AMOC and North Atlantic Ocean Decadal Variability: A Review

Abstract

The North Atlantic Ocean is vital to Earth’s climate system. Scientific investigations have identified the Atlantic Meridional Overturning Circulation (AMOC) as a significant factor influencing global climate change. This circulation involves ocean currents that carry relatively warm, salty water northward in the upper layers, while transporting colder, less salty water southward in the deeper layers. The AMOC relies on descending water at deep convection sites in the high-latitude North Atlantic (NA), where warmer water cools, becomes denser, and sinks. A concern regarding the AMOC is that the freshening of the sea surface at these convection sites can slow it by inhibiting deep convection. Researchers have used oceanographic observations and models of Earth’s climate and ocean circulation to investigate decadal shifts in the AMOC and NA. We examined these findings to provide insights into these models, observational analyses, and palaeoceanographic reconstructions, aiming to deepen our understanding of AMOC variability and offer potential predictions for future climate change in the North Atlantic. While the influence of high-latitude freshwater is crucial and may slow the AMOC, evidence also shows a complex feedback mechanism. In this mechanism, the negative feedback from wind stress can stabilize the AMOC, partially counteracting the positive feedback effects of freshwater at high latitudes. Although some models predict significant shifts in AMOC dynamics, suggesting imminent and possibly severe deceleration, recent observational research presents a more cautious view. These data analysis studies acknowledge changes, but highlight the robustness of the AMOC, particularly in its upper arm within the Gulf Stream system. While it cannot be entirely dismissed that the AMOC may reach its tipping point within this century, an analysis of data concerning the decadal variability in the AMOC’s upper arm indicates that a collapse is unlikely within this timeframe, although significant weakening remains quite possible. Furthermore, deceleration of the AMOC’s upper arm could lead to less stable and more vulnerable North Atlantic Ocean climate patterns over extended periods.

1. Introduction

Of all oceans, the North Atlantic (NA) is considered the most important for the Earth’s climate. The Atlantic Meridional Overturning Circulation (AMOC) is crucial for the North Atlantic Ocean climate and plays a central role as a regulator of the ocean climate on decadal and longer timescales. In the late 20th century, and especially in the recent years of the 21st century, numerous modeling and observational research activities have been conducted to enhance our understanding of the AMOC and its role in driving long-term climate change. Surprisingly, despite these efforts, there is still a great deal of controversy and no consensus on the nature, operation, variability, and future of the AMOC among oceanographers and climate scientists. Therefore, it is necessary to revisit the large bulk of the results of AMOC and NA decadal variability studies to better understand what has already been established and what is still being debated and needs more attention and new research.

1.1. Climate Normals

Earth’s climate is officially described as the range of conditions through which the major components of the planetary climate system transition over a period of 30 years. The ocean plays a crucial role in the climate system in timescales ranging from decadal to centennial; therefore, it is essential to understand changes in ocean parameters to predict climate variability. One fundamental concept in climatology is the idea of “climate normals”. According to the World Meteorological Organization (WMO), a climate normal presents a baseline against which field observations can be compared. In practical terms, climate normals are the average values of the climate system’s parameters (such as temperature, salinity, and pressure) within specific periods (months, seasons, and years) over a 30-year reference period [1]. Member nations of the International Meteorological Organization (IMO-WMO) were initially required to calculate climate normals for their respective countries for the 1901–1930 period and update them every 30 years, resulting in 1931–1960 and 1961–1990 climate normals [2]. The WMO advises member countries to update their 30-year climate normals every ten years. Currently, the period 1991–2020 is the preferred interval for computing climate normals. The NCEI recently published new ocean climate normals for this interval, which are currently available on the World Ocean Atlas/Climate Normals website. The use of 30-year normals to describe the climate is currently being questioned, and it has been suggested that a more extended period should be considered as the reference interval for computing anomalies [2]. In ocean heat content calculations, the reference mean or base climatology uses the entire period from 1955 to 2006, which is 50 years, to reflect the extended time of thermocline water turnover. This is much longer than the WMO-recommended 30-year period.

1.2. Research Activities

Over the last century, climatologists and oceanographers have invested much effort and resources in compiling reliable climatological products based on observations, as well as developing new generations of ocean and climate models. Together, in situ observations and modeling provide a better understanding of the climate system’s state and variability and can be used to improve the forecasting of potential climate changes.

Numerous intensive ocean observational programs have comprehensively assessed the climatic state and long-term variability of oceans. In the North Atlantic, several studies have provided a detailed view of this part of the world’s ocean. National and international monitoring and research programs, such as the International Ice Patrol (IIP) Survey, Ocean Weather Ships (OWS), Mid-Ocean Dynamics Experiment (MODE), US-USSR POLYMODE (Polygon and MODE), World Ocean Circulation Experiment (WOCE), Climate Variability and Predictability (CLIVAR) of the World Climate Research Programme (WCRP), Rapid Climate Change Programme (RAPID), and many more, contribute to an advanced understanding of North Atlantic climate dynamics (see the review in [3]). In the 21st century, ocean observations have been greatly enhanced by the arrival of new instrumentation: Argo floats [4,5,6,7,8]. Argo floats have revolutionized observational oceanography, enabling almost global data coverage of the World Ocean—an ambition that was unimaginable only a few decades ago. It is important to note, however, that Argo floats mainly collect information from the upper 1.5 km of the ocean, whereas deeper ocean measurements still rely heavily on traditional instruments for deep ocean profiling.

1.3. Global Ocean Circulation and AMOC

The timescales of ocean circulation, which are the key to the ocean’s role in climate dynamics on millennial and shorter timescales, are very different in the upper ocean, main thermocline, and deep ocean. They can also vary from years at the basin scale to hundreds of years at a global scale. The longest timescales, being several centuries, are determined by the global thermohaline circulation or inter-basin exchange of thermocline water [9], also known as a “global conveyor” [10]. However, a practical approach to ocean climate diagnostics is to use the data obtained in the last sixty-plus years following the advent of widespread and numerous ocean observations beginning in the middle of the twentieth century, which would cover two 30-year successive periods.

There is a consensus among climate scientists that the main climate controls imposed on the Earth’s climate system by oceans are via air–sea heat and freshwater exchange. The oceans gain heat in the tropics and subtropics and release it at high latitudes. The meridional overturning circulation that regulates poleward heat transport by ocean currents facilitates energy redistribution between low and high latitudes in the oceans and eventually via air–sea interactions in the climate system. The thermohaline structure of the World Ocean reflects all three factors: ocean–air heat exchange (controlling surface water temperature), evaporation–precipitation balance (which, paired with the melting and freezing of sea ice and river runoff, controls surface water salinity), and the advection of heat and salt by ocean currents.

Since the 1960s, the Earth’s surface and almost all regions of the world have been warming, with a substantial rise in temperature in the upper ocean layers [11,12,13]. As this has been happening, concern has been growing that this ocean warming might lead to unpredictable and perhaps dire consequences in all aspects of Earth’s climate, and many have begun to believe that AMOC may play a significant, if not crucial, role in climate change. The primary concern lies in the fact that the AMOC functionality relies heavily on the upper ocean’s water density in the subpolar regions, particularly in convection sites where cooled, and consequently denser, water descends into the deep ocean. This deep convection process ensures the continuity and uninterrupted operation of the AMOC. This understanding of the nature of ocean overturning circulation was predicated on the seminal theory advanced by Stommel in the late 1950s [14,15,16]. Stommel’s main idea is that deep ocean convection occurs at only a few convection sites in the North Atlantic and Southern Ocean, where cooled and denser water descends into the abyssal ocean and then flows equatorward. The descending water was replaced by poleward-flowing water, mainly within the western boundary current, thus forming a meridional overturning circulation, as shown in Figure 1.

For more than 70 years, the basic idea of overturning circulation and deepwater formation has remained unchallenged, and hundreds of studies have been carried out to achieve a more detailed understanding of how global deep ocean circulation is initiated, maintained, and changes over decadal and longer periods (e.g., see reviews in [17,18,19,20,21], among others). Most of these undertakings are concentrated on the AMOC, exploring various aspects of deep convection and its primary driving force. Key areas of investigation include the mechanisms and locations of deep convection, its variability, influencing factors, and overall strength, aiming to elucidate the role of deep convection and the AMOC in shaping oceanic climate change.

Figure 2 presents a simplified depiction of global ocean circulation driven by deep convection, building on Stommel’s view of deepwater formation (Figure 1). It is reproduced from [22], who modified the original sketch from [23], and combines Gordon’s groundbreaking study (1986) on thermocline water mass exchange between oceans with Broecker’s visual representation of the “global conveyor” [10]. In addition, the scheme in Figure 3 explicitly focuses on the North Atlantic component of the circulation system.

While extensive studies have been conducted on various aspects of deep convection, the essence of its operation relies on a single vital factor crucial for this process: the formation of deep ocean waters in the subpolar NA and Nordic Seas regions. Deep convection can occur only when warmer and less dense water from subtropical areas reaches a potential convection site. This water must cool sufficiently to become heavier than the underlying subsurface water. Once this happens, it begins to overturn and mix with the denser water, sinking towards the deep ocean. This process initiated southward flow in the lower branch of the AMOC. It is important to acknowledge that while deep convection typically occurs when denser water forms above lighter water during winter surface cooling, alternative mechanisms for deepwater formation can also occur in subduction zones, as noted by some authors, for example, Ref. [24]. The upper ocean is primarily ventilated through subduction, which does not require denser water to form lighter water. In contrast, the deep ocean is ventilated through open–ocean convection and by cascading denser water down toward greater depths.

The unique nature of deepwater formation in high-latitude NA hinges on the intensity of processes that either enhance or diminish this phenomenon. The critical question is whether circumstances that would completely halt deepwater formation can arise. The latter scenario fuels speculation about the potential consequences of a total shutdown of the ocean’s circulation system, as dramatically portrayed in the notorious Hollywood blockbuster, The Day After Tomorrow.

Addressing this issue has predominantly involved examining results from model simulations utilizing both climate and ocean circulation models to elucidate the AMOC problem and the associated climate shifts resulting from altered conditions at deep convection locations. Numerous observational studies have examined this issue. This review aims to consolidate the findings from these models and data-processing efforts to establish a potential consensus by comparing the conclusions drawn from these two major research streams.

We caution readers that this review covers only some research papers in these three primary research areas. The vast number of publications on this topic makes it virtually impossible for any individual or research team to cover them comprehensively, especially within the constraints of a review article. Nevertheless, we tried to include as many publications as possible, with our reference list containing over 300 research papers, attesting to this effort.

This review was organized to facilitate ease of reading and understanding. We begin with a broad examination of the North Atlantic circulation, and then we explore the evolution and advancement of AMOC modeling research. Subsequently, we analyze the oceanographic data. This review discusses and provides insights into the historical and current state of the AMOC, as well as the potential changes in the warming climate of the 21st century.

2. North Atlantic Circulation: An Overview

2.1. North Atlantic Circulation System

To better understand the AMOC, it is essential to thoroughly understand the North Atlantic Ocean, including its thermohaline structure, circulation patterns, and key components that influence the AMOC and its climate effects, e.g., Refs. [22,25,26,27,28]. The significance of NA is underscored by the GS and its extensions, which form the upper branch of the AMOC. The GS is the most powerful ocean current system in the Northern Hemisphere. Extensive reviews of NA studies on the NA circulation system and AMOC functionality can be found in [3,17,19,20,21,29,30]. Evidence suggests that the recent decadal NA climate trajectory has been shaped by two intertwined processes: a gradual global warming trend and multidecadal variability superimposed on this trend [31]. As noted earlier, the transformation of warm and saline ocean water transported to deep convection sites at high latitudes was the primary driver of the AMOC. Consequently, examining changes in the upper ocean’s circulation during ongoing surface warming will help us better understand the overall stability or instability of the NA climate resulting from upper ocean warming.

A scheme of the surface NA current system, based on general knowledge of ocean currents in the region, for example, Refs. [32,33,34,35], is shown in Figure 4.

2.2. Gulf Stream System

The Northwest Atlantic Ocean current system consists of four main elements: the GS system, Labrador Sea currents, the North Atlantic Current, and the Mid-Latitude Transition Zone (MLTZ). Although not technically part of the NA, the Nordic Seas, encompassing the Greenland, Iceland, and Norwegian currents, play a crucial role in the AMOC because of their contribution to deepwater formation. This region is a significant component of the overall NA upper ocean circulation and AMOC. The NA upper ocean circulation is dominated by the GS, which forms the core of NA basin-scale circulation. The GS current system is one of the most extensively researched and mapped ocean current systems.

By the mid-1900s, oceanographers had already acquired comprehensive knowledge about the GS, particularly near the U.S. East Coast [38,39,40,41]. The GS’s westward intensification, first theoretically proved in [42], results from the Coriolis effect on spherical Earth, for example, Refs. [41,43,44]. The topography and structure of the western NA shelf-break significantly influences many aspects of the GS system. As it flows along the U.S. East Coast, the GS gains strength from three sources: the northern branch of the North Equatorial Current and two recirculation gyres. These two upper ocean recirculation gyres are cold and warm gyres north and south of the eastbound GS core, respectively. Cold water from the Labrador Sea flows westward along the coast (this flow is also known as the Slope Water Current). The warm water south and east of the GS and its continuation circulate in the Sargasso Sea and comprise the GS Recirculation Gyre [45,46], also known as the Worthington Gyre, between 75° and 55° W [46]. Both gyres feed the eastbound stream and contribute to the GS transport increase between the Florida Straits and Cape Hatteras and along the GS extension. Water transport almost doubles downstream between Cape Hatteras and approximately 55° W, resulting from the water fed by the Northern Recirculation Gyre (the Slope Water Current) and the Worthington Gyre. It strengthens before turning north at the eastern flank of the Great Banks and finally becomes the North Atlantic Current [43,47]. East of Newfoundland and the Great Banks, the North Atlantic Current meets the Labrador Current, creating the Mid-Latitude Transition Zone (MLTZ). After breaking from the continental shelf, the eastbound GS becomes the GS extension—a free baroclinic jet reaching the ocean bottom. Its structure changes from a single meandering front to multiple branching fronts with a great deal of mesoscale activity and increasingly large meanders.

As it departs from the coast, the stream position fluctuates throughout the year, shifting northward in autumn and southward in spring [32,46]. A recent analysis suggests that the central core of the GS between 70° and 50° W remains remarkably stable over many decades, with minimal changes in the jet axis position. However, the GS extension east of 50° W exhibited more substantial north–south migration over several decades, thus significantly altering the adjacent ocean climate [48,49]. At approximately 65° W, the meandering envelope measures nearly 500 km in width, which is five times the width of the jet at its separation point near Cape Hatteras. The meanders and mesoscale eddies west of the North Atlantic Current form the MLTZ, where cold and relatively fresh water from the Labrador Sea mixes with warm and salty water from the GS and North Atlantic Current. Two types of meanders exist: those generated by internal dynamic instability, similar to turbulent flow; and the quasi-stationary meanders caused primarily by bottom topography. Occasionally, large-amplitude meanders break off from the jet, forming warm- and cold-core GS rings (Figure 4). Anticyclonic warm-core rings were found north of the GS core, whereas cyclonic cold-core rings were found south of the core. These rings migrate westward and occasionally merge with the GS. The rings and meanders facilitate e heat and salt exchange across the frontal zone of the GS.

In summary, the GS, a highly stratified (vertically and horizontally) and very unstable jet current, serves as a barrier and blender for warm and cold waters along its edges. The blending or mixing of these waters is facilitated by meanders and mesoscale eddies [25]. Further mixing is performed by the so-called sub-mesoscale streamer transitioning between eddy-induced mesoscale geostrophic and smaller-scale turbulent mixing. New research has shown that this type of mixing is essential at the Cold Wall of the GS, where the outer warm core of the GS contacts the cold water originating in the Labrador Sea [50,51].

2.3. North Atlantic General Circulation and AMOC

The GS system has been strongly emphasized because of its critical importance in accurately representing oceanographic observations in any model or in ocean climatology. This representation should be sufficiently comprehensive to include all of the previously mentioned essential features. The subsequent movement of warm saline water carried by the GS and its extension, the North Atlantic Current, is equally important for the AMOC function and fate. Two key factors determine the volume and characteristics of water reaching the potential convection area: the strength of the GS and North Atlantic Current, and the intensity and duration of deep convection when it occurs, as this also influences AMOC’s operation. These factors create a self-reinforcing cycle: stronger GS and North Atlantic Current transport towards the potential deepwater formation site can lead to more intense deep convection, requiring the GS system to supply more water to maintain deepwater production. Conversely, weaker convection requires less subtropical water for maintenance, resulting in a slowdown of the AMOC. Numerical ocean circulation models, either being standalone or a part of climate models, and oceanographic observations indicate that the AMOC has already been experiencing a gradual slowdown for at least several decades, e.g., Refs. [52,53,54]. Some latest reconstructions of the AMOC using numerical models, for example [55,56], corroborate these findings, indicating that the AMOC is indeed undergoing a deceleration.

