A shift in the ocean circulation has warmed the subpolar North Atlantic Ocean since 2016

The Subpolar North Atlantic (SPNA) plays a fundamental role in the global ocean circulation and the Earth’s climate system. It is a key region for surface waters to mix and sink, forming part of the Meridional Overturning Circulation (MOC) that transports heat and carbon globally. It is also the main gateway for warm, salty Atlantic waters to reach the Nordic Seas and Arctic Ocean, influencing mass and heat budgets in these vulnerable areas. Additionally, the SPNA likely affects atmospheric regimes (e.g., storm tracks), continental heat waves, and may trigger multicentennial climate variability through coupled ocean–atmosphere–ice processes in the North Atlantic. The climatic imprint of the SPNA is tightly linked to upper-ocean temperature, which responds to intermittent atmospheric forcing and decadal-scale circulation changes and associated heat transport. A circulation-driven warming-to-cooling reversal occurred in the mid-2000s, with impacts seen in East Atlantic sea level. An observationally based reconstruction of meridional heat transport at 45°N predicted a new advection-driven reversal beginning in 2016, with temperatures in the early 2020s approaching their 2005–2006 maximum. This study employs new observational and statistical approaches to document the recent and ongoing warming and demonstrates that an increased poleward transport of warm subtropical waters is the dominant mechanism.

Background literature indicates that SPNA heat content variability is central to the Atlantic MOC and broader climate impacts. Prior work has identified: (1) a mid-2000s SPNA warming-to-cooling reversal linked to decadal ocean circulation and heat transport changes; (2) basin-scale connections to sea level variability along European shelves; (3) the role of air–sea heat flux anomalies in episodic events such as the 2014–2015 cold anomaly; and (4) observational reconstructions predicting an advection-driven warming phase commencing around 2016. Studies of gyre dynamics and North Atlantic Current (NAC) branch variability highlight contractions/expansions of the subpolar and subtropical gyres that modulate the relative contributions of warm subtropical versus cold subpolar waters feeding the eastern SPNA. Lagrangian and reanalysis-based assessments emphasize the importance of horizontal advection and the coherence of surface geostrophic flows with interior pathways for driving decadal temperature changes.

Data and regions: The study focuses on the eastern SPNA (10°W–35°W, 40°N–65°N). It uses ISAS-15 gridded temperature/salinity fields (2002–2019; Argo-based with additional profile types), DUACS DT2018 satellite altimetry for sea surface height (SSH) and geostrophic velocities (¼° grid, 1993–2019), NCEP/NCAR reanalysis heat and momentum fluxes (2.5° grid, 1948–2019), the ECCOv4r4 ocean state estimate (1992–2017) for 3D circulation, and the NAO index. Diagnostics: Upper-ocean temperature anomalies were computed over depth ranges including 0–100 m and integrated to 2000 m for ocean heat content (OHC). Mean kinetic energy (MKE) was derived from geostrophic surface velocities. Spatial contrasts were analyzed for 2017–2019 minus 2014–2016. Passive tracer advection–diffusion model: A 2D advection–diffusion equation for passive tracer concentration C was solved on a subset of the AVISO grid with forward time-stepping and upstream differencing for advection, and an explicit 2-level scheme (time step ~1500 s). Tracer sources were set to unity in two boxes: western subtropical (60°W–75°W, 35°N–45°N) and western subpolar (50°W–60°W, 55°N–65°N), representing the principal upstream regions feeding the NAC. Mesoscale eddy diffusivities were imposed from observation-based estimates. Surface Ekman velocities were not included to focus on the geostrophic component representing the first baroclinic mode and main thermocline motions. Experiments were run over overlapping N-year means of the altimetry-derived velocity fields with N=2, 3, and 4 years, forming an ensemble to represent typical surface advective timescales. The resulting final concentrations CSTG and CSPG were used to compute the proportion of subtropical-origin water PSTG=CSTG/(CSTG+CSPG). A companion 3D tracer experiment with ECCOv4r4 added vertical advection and set unit source concentrations over the full water column in the same boxes; tracer fields were averaged above the main pycnocline (σθ=27.6 kg m−3). Advection-driven temperature reconstruction: To quantify drivers of eastern SPNA temperature variability, advection-driven temperature anomalies were decomposed as θ′ADV = P′STG(θSTG − θSPG) + P′STG θ′STG + P′SPG θ′SPG, associating changes in the relative proportion of source waters and their initial temperatures (with appropriate N-year lags) to reconstruct observed anomalies in the upper ocean (above σθ=27.6). Statistical clustering (machine learning): A profile classification model (PCM) using Gaussian Mixture Models (K=2) was trained on the 2002–2019 time-mean ISAS temperature and salinity profiles deeper than 2000 m. Variables were standardized by depth level and reduced via PCA before EM fitting. The trained PCM classified each year’s profiles into two classes representing stratified subtropical (KSTG) and more homogeneous subpolar (KSPG) water masses, providing spatial probabilities PK(x,y,t). The OHC of the 0–700 m eastern SPNA was decomposed into class contributions OHC_K(t) = ∬ PK(x,y,t)[∫0 to 700m ρw Cp θ dz] dxdy, enabling attribution of OHC variability to changing class proportions and properties. Validation and uncertainty: Ensemble spread across N=2–4 years quantified uncertainty in tracer-derived indices. The correlation between temperature anomalies and SSH validated thermosteric control and supported advective mechanisms. ECCOv4r4 experiments assessed vertical coherence between surface geostrophic pathways and interior spreading, acknowledging known low-resolution model limitations in the region.

