The downward spiralling nature of the North Atlantic Subtropical Gyre

The AMOC, a key regulator of Earth’s climate via meridional heat transport, has likely weakened during the 20th century and is projected to further decline. While the Gulf Stream and the North Atlantic Subtropical Gyre (hereafter the Gyre) are integral to the northward branch of the AMOC, the extent to which the Gyre circulation conditions water to join the upper limb of the AMOC has remained unclear. Prior work suggests the transition between subtropical and subpolar gyres occurs mainly at depth and that North Atlantic Subtropical Mode Water (NASTMW) may link these systems. This study addresses: To what extent does recirculation within the Gyre densify waters and determine their pathways to the northern AMOC? The purpose is to quantify the role of repeated subtropical gyre circuits in setting the density, depth, and timing of waters that ultimately flow northward, thereby informing AMOC strength and variability and its climate impacts.

Multiple studies document a weakening AMOC in recent decades and over the last millennium and project further decline under climate change. The Gulf Stream and subtropical gyre are recognized as part of the northward AMOC limb. Observations and Lagrangian studies indicate that inter-gyre exchanges often occur at depth, with re-circulation within the Subtropical Gyre contributing to connectivity with the Subpolar Gyre. NASTMW forms on the warm flank of the Gulf Stream through winter convection and is implicated as a conduit linking the gyres at depth. Ocean heat transport depends strongly on wind-driven circulation, and warming is expected to reduce subtropical water density via warming and salinification, potentially altering gyre circulation and AMOC pathways. Coarse-resolution models typically capture large-scale means but parameterize eddies; higher resolution can alter Gulf Stream dynamics and mixing, with implications for Lagrangian pathways and heat uptake.

The study uses Lagrangian trajectory simulations with the mass-conserving TRACMASS model driven by monthly mean mass flux, temperature, and salinity fields from the EC-Earth Earth System Model. Two configurations were analyzed: (1) Low-resolution (LR) EC‑Earth‑Veg v3.3.1 with NEMO 3.6 ocean at 1° horizontal resolution and 75 vertical levels (not eddy-permitting), atmospheric IFS coupled to LPJ-GUESS; period 1850–2014. Vertical mixing uses a TKE scheme; eddy tracer fluxes parameterized (Gent–McWilliams). (2) High-resolution (HR) EC‑Earth3P‑HR (HighResMIP) with 1/4° ocean and 75 levels, monthly means 1950–2014. Trajectories were initialized at all depths and longitudes along 17°S where velocity is northward, over one year to include seasonality. In the LR case, 278,361 trajectories represented 44.7 Sv; termination criteria: reaching 58°N (used in analysis) or returning to 17°S. Of these, 11 Sv (68,064 trajectories) reached 58°N; 3.3 Sv went directly north without gyre spiralling; the remainder executed one or more subtropical gyre circuits before proceeding north. Integration was extended to 1000 years by looping the 1850–2014 fields; only 0.1 Sv remained within the domain at the end. Trajectories reaching the sea surface and evaporating were excluded to maintain mass conservation between boundaries; 0.9 Sv evaporated. For the HR run, 10 Sv reached 58°N across 83,337 trajectories. Lagrangian meridional overturning and barotropic stream functions were computed from conserved volume transports to distinguish direct northward flow from gyre-spiralled waters, in both depth and potential density (σθ/σo) coordinates. A mean gyre trajectory (the “downward spiral”) was constructed by segmenting the gyre using a 12-sector colour-clock centered in the gyre; for each sector and for each circuit count (1st, 2nd, …), mean latitude, longitude, depth, temperature, salinity, and density were computed over all transits, enabling a representative 3D pathway in geographic, T–S, and density–space. Lagrangian divergences of heat, salt, and density were computed gridwise from trajectory in/out properties and transports, then converted to fluxes (W m⁻², kg s⁻¹ m⁻²) by dividing by grid-cell area and multiplying by seawater density and heat capacity where appropriate. Divergences were accumulated separately within the mixed layer (defined by a density criterion from NEMO) and below, to attribute changes to air–sea interaction versus interior mixing. Diagnostics included mapping divergence fields and streamlines for the subset of waters that recirculated in the subtropical gyre at least once.

