Supplementary Materials for
Global reorganization of deep-sea circulation and carbon storage after the
last ice age
Patrick A. Rafter
et al.
Corresponding author: Patrick A. Rafter, prafter@uci.edu
Sci. Adv.
8
, eabq5434 (2022)
DOI: 10.1126/sciadv.abq5434
This PDF file includes:
Materials and Methods
Figs. S1 to S10
Tables S1 and S2
References
Supplementary Materials for
Global
reorganization of deep
-
sea circulation and carbon storage after the last ice
age
Patrick A. Rafter 1, William R. Gray 2, Sophia K.V. Hines 3, Andrea Burke 4, Kassandra M.
Costa 3, Julia Gottschalk 5, Mathis P. Hain 6, James W.B. Rae 4, John R. Southon 1,
Maureen
H. Walczak 7, Jimin Yu 8,9, Jess F. Adkins 10, Timothy DeVries 11
1. University of California Irvine, Irvine, CA, USA
2. Laboratoire des Science du Climat et de l’Environnement (LSCE/IPSL), Université
-
Paris
-
Saclay, Gif
-
sur
-
Yvette, France
3. Woods
Hole Oceanographic Institution, Woods Hole, Massachusetts, USA
4. University of St. Andrews, St. Andrews, Scotland, UK
5. Institute of Geosciences, Kiel University, Kiel, Germany
6. University of California Santa Cruz, Santa Cruz, CA, USA
7. Oregon
State University, Corvallis, OR, USA
8. Pilot National Laboratory for Marine Science and Technology (Qingdao), Qingdao,
266237, China
9. Australia National University, Canberra, AU
10. California Institute of Technology, Pasadena, CA, USA
11. Department
of Geography and Earth Research Institute, University of California, Santa
Barbara, CA, USA
*Corresponding author. Email:
prafter@uci.edu
This PDF includes:
Materials and Methods
Figs. S1 to S10
Tables S1 to S2
Sup
plementary Materials and Methods
Reporting seawater and fossil radiocarbon concentrations (
14
C/C)
The concentration of radiocarbon (
14
C) in seawater can be expressed as
14
C/C, reflecting
the relationship between
14
C and total C atoms in seawater. We use this to refer to the
concentration of radiocarbon in seawater, which is often described here as both delta
notation and as an age referenced to 1950 (the BP in years BP). Delta notation corrects the
14
C/C for
13
C fra
ctionation and decay since the year of collection (for seawater, this can be
the year the measurement was made, but for fossils this would be the calendar age). Thus,
the age model for sediment cores is critical for discussing any paleoceanographic data, b
ut
it is even more important when calculating
Δ
14
C. Here is the equation for calculating
Δ
14
C:
Δ
14
C = (exp^(
14
C age/
-
8033)) / (exp^(calendar age /
-
8267))
–
1 * 1000
The 8033 is the average decay rate derived from the original “Libby half
-
life” of 5568 y
ears
(
62
)
, while the 8267 is the average decay rate for the more accurate half
-
life 5730± 40
years
(
63
)
.
The
14
C age is calculated from the Fraction Modern or (Fm; the amount of
14
C relative to
1950, as measured in all radiocarbon laboratories) as:
14
C a
ge =
-
8033 * LN(Fm)
Compiling the benthic foraminifera and deep
-
sea coral
14
C/C dataset
Seawater
14
C/C can be reconstructed using marine macro
-
fossils such as deep
-
sea corals,
which are sporadically distributed both spatially and temporally, but can be da
ted with U
-
series measurements to provide an absolute calendar age
(
64
)
. Marine microfossils of
foraminifera are more widely distributed in space and time, but rely on a variety of other
methods for determining the calendar age.
Building upon the publish
ed compilation from Zhao et al.
(
21
)
, we searched the literature,
as well as two known databases (NOAA’s NGDC and PANGAEA) for all published benthic
foraminifera and deep
-
sea coral
14
C/C data. Even though this study only deals with sites
>=27.5
γ
n
(neutral
density in kg m
-
3
; see text), we have made the entire dataset (even for
sites in adjacent seas and shallower than used in this study) available at unique DOI that
cites every contributing author (i.e., all who have measured benthic foraminifera or deep
-
se
a coral
14
C/C). The dataset can be found at:
doi.pangaea.de/10.1594/PANGAEA.946522.
The location of all published and new deep
-
sea coral and foraminifera
14
C/C measurements
is shown in Fig. 1A and the contribution of each compilation study is illustrated in Fig. S3.