The subpolar gyre, particularly its northern section connected to the Nordic Seas, is a crucial component of the North Atlantic circulation that significantly influences AMOC dynamics and changes (refer to the diagram in Figure 4). The North Atlantic Current system, comprising three primary branches (northern, central, and southern), dominates the eastern and central regions of the subpolar gyre [57]. The currents on the north branch flow towards the Irminger Sea, while those of the central branch move in the direction of the Greenland–Scotland Ridge. The southern branch’s waters progress towards the northern boundary of the subtropical gyre.

The North Atlantic Deep Water (NADW) forms the principal element of the AMOC, originating in the deep convection zones of the Labrador, Irminger, and Nordic Seas. These regions collectively constitute the deep ocean branch of the AMOC (Figure 5). This diagram offers a simplified representation of meridional overturning. NADW moves southward through various intricate pathways, including advection in the Deep Western Boundary Current (Figure 3), recirculation in deep gyres, and mixing processes. The sketch demonstrates the critical features of the AMOC, including stratification of its upper and lower branches. Another element of the AMOC system is the Antarctic Bottom Water (AABW) formed by dense cold water sinking near Antarctica and flowing northward. A portion of Antarctic Bottom Water (AABW) from the Weddell Sea moved northward into the Atlantic Ocean below the NADW. Squeezed from below, the NADW occupies deep layers down to approximately 4000 m, whereas the AABW fills the abyssal ocean below this depth. NADW and AABW comprise the deep ocean component of global thermohaline overturning circulation, as shown in Figure 1 and Figure 2.

The NA portions of NADW are formed at open ocean deepwater convection sites and feed the deep layers directly. In contrast, the NADW formed in the Nordic Seas joins the main NADW southward flow through dense water overflow across the Greenland–Iceland–Scotland Ridge [58]. NADW comprises two layers of water. The deep open ocean convection during winter produces Labrador Sea Water, formed mainly in the Labrador Sea and, as suggested by new research, also to some extent in the Irminger Sea basin [59,60,61]. While sinking, this deepwater can reach a depth of 2000 m. Another portion of NADW forming in the Labrador Sea is called Classical Labrador Sea Water. The lower water mass of NADW, reaching a depth of approximately 4000 m, forms from the overflow of the Greenland–Iceland–Scotland Ridge and consists of the Iceland–Scotland Overflow water and Denmark Strait overflow water.

Regarding the Labrador Sea Water, some recent observation programs indicate that some fraction of the NADW, and partly the Labrador Sea Water, is formed by open ocean convection originating from the Irminger Sea [60,62,63,64]. Furthermore, some authors contend that the AMOC lower limb is primarily derived from the Iceland Basin and Irminger Sea, and that NADW formation in these regions may be crucial for the overall functioning of the AMOC [63,65,66].

3. Internal and External Controls of AMOC

3.1. Primary Factors Influencing AMOC

Two primary factors, largely independent of their direct effects, influence the critical components of the feedback loops controlling AMOC. The first factor is wind stress over the North Atlantic, which is crucial for determining various aspects of the GS system. This influence occurs through two mechanisms: directly driving surface currents, and indirectly through the Ekman pumping generated by the wind stress curl. These processes collectively shape the dynamic and thermohaline structures of GS systems.

Although partially linked to wind stress patterns, another significant factor influencing the North Atlantic (NA) is the distinct processes of evaporation and precipitation. This process generates salinity patterns that contribute to the density structure of NA, and both wind stress and surface temperature influence subtropical evaporation. In contrast, subpolar NA precipitation is mainly independent of surface temperature and is an external factor that regulates freshwater exchange at the sea surface. The influx of freshwater decreases the density of the upper ocean layers by dilution. This decrease in sea surface density could play a crucial role in the formation of deepwater, potentially becoming a key factor influencing the AMOC and possibly even global thermohaline circulation.

3.2. Freshwater Exchange and Meltwater Impacts

The strength of the AMOC relies on preserving the NA surface salinity difference between the high-salinity subtropical gyre and low-salinity subpolar gyre. This contrast is primarily maintained by the moisture balance over these gyres and atmospheric moisture exchange between them in the NA. Subtropical regions experience greater evaporation (E) than precipitation (P), resulting in a positive evaporation–precipitation difference, E − P > 0, which significantly increases the surface salinity in the subtropical gyre, making the upper arm of the AMOC relatively warm and salty. Conversely, the subpolar gyre has an E − P < 0 condition, leading to fresher sea surface waters. This freshness allows for saltier and cooler water from the subtropics to become denser than subsurface water, thereby sustaining deepwater formation through convection along the isopycnal of neutrally stratified water.

Studies have shown that, over the last ~40–60 years, ocean areas with high salinity have become more saline, whereas regions with lower salt contents have grown fresher [67,68]. This phenomenon has been directly linked to the intensification of the global water cycle [69,70,71], although the rates of change vary [71]. The decreasing salinity in subpolar zones may contribute to a slowdown of the AMOC, but it remains unclear whether the increasing salinity in subtropical North Atlantic regions could offset this freshening effect and maintain AMOC strength. The intricate moisture transfer between North America and the North Pacific and the opposing exchange processes between the subtropical and subpolar regions within the North Atlantic further complicate this issue. With E dominating the Atlantic and P dominating the Pacific, amplifying this pattern would lead to further salinification and freshening of the Pacific Ocean. This salinity inter-basin dipole has been amplifying since the 1950s [67,68,71] and is primarily sustained by ~0.5 Sv of water vapor generated over the subtropical Atlantic being transported across Central America and deposited as precipitation in the equatorial Pacific [72]. Water vapor from the Pacific is not deposited as precipitation in the Atlantic because of the blocking of moisture transport by the Rocky Mountains [73]. This leads to significant differences in the freshwater balances between the two oceans, especially in the subpolar regions.

Models have shown that the inter-basin moisture transport—termed the “Atmospheric Bridge”—and resulting near-surface salinity (NSS) contrasts between the Pacific and Atlantic are a key controlling mechanism of the global thermohaline circulation [74,75,76]. Most of the precipitation in NA is sourced from the NA, with minimal moisture contributions from other ocean basins [72]. Therefore, to assess long-term ocean climate change, it is paramount to understand how salinity contrasts within the NA are built and maintained [77].

The subtropical-subpolar NA atmospheric moisture transport is an effective and rapid (on the order of days to weeks) mechanism of freshwater transport between the positive and negative evaporation–precipitation balance regions in the NA. It is one of the leading candidates for maintaining observed inter-gyre NSS contrasts and freshwater advection into the subpolar NA (SPNA) [77]. The moisture transport mechanism is shown schematically in Figure 6.

Over subtropical NA, where E > P, water vapor is produced and then diverges from the NSS maximum-salinity region. Although the strength and direction of the divergence of water vapor from the subtropical NA (STNA) vary with season [78,79], in all four seasons, it is captured and precipitated in four different regions: (i) along the Atlantic Intertropical Convergence Zone (ITCZ), (ii) in the eastern tropical Pacific ITCZ (via the aforementioned Atmospheric Bridge), (iii) off the East Coast of North America, and (iv) in the SPNA. Because the ITCZ dynamics and potential impacts on tropical surface salinity are outside our focus on extratropical NSS interconnections, only the subpolar and subtropical NA surface salinity contrasts, and their relationships to the hydrological cycle are the subjects of this study.

In the atmosphere, water vapor transported to the East Coast of North America is typically captured within eastward-/northeastward-moving mid-latitude cyclones, which normally originate from baroclinic instability caused by sharp temperature gradients off the East Coast of North America. These storms are stronger and more frequent in winter because of increased baroclinic instability. They generally follow a similar northeast trajectory (i.e., storm tracks), directing them and their associated precipitation to the subpolar and northeast Atlantic regions [80,81,82,83].

Excess E over P along the GS path and in the STNA also provides moisture for NA atmospheric rivers [84] throughout the year [85]. Alternatively, north/northwestward moisture transport from the E > P subtropical region (see schematics in Figure 6) can cause convergence in the subpolar NA, leading to precipitation, particularly during certain seasons.

There are two major pathways for water vapor produced in the STNA to be transported and deposited (as precipitation) in the SPNA: the indirect pathway (i.e., water vapor is captured in a mid-latitude cyclone and transported along the NA East Coast storm track) and the direct path (i.e., north/northwestward moisture divergence from the STNA with convergence over the SPNA), as shown in Figure 6. For the indirect path, the latency between water vapor production in the subtropical region and deposition in the SPNA is, at most, a week or two, with the direct pathway requiring only several days.

Deepwater formation can also be influenced by meltwater, mainly through influx from the Arctic. This freshwater primarily originates from melting sea ice and, to a lesser extent, from the melting of land-based ice entering the ocean near Greenland. Some models have suggested that increased freshwater flowing along Greenland’s coast can enter the SPNA region. This influx may reduce deepwater production, consequently causing slowdown of the AMOC [86].

4. Models

In climate or ocean model sensitivity studies, researchers can easily simulate the effects of freshwater or meltwater on ocean circulation. The most basic method involves “hosing” experiments, in which researchers disrupt the steady state of the AMOC by introducing surface freshwater. This can be achieved by adding freshwater fluxes or gradually lowering the sea surface salinity until deepwater formation is significantly impeded or stopped, which causes the AMOC to decelerate. These computer simulations are based on the possibility that such meltwater events may have occurred in recent geological history, particularly during the glacial–deglacial cycles following the Last Glacial Maximum (LGM), roughly 20–18 thousand years ago, and meltwater events after the LGM [87,88,89,90,91].

In contrast, observational data merely describe the disruption of deepwater formation caused by freshwater influx, providing only snapshots of ocean climate states. Although these data cannot portray the continuous process as models can, they can offer a range of ocean parameters such as temperature, salinity, and water transport. These parameters reveal the long-term climatic trajectory of the AMOC, thus providing reference states for model verification.

Conversely, another powerful method for simulating long-term ocean variability, ocean climate re-analysis, combines the strengths of models and observations. This method enables both retrospective and prospective analyses, offering further insights into the discoveries made in observational and modeling research regarding AMOC conditions and trends. For example, Ref. [92] exemplifies this approach. A notable emerging advantage of the re-analysis method is its ability to incorporate satellite measurements of the ocean’s surfaces. This presents a particularly promising opportunity to integrate the circulation and thermohaline structures into a single analysis. The following discussion of ocean re-analysis will significantly increase the volume of this review. Therefore, in the subsequent sections, we skip the ocean data re-analysis results and explore two of the three approaches in this sequence, models and in situ data analyses, to explore how these two approaches collectively contribute to a deeper understanding of AMOC functionality, current changes, and their potential trajectories in the near and distant future.

4.1. Early General Circulation Models

Since the development of ocean general circulation models [93,94,95] and coupled ocean–atmosphere models [96], numerous research papers have been published, recently been summarized in [97]. We focus here on a limited selection of modeling studies that specifically examined AMOC simulations, showing potential changes in AMOC behavior across various climate scenarios and patterns. From the early years of coupled ocean–atmosphere modeling, two studies stand out as pioneering works showing how the global conveyor, and thus its main engine, AMOC, may exist in different modes dictated by the buoyancy forces in the deepwater formation sites. Frank Bryan [98] was the first to find that salinity anomalies in the ocean’s surface layer in high latitudes can lead to significant changes in circulation patterns, with implications for understanding past climate variability and future ocean dynamics. A study using an idealized sectoral ocean basin revealed that multiple equilibrium solutions could arise from perturbations in high-latitude salinity fields, affecting the transition from symmetric to asymmetric circulation.

Manabe and Stoffer [99] discovered that, in a coupled ocean–atmosphere model, the influx of freshwater in the northern NA could hamper, or even reverse, meridional overturning in the NA. Their study identified two equilibrium states: one resembling the present-day AMOC, characterized by high-latitude sinking in the NA; and another that contrasted with the current thermohaline overturning, featuring upwelling instead of sinking in the North Atlantic. Further investigation into the stability of this reverse mode was conducted by [100,101]. Although they determined that the reverse mode was unstable, this finding did not invalidate the main conclusion of their earlier research. Their work suggested that an increase in surface freshwater in areas of deep convection could theoretically trigger a significant, albeit possibly temporary, shift in the thermohaline overturning. Additionally, it was proposed that the thermohaline circulation of the NA plays a vital role in creating surface salinity disparities between the northern regions of the NA and the North Pacific. They hypothesized that these findings might be comparable to the substantial and sudden shift between the Bølling–Allerød and Younger Dryas periods, approximately 11 Kya. Following pioneering studies of the ocean–atmosphere system using numerical models, researchers have recognized the potential of such models to enhance our understanding of past climates. This has led to the development of numerous computer-based simulations. These computational studies utilized combined ocean–atmosphere and independent ocean models to reconstruct ancient climate conditions and historical ocean circulation patterns by exploring various scenarios by altering the surface temperature and salinity parameters.

The onset of modeling studies on global thermohaline overturning circulation coincided with a new understanding of inter-basin water exchange through this circulation, as first presented in influential work by [9], followed by the advent of the concept of the global thermohaline conveyor [10]. The previous section referenced these publications when discussing the timescales of the water circulation at different ocean depths. This study examined the exchange of thermocline water between oceans, with a focus on NADW formation and distribution. NADW formation involves the movement of upper-layer water to deep ocean levels at a flow rate of 15–20 Sv (1 Sv = 10⁶ m³ s⁻¹). This water mass spreads throughout the Atlantic Ocean and is transported to the Indian and Pacific Oceans via the Antarctic Circumpolar Current. It is suggested that a balanced flow of warm upper-layer water occurs mainly within the thermocline, connecting the ocean thermoclines in a global circulation pattern. The return flow route includes movement from the Pacific to the Indian Ocean through the Indonesian Seas, across the Indian Ocean, and into the South Atlantic via the Agulhas Current. This research highlights the significance of this warm water pathway, which is considered more important than the cold water route through the Drake Passage. Changes in thermohaline forcing and wind-driven circulation can affect this flow, potentially affecting NADW formation characteristics.

4.2. Freshwater (Salinity) Control of AMOC

Around the same period, Broecker [10] introduced the concept of a global “conveyor belt.” This concept and its visualization in a schematic captivated the scientific community, gaining widespread popularity through an exciting new image replicated in numerous versions and illustrations, as exemplified in Figure 7. While [10] contended that this cartoonish image inaccurately depicts the actual structure of the global thermohaline conveyor, the fundamental concept of the conveyor being driven by salinity disparities between the NA and North Pacific remains central to our current fundamental understanding of its operation. To illustrate the conveyor belt idea, many animations have been developed to help visualize conveyor schematics, for example, the animation on the NASA website (there are other impressive visualizations on the NASA website showing ocean currents on the globe). The primary importance of the conveyor concept is due to its role in distributing salt globally, thereby sustaining the differences in salinity across various regions of the world’s oceans. It can be viewed as a “salinity conveyor belt” because the surface freshwater fluxes have a significant influence on the belt, meaning that they are essentially governed by the salinity at the sea surface.

According to Broecker [10], the conveyor is propelled by excess salt that remains in the Atlantic, owing to vapor export. On average, the Atlantic’s surface waters contain approximately one gram of sea salt per kilogram (liter) of water, which is more than the same water mass (volume) in the Pacific. It is important to recognize that the metaphor presented in Figure 7 is a gross simplification. Despite this, it effectively conveys certain concepts regarding global ocean overturning circulation that are generally considered accurate [102]. Numerous adjustments have been made to this schematic to better capture the complexities of the overturning circulation. Nonetheless, the concept of ocean interconnections illustrated in Figure 7 has played a crucial role in aiding our understanding of the fundamental aspects of global overturning circulation [102].

Several factors contribute to the strong convection experienced in the northern North American region, whereas the northern North Pacific experiences almost no convection, e.g., refs. [103,104]. First, the North Pacific receives more precipitation than evaporation and has significant river runoff, leading to substantial salinity differences. The difference in salinity between the surface waters in the northern North American region and those at similar latitudes in the northern North Pacific is even more pronounced, ranging from 1 to 3 salinity units, with approximately 34 to 37 in the NA compared to 33 to 34 in the northern North Pacific. Second, the northern North Pacific has stronger stratification due to the warmer upper ocean compared to the below-water layers, resulting in a stronger pycnocline. In contrast, the northern North American region has a weaker pycnocline. Third, the northern North American region experiences stronger heat loss to the atmosphere, owing to the dry, cold winds from North America and Greenland. Finally, once deepwater formation is established in the North American region, it tends to self-reinforce through the AMOC’s positive feedback, whereas the Pacific Ocean is a return flow basin where deepwater rise rather than sink. Because of these significant differences, when surface waters in the northern Pacific are cooled, they only descend to a depth of a few hundred meters before reaching their buoyancy limit. This prevents the formation of deepwater in the northern Pacific.