• Approximately 70% of the water flowing northward as part of the AMOC first recirculates within the North Atlantic Subtropical Gyre before continuing north; 30% proceeds directly northward.
• Waters execute between 1 and at least 7 identifiable circuits (with some up to 14), progressively cooling, slightly freshening, and becoming denser, thereby spiralling to greater depths. Entry densities in the Gulf Stream upper 100 m are σo ≈ 24–25 kg m⁻³ at ~5°N; exit densities at ~55°N exceed 27 kg m⁻³ at 250–1000 m depth.
• Table 1 metrics (LR case) per circuit for waters reaching north: Circuit 1 carries 7.7 Sv (70% of northward volume), loses 134 TW of heat, gains 0.15×10⁶ kg s⁻¹ of salt, increases mean density by 1.05 kg m⁻³, deepens by 182 m, taking ~22 years. Circuit 2: 7.4 Sv (67%), −41 TW heat, −0.39×10⁶ kg s⁻¹ salt, +0.18 kg m⁻³, +48 m, ~38 years. Circuit 3: 5.2 Sv (47%), −24 TW, −0.34×10⁶ kg s⁻¹, +0.12 kg m⁻³, +35 m, ~46 years. Circuit 4: 3.7 Sv (34%), −16 TW, −0.26×10⁶ kg s⁻¹, +0.08 kg m⁻³, +29 m, ~57 years. Circuit 5: 2.6 Sv (24%), −10 TW, −0.18×10⁶ kg s⁻¹, +0.05 kg m⁻³, +22 m, ~68 years. Circuit 6: 1.8 Sv (16%), −7 TW, −0.12×10⁶ kg s⁻¹, +0.03 kg m⁻³, +15 m, ~77 years. Circuit 7: 1.3 Sv (12%), −4 TW, −0.08×10⁶ kg s⁻¹, +0.01 kg m⁻³, +8 m, ~86 years.
• Net budgets for all spiralling waters: accumulated heat loss ≈ 0.24 PW, with 57% in the first circuit and 43% over subsequent circuits; accumulated salt loss (freshening) ≈ 1.22×10⁶ kg s⁻¹ (first circuit gains salt; subsequent circuits collectively lose 1.37×10⁶ kg s⁻¹).
• Mixed-layer Lagrangian divergences show strong cooling and salinity increase along the Gulf Stream and northern gyre flank, yielding large density increases and consistent with NASTMW formation near the Gulf Stream separation/recirculation.
• Below the mixed layer, three key regions: (1) western Gulf Stream/northern flank: strong cooling and freshening, implying interior mixing along tilted isopycnals with strong T/S gradients; (2) mid-to-eastern gyre: heat and salt gains with density decrease, likely from Mediterranean Water mixing plus regional evaporation; (3) southern gyre: cooling and freshening that are density-compensated, indicating isopycnal mixing.
• The first circuit exhibits the largest depth and density increases and is uniquely associated with salinification; deeper subsequent circuits show reduced T/S excursions, indicating diminishing surface-forcing influence and increasing dominance of interior mixing.
• High-resolution (1/4°) simulations reproduce the spiral structure and budgets, with slightly higher salinities and wider circuits; circuit times are roughly half those of the LR case due to a narrower, faster Gulf Stream and eddy effects, but transport fractions and T/S–density–depth changes per circuit are robust.

Findings demonstrate that the subtropical gyre recirculation is essential for conditioning waters to join the northern upper branch of the AMOC: repeated loops promote cumulative cooling and freshening, increasing density and depth until waters can proceed northward. The gyre’s role is twofold: (a) providing regions of strong mixed-layer heat loss (notably along the Gulf Stream) that drive densification, and (b) returning waters back to the Gulf Stream where further wintertime cooling can occur. The first circuit is primarily influenced by air–sea fluxes and vigorous near-surface mixing; deeper circuits reflect interior mixing along sloping isopycnals and lateral tracer exchange, with regional contrasts driven by Mediterranean Water influence and isopycnal mixing in the southern gyre. The spiral aligns with typical NASTMW T–S–density properties and depths, suggesting a linkage between NASTMW formation and the downward spiralling pathway. Robustness across LR and HR models indicates that large-scale dynamics governing the spiral are captured even in coarse-resolution frameworks. Climate change could modify these pathways: enhanced evaporation may increase salinity in the eastern gyre, while subpolar freshening from ice melt could counterbalance via inter-gyre mixing. Similar spiralling patterns may exist in other gyres (e.g., North Pacific, Weddell, Southern subtropical gyres), implying broader relevance for global heat and salt redistribution. Improved quantification of vertical and lateral diffusion and seasonal dependencies (e.g., wintertime cooling) is needed to refine understanding of gyre–AMOC coupling and future variability.

The study reveals a previously underappreciated mechanism: about 70% of northward-flowing AMOC waters first execute one or more circuits within the North Atlantic Subtropical Gyre, undergoing cumulative cooling and freshening that increase density and depth, enabling continuation to the northern AMOC. The gyre thus exerts strong control on AMOC pathways, strength, and northward heat transport, with significant cumulative heat loss occurring over multiple circuits. Results are robust across coarse and higher-resolution models, though circuit durations differ due to Gulf Stream structure and eddies. The work underscores the need to incorporate gyre recirculation effects in assessments of future AMOC changes. Future research should: (i) quantify mixing processes and heat/salt budgets (vertical versus lateral diffusion) along the spiral, (ii) resolve seasonal effects on densification, (iii) test for analogous spirals in other gyres, and (iv) assess how changing stratification, evaporation, and subpolar freshening alter the spiral and AMOC.

The analysis relies on model-derived monthly mean fields (EC-Earth) and parameterized eddy effects in the 1° LR configuration; even the 1/4° HR run is only eddy-permitting, and spurious mixing may influence heat uptake. Trajectory integrations loop historical forcing to 1000 years, which may not capture transient variability. The simulated northward transport (11 Sv in LR) is lower than observed transports at 26°N–41°N because only a subset of source waters was traced (excluding some northern-origin waters and those not reaching 58°N). Surface freshwater fluxes are not explicitly a source/sink in the Lagrangian mass budget (evaporating trajectories excluded), which constrains streamfunction interpretation. Inferences about the dominance of interior mixing in deeper circuits are based on divergence patterns and require dedicated process studies. Seasonal effects are suggested but not resolved beyond monthly means.