Published proxy record age model methods
All published calendar ages (a.k.a., the “age model”) were unchanged except for those based
on planktic
14
C ages. Age models using planktic
14
C ages assume a surface ocean “reservoir
age” (i.e., the offset from the contemporaneous atmospheric
14
C) and use
an atmospheric
14
C/C reconstruction to identify the calendar age. Almost none of the published data in the
compilation use the most recent atmospheric
14
C/C compilation (IntCal20)
(
20
)
, so we
calibrated all the planktic foraminifera
14
C
-
based age models to
IntCal20 (see below).
Updating the planktic
14
C
-
based age models.
The site of each proxy record using planktic
14
C measurements to construct their age model undoubtedly has a unique reservoir age for
each planktic
14
C measurement
—
reflecting both local c
irculation and the different air
-
sea
equilibrium timescales for the isotopes
(
31
,
32
)
—
but the IntCal community has provided
an estimated reservoir age (the Marine20 calibration curve
(
65
)
) that is applicable to sites
±40
°
of the equator. For sites poleward
of 40
°N and 40
°
S
, we have added an additional
reservoir age based on the original publication. The reservoir age for the new sites outside
of ±40
°
used the LGM mean and LGM error from
(
66
)
. All planktic foram
14
C
-
based age
models were updated with the R programming package Bacon
(
67
)
, which uses Bayesian
statistics to give mean values as well as errors. A sensitivity test of our reservoir age
estimates is shown in Fig. S9, where we assumed an additional +1000
-
yea
r reservoir age
for all sites poleward of the 40
°
latitude. When more than one planktic species
14
C age was
available, the more “robust” (i.e., less prone to dissolution) species were used to construct
the age model (as recommended in
(
68
,
69
)
). However, u
sing the primary datasets found in
(
70
,
71
)
, it was often the case that these robust microfossils have
14
C ages that were
older
than the benthic foraminifera
14
C ages during the Holocene. In these cases, the mean
planktic
14
C age was used to construct the
age model. The code for updating all the planktic
14
C
-
based age models in our new compilation can be found at
: 10.5281/zenodo.7112182
.
Note that there is a column in the compilation
(“Delta.R”)
for modifying the surface water
14
C reservoir age in addition
to Marine20
(
65
)
assumptions.
New radiocarbon analyses
Sediment from subarctic North Pacific, central North Pacific, and Subantarctic sites (see Fig.
S4 and detailed site descriptions below) was washed using deionized water in a 63 μm
sieve, and mixed benthic foraminifera species (notably without
Pyrgo
spp. (s
ee
(
72
,
73
)
for
more information) were selected from the >250 μm fraction. At least 10% of each sample
was dissolved using HCl to remove potential secondary calcite (precipitated post
-
deposition). Samples were graphitized following
(
74
)
and analyzed at the
Keck Carbon
Cycle Accelerator Mass Spectrometry Laboratory at University of California, Irvine
(
75
)
. All
new sample measurements and previously published are available at
:
doi.pangaea.de/10.1594/PANGAEA.946522.
Stable isotopic composition of carbon analy
ses
Samples
from Juan de Fuca Ridge sites (see detailed text below) were freeze
-
dried, and
then an aliquot was weighed into a 125 mL Nalgene bottle,
filled with
approximately 100
mL of tap water, and disaggregated on a tumble
wheel for two hours. Samples w
ere then
sieved at
63μm, and a soft brush was used to gently
break up clay clumps that did not
sufficiently disintegrate during
tumbling.
Planulina wuellerstorfi
tests
were picked from the
250
-
300um
fraction and analyzed at Lamont Doherty Earth
Observatory
(LDEO) on a
Thermo
-
Delta V Plus equipped with a
Kiel IV individual
acid bath device. The stable carbon
isotopic ratio is reported in delta notation, where:
δ
13
C = (
!"
#
!$
#
!"#$%&
!"
#
!$
#
!'"()"*)
)
−
1
,
where the standard is VPDB. This is
multiplied by 1000 to give “per mil.” Samples have
been
standardized to NBS
-
19, and all data are reported on the VPDB scale with
precision
≤0.05‰.