Analysis of Greenland ice core data [105,106] has revealed millennial-scale climate oscillations during glacial periods. These variations were characterized by 5 °C shifts in air temperature, significant changes in dust concentrations, and 50 ppm fluctuations in CO₂ levels. Based on similar-duration oscillations in surface water conditions observed in a deep-sea core at 50° N [107], it is proposed that these Greenland climate changes are likely triggered by variations in Atlantic Ocean salinity. These salinity fluctuations subsequently affected the strength of the glacial–interglacial Atlantic conveyor circulation and were viewed as a salinity oscillator operated in the glacial NA [107,108].

This foundational research is crucial for succeeding AMOC model advancements, and its significance cannot be underestimated. This provided modelers with valuable insights to focus on their numerical simulations, leading to prompt and fruitful outcomes. A seminal study [109] addressing the “conveyor on” versus “conveyor off” issue is paramount, considering the significance of initial hosing experiments that have explored the role of freshwater in AMOC dynamics. This study focuses on the freshwater forcing of Atlantic thermohaline circulation, which is crucial for transporting heat northward and significantly influencing the oceanic climate of northern NA. This study highlights that atmospheric freshwater loss primarily occurs in the subtropical South Atlantic, balanced by northward salt transport in the wind-driven subtropical gyre. However, it was found that the thermohaline circulation transports freshwater southward. The discussion was based on two contrasting views on freshwater forcing: one views it as a driver of NADW flow, whereas the other views it as a brake. Rahmstorf argued that the freshwater transport direction is linked to the stability of the conveyor belt, suggesting that, paradoxically, a net freshwater loss enhances salinity and drives NADW flow. This study employes conceptual and global circulation models to analyze these dynamics, concluding that the current state of Atlantic circulation is perilously close to a bifurcation point, which could have significant implications for future climate change. As will be highlighted later, there is currently no firm agreement within modeling communality in relation to the AMOC’s proximity to the bifurcation or tipping point.

4.3. Factors Affecting AMOC Dynamics

The broader view is that subtropical NA is the primary region for atmospheric freshwater gain (and oceanic freshwater loss), counterbalanced by northward freshwater (salt) transport in the atmosphere (ocean). The direction of freshwater transport influences the stability of the thermohaline circulation; a southward flow indicates thermal driving, which can be disrupted by freshwater forcing, potentially resulting in multiple stable circulation states. Enhanced freshwater at high latitudes of the NA input can hinder NADW formation, leading to various stable states, including one without NADW formation. The direction of freshwater transport is critical; southward flow suggests a thermally driven system susceptible to destabilization by freshwater forcing, affecting overall thermohaline circulation. England and Rahmstorf [110] provided a clear explanation of the results of their research [109] by using a bifurcation diagram illustrating the simplified freshwater exchange and general circulation models, as shown in Figure 8.

As the concept of surface salinity changes influencing the AMOC gained traction, ocean and climate scientists began investigating their impacts on ocean circulation and induced climate change. Numerous computational simulations have been conducted to mimic the effects of varying surface salinities at high latitudes in the NA Ocean. These simulations typically employed additional freshwater fluxes, either positive or negative, in some regions of the subpolar NA and Nordic Seas, either directly or indirectly. Experiments involving positive freshwater input are termed “hosing experiments,” a term which draws from the analogy of spraying the sea surface with freshwater from a hose. Numerous models have investigated the influence of high-latitude density control on thermohaline circulation [88,109,111,112,113,114,115,116,117]. This shortlist represents only a small fraction of the extensive modeling research conducted in this field. For a more comprehensive overview, readers can refer to references [23,118,119]. Stouffer [120] summarized the nature of water-hosing numerical experiments. The most recent review of hosing experiments mimicking high-latitudinal freshening of NA can be found in [121].

Figure 8. Schematic of hysteresis, with solid black lines indicating stable equilibrium climate states and dotted black lines indicating unstable climate states (details can be found in [122]). Different types of transitions are indicated by colored arrows: (a) an advective spindown related to Stommel’s salt transport feedback, (b) a convective shutdown related to Welander’s “flip-flop” feedback, (c) a transition between different convection patterns, and (d) the restart of convection. A complete hysteresis loop cycles between the “on” and “off” states of the Atlantic thermohaline circulation via transitions (a, d). The small arrows indicate movement in the phase space of the nonequilibrium states. “S” marks the Stommel bifurcation beyond which no NADW formation can be sustained—a figure adapted from Ref. [122].

Figure 8. Schematic of hysteresis, with solid black lines indicating stable equilibrium climate states and dotted black lines indicating unstable climate states (details can be found in [122]). Different types of transitions are indicated by colored arrows: (a) an advective spindown related to Stommel’s salt transport feedback, (b) a convective shutdown related to Welander’s “flip-flop” feedback, (c) a transition between different convection patterns, and (d) the restart of convection. A complete hysteresis loop cycles between the “on” and “off” states of the Atlantic thermohaline circulation via transitions (a, d). The small arrows indicate movement in the phase space of the nonequilibrium states. “S” marks the Stommel bifurcation beyond which no NADW formation can be sustained—a figure adapted from Ref. [122].

Contemporary oceanographic measurements are yet to provide evidence for extensive natural sea surface freshwater hosing, which would be strong enough to significantly reduce sea surface salinity to stop or even reverse the AMOC. Consequently, numerical freshwater hosing simulations replicate this theoretically inferred effect without the support of modern oceanographic data. Furthermore, a forceful AMOC response, including its potential cessation or reversal, cannot be validated using the current observational data. This limitation applies specifically to present-day observations. As previously mentioned, there are three significant sources of sea surface water freshening in the high latitudes of the NA: increased precipitation, melting sea ice, and melting land-based ice sheets (e.g., the Greenland Ice Sheet). The first two are capable of having some impact on the AMOC, yet they are far less than required for severe AMOC reductions or reversal. Although we can find tracks of these two factors in climate and ocean records for the past couple of centuries, they are unlikely to be of any sufficient magnitude soon, so we can only expect some possibility of an AMOC slowdown caused by precipitation and sea ice melting in the future. Similarly, major deglaciation of Greenland is unlikely, although significant ice loss can potentially hamper AMOC operations. In any case, the freshening of the subarctic NA, including possible freshwater from melting Greenland ice, yields AMOC’s slowdown by 0.46 Sv per decade since 1950 in a computer simulation using an eddy-permitting climate model [117].

4.4. Paleoceanographic Modeling of Ocean Circulation

Researchers have examined past climate events over the last 600,000 years without direct oceanographic proof that indicates the significant freshening of the high-latitude NA. They focused on large-scale deglaciation periods following glacial terminations, during which massive iceberg clusters detached from the Laurentide ice sheet. These substantial iceberg releases into the NA may have dramatically altered the AMOC and potentially reversed it. Examination of proxy data suggests that thermohaline circulation in glacial NA has indeed experienced dramatic alterations (for example, Refs. [123,124,125,126,127,128,129]). Moreover, diverse evidence indicates that the ocean circulation system has fluctuated between two or three primary states, or “modes,” with transitions between these modes occurring “abruptly” on a geological timescale [130,131,132,133,134]. Freshwater influxes in the northern latitudes of the NA substantially influenced the current circulation patterns in this basin. Some recent events, such as the “great salinity anomaly” (e.g., Dickson, 1988 #6540, Ref. [135]) can give a glimpse of what massive freshwater events of the past could look like on much grander scales.

As mentioned previously in the discussion of early climate models, it has been found that the Younger Dryas event, which occurred approximately 12,900–11,700 years before the present (BP), is particularly fascinating among Heinrich events. This event draws interest because of its timing between two swift warming periods, the Bølling–Allerød Interstadial and the Dansgaard-Oeschger (D-O) events, and thus it poses an apparent link to possible severe AMOC change. Scientists have observed a trend where the Northern Hemisphere and NA cooling coincided with Southern Hemisphere warming and vice versa. This observation gave rise to the “bipolar seesaw” concept. Although no single researcher coined the term bipolar seesaw, it evolved from a collective understanding of the phenomenon, beginning with a paper by [136], which popularized the term. This phenomenon has been discussed in numerous research publications and reviews [137,138,139,140,141]. Stocker [140] further advanced the thermal bipolar ocean seesaw hypothesis by using the simplest possible thermodynamic model to explain the relationship between DO and Antarctic Isotope Maxima (AIM) events. The proxy data record that inspired the thermal seesaw phenomenon and a schematic drawing of the concept of a bipolar seesaw are shown in Figure 9. Greenland ice core records reveal abrupt DO events with temperature variations of up to 10 °C between colder (Greenland stadial) and warmer (Greenland interstadial) climates (see the review in [138]).

After recognizing the oscillating nature of significant glacial–deglacial periods, numerous numerical studies of varying complexity conducted in the late 20th century validated the existence of multiple stable ocean circulation patterns. These studies also confirmed that alterations in high-latitude freshwater fluxes revealed in sea surface salinity SSS play a crucial role in triggering shifts between these patterns (for example, Refs. [142,143,144,145,146,147,148,149,150,151,152]).

Most modeling studies suggest that significant, swift climate changes are triggered by fluctuations in NADW formation within the northern NA and Norwegian–Greenland Seas regions. Consequently, most numerical studies have simulated freshwater inputs to alter the AMOC behavior by impeding NADW formation in these areas. These studies manipulated freshwater fluxes to replicate meltwater events, but required more direct connections to proxies to validate the freshwater variations used in their simulations.

The question remains as to whether iceberg discharge in the subpolar NA could have been the primary factor behind past AMOC disruptions and whether current or imminent iceberg calving might lead to similar disturbances. Heinrich events, when great armadas of icebergs episodically flooded the NA, provide a natural occurrence of “hosing” simulated in ocean climate models, and can provide some clues to answer this question. Zhou and McManus [153] reconstructed these ice discharges to be as high as 0.13 Sv during Heinrich event 4 and to average to 0.029 Sv over all episodes. The authors argued that iceberg calving likely did not persist long enough for icebergs alone to cause catastrophic disruption of the AMOC. Could ice mass loss from the Greenland Ice Sheet caused by climate warming disrupt large-scale ocean circulation? To answer this question, the authors found that the present-day Greenland Ice Sheet calving rates were as high as those during some of those events. However, because melting causes the Greenland Ice Sheet to recede from the coast of Greenland, where icebergs originate, iceberg discharge should not persist long enough to instigate major disruption of the AMOC as the sole cause.

Seidov et al. [87] used a different approach by employing salinity reconstructions derived from paleo-proxy data and an ocean circulation model driven by reconstructed wind stress, temperature, and salinity attributed to the LGM and a meltwater event (MWE) of 14,200–13,200 years BP. They reconstructed LGM salinity by combining various proxy datasets, including [126,154,155] for areas north of 40° N and modern data from [156] for regions south of 40° N, ensuring a gradual transition between low and high latitudes, as shown in Figure 10.

During the MWE timeframe, reconstructions by [155] were superimposed at the sea surface temperature (SST) and SSS onto the LGM data between 50° and 80° N. CLIMAP reconstructions [157] were utilized for LGM surface temperatures at low- and mid-latitudes, as this dataset remains widely accepted. In the northeastern study area, above 50° N, reconstructions from [126,155,158] were employed. The present-day wind from [159] and historical wind stress fields, calculated at the Max Plank Institute in Hamburg [160,161], were adapted from the same atmospheric circulation model instead of using modern climatological wind stress data. Additional information on this numerical study comparing the AMOC under modern, LGM, and MWE conditions can be found in [87]. This study is emphasized because it marked the first attempt to use the paleo-reconstruction of fields with severe salinity. It employed the so-called restoring boundary conditions for sea surface temperature and salinity in areas where the AMOC is most vulnerable to freshwater impacts.

Seidov and colleagues [87] demonstrated that, with the reconstructed sea surface conditions, the overturning strength and northward heat transport of the AMOC decreased significantly in the LGM and MWE scenarios (Figure 10b–d). Consequently, NADW production declined by 30% during the LGM and nearly stopped during the MWE. Notably, during the LGM, the sea surface in the GINS and Irminger Seas was considerably colder, with a somewhat lower SSS. During the MWE, the sea surface was notably warmer and the SSS was substantially lower than the present-day conditions. The LGM AMOC was significantly affected by SST, SSS, and wind stress. The LGM wind stress was far more zonal, with the North Atlantic Current in the upper arm of the AMOC extension spreading into the GINS and Irminger Sea regions being limited in comparison to its present-day extension. Consequently, the strength of the LGM AMOC was reduced by only 30%. However, when the same wind stress was applied in the model and the SSS significantly decreased, the MWE AMOC reversed. Using paleoclimate reconstructions of sea surface conditions, this study demonstrated that freshwater influx causing realistic MWE sea-surface-layer dilution (as evidenced in the reconstructed SSS) could lead to AMOC collapse or even reversal.

With advancements in computing power, climate, and ocean modeling, complex numerical simulations have been conducted to examine how the AMOC responds to freshwater influxes at high latitudes. Notably, some studies focused on global climate change, including the effects of increased atmospheric greenhouse gas concentrations, have demonstrated corresponding changes in the AMOC, which are considered crucial to the worldwide ocean conveyor system, as discussed in, e.g., Refs. [27,162,163,164,165,166].

4.5. Coupled Ocean–Atmosphere Models

With the evolution of coupled ocean–atmosphere models, scientists have begun to better understand ocean climate dynamics. These advanced models can now estimate freshwater fluxes internally, thereby eliminating the need for arbitrary external forcing. The findings from these simulations corroborated the fluctuations in AMOC strength observed in earlier ocean-only or less sophisticated climate models. These variations are influenced by oceanic processes and atmospheric conditions, and salinity anomalies play a pivotal role in driving these changes. This was demonstrated in studies such as [167,168,169] and others. Research by [168], who utilized a comprehensive ocean–atmosphere model, revealed that the AMOC is susceptible to irregular oscillations on a timescale of approximately 50 years. These irregular fluctuations appear to be triggered by density anomalies in areas where the water sinks. The AMOC exhibited sensitivity to density changes, primarily influenced by salinity, in deep convection zones, even without significant alterations in the freshwater input.

Using a coupled climate model, it has been shown that the AMOC maintains its vigor and functionality unless significant density reductions occur, typically owing to extreme freshening or warming [88,101]. When coupled models are forced to introduce freshwater fluxes into these areas, they generally demonstrate AMOC behavior similar to theoretical major meltwater events from past geological eras. Manabe and Stouffer [88,101] examined how a coupled ocean–atmosphere model responded to freshwater input in the NA that simulated the Younger Dryas event. They conducted two surface hosing experiments: one discharging freshwater into the high-latitude NA for over 500 years, and the other discharging freshwater into the subtropical NA. The northern discharge, simulating meltwater from numerous icebergs, significantly weakened the AMOC and substantially lowered the surface air temperatures across northern NA, Greenland, and parts of the Arctic. After the 500-year discharge period ended, the AMOC recovered to its original intensity within a few centuries. The subtropical NA hosing experiment showed a similar but less intense AMOC weakening than in the high-latitude NA experiment. This difference is understandable, because density depends more on temperature than salinity and becomes more sensitive to salinity changes at high latitudes, where temperatures are much lower than in the subtropics.

The northern hosing experiment, which simulated AMOC weakening, showed a notable decrease in the north NA SST, owing to the reduced northward movement of warm, salty surface water. The cooling peak observed south of Greenland was attributed to the enhanced flow of cold Arctic surface water via the East Greenland Current.

Research by [101] revealed that, immediately following the onset of freshwater discharge, both the SSS and SST experienced rapid declines and began fluctuating over periods ranging from several decades to a century. Similarly to findings from specific ocean-only models, this coupled ocean–atmosphere model highlights the crucial role of freshwater discharge location in shaping climate dynamics and its potential effects on future climate shifts. More model simulations of freshwater impacts have confirmed that, when the freshwater impact is sufficiently strong, significant changes in the AMOC, including cessation or reversal, could have occurred. When freshwater discharge stops, most models show AMOC recovery, e.g., Refs. [88,101,170,171].

Stouffer and co-authors [120] examined the modeled reaction of thermohaline circulation to hypothetical freshwater disturbances as part of the World Climate Research Program (WCRP) Coupled Model Intercomparison Project/Paleo-Modeling Intercomparison Project (CMIP/PMIP). This study involved more than a dozen institutions, providing climate models of varying complexity that were used to simulate climate responses to freshwater inputs of different intensities in northern NA. When a mild freshwater input of 0.1 Sv (1 Sv = 10⁶ m³ s⁻¹) was introduced, the average AMOC weakening across all models was approximately 30% 100 years after the start of the input. That is, all models demonstrated some AMOC reduction for minor impacts, but none predicted a complete cessation. In contrast, when a strong freshwater input of 1 Sv was applied, the AMOC rapidly ceased in all the model simulations. A notable cooling effect was observed in the NA region. The models showed some disagreement regarding the reversibility of AMOC following its shutdown. Some models predicted rapid recovery with complete AMOC restoration, whereas others showed an overshooting effect after freshwater input stopped.

