September 2025

September 2025 | Oceanography

15

also alter isotopic compositions along deep-water flow paths, par-

ticularly in poorly ventilated basins, such as the deep Pacific and

Indian Oceans (Abbott et al., 2015; Du et al., 2018, 2020). Certain

types of sediment (particularly volcanic ash and ice-rafted debris)

can also be more reactive and prone to delivering non-​conservative

additions of Nd to seawater (Wilson et al., 2013; Blaser et al., 2016;

Du et al., 2016).

The use of bulk sediment 231Pa/230Th as a circulation tracer relies

on the theoretical expectation that, while 231Pa and 230Th are pro-

duced at approximately uniform rates in the ocean (from the decay

of 235U and 234U, respectively), 231Pa is in general scavenged less

intensively by sinking particles than 230Th and hence is more sensi-

tive to circulation than 230Th (Henderson and Anderson, 2003). As

a result, the ratios of the two isotopes in sinking particles and sedi-

ment would be dependent on lateral transport of water (i.e., on the

AMOC), with stronger transport leading to lower 231Pa/230Th in

the underlying sediment. However, the 231Pa/230Th ratio of marine

particles in the water column has been found to vary with their

chemical compositions (e.g., Chase et al., 2002; Hayes et al., 2015)

and with sediment lateral redistribution (S.Y.-S. Chen et al., 2021),

complicating its use as an AMOC proxy.

One of the most widely cited reconstructions used as evidence

of AMOC change across the deglaciation is the 231Pa/230Th record

from the Bermuda Rise in the Northwest Atlantic (Figure 2e;

McManus et al., 2004). This record shows an abrupt increase in

231Pa/230Th to values close to the production ratio (which would

imply very little lateral flow out of the North Atlantic) during

HS1, and another smaller increase during the Younger Dryas. The

high 231Pa/230Th values during HS1 were attributed to a dramati-

cally weakened AMOC. Other 231Pa/230Th data from across the

North Atlantic broadly support this interpretation (Ng et al., 2018).

Compilations of benthic foraminifera δ13C from across the deep

Atlantic show low values during HS1 and an abrupt increase at the

start of the Bølling-Allerød (Figure 2g; Thiagarajan et al., 2014;

Lynch-Stieglitz et al., 2014; Lynch-Stieglitz, 2017), values that have

been interpreted as the resumption of a deep AMOC at the Bølling-

Allerød from a weaker state during HS1. Radiocarbon data from

the Northwest Atlantic also show an abrupt decrease in apparent

ventilation age at the start of the Bølling-Allerød from “older” val-

ues during HS1 and another pulse of old water at the YD (Figure 2f;

Robinson et al., 2005; Hines, 2017; Rafter et al., 2022). Compiled

εNd data are also consistent with a weakened AMOC during HS1

and the YD (Figure 2h; Pöppelmeier et al., 2019; Du et al., 2020),

although these data are less supportive of a fully collapsed AMOC.

The processes that might decouple variations in each proxy from

AMOC differ among proxies. Therefore, if these processes were the

dominant control on the deglacial variability in each record, we

would not expect them to correlate with one another. The finding

that many deglacial ocean circulation proxy records share com-

mon features at approximately the same times is apparent evidence

for changes in AMOC over the deglaciation. In other words, while

each proxy record could be explained by processes other than

circulation, the most parsimonious explanation for all the records

taken together would be that AMOC was abruptly reduced (or col-

lapsed) during HS1 and the YD.

This interpretation is also consistent with paleoclimate records

from terrestrial archives, including the oxygen isotopic composition

of Greenland ice cores (Figure 2a; North Greenland Ice Core Project

Members, 2004); the oxygen isotopic composition of Chinese spe-

leothems (Figure 2d; Wang et al. 2001; Cheng et al., 2009, 2016),

which records coeval shifts in atmospheric circulation patterns;

10

15

20

Age (ka)

HS 1

YD B/A

LGM

Holocene

-16

-14

-12

-10

εNd

-0.5

0.0

0.5

1.0

Benthic δ13C (‰)

1000

2000

3000

14C Ventilation Age (yr)

0.05

0.06

0.07

0.08

0.09

0.10

231Pa/230Th

-12

-10

-8

-6

Hulu cave δ18O (‰)

10

IRD (103 grains/g)

150

200

250

300

Atm. CO2 (ppm)

-50

-45

-40

-35

-30

NGRIP δ18O (‰)

FIGURE 2. Paleoclimate records across the deglaciation. (a) Northern

Hemisphere temperature from NGRIP δ18O of ice (North Greenland Ice Core

Project Members, 2004; Andersen et  al., 2006; Rasmussen et  al., 2014).

(b) Atmospheric CO2 from the West Antarctic Ice Sheet (Marcott et al., 2014).

(c) Ice-rafted debris concentration in the Northwest Atlantic at sites DY081-

GVY001 (solid) and EW9309-37JPC (dashed) (Zhou et  al., 2021). (d) Hulu

cave δ18O (Cheng et al., 2016). (e) 231Pa/230Th from the Bermuda Rise (thin

lines: McManus et al., 2004; Lippold et al., 2009, 2019) and across the North

Atlantic (thick line: Ng et  al., 2018). (f) Compiled deep Atlantic 14C venti-

lation age (Rafter et al., 2022). (g) Deep North Atlantic δ13C (as in Lynch-

Stieglitz et al., 2014; data from Hodell et al., 2008; Tjallingii et al., 2008;

Mulitza et al., 2008; Zarriess and Mackensen, 2011; Shackleton et al., 2000;

Skinner and Shackleton, 2004; Skinner et  al., 2007). (h) εNd from the

Blake Bahama Outer Ridge (Pöppelmeier et al., 2019). YD = Younger Dryas.

B/A = Bølling-Allerød. HS 1 = Heinrich Stadial 1. LGM = Last Glacial Maximum.

IRD = Ice-rafted debris.