Modern seawater
14
C/C statistics
Nearly all modern seawater
14
C/C measurements have taken place since the introduction of
‘bomb’
14
C/C
—
via the atmospheric testing of thermonuclear weapons
—
and are therefore
susceptible to being biased toward artificially high
14
C/C and therefore young
14
C ages
(
30
,
76
)
. While some dat
a products attempt to remove this ‘bomb’
14
C/C influence from
seawater measurements
(
77
)
, the use of this “natural”
14
C/C data product has been
questioned (i.e., see comparison and discussion in
(
78
)
). We do not use the “natural”
seawater
14
C/C data produc
t, but also note that it also does not appear to be publicly
available. Instead, in Fig. 1B and Fig. S1 we have filtered the available seawater
14
C/C data
to remove measurements made since bomb
14
C/C was introduced to the Earth system. This
is not ideal si
nce it obviously lowers the number of usable observations, but we find it to be
the most conservative approach.
To avoid the influence of bomb
14
C/C, we first removed seawater samples with a
Δ
14
C > 0 ‰,
effectively excluding measurements that are obvious
ly influenced by bomb
14
C. To exclude
seawater biased to higher
Δ
14
C by ‘bomb’
14
C/C, but still below 0‰, we further filter the
seawater
14
C/C data to exclude samples with an CFC
-
11 > 0.1 pmol kg
-
1
. Our rationale for this
value is based on simple calculati
ons (see https://water.usgs.gov/lab/software/USGS_CFC/)
showing that a seawater CFC
-
11 concentration of 0.1 pmol kg
-
1
is equivalent to waters that
equilibrated with the atmosphere at
≈
1955, which is one of the last years before the ‘bomb’
spike in atmosphe
ric
14
C/C
(
76
)
. Where there was no observed DIC
14
C age measurement
within a reasonable depth and lateral distance from the proxy site (±500 m and ±5
°
of
longitude and latitude), we used the
14
C/C from the Ocean Circulation Inverse Model
(
79
)
.
Testing the proxy
14
C fidelity
We identify the utility of the global compilation for reconstructing past changes in seawater
14
C/C in Fig. 1B and Fig. S1. This test compares modern seawater
14
C/C with all
proxy
14
C/C
over the past 4
-
kyr BP and gives a relationship of Y = (0.97±0.1)X + (
-
18±91); R
2
=0.66. The
strong correlation, slope close to 1, and Y intercept close to zero in Fig. 1B suggests that the
marine fossil
14
C/C performs well as a proxy for seawa
ter DIC
14
C/C. Additional
examinations of the proxy vs. seawater
14
C /C relationship over different time
-
intervals can
be found in Fig. S1, all showing a similar, strong relationship (see figure and caption for
detailed statistics). These age boundary test
s (Fig. S1) suggest variability of the Y
-
intercept
from
-
44
-
to
-
+54
14
C years, which should be considered a minimum error for the marine
14
C
proxy. Seawater
14
C/C deviations from modern values could explain some of the variance in
Fig. 1B, but given the in
dividual offsets from the 1:1 line of >100 years, the proxy itself is
almost certainly the first
-
order driver of the variability (consider the findings in
(
80
,
81
)
).
However, it is important to note that the potential bias (given by the Y
-
intercept) or the
individual measurement deviations from the 1:1 line are much smaller than some reported
14
C biases of >10,000
14
C years
(
82
)
. As such, we argue that
—
despite known influences on
the proxy
14
C archives from sedimentary geochemical processes
(
80
)
, sedimentar
y
bioturbation, and calendar age assumptions
(
83
)
—
the results of our tests speak to the
utility of the seawater
14
C/C proxy, especially interpretations based on a relatively large
collection of measurements.
Pre
-
treating the benthic
14
C/C dataset
The data shown throughout the manuscript (e.g., Fig. 2, 3, & 4) was pre
-
treated to remove
outlier observations in the following way. First,
14
C/C values that were above the
contemporaneous atmosphere were flagged as o
utliers and not included. Second,
14
C/C
values more than 3 sigma of the 500
-
year binned average (see below) were flagged as
outliers (same as in
(
33
)
). We iterated this second outlier identification process
(identifying a new binned mean and binned standar
d deviation plus removing the >3 sigma
outliers) four times to ensure outlier removal. The outlier data points are clearly noted in
the compilation file and can also be observed as pink diamonds in Fig. S5
-
to
-
S8. Note that
500
-
year bins with only 1 measure
ment were also considered “outliers”, are not used in the
LOESS trend estimate (see below)
—
these are also shown as pink diamonds in Fig. S5
-
to
-
S8.
Density surfaces for mid
-
depth and bottom waters
We base our interpretation of past deep
-
sea circulation on
modern deep
-
sea circulation,
which can be defined by neutral density surfaces. The “mid
-
depth” waters include modern
Pacific Deep Water (PDW) and Upper North Atlantic Deep Water (NADW) and are within
the neutral density surfaces of 27.5 and 28.0 kg m
-
3
(
23
,
84
)
. The denser “Bottom Waters”
(>28.0 kg m
-
3
) include Lower North Atlantic Deep Water and Antarctic Bottom Water
(AABW).