These authors [120] also indicated that a 1 Sv freshwater influx is highly improbable under current conditions, such as in most simulations of the ongoing CO₂ increase. However, this freshwater hosing scenario might apply to the Last Glacial Termination, as evidenced by the modeling of the Younger Dryas and subsequent meltwater-type events [88,101]. It became evident that even moderate freshwater inputs could lead to substantial AMOC deceleration. This realization prompted further investigation into the sensitivity of the AMOC to freshwater influences in a warmer climate. These modeling efforts have persisted through the initial quarter of the 21st century, and may continue in the upcoming decades.

Two key questions emerge regarding the sensitivity of the AMOC to density changes in the northern NA and Nordic Seas. The first concerns the origin of sufficient freshwater to halt NADW formation, while the second concerns whether high-latitude sea surface warming could also impede this process. As previously discussed, paleoclimate proxies suggest that meltwater events during glacial–interglacial cycles can provide ample freshwater. No substantial proxy evidence indicates the direct sensitivity of the AMOC to warming events, except for freshwater release. Therefore, determining whether ongoing global warming can slow the AMOC directly or indirectly through sea ice melting or Greenland iceberg discharge has become crucial for modern climate scientists attempting to predict future climate changes, including those related to AMOC functioning. Given that greenhouse gases, particularly CO₂, are universally viewed as potent direct causes of global warming, whether natural or anthropogenic, researchers have frequently employed coupled ocean–atmosphere models to investigate climate responses to rising atmospheric CO₂ levels, placing the AMOC at the forefront of scientific inquiry.

According to coupled model simulations, e.g., Refs. [167,169], doubling and quadrupling CO₂ concentrations would increase the temperature by 3.5 °C and 7 °C, respectively. In the scenario with doubled CO₂, the AMOC experienced some weakening, but maintained functionality similar to the current conditions. However, in the quadrupled CO₂ scenario, the AMOC transitioned to a new stable state, where it completely stopped, resulting in a deeper thermocline and significantly reduced thermocline ventilation.

Numerous simulation studies have explored the vulnerability of the AMOC, with some notable examples, including [27,162,166,172,173,174,175,176,177,178,179,180,181,182], among many others.

4.6. Climate Models

Most climate models suggest that the AMOC weakens as the climate warms, mainly because of sea-surface warming. However, they do not predict a fast and severe slowdown, complete stoppage, or reversal of the ocean conveyor in this century. However, collapse may occur in the longer run. In some estimates, the typical AMOC collapse time can reach 100 years [183]. A variety of fully coupled models have demonstrated tipping points in the AMOC by conducting experiments over longer runs (for example, see the results using the NASA GISS model [184]). While simulations of past climates indicated significant climate fluctuations within glacial cycles, primarily due to large meltwater releases from extensive ice sheet disintegration in northern North America, current global warming simulations forecast modest climate change and moderate AMOC deceleration. Furthermore, most models simulating warmer climates show that when CO₂ levels increase to four times the current concentration (with a 1% annual rise in CO₂ for 140 years), the AMOC strength decreases by 10–50% of its current level [162]. Hankel [185] contended that the significance of CO₂ change is not solely determined by its magnitude; rather, the rapid rate of change is a critical factor contributing to the weakening of the AMOC. The term “rampant rate” refers to a metric that indicates the speed at which the concentration of a substance changes over time, specifically whether its intensity increases or decreases. The author of [185] asserts that the AMOC experiences a more pronounced weakening at higher rampant rates.

In some global warming modeling experiments, the shutdowns of NADW occur in the absence of exogenous freshwater input, which is maintained for 600–1000 years, and, after the overshooting, the overturning rapidly returns to its initial state. This situation did not change when only a small amount of freshwater (0.1 Sv) was used. The AMOC was further weakened, but did not shut down, which agrees with the experiments described in [120] (note that 0.1 Sv over a couple of centuries is a reasonable amount of meltwater expected in a warming climate).

To reiterate, extreme sea surface freshening can theoretically result in complete shutdown of the AMOC. Some proxies suggest that such shutdowns might have occurred in geological history, especially during the glacial–interglacial intervals of the Pleistocene. On the other hand, surface water warming typically leads to mild or moderate AMOC deceleration. When moderate freshening occurs alongside sea surface warming, it may further decrease sea surface density and intensify AMOC slowdown, e.g., Ref. [117], but this would not cause a complete halt or reversal. The diminished northward oceanic heat transport results in the cooling of high-latitude waters. This cooling effect may overcompensate for freshening, constraining the thermally induced AMOC slowdown. The reduced northward salt transport by the AMOC enhanced the deceleration initiated by initial freshening. The interaction between the thermal and haline forces could generate fluctuations in the AMOC intensity and the corresponding heat and salt transport, which would be superimposed on the AMOC slowdown trend during continuous warming.

In a significant new study conducted by Vanderborght and colleagues [186], salt advection feedback was identified as a crucial factor that may precipitate the collapse of the AMOC. The authors contend that the stability of the AMOC is closely linked to the freshwater budget of the Atlantic Ocean, with the salt advection mechanism playing a predominant role in destabilizing the AMOC. A critical indicator of this potential instability is the sign of the overturning freshwater transport F_ovS across the latitude circle at 34° S. If F_ovS is negative, the AMOC may become unstable and potentially collapse. Salt advection feedback has been earlier identified in simple box models, such as [187], and in certain coupled ocean–atmosphere models, for instance [183] (see also [188,189]). A significant new argument presented in [186] is that the pivotal element of AMOC stability is salt advection feedback, which is determined by the direction of freshwater transport across a specific latitude in the South Atlantic Ocean.

Mecking and colleagues [190] highlight the significance of FovS in maintaining the stability of the AMOC, noting that many model studies with surface salinity biases suggest the potential for a bi-stable AMOC rather than a mono-stable one. In scenarios where FovS is excessively positive, which can occur in certain modes, a larger freshwater influence is necessary to destabilize the AMOC. However, when these biases are corrected, the AMOC may become more prone to instability.

Continuous warming and ice melting have complex effects on AMOC dynamics through meltwater distribution. Madan and co-workers [166] contended that melting Arctic sea ice is the primary driver of the freshwater impact from CO₂ increase. This meltwater flows into the Labrador and Nordic Seas, spreading across the subpolar gyre. This perspective challenges the notion that freshwater caps convective regions directly and impedes NADW production. Instead, the influx of meltwater into the subpolar gyre leads to its suppression, which in turn slows down AMOC. Madan and colleagues [166] highlighted that increased freshwater discharge into the Labrador Sea precedes AMOC weakening rather than directly capping NADW formation sites.

Similarly, Nobre et al. [177] proposed that melting sea ice in the Arctic initiates initial AMOC reduction, accompanied by decreases in the northward salt and heat fluxes. This aligns with the explanation provided in [166]. The northward heat and salt transport fluctuations can trigger the slowdown and recovery cycles of the AMOC. Nobre et al. [177] further suggested that the breakdown of stable stratification in the fresher yet colder surface waters of NA contributes to AMOC recovery after its initial alteration.

Curtis and Fedorov [178] investigated the evolution of the AMOC and its recovery timeline following a complete breakdown in numerical simulations using various CO₂ scenarios ranging from 0.5$\times$ to 16$\times$ CO₂. They aimed to better understand the gradual progression of the AMOC towards equilibrium after collapse. In a 2$\times$ CO₂ scenario, this process requires approximately 2000 years. For the 8$\times$ CO₂ scenario, the AMOC required over 10,000 years to reach a new equilibrium state after a sudden CO₂ increase. This study demonstrated that Arctic and subpolar NA freshening can impede or potentially stop AMOC recovery. The study also revealed that, even after radiative balance is achieved at the atmosphere’s upper boundary, oceanic temperature, salinity, and AMOC continue to change for millennia.

As discussed in the next section, evidence of the weakening of the AMOC can be observed in the weakening and northward shift in GS. Caesar et al. [182] identified this pattern in a high-resolution climate model responding to rising atmospheric CO₂ levels, and observed century-long temperature trends.

Bellomo and Mehling [191] conducted a study that demonstrated the intricate relationship between the AMOC and climate patterns. Their findings indicate that these patterns are influenced by the warming of sea surface temperatures and a corresponding increase in salinity at higher latitudes. This salinity boost could enable the AMOC to sustain its intensity, despite the reduction in the sea surface density caused by warming. Figure 11 shows the results of the various experiments. In the coupled climate model with quadrupled CO₂ levels, the AMOC strength declined (red curve), while the other three experiments, which involved the same CO₂ increase but with overcompensated and moderately positive salinity influx, showed different outcomes. The results are noteworthy: the additional salinity allowed for the AMOC to continue functioning as if it were unaffected by the warm water resulting from a greenhouse event.

Overall, climate models have seen remarkable advancements in recent years, leading to numerous new findings through the use of high-resolution climate models that were out of reach only a couple of decades ago. Several recent high-resolution climate simulations have specifically employed, such as the widely acknowledged Community Earth System Model version 2 (CESM2), for instance [192,193,194,195,196], to name just one. Given the extensive body of research utilizing these analogous climate models, it is impractical to cite every study. Therefore, only a few are highlighted in this review to exemplify the substantial efforts dedicated to investigating AMOC and NA circulation variability in recent years using high-resolution computer simulations.

4.7. Inter-Basin Connections

AMOC changes were not local NA events. Global thermohaline circulation depends on AMOC functionality. Many studies have focused on how the inter-basin connection would change following AMOC alterations, e.g., Refs. [75,76,78,197,198,199,200,201].

The global conveyor depends on many factors, of which the salinity despair between the Atlantic and Pacific Oceans plays the most crucial role. Seidov and Haupt [75] pointed to freshwater teleconnections and ocean thermohaline circulation as the main factors in building and maintaining the SSS disparity between these two oceans, and thus building and maintaining a global thermohaline conveyor. Freshwater transport over Central America is one of the two processes responsible for the far higher SSS in the subtropical NA Ocean than that in the Pacific Ocean. Freshwater evaporates in the subtropical NA and is transported by atmospheric flow to the subtropical Pacific Ocean in the “Atmospheric Bridge” [76]. Another source of the high SSS pool in the NA is the so-called “Agulhas Leakage”, which exports salty Indian Ocean waters into the Atlantic [76] (see below in the text). A schematic of the Atmospheric Bridge and Agulhas Leakage is shown in Figure 12.

Seidov and Haupt [75] examined the critical point of the SSS imbalance between the northern Atlantic and Pacific, which could initiate a significant shift in the global deep ocean conveyor. They determined that this asymmetry was essential for sustaining a global ocean conveyor. Their computational studies indicated that, while high-latitude freshwater influences are essential for the AMOC, inter-basin freshwater exchanges might be more potent in modifying global ocean thermohaline circulation.

Building on this work, Seidov and Haupt [200] suggested that a minor difference in freshwater distribution between the Northern Hemisphere oceans, specifically the northern Atlantic–Pacific Atmospheric Bridge illustrated in Figure 12, could be adequate for establishing and maintaining a global conveyor-like ocean thermohaline circulation. Using a highly simplified initial distribution of surface temperature, SSS, and wind stress, they demonstrated that the global conveyor could be generated and sustained solely by SSS differences resulting from the Atmospheric Bridge, referring to this simplified global thermohaline circulation as a “minimalist conveyor.”

Research by [76] addressed the resilience of the global ocean thermohaline circulation in light of the two crucial mechanisms depicted in Figure 12. These processes maintain elevated salinity in the subtropical NA and comparatively lower salinity in the North Pacific through the above-mentioned Atmospheric Bridge and the “Agulhas Leakage,” which transports salty Indian Ocean water into the Atlantic. Their findings indicated that the likelihood of an irreversible collapse of the global conveyor would increase fivefold if the Agulhas Leakage dominated over the Atmospheric Bridge in maintaining high subtropical NA salinity. Nevertheless, they recognized that when the Atmospheric Bridge is dominant, the primary process might be the influx of freshwater to the Atlantic region associated with overturning circulation. The key takeaway is that the high subtropical salinity must be sustained by both excess evaporation over precipitation and Agulhas Leakage for the global conveyor to operate in its current mode. At the same time, the Atmospheric Bridge maintains the Atlantic–Pacific SSS difference. It is worth mentioning that numerous researchers consider Agulhas Leakage as a vital factor in sustaining AMOC operations, as evidenced in studies by [202,203,204,205,206]. Nuber and colleagues argued that the high salinity of the Indian Ocean, through the Agulhas Leakage, could be a significant driver of AMOC recovery during the deglacial periods of the last 1.2 million years [207].

Furthermore, as mentioned in the Introduction, Ref. [77] found that redistributing the internal freshwater through storm patterns could substantially affect the subpolar SSS distribution. Nonetheless, its impact on global conveyors is less pronounced (Figure 6). Reagan et al. (Ref. [77]) argued that additional freshening of the high latitudes of the NA caused by storm tracks carrying freshwater from the subtropics and forming a “short circuit” of freshwater within the NA can be an essential additional factor affecting the AMOC.

Other modeling studies have uncovered further evidence of interactions between ocean basins through water channels and atmospheric influences. Numerical experiments conducted by [208] suggest that alterations in the AMOC can trigger climate fluctuations between the NA and North Pacific regions. Simulations with a sealed Bering Strait demonstrated that introducing extra freshwater into the NA weakened the AMOC, but strengthened the Pacific Meridional Overturning Circulation (PMOC). This change in the NA and North Pacific Ocean relative strength of meridional overturning results in a cooler NA, owing to decreased northward heat transport by the AMOC. At the same time, the North Pacific warms as the PMOC carries warmer water from tropical and subtropical areas to the subpolar North Pacific. Research by [209], utilizing a comprehensive coupled climate model, revealed that this Pacific–Atlantic seesaw effect linked to AMOC changes only occurs when the Bering Strait is closed, which corresponds to glacial periods when sea levels are low because of water trapped in continental ice sheets. Strait closure interrupts oceanic communication between the North Pacific and the Atlantic. Consequently, when the AMOC collapses, the NA experiences cooling, whereas the North Pacific warms because of PMOC intensification. The enhanced PMOC transported warmer saline subtropical water into the North Pacific, leading to contrasting climate shifts in these two ocean basins.

Regarding the Arctic freshwater throughflow from the North Pacific to the NA, Ref. [210] showed that, as the AMOC weakens because of freshwater in the northern NA, the Bering Strait throughflow weakens and can even reverse its direction. Weakening of the throughflow leads to reduced exports of Arctic freshwater into northern NA. It can even be the case that the Arctic receives freshwater from northern NA and helps the AMOC recover relatively quickly after the end of freshwater hosing.

Research conducted by [211] revealed that the breakdown of the AMOC can strengthen Pacific trade winds and Walker circulation by generating excess heat in the tropical South Atlantic. This tropical warming leads to unusual atmospheric convection, which increases subsidence over the eastern Pacific and amplifies the Walker circulation and trade winds. The study also identified additional long-reaching effects, including weakening of the Indian Ocean and South Atlantic subtropical high-pressure systems and intensification of the Amundsen Sea Low.

4.8. Northern and Southern Hemisphere Interplay

The concept of the AMOC as a universal force behind global thermohaline circulation, both in its present state and under different climate scenarios, prompts inquiries into how freshwater impacts the conveyor in the northern and southern hemispheres, as well as the competitive dynamics between the two. Stommel’s concept of abyssal circulation suggests that deepwater formation occurs at high latitudes of the NA and convection sites in the Southern Ocean near Antarctica’s shelf. As discussed above, freshwater flux can halt deepwater formation in the NA, sometimes severely impacting the AMOC and the global thermohaline conveyor. The possibility of similar disruption in the Southern Ocean remains a topic of inquiry.

Seidov and Haupt [212] studied the climate’s reaction to a hypothetical freshwater influx into the Southern Ocean by employing a coupled atmosphere–ocean model. Their findings revealed that surface waters surrounding Antarctica became fresher and cooler in response to this input. These studies were juxtaposed with freshwater addition experiments using ocean-only simulations in both hemispheres [75,212]. Due to a fundamental distinction in altering sea surface salinity (SSS) from the northern and southern hemispheres, an oceanic bi-polar seesaw fails to materialize in the ocean. Controlled ocean-only experiments with mixed boundary conditions and comparable short-term southern freshwater effects were aligned with coupled experimental results. These findings suggest that the concept of ocean bi-polar seesaws should be employed with caution.

Stouffer and co-authors [201] delved deeper into this issue by employing a coupled ocean–atmosphere model for their research. They simulated the addition of 1 Sv of freshwater to the ocean surface for over 100 model years, followed by its removal. This study examined two cases: one in which freshwater was introduced to the Atlantic Ocean between 50° and 70° N, and another where it was added south of 60° S in the Southern Ocean. Their findings revealed that the freshening of the subpolar NA surface waters weakened the AMOC and its associated northward oceanic heat transport. Conversely, when Antarctic surface waters were freshened, the AMOC remained relatively stable, with only a minor weakening observed near the end of the integration. This slight weakening was attributed to a fresh sea surface anomaly spreading from the Southern Ocean to other regions of the World Ocean. The study proposed two potential mechanisms for AMOC weakening: first, a reduction in the SSS difference between the NA and North Pacific; and second, the prevention of surface water sinking in the high latitudes of NA due to the Southern Ocean freshwater impact, resulting in a weakening of the entire thermohaline circulation.

