Decadal changes in Atlantic overturning due to the excessive 1990s Labrador Sea convection

The Atlantic Meridional Overturning Circulation (AMOC) is a complex three-dimensional current system transporting warm, buoyant waters northward and colder, denser waters southward. Understanding the mechanisms driving AMOC variability, particularly on decadal timescales and beyond, is crucial due to its significant climate impacts. Decadal AMOC changes in the subtropical North Atlantic have been linked to density anomalies originating in the subpolar gyre and propagating southward. Modeling studies suggest that meridionally coherent, low-frequency AMOC anomalies arise in response to low-frequency surface buoyancy forcing in the subpolar region. While model simulations show varied results regarding multi-decadal trends, they generally agree on AMOC strengthening during the 1980s and 1990s, followed by a decline in subsequent decades. These decadal changes align with continuous observations in the subtropical North Atlantic since 2004 and various reconstructions for past decades. Previous research focused on Labrador Sea deep winter convection as a primary conduit for surface buoyancy flux anomalies into the deep ocean. The impact of positive and negative North Atlantic Oscillation (NAO) indices on subsurface density changes and their effect on the AMOC were emphasized. However, the causal link between Labrador Sea convection and decadal AMOC variability is debated due to inconsistencies with observations. Overturning circulation from hydrographic sections across the Labrador Sea appeared weak even during the intense 1990s convection; multi-year records showed no convection intensity changes in the Labrador Sea's deep western boundary current; and the observed AMOC weakening in the subtropical North Atlantic since 2008 occurred in the lower North Atlantic Deep Water, with no detectable decadal change in Labrador Sea Water (LSW) transport. Furthermore, initial results from the Overturning in the Subpolar North Atlantic Programme (OSNAP) (2014–2018) indicated that most downwelling occurs in the Irminger and Iceland basins, with minimal contribution from the Labrador Sea. This study aims to reassess the role of Labrador Sea buoyancy fluxes in interannual to interdecadal AMOC changes and their geographical distribution, focusing on the period of enhanced winter heat loss associated with the prolonged positive NAO phase from the late 1980s to the mid-1990s, which led to the formation of exceptionally dense LSW.

Numerous studies have investigated the mechanisms driving AMOC variability, focusing on the role of deep convection in the Labrador Sea and the influence of the North Atlantic Oscillation (NAO). However, observational data have presented challenges to the established understanding. While some studies highlighted a strong correlation between Labrador Sea convection and AMOC strength, others found inconsistencies between observed changes in convection intensity and AMOC variability. The lack of a clear observational link between Labrador Sea convection and decadal AMOC changes motivated this study to re-examine this relationship using high-resolution modeling techniques, aiming to resolve the discrepancies between model simulations and observational data. The paper reviews previous studies linking decadal AMOC changes to density anomalies originating in the subpolar gyre and propagating southward. The debate surrounding the role of Labrador Sea convection is presented, referencing prior work that suggests alternative mechanisms might be primarily responsible for AMOC variability. The inconsistency between model outputs and observations related to the impact of Labrador Sea convection on the AMOC provided the crucial impetus for the current investigation.

This research utilized the global ocean-ice model VIKING20X, a high-resolution configuration of the NEMO system with a refined grid spacing of 0.05° over the Atlantic Ocean. This high resolution enabled the model to realistically represent energetic mesoscale flow patterns and eddy activity in the Labrador Sea, crucial for simulating the extent and intensity of deep winter convection and the complex structure of the deep boundary current system. The model's finer representation of steep topographic slopes also improved the simulation of abyssal circulation aspects, such as outflows from the Nordic Sea that contribute to the AMOC's lower limb. Two main experiments were conducted: a hindcasting experiment (CTRL) using the full set of ocean forcing data from JRA55-do products (1958–2019); and a sensitivity experiment (SENS), where year-to-year variations in Labrador Sea buoyancy forcing were artificially constrained. SENS branched off from CTRL in May 1980, using the same forcing as CTRL except for Labrador Sea heat fluxes (53°N–65°N, 45°W–64°W), which were replaced by repeated fluxes from May 1980 to April 1981. This allowed isolating the effect of the exceptional 1990s Labrador Sea forcing on the AMOC. The robustness of the LSW formation and overturning variability simulated in CTRL was assessed using three additional hindcast experiments with varied atmospheric forcing, freshwater fluxes, continental runoff, and initial conditions. The analysis involved examining changes in LSW formation, decadal overturning changes in the subpolar basins, geographical deconstruction of the AMOC's downwelling limb, the remote effect of the excessive 1990s Labrador Sea cooling period, and the spreading of LSW density anomalies. Overturning transports were calculated using the maximum streamfunction in density space, consistent with OSNAP array procedures. Mixed layer depth (MLD) was defined as the depth where potential density changed by 0.01 kg m⁻³ relative to 10 m depth. Volume per potential density in the Labrador Sea was calculated by binning grid box volumes based on density. Potential density anomalies were analyzed using the vertical extent of the dense LSW class renewed during the 1990s convection period. The study also utilized data from the RAPID AMOC monitoring project and observational data on Labrador Sea Water thickness anomalies. The NEMO code, model configurations, and additional information are available through GEOMAR's institutional repository.