Proxy data trend and uncertainty estimates
We pull out the basin wide trends (and the associated uncertainties) from the data
using
two independent methods: binning and LOESS regression
.
Proxy
14
C/C measurements were binned in 500
-
year intervals (based on calendar age).
This follows prior arguments that the binning width should reflect the mean calendar age
error
(
21
)
. The aver
age calendar year error of the compilation is ±250 years, but this does
not take into account various assumptions including changes in surface ocean
14
C reservoir
ages (used in planktic
14
C
-
based age models), errors associate with the different age model
a
ssumptions
(
6
)
, and more. To account for these additional, unknown errors, we assumed a
binning width of 500
-
years. After subsetting the data at 500
-
year time steps, we remove
the three sigma outliers (pink diamonds in Fig. S5
-
to
-
S8) and calculate the mean
, standard
deviation, and standard error of each bin. The mean bin values are shown as black symbols
in Fig. S5
-
to
-
S8 and the standard error is shown by vertical lines. These binned values are
different than the LOESS estimated
14
C/C trends.
Fossil
14
C da
ta were fit with a non
-
parametric regression (LOESS, a locally estimated
smoothing) as a function of calendar age in R (R core team), with the smoothing parameter
(α) optimized by Generalized Cross Validation (GCV). Note, the LOESS regressions are fit to
t
he pre
-
treated (see above), unbinned
14
C data (with 3
-
sigma and other outliers removed).
The most likely fit to the
14
C data and uncertainties were calculated with a Monte
-
Carlo
approach (1000 iterations); this includes a bootstrap resampling of the datas
et
(
85
)
, and
Monte
-
Carlo resampling of the calendar age and
14
C analytical errors, as well as
propagation of these errors through to the ∆
14
C
and ventilation age estimates in a manner
which accounts for the covariance of the calendar age
uncertainties with the ∆
14
C and
ventilation age uncertainties. We report the uncertainties as the 5th, 32nd, 68th, and 95th
percentiles of these LOESS fits (with the best fit taken as the 50th percentile), including a
Monte
-
Carlo resampling of the SE of th
e LOESS fit during each iteration.
The code for creating the basin mean LOESS trends for
14
C ventilation age can be found at:
https://github.com/patrickrafter1/14C
-
compilation
-
2022.
Site bias adjustments to calculate the basin mean
14
C/C
Note that using
relatively sparse sites for reconstructing a basin mean could be biased to
sites that have a higher or lower value than the actual basin mean for those density
surfaces (see above). In consideration of this potential biasing, we adjust each record based
on
the difference between the modern seawater
14
C bathing the site and the modern mean
value for that basin. In other words, each fossil
14
C measurement was adjusted to account
for seawater
14
C/C differences between the basin average and the proxy record loc
ation. It
is this “site bias corrected” data that was used to construct the trends in Fig. 3 & Fig. 4.
These same trends, binned mean values (black circles for mid
-
depth and black squares for
bottom waters), binned standard error (lines), all accepted obse
rvations (circles, color
-
coded to original compilation), and outlier data points (pink diamonds) can be observed in
Fig. S5 & S6. Note that a first
-
order test of this approach can be observed by the
approximate overlap between most late Holocene reconstruc
ted and modern observed
14
C/C (symbols to the left in all figures).
We also show the “without site bias correction” trends and datapoints in Fig. S7 (shown as
the uncorrected
Δ
14
C) & Fig. S8 (shown as the uncorrected
14
C ventilation age). These
“uncorrect
ed” trends show that the adjustments do not make significant changes to the
reconstructions, although they slightly increase the standard error and worsen the overlap
between the trend and the observed, ship
-
based dissolved inorganic carbon (DIC)
14
C/C
(sy
mbols in Fig. S5
-
to
-
S8). One notable difference is for the Southern mid
-
depth
14
C/C,
which is without the site/ depth bias correction gives a younger age than the observed
modern value (circle to the left in each plot).