Over the past decade, the AMOC theme has been further examined from various perspectives in numerous modeling studies. Berglund and others [213] discovered a novel aspect of circulation within the NA Subtropical Gyre (NASG) linked to the upper arm of the AMOC in the NA. Their research revealed that up to 70% of the northward-flowing AMOC upper layer water circulated at least once in the NASG before continuing its northward journey. These circulations are essential to increasing water density through air–sea interactions and interior mixing, allowing for it to escape the NASG latitudes and join the northern upper branch of the AMOC towards convection sites. Consequently, alterations in NASG dynamics could potentially impact AMOC pathways and strengths.

Kostov et al. [214] investigated the links between the Labrador Sea and subtropical part of the AMOC. They emphasized the importance of the North Atlantic Current for north–south connectivity in the AMOC and meridional transport of Lower NADW (LNADW).

In a recent study [215], it was shown that the AMOC is affected by freshwater discharge from both the Greenland and Antarctic ice sheets and, as an interhemispheric teleconnection bridge, exacerbates the opposing ice sheet’s retreat via the bi-polar seesaw. These results highlight the key role of ice-sheet–climate interactions via freshwater flux in future ice sheet retreats and associated sea-level rise.

Lin et al. [164] discussed AMOC weakening through subsurface warming in the Labrador Sea. In response to CO₂ increase, surface warming was mixed with the deeper Labrador Sea in the models with more vigorous upper ocean mixing. This subsurface warming and the corresponding decrease in density drive AMOC weakening through advection from the Labrador Sea to the subtropics via the Deep Western Boundary Current. Time series analysis shows that most CMIP6 models agree that a decrease in subsurface Labrador Sea density leads to weakening of the AMOC in the subtropics by several years. In addition, idealized experiments conducted in an ocean-only model show that subsurface warming over 500–1500 m in the Labrador Sea leads to pronounced AMOC weakening several years later, whereas warming that is too shallow (<500 m) or too deep (>1500 m) in the Labrador Sea causes insignificant AMOC weakening.

Wind stress over the NA, in tandem with buoyancy forces (which also depend on the wind stress via Ekman pumping), is the leading factor in surface circulation. Lohmann and co-authors [216] modeled the response of the AMOC to reduced or enhanced wind stress forcing and found that, under reduced wind stress forcing, the AMOC’s strength decreased significantly. In contrast, under enhanced wind stress forcing, the AMOC’s strength increased only in the first decades and then decreased and returned to a state close to the reference simulations with the present-day wind stress strength.

The primary concern of all modeling efforts is how oceanic northward heat transport is facilitated by the AMOC changes with AMOC fluctuations. The current knowledge of oceanic heat transport is well established. Many publications on heat transport estimates largely agree with each other and with the results of the present-day climate models. One often-cited estimate is based on atmospheric re-analysis [217]. The implied oceanic heat transport was calculated as the residual between the total energy balance at the top of the atmosphere and computed atmospheric heat transport. The analysis indicated that atmospheric transport contributed a larger portion of the required poleward transport than oceanic transport, accounting for 78% in the Northern Hemisphere and 92% in the Southern Hemisphere. The implied ocean heat transport from the global ocean and three oceans is shown in Figure 13 as an example of the general shape and intensity of the estimated present-day oceanic heat transport [217]. These estimates mostly agree with others, for example, Ref. [218], and model results, for example, Refs. [219,220].

The striking feature in Figure 13 is that the Atlantic Ocean heat transport in both re-analyses is near 0.5 PW across the equator—a feature often referred to as the Atlantic Ocean “heat piracy” [221]. An equivalent amount of northward water transport must compensate for the southward NADW outflow into the Southern Hemisphere in the deep ocean in either the upper or intermediate layers. The return upper ocean flow provides positive northward heat transport across the equator. This so-called NA oceanic heat piracy [221] offers sufficient warmth to maintain the current level of NADW formation by cooling warm and salty surface water to make them sink, thus keeping the present-day world ocean in an interhemispheric heat balance that is positive in the NA. Heat piracy depends on the AMOC state. During a glacial period, for example, at the LGM, NA heat piracy was reduced as the AMOC weakened, and at some points, it even stalled. During meltwater events, such as the MWE of 14,200–13,200 years BP, it may even have the opposite sign, i.e., becomes South Atlantic piracy (see Figure 10 and Refs. [87,221]). The NA heat piracy concept is schematically illustrated for these three periods (present-day, LGM, and MWE) in Figure 14 (based on Ref. [221]).

The structure and dynamics of the AMOC in the models depended on the resolution of the numerical grid. Hirschi and others [222] reviewed new insights into AMOC provided by high-resolution models. The latest generation of ocean and climate models now allow for resolutions sufficient to simulate ocean circulation that explicitly includes mesoscale eddy motion, for example, Refs. [223,224,225]; more studies are reviewed in [222]. High-resolution simulations employing high-resolution oceans have revealed marked improvements in modeled climate change accuracy, e.g., Refs. [226,227,228].

Moreover, opinions on the essential deep convection sites for AMOC dynamics and intensity arising in computer simulations may differ; data analyses also differ in this area. Some authors argue that convection in the Labrador Sea is the leading factor in AMOC variability, with the Labrador Sea Water playing an important role in models and data analyses, for example, Refs. [61,164,229,230], while others say that the key locations are in the Irminger Sea, for example, Refs. [63,231,232].

4.9. AMOC Modus Operandi

The details of the locations of the deepwater formations are not the main topic of this paper. Instead, we focus on whether models, data analyses, and re-analysis studies can confidently predict the speed and intensity of further slowdown of the AMOC. The critical outcome of the ocean circulation models, stand-alone or as part of climate models, implies that the AMOC in the past, at least in the last glacial cycle, could, in principle, exist in three basic modes: (a) intense (i.e., present-day), (b) glacial-weakened, or (c) where meltwater ceased to exist (or even reversed overturning), as shown in Figure 15.

It is important to note that the glacial-weakened mode resembles the “warm world” regime of developing global warming, featuring a warmer upper AMOC branch and increased meltwater in Nordic Seas and other deep convection areas. The “warm world” regime is distinguished by moderate freshening in the high-latitude NA and maintains AMOC functionality. In contrast, the meltwater mode is characterized by an extremely weak or non-existent AMOC, owing to extensive high-latitude freshening.

It is crucial to understand that, although Figure 15 offers a basic diagram and might seem oversimplified, recent studies indicate that during the LGM (as depicted in the central panel of Figure 15), the volume of AABW was considerably greater than it is today [233]. This increase occurred because the volume of NADW decreased, allowing for the AABW to occupy a much larger portion of the deep ocean. Extending this idea to the potential collapse of the AMOC during meltwater events, the volume of the AABW can expand even further, as illustrated in the lower panel of Figure 15.

When assessing modeling efforts to simulate the AMOC’s historical, current, and future operations, it is important to note that the main factor contributing to the AMOC’s decline is most likely the increasing freshwater content in the surface waters of the high-latitude NA and Nordic Seas. Most paleoceanographic simulations agree that during excessive interglacial meltwater events in these areas, the AMOC was prone to enter an inactive state characterized by minimal or even reversed flow, e.g., Refs. [23,87,91,120,221,234,235]. However, there is no firm agreement regarding the fate of the AMOC under the ongoing global warming trend. Although practically all models confirm the AMOC’s slowdown, some predict a substantial decline in the AMOC (for example, Weijer et al. predicted substantial AMOC deceleration in the CMIP6 model [236]), whereas others argue that the slowdown will be moderate. Only a few models predict the complete collapse of the AMOC in a warming world, usually when prescribing a steep warming trend, for example, with less realistic quadrupling, compared to the more realistic doubling, of the CO₂ concentration in the atmosphere.

To emphasize most model results, Intergovernmental Panel on Climate Change (IPCC)’ assessments, based on the Climate Model Intercomparison Project (CMIP) model simulations, suggest that, although the AMOC is expected to slow over the coming centuries, its complete collapse by the end of this century is unlikely [237]. This conclusion is supported by IPCC’s prediction that the Gulf Stream, the primary component of the AMOC’s upper arm in the NA, is not anticipated to change significantly. Consequently, although the AMOC is expected to continue slowing, it is not projected to break down completely, as illustrated in Figure 16 (from [237]).

Caution should be exercised regarding the terminology used to describe the potential “collapse” of the AMOC. The term “collapse” itself is somewhat ambiguous. Does it imply a complete cessation of AMOC activity, or could a significantly reduced but still non-zero AMOC still be classified as a collapse? If the latter is true, then the interpretation of the right panel of Figure 16 becomes contentious, as it depicts a scenario of markedly diminished overturning, which might be construed by some as a “collapse.” Regrettably, the current state of climate modeling does not offer a consensus on this matter, and a precise definition of what constitutes an AMOC collapse remains elusive. A recent study by Baker and co-authors [238] indicated that the AMOC continues to operate in a “conveyor-on” mode even during severe climate events, demonstrating resilience to extreme greenhouse gas-induced warming and significant freshwater influx. This finding was based on the results of an extensive array of 34 climate models. Unexpectedly, the winds in the Southern Ocean acted as stabilizing forces, preventing the collapse of the AMOC. Notably, as elaborated further in the text, wind stress and the resulting Ekman pumping in the NA can also function as stabilizing factors to maintain the current operation of the AMOC.

As anticipated, the stability of the GS is pivotal for operational functioning of the AMOC in a warming climate. This is evident in Figure 16 of the IPCC report [237]. However, upon closer examination of the observational analyses, it became evident that the GS does not exhibit the dramatic changes depicted in Figure 16, particularly over the past six decades. Instead, it demonstrates the remarkable resilience of the GS-to-sea-surface warming in the NA [48]. Nevertheless, the changes shown in Figure 16 cannot be entirely dismissed in the coming decades.

4.10. Searching for Consensus

To evaluate the current state-of-the-art in AMOC prediction, it is essential to conduct a thorough analysis of numerous model simulations to identify a consensus. However, it is also important to consider outliers (it is well known that a majority’s opinion, as history has spectacularly demonstrated, is not necessarily true). Within this approach, it is important to note that while the IPCC’s assessment indicates that a complete collapse of the AMOC is unlikely by the end of the century, certain researchers have suggested that AMOC shutdown is possible in a persistently warming climate. In a recent paper [183], the reaction of the AMOC to high-latitude freshening was examined by employing a modified approach to freshwater hosing to determine how the AMOC nears its tipping point. Using a version of the Community Earth System Model, they demonstrated AMOC’s approach to its initial tipping point by incrementally increasing freshwater input, revealing the significant climate impact of AMOC flipping (a model description can be found in [239,240]). These researchers contended that a physical indicator exists to provide an early warning of the AMOC near its tipping point, precisely, the critical value of freshwater transport changes in the upper 500 m at 34° S.

This early warning signal was detected in the model within the South Atlantic rather than in the NA. They caution that the AMOC may be approaching its tipping point, given its weakening since 1950, and is now at its lowest strength in the millennium. However, after gradual freshwater hosing, the final freshwater input appears unrealistic (this can be caused by model biases; correction for some of those biases can yield a more sensitive AMOC [241]). The amount of meltwater required to push the AMOC beyond its tipping point seems significantly higher than what is plausible, considering the rates of meltwater injection in Northern NA, even in a warm climate with quadrupled CO₂ levels. This may be compared to the first meltwater event of approximately 14,200–13,200 14C years B.P., known as the D-O-related Meltwater Phase 1A. For example, Kim et al. [242] called this type of AMOC approaching to a tipping point a “slow and soft passage through a tipping point.” As mentioned above, an ocean-only model based on proxy salinity data from the Northern NA and Nordic Seas revealed AMOC flipping during this meltwater event following glacial termination [87].

A new warning regarding a potential tipping point in the AMOC’s climate trajectory was issued in a recent study by [243]. This study employed a statistical analysis of the AMOC’s fingerprint to forecast the timing of a tipping point without relying on ocean circulation or climate models. The authors predicted that the collapse of the AMOC could occur between 2025 and 2095, implying an imminent event by the end of this century. They argued that early warning signs indicate that the AMOC is approaching a tipping point, as the system approaches a critical parameter linked to freshwater signals in the NASG. These signs include heightened variance (indicating reduced resilience) and increased autocorrelation (signifying the critical slowing of the AMOC system). These statistical markers indicate the system’s proximity to a tipping point, with significant trends in the mean, variance, and autocorrelation of the AMOC indices, indicating a potential shutdown. While many studies, including [237], only predict a slowdown of the AMOC without an imminent collapse in the 21st century, Ref. [243] make their prognosis highly confident that the AMOC will reach a tipping point around the mid-century. Nevertheless, the predictions in [243] have faced some criticism, notably from Ben-Yami and colleagues [244], as well as from van Westen et al. [183], who contended that the results might overstate the potential collapse of the AMOC around the mid-century due to the methodology used.

Figure 17 presents the simplest schematic representation of the feedback loops within the AMOC system to summarize the key findings from the ocean and climate model evaluations. Extensive research has identified these processes using computational models to examine the AMOC dynamics and their effects on climate patterns.

The AMOC feedback loops shown in Figure 17 demonstrate that both negative and positive feedback mechanisms affect stability. The negative feedback process begins with enhanced heat transport to the Arctic, accelerating the ice melting in the sea. This phenomenon decreases the surface salinity and density of NA waters at high latitudes, thereby weakening the NADW formation. As a result, the diminished NADW formation causes additional freshwater to flow from the Arctic into the Nordic Seas, further reducing salinity and density. This chain reaction inhibits deep convection in crucial areas, such as the Labrador and Irminger Seas, ultimately weakening AMOC intensity and decreasing the poleward heat transport. This self-sustaining cycle contributes to a slower circulation pattern, as shown in the right half of Figure 17.

In counterbalance, reducing heat transport towards the poles initiates a positive feedback cycle, leading to decreased ice melting in the Arctic region. Fewer freshwater flows out of the Arctic, enabling the surface salinity and density in the subpolar NA to bounce back (left half of the diagram in Figure 17). The increased surface density promotes deep convection in the Labrador and Irminger Seas and strengthens NADW formation in the Nordic Seas. As this formation intensifies, the AMOC becomes stronger, restoring poleward heat transport and further accelerating the process. This self-reinforcing mechanism contributes to the robustness of the AMOC and its ability to sustain a strong circulation under favorable conditions.

These regulatory mechanisms highlight the sensitive balance among freshwater input, salinity levels, and heat transfer in maintaining AMOC operation. The intricate interplay between these elements determines whether the AMOC remains robust or shifts to a weakened condition, which has significant implications for global climate regulation.

Most models suggest a gradual decline in the strength of the AMOC without reaching its lowest point in the coming decades, at least by the end of this century [243]. This indicates that the AMOC is only in the initial stages of deceleration, which may or may not culminate in complete breakdown. Metaphorically, the current position of the AMOC is approximately at the beginning of the slowdown curve depicted in Figure 17. Nevertheless, certain model studies point towards a more imminent collapse of the AMOC, such as those by [183,243].

To make the situation even more confusing, the latest examination of 24 Earth System Models from the Coupled Model Intercomparison Project Phase 6 (CMIP6) revealed a contrasting scenario [245]. This study contends that these temperature anomalies are unreliable indicators for AMOC reconstruction. Terhaar et al. [245] discovered a strong correlation between air and sea heat flux anomalies north of any given latitude in the NA (between 26.5° N and 50° N) and the AMOC anomaly on an annual scale. Notably, these air–sea heat flux anomalies were primarily influenced by atmospheric variability rather than the AMOC itself. The researchers asserted that, despite considerable variability across all latitudes, the decadally averaged AMOC at 26.5° N did not show a weakening trend from 1963 to 2017. While some models suggest that the AMOC may not experience a notable decline for several decades [245], other recent model reconstructions indicate a gradual weakening of the AMOC over extended periods, as opposed to the findings of shorter-term models or observational studies, as highlighted in subsequent research [55,56].

4.11. Model Conclusions

A review of the modeling efforts spanning over five decades that aimed toward understanding AMOC variability and its tendencies to strengthen or weaken over extended periods has yielded a broad range of outcomes. While these studies have provided important insights into the behavior of the AMOC on a centennial and longer timescale, they have failed to definitively answer the most critical question: What is the fate of the AMOC in the coming decades, particularly by the end of the 21st century? The only conclusive finding is that the freshwater regime of the subpolar NA and Nordic Seas plays a pivotal role in significantly weakening the AMOC. The models consistently reconstructed climate change during glaciation and deglaciation cycles, showing that massive meltwater influx into the subpolar NA almost certainly led to substantial AMOC degradation (followed by recovery). However, model estimates become confusingly variable when dealing with more subtle signals of present-day changes in the thermal and haline structures and dynamics.