The study's key findings include: 1. **Suppression of Interannual Variability:** The sensitivity experiment (SENS) effectively suppressed interannual variations in Labrador Sea deep winter convection intensity, removing the exceptionally strong convection events of the early 1990s while retaining a weaker convection level consistent with negative NAO phases. 2. **Curbed Dense LSW Production:** SENS curtailed the production of the exceptionally dense LSW vintage observed in the CTRL experiment during the intense 1990s convection period. This highlights the strong connection between intense Labrador Sea convection and the formation of this unique water mass. 3. **Dominant Northeastern Basin Downwelling:** The majority of downwelling contributing to the AMOC occurs in the northeastern subpolar North Atlantic (Irminger and Iceland basins), rather than the Labrador Sea. The model results, although slightly weaker than OSNAP observations, confirm this pattern. 4. **1990s AMOC Peak Driven by Labrador Sea Forcing:** The increase in AMOC transport from the 1970s to its peak in the mid-1990s in CTRL is largely attributable to the heat flux changes in the Labrador Sea during that period. The SENS experiment, without the amplified Labrador Sea forcing, shows a significant reduction in this peak, confirming the crucial role of this region. 5. **Negligible Labrador Sea Impact on Interannual AMOC Variability:** Despite the major effect on the inter-decadal AMOC peak, the Labrador Sea's influence on interannual AMOC variability is negligible. The interannual transport variability in both CTRL and SENS is similar, suggesting other factors such as Irminger and Iceland Basin buoyancy forcing play the primary role in the interannual variations. 6. **Rapid Spread of Density Anomalies:** The exceptionally dense LSW formed in the Labrador Sea during the 1990s spread rapidly into the Irminger Sea via the mid-depth circulation, influencing the downwelling in the northeastern basins and contributing to the AMOC increase. 7. **Modulation of Abyssal Water Density:** The study suggests that Labrador Sea density anomalies modulate the density of abyssal waters along the East Greenland continental slope, the source of lower North Atlantic Deep Water. This could explain the observed decline in the lower deep-water AMOC limb transport after 2008 while the Nordic Seas overflow remained largely unchanged. 8. **Entrainment of Intermediate Waters:** The deep western boundary current off Greenland plays a key role in the AMOC. The model indicates that this current continuously entrains intermediate waters along its path in the Irminger Sea, significantly increasing the transport of dense water masses. The increase in the boundary current's density, particularly in the 1990s, correlates closely with AMOC overturning changes, suggesting that the intensified diapycnal mixing in this region plays a crucial role in shaping the AMOC’s interdecadal variability.

The findings challenge the prevalent view that Labrador Sea convection is the primary driver of decadal AMOC variability. While the exceptionally dense LSW produced during the strong 1990s NAO positive phase significantly impacted the AMOC, its influence on interannual variability appears minimal. The rapid propagation of density anomalies from the Labrador Sea into the Irminger Sea, and the subsequent entrainment into the Greenland deep boundary current, highlights a previously underappreciated pathway for influencing the AMOC. This pathway suggests that LSW density anomalies primarily affect the lower, rather than the upper, portion of the AMOC's deep-water limb. This offers a potential explanation for observations showing a decline in the lower deep-water range at 26.5°N after 2008 despite consistent Nordic Seas overflow. The results support previous findings suggesting a dominant influence on the AMOC by large Labrador Sea buoyancy forcing anomalies on multi-decadal timescales. The study's findings support classical understandings of the deep circulation in the subpolar North Atlantic, highlighting the importance of the dense water outflow and entrainment of intermediate waters along Greenland’s continental slope. The results suggest that AMOC multidecadal variability is influenced by interactions between Labrador Sea convection, the Irminger Sea, and the East Greenland continental slope.

This study demonstrates that while the exceptionally strong Labrador Sea convection during the 1990s significantly impacted the AMOC, its influence is primarily on interdecadal timescales and operates via a remote mechanism involving the rapid propagation of density anomalies into the Irminger Sea. Interannual variability of the AMOC is less affected by the Labrador Sea convection. These results highlight the importance of considering the complex interplay between different subpolar North Atlantic basins in understanding AMOC variability. Future research should investigate the sensitivity of this remote influence mechanism to different model resolutions and parameterizations. Moreover, improving the representation of small-scale processes governing deep formation in global climate models is crucial to advance our understanding of long-term AMOC variability.

The study primarily relies on model simulations, which may have limitations in representing all aspects of ocean processes. While the model incorporates high resolution to capture mesoscale eddies and other important features, the inherent uncertainties associated with model parameterizations and forcing data could affect the results' accuracy. Further validation using independent observational data sets would enhance the robustness of the findings. The artificial design of the sensitivity experiment (SENS) may affect the results. The choice of a particular year for repeated forcing could influence the outcome and the generalization of results. A more systematic study investigating other potential years might be beneficial.