Sensitivity testing the basin mean
14
C ventilation age trends
In Fig. S9 & Fig. S10, we test the sensitivity of the assumptions used in creating the
14
C
ventilation trends in Fig. 3 & Fig. 4. The site
-
bias / depth
-
bias corrections with our assumed
depth surface separations are shown in pa
nels (A) to (C) in both figures (see text above for
details and rationale for this approach). Fig. S9 (D
-
to
-
F) shows the basin
-
averaged
14
C
ventilation trends without the depth
-
bias correction. In Fig. S9 (G
-
to
-
I), we show the
14
C
ventilation trends assumi
ng that an additional 1000
-
year reservoir age at sites poleward of
40
°
N and 40
°
S may be a better assumption for planktic foraminifera
14
C
-
based age models.
Another concern when using benthic foraminifera
14
C/C measurements is that bioturbation
may influenc
e the observed
14
C age and in Fig. S9 (J
-
to
-
L), we create basin
-
averaged
14
C
ventilation trends only using sediment cores with sedimentation rates >2 cm kyr
-
1
. In Fig.
S10, we provide additional sensitivity tests assuming different density
surfaces: (A
-
to
-
C)
uses our default 27.5
-
to
-
28 kg m
-
3
and >28 kg m
-
3
density surfaces; (D
-
to
-
F) uses the 27.4
-
to
-
28 kg m
-
3
and >28 kg m
-
3
density surfaces; (G
-
to
-
I) uses the 27.6
-
to
-
28 kg m
-
3
and >28 kg
m
-
3
density surfaces; and (J
-
to
-
L) uses the 27.5
-
to
-
2
8.1 kg m
-
3
and >28.1 kg m
-
3
density
surfaces.
While these different assumptions adjust the finer details of the estimated basin
-
scale
14
C
ventilation age, it appears that our main
14
C ventilation age discussion points
—
older
14
C
ages of all bottom waters,
the inversion of Pacific mid
-
depth and bottom water ages, the
different LGM variability in all basins
—
are still supported by these varied sensitivity tests.
T
-
tests of the flipped Pacific
14
C ventilation ages
A simple t
-
test of our results lends further
support for a significant difference between
Pacific mid
-
depth and bottom water
14
C ventilation ages during the LGM, for both our
default case and these sensitivity tests. For example, the estimated Pacific trends from our
default assumptions (Fig. S9 Pan
els A
-
to
-
C) indicate there is a significant difference
between Pacific mid
-
depth and bottom water
14
C ventilation ages from 23
-
to
-
18
-
kyr BP (t
-
test results: t =
-
4.8345, p
-
value = 2.28e
-
06). Contrast this with a similar analysis of Pacific
mid
-
depth and bo
ttom water
14
C ventilation ages from 6
-
to
-
0
-
kyr BP indicating no
significant difference (Welch’s t
-
test results: t = 0.13002, p
-
value = 0.8968). Looking to one
of the more different results of our sensitivity tests, the estimated Pacific
14
C/C from the
”+1
000
-
year reservoir age sensitivity test” (Fig. S9; panels G
-
to
-
I) also indicates a
significant difference between Pacific mid
-
depth and bottom water
14
C ventilation age
during the LGM (t =
-
2.9036, p
-
value = 0.004031), but no significant difference during
the
late Holocene (t = 0.32647, p
-
value = 0.7449).
Estimating average bottom water
14
C/C during the LGM
The volume of Atlantic, Southern, and Pacific mid
-
depth and bottom waters was estimated
from the Ocean Circulation Inverse Model (OCIM)
(
79
)
, assuming
glacial changes in volume
are negligible for these calculations. Note that the Indian Ocean contains so few values that
it was not used in this calculation. The
14
C ventilation age during the LGM and percent of
global ocean volume for each ocean basin and
density range are as follows. Mid
-
depths:
Atlantic bottom waters are estimated to be assumed to be 22%, the Southern Ocean is
assumed to be 16%, and the Pacific is assumed to be 46% of global waters along these
density surfaces. Bottom waters: Atlantic wa
ters are assumed to be 19%, the Southern
Ocean is assumed to be 21%, and the Pacific is assumed to be 46% of global waters along
these density surfaces. The global mid
-
depth and bottom water
14
C ventilation age was
estimated for each 500
-
year bin over the
past 25
-
kyr using LOESS (see above) and this
time
-
interval was weighted by the ocean basin volume percent. The average and standard
error for relevant time
-
intervals are shown in Supplementary Tables 1 and 2, along with
their difference from modern DIC
14
C
ventilation age (see above).