Although modeling approaches offer valuable advances in understanding the functioning and maintenance mechanisms of the AMOC, they lack the natural constraints of observational data and analysis. Unfortunately, the current volume of comprehensive gridded data spanning centuries or more is insufficient to provide meaningful spatiotemporal coverage of at least the NA suitable for a century-long AMOC retrospective study. However, we already have access to several decades of direct measurements and ocean climate reconstructions, which can either support or challenge the assessments and predictions derived from modeling efforts.

5. AMOC Monitoring, In Situ Observations, and Ocean Data Analyses

5.1. Monitoring and Data Analyses Versus Models

Numerical modeling has been the primary source of our knowledge of ocean circulation and its effects on climate for decades. The fundamentals of thermohaline- and wind-driven ocean circulation were established in the 1960s and the 1970s, respectively. However, the creation of detailed maps depicting ocean circulation and thermohaline structure remained elusive until the late 1980s, and was hampered by insufficient observational data coverage and model resolution.

This situation has transformed with the advent of Argo profiling float instrumentation and satellite oceanographic observations, which have enabled the more precise mapping of ocean thermohaline and its broad-scale dynamics. This progress coincided with enhancements in the spatial resolution of the ocean models, leading to significant advancements in the field. However, a practical approach to ocean climate diagnostics is to use the data obtained in the last sixty-plus years following the advent of widespread and numerous ocean observations beginning in the middle of the twentieth century, which would cover two 30-year successive periods.

Examining observations and data analyses related to AMOC variability is typically less complex than reviewing modeling studies. This is mainly because direct AMOC monitoring is more targeted, and combined ocean climate data analyses tend to produce more consistent conclusions than those derived from models. Furthermore, several comprehensive reviews of AMOC monitoring have provided detailed and in-depth analyses of field observations. Consequently, the following section offers only a brief overview of AMOC monitoring, with greater emphasis on the decadal AMOC variability inferred from ocean climatologies and other ocean data analyses. These analyses encompass the entire NA or its regions, which is crucial for understanding the AMOC dynamics on decadal and longer timescales.

5.2. Monitoring Along Basin-Wide Sections

The AMOC has already had approximately 20 years of direct observations, and there are some reviews of different specifications and levels of complexity of the analysis, for example, Refs. [19,20,246,247,248], to name just a few. The primary tactic for monitoring the AMOC is to assess transport across selected sections of the Atlantic Ocean. These sections, both in the Southern and Northern hemispheres, are shown in Figure 18a, while Figure 18b depicts two major sections in the NA.

Bryden et al. reported a 30% decline in AMOC transport at 26.5° N from 1957 to 2004. This is the first direct observational evidence of slowdown [52]. Their analysis showed a reduced southward flow of NADW and a northward transport in the GS. Rahmstorf et al. [249] combined paleo-data, sea surface temperature trends, and modeling to suggest a 15% weakening of the AMOC since the mid-20th century. Subsequently, Ref. [182] refined evidence of the weakening of the AMOC, linking it to direct observations and proxy records.

Concurrently, direct observations of the AMOC have been expanding, with increased data collection at key Atlantic Ocean sections, particularly at the RAPID and OSNAP arrays. The RAPID-MOCHA array (Rapid Climate Change-Meridional Overturning Circulation and Heatflux Array), positioned at 26.5° N, was established in 2004, but its data became accessible around 2008 [247,250]. The array continuously provided detailed measurements of AMOC fluctuations, revealing significant annual and seasonal variations. However, no distinct long-term patterns were observed. Furthermore, contrary to claims of AMOC decline, an analysis of the 30-year record from the RAPID-MOCHA section did not indicate any decline [251].

The OSNAP array was established to monitor the heat, mass, and freshwater flowing continuously across the subpolar NA basin. This observational system comprises two segments: one stretching from southern Labrador to southwestern Greenland, and another from southeastern Greenland to Scotland [252,253]. A recent study examining the Deep Western Boundary Current (DWBC) in SPNA using OSNAP data provided further evidence of AMOC stability [254]. Notably, despite a significant 26% reduction in DWBC transport since 2014, the AMOC has maintained relative stability, challenging the notion of its decline. On the other hand, different studies have insisted that there has been a substantial slowdown of the AMOC since 2004, but, in the last decade, the decline has reached a plateau, for example, Ref. [251]. This study reconstructed the AMOC from 1981 to 2016 and found that, while there was a significant weakening between 2004 and 2012, the overall time series did not show a long-term decline (see also [255]). Observational evidence shows that increasing the atmospheric CO₂ concentration affects the NA heat fluxes and precipitation rate and weakens the AMOC, which is consistent with numerical simulations. The inferred weakening, starting in the late 19th century, earlier than previously suggested, is estimated at 3.7 ± 1.0 Sv over the 1854–2016 period, more significant than previously estimated. Lee et al. argued that anthropogenic factors may have been essential for the most recent slowdown in the late 20th century [255].

In ongoing discussions, a significant observation emerges from comparing the RAPID section monitoring at 26° N with data analysis from the OSNAP array. Initially, when the RAPID array was deployed, the overturning was generally considered uninterrupted. It was anticipated that changes in overturning measured at one latitude would correspond to those observed at other latitudes. However, subsequent modeling studies and comparisons between the RAPID and OSNAP arrays have led to a revised understanding. Current assumptions suggest that AMOC variations on interannual timescales are only coherent across limited north–south distances [253]. In this regard, Ref. [256] suggested that the observed lack of meridional coherence in the AMOC could be caused by Ekman pumping, which influenced the vertical transport of the AMOC in the upper 1 km.

In addition to common knowledge about the Nordic and Labrador Seas, which play a major role in AMOC dynamics, the eastern subpolar NA has recently been recognized as another key driver of the AMOC, generating a significant portion of the NADW. Some researchers, for example, Refs. [66,257,258], have highlighted the importance of the eastern and central-eastern regions in AMOC dynamics. Sarafanov et al. [258] reported that the contribution to the AMOC lower limb at 59.5° N from the overturning of upper ocean waters in the Irminger Sea basin, south of the Greenland–Scotland Ridge, exceeds that of the Nordic Sea overflows by one-and-a-half times. Moreover, the AMOC variability reconstructed using the Irminger Sea density revealed a strong NA Oscillation influence on subpolar overturning across various timescales [65]. Despite the prominent role of the Irminger Sea, the Nordic Sea overflow remains crucial for AMOC dynamics, for example, Ref. [259]. In essence, the Irminger and Nordic Seas are now believed to be key to the sustainability of the AMOC.

5.3. AMOC Tipping Point

According to [20], the AMOC at 26° N, where the RAPID array was located, exhibited a downward trend from 2004 to 2012. However, this decline was not continuous, which is consistent with the subsequent analysis by Lozier [253] (mentioned earlier). The key observation is that AMOC variability encompasses both periods of decrease and increase, rather than solely decline or stability. Notably, the AMOC at 26° N experienced a significant reduction of approximately 30% between 2009 and 2010. Nevertheless, the period after April 2008 remained relatively constant, with an average transport only approximately 2.7 Sv lower than that observed between April 2004 and April 2008 [260]. Furthermore, Lee et al. [255] suggested that the interaction between anthropogenic and natural signals affecting the AMOC may explain the pause in its slowdown after the early 2010s. They proposed that the natural component of the AMOC intensified owing to the emergence of a strong positive NA Oscillation, counteracting anthropogenically driven weakening.

Unlike some recent studies, Ref. [182] researched the AMOC dynamics by combining a high-resolution model and observing long-term temperature trends. Their findings indicate a substantial weakening of the AMOC by approximately 15% since the mid-20th century. Although this decrease is not as severe as that in [52], it still represents a notable reduction in AMOC strength. The authors attributed this weakening, which they believe is likely anthropogenic, to cooling in the subpolar gyre area and warming in the GS region. These changes could potentially affect the European weather patterns, sea levels, and droughts. As global warming continues, it is expected to further diminish the AMOC, possibly pushing it toward a critical threshold. The cooling of the subpolar gyre is viewed as an indicator of AMOC weakening, and the ongoing decrease in AMOC intensity is seen as approaching this tipping point. It is important to acknowledge once more the presence of a significant feedback mechanism, specifically salt advection feedback, which may play a crucial role in regulating the AMOC tipping point [186].

According to a review analysis by [261], studies involving brief freshwater introductions suggest that numerous climate models operate in a mono-stable state, placing them relatively distant from the tipping point. This does not negate the existence of a tipping point or bi-stable regime in these models; rather, it indicates that they are not currently in such a state because of their present climate (likely inaccurate). Research by Dima et al. [262], who examined worldwide sea surface temperature shifts since the nineteenth century, indicated a weakening of the AMOC and global conveyor belt since the late 1930s, with a sudden change in the overturning pattern occurring around 1970. Additional research verified that AMOC deceleration results in subpolar NA cooling, initially termed the “warming hole” [263] and later referred to as the “cold blob.” As illustrated in Figure 19, this cold blob is now recognized as a marker of AMOC slowdown [182].

A complete collapse of the AMOC significantly expanded the cold blob, resulting in a chilled airflow over the subpolar NA. This would cause severe, long-lasting cooling across Northern Europe and trigger numerous unforeseen effects. While the AMOC is just one of several potential tipping points, it is by far the most powerful, as highlighted by the authors of The Global Tipping Points Report [264]. According to these authors, the AMOC remains the most significant threat, despite other turning points.

In light of the discrepancies between certain models and field analyses, we find ourselves at a critical juncture, where we must consider two perspectives: the IPCC’s assertion that an AMOC collapse is unlikely, and persistent concerns about the potential for AMOC to reach a tipping point and collapse. Although the exact middle ground remains uncertain, it is important to recognize that most conclusions and arguments stem from either models or the AMOC dynamics observed in specific sections. These localized datasets may not fully capture comprehensive AMOC evolution patterns. A more thorough assessment can be achieved by examining the broader picture and utilizing all available ocean climate data to reconstruct ocean climate change across the entire NA.

5.4. Analyses of Gridded Ocean Climatologies

In recent decades, temperature and salinity data coverage of the World Ocean has vastly expanded. This improvement has allowed for the reconstruction of these essential parameters on a regular grid that covers nearly the entire ocean. The reconstruction features a minimum one-degree spatial resolution across over hundred depth levels, extending from the surface to a depth of 5 km. For example, comprehensive information is available in the NOAA WOA [265,266].

The WOA now offers an even higher global spatial resolution of one quarter of a degree. While data in most areas do not fully support this high-resolution coverage, several key regions have sufficient data to reconstruct the ocean climate on decadal and longer timescales at a one-quarter-degree resolution. In some areas, reconstruction is possible at one-tenth of a degree, matching the resolution of eddy-permitting and eddy-resolving ocean and climate models. The NA is one such region where this level of detail is achievable. As a result, recent research has focused on tracking NA’s ocean climate trajectory of the NA over these extended periods.

Seidov et al. examined approximately 60 years of observations as ten-year climate datasets for in situ temperature in the NA [31,267], utilizing temperature with a quarter-degree resolution from the WOA 2013 [268]. Their research concentrated on temperature and ocean heat content (OHC), comparing two 30-year climate periods. OHC, which indicates the heat stored in oceans, shows the most significant variability in the NA, e.g., Ref. [11]. Many studies have investigated OHC changes, a decisive indicator of ocean warming or cooling, using various databases, including [269,270,271,272,273,274] before [31]. However, Ref. [31] differed in analyzing 30-year OHC climates and their comparisons, thus showing ocean climate change based on its definition. This aligns with the WMO’s definition of climate [1], which is most applicable to oceans, the most inertial component of Earth’s climate system over decades and centuries.

Figure 20 shows the 30-year climate shift as OHC differences between 1985–2012 and 1955–1984 time intervals in the 0–300, 0–700, 300–700, and 700–2000 m depth layers. For better visualization and interpretation, the OHC maps in Figure 20 represent the “volumetric density” of the OHC rather than the actual OHC values. The OHC volumetric density is simply the areal density divided by the depth (or thickness) of the analyzed layer.

In the NA, most regions experienced warming in the top 300 m of the ocean (Figure 20a). Exceptions were found in the central subpolar gyre and the easternmost Labrador Sea, where cooling occurred in the 0–300 m layer. This localized cooling is not evident when examining global ocean warming trends [11]. Although previous studies, such as [275], have identified this cooling, Ref. [31] provided a more comprehensive analysis of true climate change, defined as two 30-year averages. High-resolution grid mapping of OHC reveals a detailed structure of the OHC climate shift, while also showing a coherent pattern of OHC change concentrated in a narrow band of high positive values along the GS and its extension. The observed warming–cooling pattern in the NA subpolar gyre aligns with the concept of a cold blob or warm hole, suggesting a significant reduction in AMOC strength and its deceleration.

It is important to acknowledge that, in the study of AMOC trends, the time intervals encompassed by data and model simulations are crucial, as records shorter than a decade offer limited value for such research. For instance, Lobelle and colleagues [276] determined that the 14-year RAPID series represents the lower-limit interval, and that a time window of at least twice as long (between the medians of 24 and 43 years) is necessary for the detectability of AMOC trends.

5.5. Gulf Stream Resilience

The GS, which forms the upper branch of the AMOC along the coast of North America, is thought to affect the temperature variations in the subpolar area. This prompts the question of whether GS alterations are significant enough to be observed as a reduction in AMOC intensity, and, more crucially, as shifts over a decade. Consequently, understanding the long-term variability of the GS path is essential for understanding the role of the ocean as a climate influencer. Seidov et al. [48] examined the decadal fluctuations of the GS north wall (GSNW), defined by the 15 °C isotherm at a 200 m depth. This isotherm is regarded as the primary indicator of the Gulf Stream’s route, as noted by several researchers [277,278,279,280,281]. The results show the remarkable resilience of the GS over the entire 53 years of tracing, using the WOD 2013 data. Figure 21 illustrates how coherent and stable the GS remains within the longitudes from 75° W and 45° W—that is, its main jet before it veers into more dispersed “spaghetti” east of the Gran Banks.

Speculating whether GSNW migration may be a factor in AMOC fluctuations is enticing, especially given the AMOC monitoring along the RAPID array (see above). Nevertheless, research conducted by [48] suggested that linking GSNW position shifts to AMOC decadal variability is challenging. To comprehend this connection, it is key to explore how GSNW and AMOC decadal variability might be interconnected. Considering the notable stability of the GSNW, Ref. [48] proposed that the influence of GS on the AMOC over decadal and longer periods could result from (a) significant decadal fluctuations in GS volume transport within a rigid and resilient jet between 75° W and 50° W (although some authors question this possibility), (b) meandering of the GS extension and the North Atlantic Current east of 50° W, or (c) a combination of both. Specifically, if the AMOC’s decadal variability, which is primarily governed by ocean–atmosphere interactions in the NA subpolar gyre (as generally explained in the texts by [282,283]), is indeed affected by GS decadal changes, it can only occur through the mechanisms mentioned above, rather than through decadal variations in the GSNW position, which we have demonstrated to be relatively minor in the robust zone. However, the latter may become more significant in the extension zone, although proving this remains challenging without direct measurements of AMOC variability in this area, which has yet to be conducted.

The OHC metric is one of the most important indicators of the ocean’s impact on climate change. While OHC generally aligns with temperature patterns, NA has experienced the most significant OHC shifts over the past 50 years, e.g., Refs. [11,275]. Researchers have extensively examined OHC variations [269,270,271,272,273,274], derived projections from climate models [269,284], and evaluated them through re-analysis efforts [92,285,286,287].

5.6. Ocean Heat Content and Eighteen Degree Water

Eighteen-degree water (EDW) represents another crucial aspect of heat and salt accumulation in the GS region, with its volume increasing because of subtropical water heaving, e.g., Refs. [45,288,289,290,291]. This heaving phenomenon, likely resulting from changes in long-term wind stress patterns [275,292], was observed during the analysis of NA climate change using WOA13 data at 1/4° × 1/4° grid resolution [267,292]. The lower EDW boundary extended for approximately 100–120 m to a deeper depth below and southeast of the main GS core (see Figure 8 in [267]). When analyzing decadal changes in the EWD volume in NA, its ~60-year variations, from a ~20% decrease in 1965–1974 to a ~15% increase in 1995–2004 and 2005–2017, are illustrated in Figure 22 [293]. As subtropical water heaving is considered a potential cause of localized heat trapping over multidecadal periods, it is vital to map the geographical distribution of these heat-trapping areas using in situ climatological data. Figure 23 (reproduced from [267]) indicates that the Sargasso Sea, located southeast of the path of the Gulf Stream in the EDW region, experiences the highest heat and salt gain rates.

The maximum heat gains below 200 m southeast of the GS extension could be attributed to the vertical shift in the isopycnal surfaces that make up EDW heaving (e.g., Refs. [294,295]). This increase in warm and salty water east of the GS may help maintain the jet by preserving the density gradients across it, even if the downstream dynamics have changed. However, its role in GS resilience remains uncertain, and has not yet been confirmed using data analysis models. The hypothesis that a combination of warm subtropical water heaving and AMOC slowdown might be responsible for excessive heat accumulation in the GS system, and that this accumulation could be patchy because of the high mesoscale dynamics in the area [48,267], cannot be decisively proven or disproven. Nevertheless, this presents a plausible explanation that is worthy of further investigation.

