to ~47°N); four recirculated eastward toward the Iceland Basin (shallower ~2000 m; likely influenced by the deep-reaching NAC); only one had potential to turn north along the Reykjanes Ridge boundary current but ceased near 53.3°N due to topography.
• Float temperatures at CGFZ crossings were 2.9–3.3 °C, consistent with ISOW (σθ >27.80 kg m⁻³; S >34.94; T 2.4–3.4 °C).
• AVISO ADT maps showed floats encountering NAC meanders/eddies: trapping at meander crests followed by anticyclonic deflection southward, demonstrating deep-reaching NAC influence on ISOW pathways.
• Simulated FLAME floats (n=3593; 2-year integrations) reproduced three dominant branches and local trapping: 59 ± 9% west–northwestward interior pathway (preferentially initiated below 2600 m); 19 ± 8% southward along western MAR (no strong launch-depth preference); 6 ± 5% eastward (primarily from shallower ISOW); 9 ± 6% local recirculation near CGFZ; only 5 ± 4% continuously followed the deep boundary current along the western flank of the Reykjanes Ridge.
• Relative to observed percentages (8/21 ≈ 38% for both west–northwestward and southward), FLAME overestimates the west–northwestward and underestimates the southward branch, partly due to limited observed sample size and stronger eddy/meandering activity during 2014–2016 (observational period).
• NAC eddy/meander impact: Surface (10 m) and deep (2500 m) EKE time series over the region were significantly correlated (r = 0.60 monthly; r = 0.84 annual), indicating deep-reaching influence. During high-EKE periods, the west–northwestward branch decreased to 48 ± 11% and the southward branch increased to 31 ± 10%. During low-EKE periods, westward increased to 68 ± 10% and southward decreased to 12 ± 7%. Eastward, local, and boundary-following fractions changed by <3%.
• Over 10 years, simulated ISOW trajectories spread across the subpolar North Atlantic with some reaching ~30°N. Export across 45°N was concentrated west of the MAR: a major branch along the Deep Western Boundary Current (west of ~44°W) and an interior branch between 44°W and 35°W, with mean ages ~7.1 ± 2.0 years west of 40°W. A distinct export branch along the western flank of the MAR (35°W–28°W), fed by the southward pathway, had a shorter mean transit time to 45°N of 5.9 ± 1.9 years, implying a relatively faster export route to subtropical latitudes.
• FLAME’s mean ISOW transport through CGFZ was −0.9 Sv (σ ≈ 0.4 Sv annually), smaller than observed estimates (−1.7 to −2.4 Sv), though mesoscale variability complicates long-term means.
• Collectively, results necessitate revising the historical depiction: most ISOW does not follow a single boundary current west of the CGFZ but partitions between interior west–northwestward and southward MAR-flank pathways, with partitioning modulated by NAC eddy/meander activity.

The findings substantially revise the canonical view of ISOW spreading. Rather than predominantly turning northward along the western flank of the Reykjanes Ridge, most ISOW leaving the CGFZ either follows interior west–northwestward pathways into the central Irminger and Labrador Seas or proceeds southward along the western flank of the Mid-Atlantic Ridge. The observed and modeled southward branch represents a newly documented route that can expedite export to subtropical latitudes. The influence of NAC eddies and meanders, which reach ISOW depths, modulates the partitioning: enhanced eddy/meander activity traps ISOW near meander crests and diverts it southward; reduced activity favors continued westward spreading. This mechanism mirrors previously observed transport variability within the CGFZ and establishes a clear connection between upper-ocean mesoscale variability and deep water mass pathways. The presence of ISOW signatures along the boundary west of the Reykjanes Ridge can be reconciled by leakage of ISOW through Reykjanes Ridge gaps north of the CGFZ (e.g., Bight Fracture Zone ~57°N), feeding boundary currents independently of the CGFZ outflow. These revised pathways, together with interior routes identified for other deep waters (e.g., Labrador Sea Water), imply that export of climate signals (heat, carbon, tracers) from high latitudes is more diffusive and spatially distributed than previously assumed, with implications for observing system design and for interpreting deep ocean inventories. Analogous dynamics in the South Atlantic (e.g., Agulhas rings influencing NADW) reinforce the broader significance of mesoscale–deep circulation coupling.

This study re-draws ISOW pathways downstream of the CGFZ, demonstrating two dominant routes—an interior west–northwestward branch into the central Irminger and Labrador Seas and a newly identified southward branch along the western flank of the Mid-Atlantic Ridge—while a continuous boundary-following path along the Reykjanes Ridge is relatively minor. The partitioning between these branches is influenced by deep-reaching NAC eddies and meanders; higher eddy activity enhances southward diversion and suppresses westward spreading. The southward MAR-flank pathway offers a comparatively faster export route to subtropical latitudes. These insights call for updates to schematic depictions of North Atlantic deep circulation and inform strategies for monitoring AMOC variability and tracer transports. Future work should expand Lagrangian observations (including Deep-Argo), extend and overlap observational and model periods, improve model representation of overflow transport through CGFZ and ridge gaps, and quantify interannual–decadal variability in pathway partitioning and its climate impacts.

• Observational sample size is small (21 RAFOS floats), limiting statistical robustness of pathway fractions.
• Temporal mismatch between observations (2014–2018) and FLAME simulations (1990–2004) may introduce differences in NAC state and mesoscale activity.
• FLAME underestimates mean ISOW transport through CGFZ (−0.9 Sv vs observed −1.7 to −2.4 Sv), potentially biasing pathway partitioning.
• Mooring-based mean sections (2010–2012) do not overlap with float launch times; water mass properties and NAC configuration may differ.
• Model thresholds for ISOW (e.g., σθ = 27.82 kg m⁻³; variable salinity thresholds) and velocity field recycling to extend trajectories may affect realism.
• HYCOM results, while broadly consistent, show detailed differences, underscoring model dependency of some pathway features.
• ADT/EKE-based inference assumes surface EKE is a reliable proxy for deep eddy activity (supported by correlations but still an approximation).