Differences in LGM time
-
slice depth profile and global average
14
C ventilation
estimates
We used our
14
C/C compilation to reconstruct the
14
C ventilation age of the Pacific with
depth during the LGM in Fig. S2 (panel C) and this “time
-
slice” view of the dataset is one
piece of evidence
—
along with an LGM time
-
slice of
d
13
C and the full glacial
-
interglacial
time
-
series in Fig. 3, Fig. 4, Fig. S3
-
to
-
S8
—
indicating that the oldest
14
C ages are found in
the deepest Pacific Ocean during the LGM. However, we only have one observation for the
Pacific Ocean >4000 m (see our Fig. 2D & Fig. S2), which differs from earlier work
(
33
)
including 6 “Pacific / I
ndian” observations >4000 m (their Figure 3B). Breaking down the
data used to construct the LGM Pacific time
-
slice in
(
33
)
(using their compilation available
online), their Figure 3B below 4000 m includes 1 measurement from the Central Equatorial
Pacific a
nd 5 measurements that are south of 40
°
S (off of southern New Zealand). Of these
observations, only the Central Equatorial Pacific measurement falls within our definition of
the Pacific Ocean (between 40
°
S and 60
°
N). Thus, the difference in LGM Pacific dep
th
-
profile of
14
C ventilation age shown here (Fig. S2) and in prior work (Figure 3B in
(
33
)
) is
caused by our geographic definition for the Southern Ocean (see Fig. 1A). Considering the
much larger volume of the Pacific Ocean, our decision not to include t
hese Southern New
Zealand observations likely contributed to our significantly older estimated global average
14
C ventilation age than in prior work
(
6
,
33
)
.
New proxy
14
C/C measurements from the Pacific Ocean
We have generated several new records of mixed benthic foraminifera
14
C/C from the
Pacific Ocean and extended a recently published mixed benthic foram record
(
86
)
from a
site in the Southern Ocean (see Fig. 1A). No
Pyrgo
spp. were included in these
measure
ments per earlier studies
(
72
,
73
)
. All new measurements were measured in the
Keck Carbon Cycle AMS Laboratory at UC Irvine according to the preparation and
measurement protocols found in
(
74
,
75
)
. The results of our new measurements (and
comparisons with
trends from all sites within that seawater density range) can be seen in
Fig. S7. Below we provide additional details for each site.
Southern Ocean site MD97
-
2106.
New benthic foraminifera
14
C measurements from
Southern Ocean site MD97
-
2106 (45.2
°
S, 146.3
°
E; 3310 m water depth) were added to
published results
(
86
)
to extend the record back to 25
-
kyr BP. The age model for these new
measurements is based on planktic
14
C ages on the Marine20 calibration with additional
reservoir age of 256±219 years, derived
from average values for that time and location
published in
(
66
)
. Among the new insights the core brings is the decrease in
14
C ventilation
age during HS2 as seen in another Southern Ocean Bottom
-
Water record (MD07
-
3076 at
3770 m water depth) in the Atlant
ic Sector of the Southern Ocean
(
5
)
.
North Pacific site VINO19
-
4 GGC
-
37.
We also have new benthic foraminifera
14
C
measurements from Subarctic Pacific site VINO19
-
4 GGC
-
37 (50.4
°
N, 167.7
°
E; 3300 m
water depth). The age model here uses a combination of stratigraphic matching and
planktic
14
C ages calibrated using Marine20. We applied an additional reservoir age
correction of 400±400 years applied to the Marine20 because of the core’s higher
latitude
(see above). One stratigraphic tie
-
point is used where the large decline in planktic
d
18
O (see
original publication
(
35
)
) is tied to the onset of the BA/ACR at 14.7
-
kyr BP. The two
D
14
C
spikes to high values observed during the middle of the BA/A
CR and
≈
10
-
kyr BP are
removed as 2
-
sigma outliers, although earlier studies have suggested brief, but deep
penetration of surface waters in the Subarctic Pacific
(
43
)
. Regardless, the high resolution
of our measurements diminishes the importance of these t
wo seemingly
-
anomalous data
points
Tropical Pacific site ML1208
-
01PC.
Our third new benthic foraminifera
14
C/C record is
from site ML1208
-
01PC (21.2
°
N, 158.5
°
W; 2960 m water depth). The age model for this
sediment core west of Hawai’i was constructed usin
g a combination of stratigraphic
matching (increasing sediment density change beginning at 23
-
24 cm was assumed to be
18
-
kyr BP) and planktic
14
C
-
based age model calibration (to MARINE20) at planktic
foraminifera abundance maxima. The benthic foraminifera
14
C ventilation age has a larger
degree of variability around the Pacific Bottom
-
Water trend (blue line), which likely speaks
to the impacts of bioturbation on this relatively slowly accumulating sediment core.