5.7. Wind Stress Impact on AMOC

The connection between the OHS and EDW oscillations in relation to wind stress curl (WSC) and the resulting Ekman pumping is the key element that may be one of the most significant factors affecting AMOC stability. Figure 24a displays an approximate line indicating where WSC is equal to zero. As noted by Mishonov et al. [36], this zero-WSC position remains relatively constant over time and space, as illustrated by the Hofmuller diagrams in Figure 24b. These diagrams show only a slight north–south shift in the zero WSC position at longitudes of 60°, 40°, and 20° W from 1980 to 2019. Figure 24c depicts the WSC variations during the 2005–2017 and 1985–1994 decades.

Some authors, such as [188], contend that wind-driven transport is ineffective in altering the salinity across the entire Atlantic Ocean. However, our data analysis suggests that WSC regulates the dynamics of the gyres through Ekman pumping, thereby indirectly influencing the temperature and salinity structures within the gyres. This process may ultimately impact the AMOC’s functionality. Notably, salt advection, as proposed by Vanderborght et al. [186], may serve as a more effective mechanism for controlling salinity. Nonetheless, the wind stress undoubtedly contributed to the overall process.

The WSC determines upward or downward Ekman pumping areas, which are the vertical movements at the Ekman mixed-layer’s base. The negative WSC regions experienced downward pumping, whereas the positive WSC areas underwent upward pumping or suction. The subtropical (anticyclonic) gyre primarily experienced downward Ekman pumping, whereas the subpolar (cyclonic) gyre experienced upward pumping (see Figure 4 for a schematic representation). The WSC = 0 line roughly separates these two gyres (Figure 24a). WSC differences highlight areas where pumping, either upward or downward, strengthened or weakened between 2005–2017 and 1985–1994 (Figure 24c).

Given that the WSC structure suggests a strong correlation between the upper arm of the AMOC (stretching from southwest to northeast) and the WSC zero line over the NA, Ref. [36] linked NA dynamics to the overall WSC structure and Ekman pumping geometry induced by the WSC. They contended that, because the wind-generated circulation gyre geometry remained relatively stable over an extended period, wind component variability can be disregarded when examining ocean circulation changes driven by surface thermohaline transformations. As previously discussed, the most influential thermohaline transformation affecting AMOC variability is the freshwater regime shift in the NA subpolar gyre and Nordic Seas.

An additional effect of Ekman pumping on AMOC has recently been discussed. It has been argued that Ekman pumping can have an impact on the vertical structure of the AMOC and, thus, its meridional coherence. As mentioned above, according to [256], the observed lack of meridional coherence in the AMOC (see above) can be attributed to Ekman pumping, which influences vertical transport across the 1000 m depth surface. Although this review did not further examine this aspect, it could be beneficial for interpreting certain elements of AMOC monitoring.

5.8. Decadal Averages of the In Situ Data

The NA Ocean experiences nonuniform warming across space and time [31,296]. Based on the apparent geometric structure of the WSC illustrated in Figure 24a, Ref. [36] segmented the NA into five distinct areas: (1) the Western NA associated with the GS region, (2) the Eastern Subtropical NA, (3) the Western Subpolar NA associated mainly with the Labrador Sea, (4) the Eastern Subpolar NA, and (5) the Nordic Seas (refer to Figure 1). These five regions exhibit fundamentally different processes, necessitating separate analyses.

Mishonov and co-authors [36] employed two techniques to compute ocean current velocities. The first approach utilized the traditional dynamic method, assuming a stationary reference level at 1500 m depth for the entire 1955–2017 timeframe, using WOA18 seawater density data [265,266,297]. The second technique, applied to the 1985–2017 period, calculated the current velocity without the stationary depth assumption. Instead, it referenced the surface current velocity derived from sea surface height (SSH) data obtained from the Simple Ocean Data Assimilation re-analysis (SODA) version 3.4.2 (http://www.soda.umd.edu) [92] (last accessed on 31 August 2025). As the SSH data in SODA were unavailable before 1980, only the last three decades have been free from stationary assumption constraints.

Consequently, these three decades alone have allowed for a comparison between the two velocity calculation methods, with the SSH-referenced dynamic method proving to be significantly more precise. Using these velocities, Mishonov et al. [36] created maps of currents and their kinetic energy between the surface and 1500 m depth and estimated water transport across various sections in different regions (boxes 1–5, as mentioned earlier). In other words, they mapped and analyzed the upper arm of the AMOC for over 60 years.

The eastern subpolar region of the NA and Nordic Seas (boxes 4 and 5 in the nomenclature from [36]) are likely to be the key factors influencing AMOC variability. Given that the salinity and density of the ocean’s near-surface layer are decisive for the AMOC, it is essential to examine Figure 25, which illustrates the decadal fluctuations of these two parameters in these specific areas from 1955 to 2017 at different depth layers. Surprisingly, salinity in these critical regions has not decreased since 2005, as expected in the AMOC, supposedly a slowdown phase found in a number of studies (see above). In contrast, it increased noticeably in the eastern subpolar NA and the Nordic Seas. However, unexpectedly, despite the increase in salinity, the density in both regions decreased, which could only occur because the sea surface temperature increased. The latter, in turn, might occur because of an amplified inflow of warm and salty water of GS origin, that is, increased intensity of the AMOC or a northward shift in GS and the North Atlantic Current, as argued in [48,49]. As mentioned above, the wind stress variability was very minor and, therefore, is excluded from the analysis; therefore, the only reason for the changes in these two regions is the thermohaline variability caused and sustained by the upper ocean circulation.

5.9. Decadal Variability of Water Transport in Upper Arm of the AMOC

The complexity of the total water transport throughout the NA increases when the upper arm of the AMOC enters the Irminger Sea on its way to the Nordic Seas. Contrary to expectations, temperature and salinity anomalies indicate the increased presence of warmer and saltier water in regions traditionally associated with freshwater inflow from the Arctic, particularly at convection sites within the Nordic Seas and the eastern subpolar NA. This phenomenon might be attributed to the density reduction caused by warmer water, despite its higher salinity. Both regions exhibited significant negative density anomalies, suggesting diminished NADW formation and a slower AMOC, as shown in Figure 26.

Utilizing SODA SSH to compute the reference velocity at the surface offers a more reliable estimate of water transport than the dynamic method with a depth of no motion. Water transport across 50° N indicates an AMOC upper-arm decline that predates the onset of the 21st century, possibly even beginning as early as the decade of 1985–1994 (and possibly even earlier).

To summarize the analysis in [36], it can be stated that none of the climate parameters—temperature, salinity, density, transport, or kinetic energy differences—between the decades suggest dramatic changes in the NA climate over the past 60 years. The only differences between the AMOC changes and temperature (salinity and density) are as follows: (i) the evident slowdown of the upper arm of the AMOC manifested in weakening the northward water transport within the upper 1500 m across 50° N in the last decade (Figure 26); and (ii) continued warming throughout the NA, except in the western part of the subpolar gyre.

In line with a recent analysis by [49], the present circumstances may not fully represent future outcomes. An alternative perspective proposes that, rather than freshwater influx in the Nordic Seas and subpolar NA causing the AMOC slowdown, the AMOC might not yet be experiencing a genuine slowdown due to freshwater input. Instead, it could decelerate because warmer surface water is transported northeastward by the upper branch of the AMOC. This implies that, if we are to observe an actual AMOC slowdown comparable to or reminiscent of past deglacial meltwater events, the meltwater phase may still be forthcoming.

Importantly, caution is warranted when considering the future of NA circulation and climate with regard to the AMOC dynamics. From 2005 to 2017, an unprecedented northward movement of warm surface GS waters during the summer months was observed in the data analysis [49], potentially indicating a decline in the resilience of the GS system. It remains unclear whether this trend will continue, intensify, or diminish in the coming years. Although they may seem improbable, scenarios involving a significant slowdown or collapse of the AMOC, as proposed in [243,298] and other publications (see above), cannot be entirely ruled out. Furthermore, it is impossible to state with absolute certainty that sea surface temperature and density trends will persistently move towards warmer and less dense surface ocean water in the foreseeable future. Regardless of these trends, the fate of AMOC remains unclear. Nevertheless, the above analysis suggests that the current NA Ocean circulation and climate maintain relative stability, despite the observed surface warming and potential recent slowing of the AMOC.

To reiterate, resilience of the GS is not guaranteed to continue in the near or distant future. However, what was noticed in [49] is that the Slope Water, the region between the North Wall of the GS and the North American coastline, has been warming at an increasing rate. None of the AMOC fingerprints—temperature, salinity, and density of the upper 1500 m—indicate that the NA climate trajectory will remain on the current upward slope, and may drift toward even warmer and lighter upper ocean water. If this happens, the AMOC may slow down, even without a massive freshwater incursion from the north to the subpolar NA.

5.10. Ocean Climate Data Analyses

Bashmachnikov et al. [299] investigated the interannual variability in the deep convective intensity in the Greenland Sea from 1993 to 2016. All metrics show that the intensity of convection increased during the 2000s. This does not agree with the results discussed in [36], who argued that, despite the salinity in the eastern subpolar NA and Nordic Seas, the warmer surface water in the last 30 years has yielded less dense surface water, thus decreasing NADW formation and resulting in AMOC slowing. At the same time, Ref. [299] demonstrated that the increased deep convective intensity in the Greenland Sea is associated with increased upper ocean salinity. The authors suggested that concurrent variations in oceanic heat release to the atmosphere effectively abate the variability in advected heat. However, Ref. [36] showed that water temperature in the region increased in the 2000s (Figure 25a,b). The discrepancy between the two analyses may be explained by interannual variability, which has superseded the overall decadal variability.

Recently, Ref. [300] found that the thermocline in the eastern subpolar gyre separating the deep and upper water shallows by as much as 200 m from 2008 to 2018. They attributed this change to a reduction in the northward flow of upper waters into the subpolar gyre. However, such a change is also consistent with the convergence of deep flows filling the subpolar reservoir of deepwaters (see the discussion in [300]). According to [36], such shoaling can result from warmer, albeit saltier, water in the eastern subpolar NA, and therefore lessened convection due to negative-density anomalies (Figure 26).

Chafik et al. argued that enhancing the NA current and associated meridional heat transport increases the subtropical heat influx into the eastern subpolar NA [301]. They suggested that the recent warming phase since 2016 is primarily associated with the observed mechanism of changes in the deep western boundary density, an essential element in these interactions.

In a recent review, Jackson et al. synthesized an understanding of the decadal variability in the AMOC derived from observations, ocean re-analysis, forced models, and proxies [26]. They concluded that there is evidence of AMOC strengthening and weakening after the 1980s; however, the magnitude of these changes remains uncertain. In the subpolar NA, the AMOC strengthened until the mid-1990s and weakened until the early 2010s, as shown in Figure 26. They also acknowledge, as we do, the possibility of subtle evidence suggesting a future AMOC strengthening. However, in the subtropics, they found evidence of AMOC strengthening from 2001 to 2005 and strong evidence of weakening from 2005 to 2014. As many other authors have noted, Ref. [26] emphasized that significant interannual and decadal variability complicates the detection of ongoing long-term trends, but does not negate the alleged weakening associated with anthropogenic warming.

5.11. AMOC Fingerprints and Stability Assessments

To better understand the seemingly controversial results of AMOC weakening in times of warmer but saltier water in the eastern subpolar NA, while the logic requires fresher (and perhaps cooler) water to slow the AMOC, Figure 27 from [302] presents the schematics of such a mechanism. Focusing on the major shift in the climate model HadGEM3-HH within the model time interval between 2010 and 2040, they analyzed near-surface freshwater pathways. Figure 27 summarizes the drivers of the changes in upper ocean heat and freshwater budgets, T-S variability, and stratification in this model, contrasting conditions in the 1990s and the 2040s. Lagrangian analysis in [302] revealed that, by 2040, high-salinity inflow to the subpolar gyre from subtropical latitudes has declined relative to fresh inflow from higher latitudes, lowering SSS. Meanwhile, surface net heat fluxes tended to be positive from 1990 to the 2040s, increasing SST. The consequence is reduced surface density and stronger stratification. As the stratification strengthened, the surface density decreased, and the stratification strengthened.

To address the cold blob and AMOC fingerprint issue, Ref. [182] identified a pattern of cooling in the subpolar Atlantic Ocean and warming in the GS area as indicators of AMOC change. The OHC 30-year change in the upper 300 m (Figure 20a) shows significant cooling in the western subpolar NA. However, Figure 20b, illustrating the eastern subpolar NA, primarily displays positive 30-year temperature anomalies of the OHC in the upper 300 m layer, with intensified warming in the Nordic Seas (evident in Figure 25). This creates a conundrum: the cold blob in the western subpolar NA coinciding with GS region warming suggests an AMOC slowdown but simultaneous warming in the eastern subpolar NA and Nordic Seas concurrent with GS warming and AMOC deceleration presents a conflicting scenario. The dominant perspective holds that the primary AMOC drivers are in the Irminger Sea and, to a lesser degree, Nordic Seas, with deep convection in the Labrador Sea playing a minor role. Conversely, the existence of a cold blob seems to indicate the AMOC’s decline. These contrasting views raise questions regarding what constitutes the AMOC’s slowdown fingerprint. Despite extensive observational and modeling efforts, this issue remains unresolved.

In the broader context of SPNA circulation and AMOC, Chafik and Lozier [303] contended that the two primary metrics, namely, the AMOC fingerprint and GYRE INDEX (a measure of the strength and variability of the NA Subpolar Gyre), are not exclusive indicators of the AMOC’s upper arm dynamics. Both metrics capture the variability in upper ocean heat content, which can arise from diverse mechanisms. Consequently, they are not individually indicative of the state of the AMOC and cannot be used to unambiguously predict its dynamics and subpolar temperature fluctuations.

Rahmstorf provided a comprehensive analysis of the potential instability of the AMOC, with particular emphasis being placed on its possible proximity to a critical threshold [261]. This study identifies two primary tipping points that threaten the AMOC: one related to salt feedback, which has been previously explored; and the other connected to convective mixing. As climate change progresses, the AMOC faces an increased risk of abrupt changes that could lead to drastic short- and long-term climate shifts, resulting in substantial ecological and societal impacts. Some scientists interpret recent observations, including the previously mentioned cold anomaly, the above-mentioned cold blob, in western NA, as indicators that the AMOC may already weaken. This review also reaffirms the prevailing perception that AMOC weakening correlates with increased freshwater input from melting ice and precipitation, emphasizing the elevated sea levels along the North American East Coast, modified weather patterns, substantial ecosystem disruptions, and the pressing need to address fossil fuel emissions. The outcomes could be severe if the AMOC reaches and surpasses the tipping point or threshold. Regional climate shifts may involve considerable cooling in NA, particularly affecting Iceland, Britain, and Scandinavia. This cooling would also intensify the temperature difference between Northern and Southern Europe (see more on the related AMOC slowdown impacts on the climate in Europe and worldwide in, for example, Refs. [183,211,304,305,306]).

The chain of events following the disruption of the AMOC upper arm would not stop there. A weakened GS would increase the coastal sea levels in northwestern NA. Another potential consequence of the AMOC breakdown could be significant shifts in tropical precipitation patterns, potentially causing droughts in the northern tropics of the Americas and Asia, with possible effects on the Afro-Asian monsoon region. Marine life could face dire consequences due to a substantial decrease in oxygen supply to the deep ocean as the overturning circulation ceases. Additionally, on a larger scale, with yet unknown impacts, European agriculture and possibly other regions might encounter significant challenges due to temperature drops and changes in rainfall patterns. As with many different discussions and modeling efforts, potential AMOC alterations are linked to global climate trends accelerated by greenhouse gas emissions [261,306]. Figure 28 shows the yearly average SST and precipitation changes compared to their current distributions, as provided by [241], after [261].

Figure 28 provides a glimpse into the potentially catastrophic effects of AMOC surpassing its critical threshold. As previously noted, neither modeling efforts nor observational studies conducted over the past five decades have yielded definitive predictions regarding our proximity to this alarming tipping point. In the subsequent discussion, we determine whether the current situation in both domains can offer a more precise outlook, at least for the upcoming decades. Additionally, based on our recent research, we explored a potential stabilizing mechanism by which wind stress might affect the AMOC system.

6. Discussion

Summarizing our examination of the accomplishments related to the significance and potential future of the AMOC, which is limited to a small subset of the vast body of AMOC-related studies, we reached an astonishing conclusion. Despite extensive observational and modeling efforts, the outcome fails to provide a definitive and unambiguous answer regarding the system’s prospects, which is undoubtedly the most sought-after information on this research topic. Model predictions for the AMOC vary widely, ranging from complete breakdown to near-total stability within the next century. However, it is undisputed whether significant climate change would ensue if the former scenario occurred. The primary and undebated cause is the possible substantial freshening of the Nordic Seas and subpolar NA.

