September 2025

Oceanography | Vol. 38, No. 3

14

mechanisms of past AMOC changes as inferred from paleoceano-

graphic reconstructions and modeling studies, and (2) the implica-

tions of these changes for future AMOC variability.

PALEOCEANOGRAPHIC PROXIES

OF THE AMOC

Paleoceanographic data provide an avenue for extending the rel-

atively short instrumental record and for documenting the state

of the ocean during periods of past climate change. In particu-

lar, they provide a source of empirical information for assessing

the capacity for AMOC to undergo a drastic state change, such as

depicted schematically in Figure 1. We focus on the most recent

glacial-interglacial transition (also called the last “deglaciation”

or “Termination  I”), which occurred following the Last Glacial

Maximum (LGM; ~22–18 ka; ka = thousands of years ago) and

ended at the start of the Holocene (10 ka), the current interglacial

period (see Lynch-Stieglitz, 2017, for a broader review of AMOC

proxy data during the last glacial period). During the deglacia-

tion, several abrupt cooling and warming events occurred in the

circum-North Atlantic that have been linked with, respectively,

AMOC decrease and increase through its role in transporting heat

to the high-latitude North Atlantic. After describing the deglacial

sequence of climatic events, we review the evidence that led to the

widely held view that deglacial climate oscillations were linked

to AMOC changes.

The first event, called Heinrich Stadial 1 (HS1; 18–14.7 ka), was a

North Atlantic cold interval notable for high iceberg discharge and

thought to be associated with reduced AMOC strength (Heinrich,

1988; Bond et al., 1992, 1993; Broecker et al., 1992; Broecker, 1994;

Hemming, 2004). Following HS1, the North Atlantic warmed

abruptly at the beginning of the Bølling-Allerød (BA, 14.7–12.6 ka),

thought to be associated with rejuvenation of the AMOC (T. Chen

et  al., 2015). The BA was followed by another cold period, the

Younger Dryas (YD, 12.9–11.6 ka), which is also thought to be

associated with a weak AMOC (Broecker, 2003). Finally, the YD

concluded with another abrupt warming, at the beginning of the

Holocene, the relatively stable current warm period.

Paleoceanographic proxies used to make inferences about the

strength and/or structure of the AMOC (and/or the associated deep

counter-rotating cell) are often classified into two basic categories:

water mass proxies and kinematic proxies. Water mass proxies are

thought to record the distinct isotopic or chemical signature of dif-

ferent deep water masses, in particular, northern-sourced NADW

and southern-sourced AABW. Examples of water mass proxies are

the stable carbon isotope ratio (δ13C) of fossil benthic foramin-

ifera (W.B. Curry et al., 1988; Duplessy et al., 1988; W.B. Curry and

Oppo, 2005; Eide et al., 2017), the cadmium/calcium concentra-

tion ratio of fossil benthic foraminifera, from which the seawater

Cd concentration (CdW) is estimated (Boyle, 1988; Marchitto and

Broecker, 2006; Oppo et al., 2018), and the authigenic neodym-

ium isotopic composition (εNd) of sediments and deep-sea corals

(Frank, 2002; Goldstein and Hemming, 2003; Du et  al., 2020).

Kinematic proxies are assumed to be more sensitive to flow rate

than water mass proxies. Examples include the radiocarbon age

of fossil benthic foraminifera and deep-sea corals (Keigwin, 2004;

Robinson et al., 2005), the protactinium-231 to thorium-230 activ-

ity ratio of bulk sediment, 231Pa/230Th (Yu et al., 1996; McManus

et al., 2004), and the mean size of sortable silt, SS

— (McCave et al.,

1995, 2017; McCave and Hall, 2006). Note that, albeit conceptually

useful, the distinction between water mass and kinematic proxies is

not without ambiguity: all water properties derived from measure-

ments in the sediment or deep-sea coral are affected by the flow

rate, which would make them “kinematic,” and kinematic proxies

reflect to some degree the composition of water masses.

All proxies are imperfect in the sense that proxy values may be

sensitive to multiple factors, other than the effects of water mass

composition and circulation rate, and each of them has limita-

tions that are necessary to consider when interpreting paleoceano-

graphic records. Some of the water mass tracers (δ13C of dissolved

inorganic carbon and CdW) are functions of biological activity.

The differences in composition between northern- and southern-​

sourced deep water reflect regeneration of dissolved inorganic car-

bon and nutrients in the deep ocean as organic matter from the

surface is remineralized at depth. Thus, changes in biological activ-

ity can alter the spatial distribution of these tracers independently

of water mass or circulation rate change. The δ13C of dissolved

inorganic carbon is also affected by air-sea gas exchange (Lynch-

Stieglitz and Fairbanks, 1994; Lynch-Stieglitz et al., 1995).

Radiocarbon measurements on benthic foraminifera or deep-

sea coral samples are corrected for isotopic fractionation (includ-

ing biological fractionation), so biological activity should not

affect the distribution of these measurements. However, radio-

carbon is still a complicated tracer, because surface waters that

sink to depth in high-latitude regions are characterized by dif-

ferent initial radiocarbon values (Key et al., 2004). It takes about

a decade for the carbon isotopic ratios in the ocean mixed layer

to equilibrate with the atmospheric values (Broecker and Peng,

1974; Lynch-Stieglitz et  al., 1995; Sarmiento and Gruber, 2006;

Jones et al., 2014). This equilibration time is longer than the resi-

dence time of surface waters in deep-water formation regions, par-

ticularly in the Southern Ocean (Bard, 1988). Processes such as

upwelling and the presence of sea ice, which reduces air-sea gas

fluxes (Prytherch et al., 2017), can lead to large differences between

the radiocarbon activity, or age, of the surface waters and that of

the atmosphere (“surface reservoir age”). Therefore, radiocarbon

records from benthic foraminifera and deep-sea corals reflect

both the water mass transit time from the surface (due to en route

radioactive decay) and the surface reservoir age. Some recent work

(Muglia and Schmittner, 2021) suggests that surface reservoir age

is the primary driver of deep radiocarbon distributions in the

Atlantic Ocean, thus making Atlantic radiocarbon values more a

water mass tracer than a kinematic tracer.

For neodymium isotopes, deep-water values are thought to be

dominated by conservative mixing, but sedimentary sources can