Figures and other acknowledgments
. Some figu
res were initially generated using Ocean
Data View
(
87
)
and utilize color palettes developed by
(
88
)
.
Figure S1: Examining modern and marine fossil late Holocene fossil
14
C/C.
The x axis
in all plots is the modern DIC Δ
14
C at the proxy site and the Y axis is the difference between
the contemporaneous atmosphere and fossil
14
C (“
14
C ventilation age”) for va
rying calendar
age boundaries. Shaded regions are 1 and 2 sigma confidence intervals for linear
regression. Color shading is the ocean basin of the proxy measurement (green is Atlantic,
light blue is Southern, and brown is Pacific; see Fig. 1A for legend).
The statistics for these
analyses are as follows: (A) Measurements from the last 6
-
kyr BP: Y = (0.99±0.1)X + (
-
44±93); R
2
= 0.60. (B) Measurements from the last 4
-
kyr BP: Y = (0.97±0.1)X + (
-
18±91); R
2
= 0.66. (C) Measurements from the last 2
-
kyr BP: Y =
(0.96±0.1)X + (54±116); R
2
= 0.68. (D)
Measurements from the last 1
-
kyr BP: Y = (1.03±0.1)X + (32±113); R
2
= 0.82. The symbols
in the legend signify the age model used for the study: Triangles for deep
-
sea coral and U
-
series dates; Circles for wood or teph
ra dating; Square for planktic foraminifera
14
C
calibrated to the atmosphere; Diamonds for a variety of stratigraphic matching (e.g.,
oxygen isotopes matched to deep
-
sea compilations); and upside
-
down triangles show other
methods (e.g., plateau
-
tuning of
1
4
C/C).
Table S1: The volume
-
weighted mid
-
depth water (between the 27.5 and 28 kg m
-
3
neutral
density surfaces)
14
C ventilation ages is slightly older during the LGM and are within error
of modern values only mid
-
way through the deglacial global warming.
M
odern
14
C/C
observations are from
(
30
)
and are rounded to the 10. Reconstructed
14
C ventilation values are
rounded to the 50. The difference between modern and reconstructed
14
C ventilation ages and
their propagated errors are also shown (right side).
vol
ume
-
weighted average
14
C ventilation
age for Atlantic, Southern, and Pacific
Ocean mid
-
depth water
14
C
ventilation
age (years)
14
C
ventilation
age standard
deviation
(years)
14
C
ventilation
age difference
14
C ventilation
age difference
(propagated
standard
deviation)
PRE
-
INDUSTRIAL (observed)
1290
110
0
--
HOLOCENE (10
-
to
-
0
-
kyr BP)
1300
150
0
150
YOUNGER DRYAS (12.8
-
to
-
10
-
kyr BP)
1250
100
-
50
150
BA/ACR (14.7
-
to
-
12.8
-
kyr BP)
1300
100
0
150
HS1 (18
-
to
-
14.7
-
kyr BP)
1500
100
200
150
LGM (23
-
to
-
18
-
kyr BP)
1750
150
450
150
Table S2: The volume
-
weighted bottom water (below the 28 kg m
-
3
neutral density surface)
14
C ventilation ages is much older during the LGM and within error of modern values mid
-
way through the deglacial global warming.
Here, Atlantic waters are assumed to be 19%, the
Southern Ocean is assumed to be 21%, and the Pacific is assumed to be 46% of
global waters along
these density surfaces.
Modern
14
C/C observations are from
(
30
)
and are rounded to the 10.
Reconstructed
14
C ventilation values are rounded to the 50. The difference between modern and
reconstructed
14
C ventilation ages and their propag
ated errors are also shown (right side).
volume
-
weighted average
14
C ventilation
age for Atlantic, Southern, and Pacific
Ocean bottom water
14
C
ventilation
age (years)
14
C
ventilation
age standard
error (years)
14
C
ventilation
age difference
14
C
ventilati
on
age difference
(propagated
standard
deviation)
PRE
-
INDUSTRIAL (observed)
1350
90
0
--
HOLOCENE (10
-
to
-
0
-
kyr BP)
1300
150
-
50
150
YOUNGER DRYAS (12.8
-
to
-
10
-
kyr BP)
1500
150
150
150
BA/ACR (14.7
-
to
-
12.8
-
kyr BP)
1450
150
100
150
HS1
(18
-
to
-
14.7
-
kyr BP)
2050
150
700
150
LGM (23
-
to
-
18
-
kyr BP)
2350
150
1050
150
Figure S2: Depth
-
binned LGM seawater
14
C ventilation age (colors) is significantly
older than pre
-
industrial seawater carbon isotopic composition (black).