- A sharp, surface-intensified and large-scale warming emerged in the eastern SPNA beginning in 2016 after a decade-long cooling since 2006. - The warming pattern (2017–2019 minus 2014–2016) shows upper 0–100 m temperature increases up to ~0.6 °C across the Iceland Basin, Iberian plain, and eastern Newfoundland Basin, with consistent positive SSH anomalies over most of the eastern SPNA. - Strong agreement between upper-ocean temperature anomalies and SSH (r ≈ 0.82 with 0–100 m temperature time series) indicates a thermosteric, advective origin. - Regional 0–2000 m OHC increased by ~8.5 × 10^21 J from April 2016 to December 2019, with ~80% of the gain above 700 m. Air–sea heat flux anomalies account for roughly one-quarter of this increase, implying dominant ocean heat transport convergence. - MKE anomalies indicate an intensification of the subtropical portion of the NAC south of 45°N and west of 30°W, with downstream strengthening in the northern NAC branch and weakening in the southern branch, consistent with a gyre expansion/contraction seesaw. - Passive tracer experiments show that increases in the subtropical-origin proportion PSTG in the eastern SPNA closely track temperature anomalies, capturing the 2006 and 2016 trend reversals. The empirical sensitivity is estimated at ~0.05 °C per 1% change in PSTG. - Spatial PSTG increases align with observed warming, and a northward shift of the 50% PSTG isoline since 2015 indicates enhanced penetration of subtropical waters into the Iceland Basin. - Decomposition of advection-driven temperature anomalies shows that variability in the relative proportion of source waters (P′STG term) dominates over advection of remote temperature anomalies in explaining observed trends. - ECCOv4r4-based 3D tracer experiments support vertical coherence of subtropical tracer penetration below the Ekman layer to the main pycnocline and confirm consistency between surface geostrophic pathways and interior spreading. - Machine-learning PCM (2 classes) reproduces the spatial dichotomy between subtropical (KSTG) and subpolar (KSPG) waters and shows that since the mid-2010s, OHC and areal proportion contributions of KSTG increased while those of KSPG decreased, accounting for the 2016 warming reversal. - The circulation shift is consistent with a strengthening of MOC heat transport across 45°N and likely pending recovery at 26°N, implying broader AMOC impacts.

The findings demonstrate that the 2016 cooling-to-warming reversal in the eastern SPNA is primarily circulation-driven: enhanced northeastward advection of warm, saline subtropical waters by the NAC increased the subtropical-origin fraction of waters entering the SPNA, raising regional temperatures and OHC. The strong correlation between upper-ocean temperature, SSH, and PSTG, along with the MKE seesaw of NAC branches, confirms an advective mechanism rather than dominant local surface heating. This circulation shift is likely to have downstream and vertical consequences: embedded within the cyclonic subpolar gyre, the warming is expected to propagate westward into the Irminger and Labrador Seas and to greater depths through boundary processes, potentially altering water mass transformation, convection depths, and near-future AMOC strength and property transports. Given the SPNA’s long memory and similarities to the mid-1990s circulation regime change, the advective warming may persist for several years. At the climate scale, SPNA heat content variability influences the Atlantic Multidecadal Variability (AMV). While some have anticipated a transition to a negative AMV phase due to post-2006 SPNA cooling, the ongoing advective warming injects uncertainty into such projections. Whether this reversal will dampen the AMV downturn remains unresolved, but the mechanism identified here provides a physical basis for updated decadal outlooks and emphasizes the need to integrate ocean circulation indicators into climate indices and predictions.

The study provides observational evidence that a shift in ocean circulation since 2016 has triggered a surface-intensified, basin-scale warming of the eastern SPNA by enhancing the penetration of warm subtropical waters via the NAC. Two independent, observation-based approaches—passive tracer advection/diffusion driven by altimetry-derived geostrophic currents and machine-learning clustering of T–S profiles—both attribute the 2016 trend reversal primarily to changes in the relative proportion of subtropical versus subpolar source waters, with advection-driven processes dominating over local air–sea heat fluxes. Main contributions include: (1) documentation of the 2016 warming onset and quantification of associated OHC increases; (2) identification and mechanistic attribution to circulation shifts and NAC branch reorganization; (3) demonstration of vertical coherence of advective pathways and confirmation through a statistical water-mass framework. Future research directions include: (a) resolving the partitioning and causality between atmospheric forcing (e.g., NAO) and oceanic circulation adjustments leading to the shift; (b) assessing the westward and downward propagation of the warming signal and its impacts on subpolar convection and MOC variability; (c) refining predictive indicators for SPNA heat content and AMV by integrating dynamic circulation metrics; and (d) leveraging higher-resolution ocean models and additional observations to better resolve key features (e.g., Northwest Corner) and mesoscale processes.

- The surface tracer experiments omit Ekman velocities, potentially over-permitting cross-gyre surface exchange of subtropical-origin tracers; results focus on geostrophic, first baroclinic mode dynamics. - The eddy diffusivity field is time-invariant, so diffusive variability reflects only changes in advective pathways; parameterized mixing contributes less than advection but adds uncertainty to PSTG magnitude (~4% reduction on average in the box). - ECCOv4r4 is a coarse-resolution state estimate with known regional deficiencies (e.g., inadequate representation of the Northwest Corner), limiting fidelity of some interior pathways despite overall coherence with surface-geostrophic results. - The choice of advective timescales (N=2–4 years) is based on tracer-based estimates and literature; ensemble spread captures uncertainty but exact lags remain approximate. - ISAS data for 2017–2019 were in near-real-time mode rather than delayed-mode QC; large-scale anomalies are robust but small-scale features may be affected. - The PCM uses K=2 clusters for interpretability; other cluster counts can capture additional structure but were not the focus here. - The study period ends in 2019 (altimetry) and 2017 (ECCO), so inferences about persistence into the 2020s rely on mechanistic reasoning and ancillary indicators rather than direct observation beyond those dates.