6.1. Mechanisms Behind AMOC Weakening

The mechanism behind AMOC weakening is based on the fundamental principle that freshwater at a constant temperature is less dense than saltier water. Vertical mixing, which facilitates deep convection, ceases when it occurs, disrupting the continuity of the AMOC. The two extreme states of the AMOC, on and off, depend on whether the warm and salty water arriving in the subpolar NA can be sufficiently cooled by contact with the colder atmosphere, causing the surface water to sink to great depths and to flow southward to form the lower branch of the AMOC. If fresher water intrudes at the top of warm and salty water, cooling may become insufficient to sustain deep convection.

The properly functioning meridional overturning circulation of the AMOC carries sufficient heat and sea salt northward to keep the subpolar NA warm enough to heat the colder atmosphere, as it currently does. The breakdown of the AMOC, which would disrupt these processes, could result in severe climatic and environmental impacts. From this perspective, the variations among climate models are concerning, with some predicting AMOC collapse by the end of this century. However, IPCC’s assessment, based on an evaluation of most models, suggests that AMOC collapse is improbable before the end of the century, which is somewhat reassuring despite lingering uncertainties.

Contrary to expectations, observational data and continuous AMOC monitoring have not resolved the model discrepancies. The AMOC began weakening at the turn of the century, but this trend subsequently slowed and halted, resulting in a “pause” in its decline. While recent evidence suggests that the weakening has resumed, there is no certainty that another pause will not occur or that the AMOC might not accelerate instead of slowing down.

Moreover, there is a significant disagreement in observational studies regarding which region of the subpolar Atlantic (including the Nordic Seas) is key for AMOC stability. Some authors emphasize the importance of the Irminger Sea, whereas others highlight the Nordic Seas. However, some maintain that the Labrador Sea remains relevant. Some authors, for example, Refs. [182,261], consider a cold blob in the Subpolar Atlantic as an AMOC indicator. However, Ref. [31] argued that this cold blob is more likely to be located in the western subpolar Atlantic than in the eastern part; that is, not in the Irminger Sea, where freshwater is thought to have the most significant impact on the AMOC.

Given these points, we propose an alternative hypothesis regarding the potential AMOC deceleration that complements the established view of the role of freshwater. A significant difference was evident when we compared the present NA surface temperature distribution with that reconstructed for the LGM. In modern NA, subpolar gyre isotherms slope approximately 45° relative to latitudinal circles (Figure 29a), with surface geostrophic currents roughly following the isotherm slope (Figure 29b).

The position of the 10 °C isotherm, which can serve as a diagnostic of the subpolar front position, aligns with the general direction of the North Atlantic drift, which approximately coincides with the slope of the zero line of the WSC. The 10 °C isotherm approximately divided the upward and downward Ekman pumping areas, as shown in Figure 24a. The 10 °C isotherm in winter follows the Subpolar Front position and is often used as a proxy for the boundary between the subtropical and subpolar waters. In summer, this proxy is closer to the 12–14 °C isotherms. Nevertheless, the present-day subpolar front is oriented from southwest to northeast, beginning from 50° N and ending northwest of the British Isles and continuing into the Norwegian Sea.

6.2. A Comparison of the Present-Day and Glacial-Interglacial AMOC

The AMOC has undergone significant changes since the last deglaciation, e.g., Refs. [23,235]. In the previous ice age, the LGM, specifically the subpolar NA, cooled sharply. The CLIMAP (Climate: Long-range Investigation, Mapping, and Prediction) project [157] produced detailed reconstructions of Earth’s surface temperatures during the Last Glacial Maximum (approximately 21 Kya; Kya means one thousand years ago) (see the review in [307]). These reconstructions, as well as more recent ones, demonstrate a substantial shift in the subpolar front and its much more zonal nature during the LGM [308,309].

Figure 30 shows the CLIMAP reconstruction of the Atlantic Ocean’s surface temperature compiled by the CLIMAP project members [157,310] (Figure 30a) and a more recent Glacial Atlantic Ocean Mapping (GLAMAP), for example, Refs. [309,311] (Figure 30b). Both reconstructions show that the 10 °C and warmer isotherms are mostly oriented latitudinally. The same pattern of wind stress displacement was shown by LGM climate models (see, e.g., ref. [160,312]).

The wind stress data from [160,161] were used by [87] to interpolate the wind stress to a one-degree grid in the NA. These maps show a substantially increased zonality of the wind stress over the subpolar NA, leading to an increased zonality of the subpolar front represented by the 10 °C isotherm. More recent simulations have largely confirmed the results of the above-mentioned numerical studies, for example, Ref. [313], who argued that there was a deeper and stronger subtropical NA gyre during the LGM than it is today. They suggested that the subtropical gyre, including the Gulf Stream, was deeper and stronger during the LGM than at present, which is attributed to the increased glacial wind stress curl, as supported by climate model simulations, as well as greater glacial production of denser subtropical mode waters. Similarly, Ref. [87] noted that the upper ocean wind-driven subtropical gyre during the LGM was more potent than that at present. This supports our argument that increased glacial Ekman pumping might have led to weaker overturning, as a more zonal North Atlantic Drift would provide less warm and salty water needed for deep convection. As a result, the LGM’s AMOC was weaker, yet still functioning, while meltwater inflow following deglaciation might have either entirely shut down or severely weakened the AMOC. Moreover, Ref. [314] asserted that the deglacial warming of high-latitudinal NA, between 40° N and 65° N, occurred in three discrete steps, each of which was characterized by the latitudinal orientation of the subpolar front, with the coldest glacial front attributed to the LGM oriented along the parallels, as shown in Figure 30, and the warmest with more latitudinal stretch of the front, oriented in the south–southwest to north–northeast direction.

The AMOC behavior was best modeled for the LGM and present-day period using high-latitudinal meltwater hosing to reproduce the AMOC’s response (weakening). Intense modeling efforts have been made to simulate the LGM climate and meridional overturning circulation in NA, such as the Paleoclimate Modeling Intercomparison Project (see, for example, a brief overview in [91]). Knowing the difference between the LGM and present-day AMOC dynamics and geometry, it can be argued that the upper arm of the AMOC and the associated subpolar front at present flow from southwest to northeast, compared to west–east during the LGM. It can also be concluded that the present-day AMOC is more similar to the warm period after the LGM, called Bølling–Allerød, which occurred as abrupt warming in the Northern Hemisphere between 14.5 and 12.9 Kya.

Skinner and co-authors [315] analyzed the AMOC’s change at the onset of Bølling–Allerød and argued that the AMOC resumed its present-day intensity and NADW formation increased, causing more vigorous ventilation comparable to modern NA ventilation. Based on the new AMOC index reconstruction, Ref. [316] argued that the ocean climate change in the Early Holocene (10–5 kyr BP) could be associated with the AMOC’s acceleration trend, followed by a general weakening trend from approximately 6–7 to 2 kya. They found that the Late Holocene was marked by two fluctuations, with maxima near 4.2 and 5 Kya. Thus, the migration of the subpolar front may mark the transition from the weakened mode of the LGM to the weakened mode of the mid-Holocene and then to the strengthened mode, as shown in Figure 31.

The transition between cold and warm phases in Earth’s climate can be traced through the migration of the subpolar front, as reconstructed from paleoclimate proxies [314]. The curves in Figure 31, which loosely follow the sketch in [289], illustrate the shifting position of the North Atlantic subpolar front from the LGM to the Holocene Optimum.

During the LGM (approximately 20–18 Kya), the subpolar front experienced its most pronounced southward shift, marking a prolonged cold phase. As the climate warmed, the front gradually migrated northward, reaching a transitional position during the Bølling–Allerød Interstadial period (approximately 13–11 Kya). This warming period was followed by a rapid return to colder conditions during the Younger Dryas (11–10 Kya), before another warming phase led into the Early Holocene (10–9 Kya). The final progression of the subpolar front brought it to its Holocene Climate Optimum position, with a warm phase spanning approximately 9–6 Kya.

Zahn used an analogy of an open and closed “door” to describe the warm and cold phases of the SPNA regimes since the Last Glacial Maximum (LGM) [292]. Figure 31 illustrates this analogy by depicting two positions of the subpolar front: the cold NA regime (door fully closed), representing the LGM; and the warm regime, which is warmer than today (door fully open). A sketch based on [292] visualizes the climate transitions between these two extremes after the LGM, culminating in a moderate present-day state. These cold and warm regimes are linked to various positions of the subpolar front, which may fluctuate between the climatic edges. Our current situation is one such scenario, with the AMOC functioning as it does today. The present-day subpolar front position is indicated by the thick dashed curve in Figure 31. The opening and closing of the “climate door” are depicted by thin dotted lines following a sketch in [292]. In warmer regimes, if sea ice melts rapidly during door-opening phases, the AMOC may weaken or even collapse.

Wind-driven changes may have also played a role in these climate transitions by influencing the position of the subpolar front. Variations in wind stress and sea surface temperatures likely contributed to front migration and overall shifts in thermohaline circulation. For a more detailed analysis of the subpolar front migration during the deglaciation, see [314,317,318].

The shifts between these climate states were closely linked to changes in the AMOC. During the LGM, the AMOC was significantly weaker because of the reduced inflow of relatively warm and salty water into the ice-covered convection regions of subpolar NA. The deglaciation triggered a brief warming period around 14 Kya, marked by massive meltwater discharge (Meltwater Pulse 1A), further weakening the AMOC by freshening the surface waters of the NA.

This weakening of the AMOC contributed to the abrupt cooling event of the Younger Dryas as the cold conditions temporarily returned. However, continuous ice melting has already reduced the extent of sea ice in the subpolar NA, allowing for the gradual expansion of the surface ocean circulation. The retreat of the ice sheets modified the pressure gradients, allowing for the Azores High to expand and further stabilize the warm conditions in the NA. As the influx of meltwater decreased and saltier waters returned to the region, deepwater formation resumed, ultimately restoring the AMOC to its “conveyor on” mode.

6.3. Wind Stress as a Stabilizing Factor

In summary, we can conclude that the warm upper branch of the AMOC is controlled by two competing processes, freshwater effects in regions of deep convection and the wind stress curl (WSC), which is associated with Ekman pumping in the NA. These mechanisms work in accordance with the regulation of the upper arm of the AMOC, affecting the subpolar front intensity and position, and, as a result, controlling deepwater formation. Therefore, the position of the subpolar front, which is roughly associated with the 10 °C isotherm at the sea surface, may serve as a diagnostic of the tendencies in the upper arm of the AMOC in the subpolar NA. The WSC acted as a stabilizing force for the southwest–northeast flow, thereby providing negative feedback. Conversely, the freshwater feedback destabilizes this flow by hindering the heaving of incoming subtropical warm and salty water in the presence of freshwater. We can identify two indicators of AMOC change, the cold blob in the western subpolar NA and the WSC’s zero line, which roughly delineates the subpolar front (approximated by the 10 °C isotherm). This zero line supports northeastward flow by maintaining density gradients through opposing Ekman pumping.

Although the prevailing view suggests that a complete breakdown of the AMOC is improbable, the global thermohaline conveyor model can be adjusted to incorporate new findings on stabilization mechanisms, with wind stress acting as a counterbalancing force to maintain AMOC functionality. Through an examination of simulations from 34 climate models, Baker et al. indicated that strong wind stress over the Southern Ocean is a key factor in AMOC resilience [163]. Although they emphasized the importance of wind stress in the Southern Ocean, we believe that both Ekman pumping in the North Atlantic and wind-driven upwelling in the Southern Ocean play crucial roles in stabilizing the AMOC and preventing its collapse. Nonetheless, the analysis by Baker et al. [238] necessitates an update to the global thermohaline conveyor model, highlighting downwelling sites in the North and South Pacific. These authors noted that a global overturning system emerges in the Pacific Ocean, supplying deepwater to the Southern Ocean.

In his review of Baker et al.’s work, Hu presented a revised depiction of the global conveyor, reflecting this updated understanding of global meridional overturning without imminent collapse of the AMOC [319].

Figure 32 (reproduced from [319]) shows two modes of global conveyor operation: the current mode (upper panel); and a hypothetical mode with reduced AMOC (likely due to greenhouse-induced warming and/or freshwater impact) but without AMOC collapse (bottom panel), owing to Pacific Ocean downwelling and wind-induced upwelling in the Southern Ocean balancing AMOC reduction.

According to [238], the necessity for upwelling in the Southern Ocean to be counterbalanced by downwelling in either the Atlantic or Pacific implies that downwelling in the SPNA cannot be entirely halted (refer to [319]).

7. Conclusions

Unfortunately, despite extensive research into the variability in the AMOC across different timescales, we still lack a comprehensive explanation of its variability, and our ability to predict its fate is limited. Moreover, our inability to predict the future of the AMOC, which is closely linked to climate change even at the end of this century, is particularly frustrating. In the realm of modeling, the results of model simulations exhibit significant discrepancies, sometimes even contradicting each other. Some models suggest that the anticipated changes are modest, implying a stable AMOC and climate, albeit different from the current state, but not catastrophic. Conversely, others caution that the AMOC’s resilience may severely diminish, potentially leading to dramatic climate change in Europe, if not globally. The IPCC’s analysis offers some reassuring insights, suggesting that, based on the findings in most climate models, we should not anticipate catastrophic changes in the AMOC, at least until the end of this century. Thus, the IPCC implies that a climate disaster linked to the AMOC should not be expected to be a crucial tipping point in the near future. In other words, we can feel somewhat at ease, as it seems that we still have a long way to go to reach this critical threshold. However, the following question persists: Is it wise to rely on reassurance to relax? Indeed, the IPCC’s conclusion was drawn from the majority of the model simulations. However, as history has vividly shown, the majority might not always be correct, and sometimes the minority, or even a single perspective, can prove to be more accurate than the majority.

Although observational data provide a clearer understanding of AMOC fluctuations over the past several decades, they fail to fully address the critical question around the imminent deceleration of the AMOC in the foreseeable future. Moreover, research has revealed that the AMOC slowdown observed since 2004 is not continuous, with a significant interruption in the deceleration process identified between 2010 and 2015. This pause was not expected, based on previous trends.

Unlike the widely varying findings concerning the overall stability of the AMOC, studies on the steadiness of the GS have yielded more consistent results. It has been argued that the overall stability of the structure of the important stabilizing factor, WSC, remains unchanged [36,48,49]. The discovered resilience of the GS further supports the contention of the stability of the upper arm of the AMOC in the context of ongoing ocean warming [48]. From 1955 to 2005, the GS exhibited a property known as “stiffness,” maintaining a remarkably consistent path [33,320]. Although the annual position of the GS North Wall showed an increased latitudinal spread between 2005 and 2017, the decadal average of the GS route did not show significant changes. This suggests that the variability in thermohaline circulation is largely due to fluctuations in the density gradients maintained by the currents and Ekman pumping.

Furthermore, since 1965–1974, the subpolar gyre and Nordic Seas have consistently experienced a decline in water density. This potential reduction in deepwater formation could lead to a subsequent slowdown in the AMOC. Building on previous research [31,58,321], we argue that GS and its extension significantly influence the climate of the subtropical gyre toward resilience, rather than the instability of the NA gyres. This stabilizing effect implies that, even if the AMOC undergoes significant changes soon, its impact may not be as severe as previously anticipated. While this evidence is not decisive, it provides some reassurance, particularly considering that the WSC structure has remained relatively stable for at least three decades, without any indications of imminent change.

This stability may contribute to the relatively modest fluctuations in water transport within the GS extension. According to [36], substantial evidence supports the idea that a combination of factors, including variations in transport or kinetic energy and meandering of the GS extension, collectively influence potential AMOC slowdown.

While evidence suggests that the AMOC is not yet close to its critical point, our study of AMOC indicators, including temperature, salinity, and density in the upper 1500 m, along with WSC stability, does not necessarily imply sustained resilience of the NA Ocean climate if it transitions to a warmer, less dense upper ocean condition. It remains unclear whether a slowed upper arm of the AMOC will lead to less stable and more vulnerable NA ocean climate patterns. Furthermore, we cannot predict with certainty whether sea-surface temperature and density trends will progress towards significantly warmer and less dense surface ocean waters in the immediate future. Irrespective of these trends, the future of the AMOC remains uncertain.

Therefore, it is advisable to exercise a degree of caution. Although some studies have indicated that the current North Atlantic circulation and climate are relatively stable, it is undeniable that the AMOC is experiencing slowdown. Numerous observations and advanced high-resolution models have confirmed this finding with absolute certainty. However, at present, it is not feasible to favor either a resilient or unstable AMOC scenario within this century, yet substantial progress has already been made in addressing this critical issue. The possibility that the AMOC may eventually reach its tipping point cannot be dismissed, which urgently necessitates significant efforts to monitor and model AMOC development. It is imperative to sustain these efforts to timely detect a potentially catastrophic collapse of the AMOC in both the near and distant future.