September 2025

September 2025 | Oceanography

17

FRESHWATER MECHANISMS FOR DRIVING

ABRUPT CHANGES IN THE AMOC

Given that geologic reconstructions suggest that HS1 and the YD

were not times of accelerated melting of Northern Hemisphere

ice sheets and elevated freshwater fluxes to the North Atlantic

(Figure 3a), alternative mechanisms for AMOC weakening at

these times must be sought. The mechanisms driving AMOC

reduction at HS1 and the YD need not have been the same, and

paleoceanographic data are consistent with different magnitudes

of AMOC change at each event, with HS1 thought to be the larger

and longer reduction of the two (Ng et al., 2018).

Heinrich events were associated with massive iceberg dis-

charges from the Laurentide Ice Sheet (Ruddiman, 1977; Heinrich,

1988; Broecker, 1994; Hemming, 2004), so it is possible that fresh

water from melting icebergs played an important role. Unlike the

deglacial meltwater that enters the ocean directly in liquid form,

icebergs can travel much farther from the coast before they dis-

integrate (Fendrock et al., 2022). The paths of large icebergs orig-

inating from terrestrial ice sheets can be tracked by ice-rafted

debris (IRD), which consists of coarse grains of continental origin

that are embedded in the icebergs and deposited on the seafloor

as the icebergs melt. In general, IRD is found most prominently

in marine sediment cores collected from between ~40°N and

50°N in the Atlantic Ocean (Ruddiman, 1977); however, smaller

amounts of IRD have been found much farther north, including

in the Nordic Seas and south of Iceland (e.g., Elliot et al., 2001;

Thornalley et al., 2010), and to the south on the Bermuda Rise

(Keigwin and Boyle, 1999). Therefore, the supply of fresh water

from melting icebergs could be a mechanism for reducing the

AMOC (Broecker, 1994). This hypothesis is supported by low

δ18O values measured in fossil planktonic foraminifera (indica-

tive of low salinity) from Heinrich layers within the main IRD

belt proposed by Ruddiman (Bond et al., 1992; Hemming 2004).

However, as yet, clear evidence of low salinity farther north has

not been found.

The Younger Dryas is another period of IRD deposition in the

North Atlantic (e.g., Zhou and McManus, 2024) and the Arctic

Ocean (e.g.,  Hillaire-Marcel et  al., 2013; Lakeman et  al., 2018),

although IRD fluxes at this time appear to have been smaller than

at HS1 (e.g., Zhou and McManus, 2024). While the YD is not asso-

ciated with a time of widespread ice sheet melting according to

sea level data and ice sheet models (Tarasov and Peltier, 2002;

Tarasov et al., 2012; Lambeck et al., 2014; Briggs et al., 2014), it

is notably associated with the abrupt draining of Lake Agassiz, a

proglacial lake formed by the melting Laurentide Ice Sheet that

sat at the boundary of Minnesota, North Dakota, Ontario, and

Manitoba (Broecker et al., 1989; Teller et al., 2002). It was initially

thought that Lake Agassiz drained east at the YD, directly into the

North Atlantic via the St. Lawrence River (Broecker et al., 1989;

Clark et al., 2001), but direct evidence for this has been elusive,

and more recent studies suggest that the lake instead drained north

into the Arctic via the Mackenzie River (Tarasov and Peltier, 2005;

Murton et al., 2010; Keigwin et al., 2018; Süfke et al., 2022). This

result is supported by model simulations, which show that fresh

water discharged into the ocean from the St. Lawrence River does

not immediately spread offshore but is instead transported away

from the subpolar North Atlantic in boundary currents, into the

subtropical gyre. Meltwater from the Mackenzie Valley into the

Arctic Ocean is more likely to reach deep-water formation regions

directly, regardless of whether the Canadian Arctic Archipelago is

ice-covered or open (Condron and Winsor, 2012).

The focus has often been on deep convection regions when it

comes to deglacial freshwater-driven perturbations of the AMOC,

whether the fresh water is delivered by icebergs or directly in liq-

uid form; however, recent physical oceanographic observations

and modeling indicate that capping water mass transformation

along the boundary currents or reducing the zonal density gradi-

ent across the mid-latitude North Atlantic (Buckley and Marshall,

2016; Yeager et  al., 2021; Chafik et  al., 2023; Frajka-Williams

et al., 2023) may be more important for disrupting the AMOC.

During the deglaciation, meltwater introduced to the western sub-

polar North Atlantic could have been entrained offshore along

the northern flank of the western boundary currents that consti-

tute the upper limb of the AMOC, including the Gulf Stream and

the North Atlantic Current (the eastward extension of the Gulf

Stream). This entrained meltwater could significantly alter the

density gradients across these powerful currents and hence reduce

their strength and associated heat transport (e.g.,  Yeager et  al.,

2021; Madan et al. 2024).

0.0

0.2

0.4

0.6

0.8

Freshwater Flux (Sv)

Younger

Dryas

Heinrich

Stadial 1

GLAC-1D

TraCE-21k

10

15

20

25

AMOC (Sv)

-10

-12

-14

-16

-18

-20

Age (ka)

iLOVECLIM

TraCE-21k

B/A

FIGURE 3. Transient model simulations of AMOC across the deglaciation.

(a)  Freshwater flux from ice sheet model simulation GLAC-1D with fresh-

water flux time series in the TraCE-21k model (Liu et  al., 2009; Bouttes

et al., 2023). (b) AMOC strength calculated as the maximum streamfunction

between 20°N and 50°N below 500 m from TraCE-20k (Liu et al., 2009) and

iLOVECLIM (Bouttes et al., 2023).