The pre
-
industrial average seawater
14
C ventilation ages with depth for (A) Atlantic, (B) Southern,
and (C) Pacific Oceans are shown by the
vertical black lines, where the grey envelope is 1
-
standard error (data from the GLODAP v.2
(
30
)
). Colored lines show average seawater
14
C
ventilation ages with depth during the LGM (23
-
to
-
18
-
kyr BP). Circles in (A)
-
to
-
(C) are
LGM proxy observations (23
-
to
-
18
-
kyr BP). The different color envelopes are 1
-
and 2
-
Standard Error. We color
-
code the LGM observations based on their compilation study (see
Legend). In (D), the pre
-
industrial average seawater
d
13
C with depth for the Pacific
(GLODAP v.2; black with gr
ey envelope) is compared with new and newly
-
compiled North
Pacific benthic foraminifera
d
13
C from the LGM (brown line with 1 and 2 Standard Error
envelopes)
(
35
–
37
)
. The upper and lower dashed lines in all figures represent the modern
regional mean depth o
f the 27.5 and 28.0 neutral density surfaces, respectively
(
89
)
. See
Supplementary Material (SM) for benthic foram
d
13
C methods. Basin average trends with
depth were estimated using LOESS (see SM for methods).
Figure S3: The spatial distribution for the different benthic foraminifera and deep
-
sea coral
14
C/C compilations.
S
ites
in the reference
(
33
)
compilation
are
shown in (A)
and
sites
in the
reference
(
21
)
compilation are shown in (B). There is some overlap i
n the
location of the reference
(
33
)
compilation sites (A) and the reference
(
21
)
compilation sites
because only LGM
-
aged observations were included in (A). In addition to creating 3 new
glacial
-
interglacial
14
C/C records (diamonds here and in Figure 1), o
ur new compilation
includes the sites shown in (C). All available
14
C/C observations are shown in (D). The
number of unique sites and observations for each compilation is shown in each plot.
Different compilation studies report observations from the same s
ite, so that the sum of the
unique sites in panels A
-
to
-
C do not equal the unique sites in Panel D.
Figure S4. New
14
C/C measurements on cores from the Pacific and Southern Oceans.
The
Δ
14
C for three new sites in: (left) the Southern Ocean
(MD97
-
2106;
-
45.2
°
N, 146.3
°
E;
3310 m), (center) the Subarctic Pacific (VINO 19
-
4 GGC
-
37; 50.4
°
N, 167.7
°
E; 3300 m), and
(right) Subtropical North Pacific (ML1208; 21.2
°
N, 158.5
°
W; 2960 m). Each is compared to
atmospheric
Δ
14
C (grey line (
5
)) and LOESS
-
avera
ged deep water
Δ
14
C for each basin
(neutral density (
γ
n
) >28; see Fig. 3 and main text). Symbols to left are the average modern
seawater
Δ
14
C values for each ocean basin with
γ
n
>28 kg m
-
3
. Pink diamonds are not used
to create the basin averages, either ha
ving a value that was >3 standard
-
deviation units of
the binned mean value or the only value within that 500
-
year bin. HS2= Heinrich Stadial 2;
LGM=Last Glacial Maximum; HS1=Heinrich Stadial 1; BA=Bølling
-
Allerød; YD = Younger
Dryas.
Figure S5: Mean seaw
ater
Δ
14
C trends for Atlantic, Southern, and Pacific Oceans (left
-
to
-
right) with site bias correction.
Here we show the LOESS trends of
D
14
C (in per mil)
from Fig. 3 plus all the observations (grey circles), sorted by their published compilation
(see legend). Black symbols and lines are binned mean values and 1
-
standard error,
respectively. Pink diamonds are observations that are >3
-
standard
deviation the binned
mean or bins with only one observation (note that these are not used in the LOESS
estimate). Panels (A
-
C) are for mid
-
depth density surfaces and panels (D
-
F) are for bottom
water density surfaces. Panels (G
-
I) are the histograms for th
e observations used to create
the LOESS trends, dotted outline bars are mid
-
depth and solid outline bars are bottom
-
waters. The depth
-
bias correction involves an adjustment applied to each observation in
consideration of site biasing relative to the basin
average for that depth range (see
Methods). LOESS error envelopes show 68% and 95% confidence interval from
bootstrapping/Monte
-
Carlo simulation (see Methods)