LETTERS
PUBLISHED ONLINE: 13 JULY 2009 | DOI: 10.1038/NGEO578
Carbon dioxide forcing alone insufficient to explain
Palaeocene–Eocene Thermal Maximum warming
Richard E. Zeebe
1
*
, James C. Zachos
2
and Gerald R. Dickens
3
The Palaeocene–Eocene Thermal Maximum (about 55 Myr ago)
represents a possible analogue for the future and thus
may provide insight into climate system sensitivity and
feedbacks
1,2
. The key feature of this event is the release
of a large mass of
13
C-depleted carbon into the carbon
reservoirs at the Earth’s surface, although the source remains
an open issue
3,4
. Concurrently, global surface temperatures
rose by 5–9
◦
C within a few thousand years
5–9
. Here we use
published palaeorecords of deep-sea carbonate dissolution
10–14
and stable carbon isotope composition
10,15–17
along with a
carbon cycle model to constrain the initial carbon pulse to a
magnitude of 3,000 Pg C or less, with an isotopic composition
lighter than −50. As a result, atmospheric carbon dioxide
concentrations increased during the main event by less than
about 70% compared with pre-event levels. At accepted values
for the climate sensitivity to a doubling of the atmospheric
CO
2
concentration
1
, this rise in CO
2
can explain only between
1 and 3.5
◦
C of the warming inferred from proxy records. We
conclude that in addition to direct CO
2
forcing, other processes
and/or feedbacks that are hitherto unknown must have caused
a substantial portion of the warming during the Palaeocene–
Eocene Thermal Maximum. Once these processes have been
identified, their potential effect on future climate change needs
to be taken into account.
The magnitude of future global warming from anthropogenic
CO
2
forcing remains unknown because of uncertainties in pre-
dicting climate system feedbacks
1
. Studying past episodes of global
warming and rapid carbon release such as the Palaeocene–Eocene
Thermal Maximum (PETM) may help to reduce those uncertainties
or at least isolate the possible sources
2
. The onset of the PETM
was marked by a global increase in surface temperatures by
5–9
◦
C within a few thousand years
5–9
. At nearly the same time, a
substantial carbon release occurred, as demonstrated by a large drop
in the
13
C/
12
C ratio of surficial carbon reservoirs. The carbon release
led to ocean acidification and widespread dissolution of deep-sea
carbonates
10,18
. Different sources for the carbon input have been
suggested, which has led to speculations concerning the mechanism.
Some, such as volcanic intrusion, imply that the carbon drives the
warming. Others, such as the destabilization of oceanic methane
hydrates, imply that the carbon release is a feedback that can exacer-
bate warming
3,4,19
. Remarkably, however, even the lower estimates
for the carbon release during the onset of the PETM (∼1 Pg C y
−1
)
and over the past 50 years from anthropogenic sources seem to be
of a similar order of magnitude (see the Methods section). The
PETM may therefore serve as a case study for the consequences of
the carbon dioxide released at present by human activities.
We have used deep-sea carbonate dissolution records
10–14
and
stable carbon isotope records across the PETM (refs 10, 15–17)
1
School of Ocean and Earth Science and Technology, Department of Oceanography, University of Hawaii at Manoa, 1000 Pope Road, MSB 504, Honolulu,
Hawaii 96822, USA,
2
Earth and Planetary Sciences Department, University of California, Santa Cruz, California 95064, USA,
3
Department of Earth
Sciences, Rice University, Houston, Texas 77005, USA. *e-mail: zeebe@soest.hawaii.edu.
in combination with carbon cycle modelling
2,18,20
to constrain
the mass of the PETM carbon input (Fig. 1). The observed drop
in the stable carbon isotope composition (δ
13
C) of the surficial
carbon reservoirs is about 3h. However, the δ
13
C signal alone is
insufficient to determine the mass and δ
13
C value of the carbon
input. In this study, the input mass is estimated from carbonate
dissolution records. The δ
13
C composition of this carbon was then
constrained by requiring the model outcome to match observed
marine δ
13
C records at the given input mass (see Supplementary
Information). For our model simulations, we used the long-
term ocean–atmosphere–sediment carbon cycle reservoir model
LOSCAR (see refs 2, 18, 20 and Supplementary Information).
To simulate the observed time-dependent profile (that is,
magnitude and duration) of the carbon isotope excursion (CIE)
during the PETM main phase (Fig. 1b,c), we assumed a large
initial input pulse followed by further smaller pulses and a low,
continuous carbon release during the main event (Fig. 1a). Without
the further release, the model was unable to reproduce the CIE
duration because δ
13
C values returned to pre-excursion values too
quickly (Fig. 1b, dotted green line). A pulsed carbon release (rather
than a single input peak) is consistent with δ
13
C records from most
marine and terrestrial sections
21,22
.
The prolonged carbon release is also important to simulate the
observed duration of deep-sea carbonate dissolution (Fig. 1d–f).
For example, the carbonate records from Walvis Ridge in the
Atlantic Ocean show that wt% CaCO
3
values at various palaeowater
depths return to pre-excursion values only after more than
70 kyr (Fig. 1e). The extended duration of the dissolution event
could not be reproduced in the model without the continued
carbon release (Fig. 1d).
The size of the carbon input, on the other hand, is determined
by the magnitude of CaCO
3
dissolution or shoaling of the calcite
compensation depth (CCD) in the different ocean basins (Fig. 1f,g).
Note that for the quantification of the carbon input, the position of
the CCD before the event is as critical as the actual shoaling. For
example, the late Palaeocene CCD was about 1–1.5 km shallower
than today in all ocean basins, including the Pacific basin, which
was much larger than today (see refs 23, 24 and references therein).
Just before the event, Ocean Drilling Program core sites 1208
(Pacific Shatsky Rise) and 1221 (Equatorial Pacific) at 3,350 m and
3,200 m palaeowater depth, respectively, were located very close
to the CCD, indicating a Pacific pre-event CCD shallower than
3,500 m. This depth is consistent with other reconstructions
11,23
(see
Supplementary Fig. S5). In the late-Palaeocene Pacific basin, the
erodible sediment CaCO
3
inventory in the depth range 3.5–4.5 km
would have been ∼2,000 Pg C, which has the capacity to neutralize
∼2,200 Pg C of CO
2
(see the Methods section). Setting the pre-
event CCD at a depth below that indicated by observations
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 1
© 2009 Macmillan Publishers Limited. All rights reserved.
LETTERS
NATURE GEOSCIENCE DOI: 10.1038/NGEO578
0
0.5
3,000 Pg C
1,480 Pg C
¬2
0
2
LA
DA
DI
DP
Bulk 690
Bulk 1263
Bulk 1265
Bnthc 401
Bnthc 690
Subb. 690
0
50
100
1,000 m
1,500 m
2,000 m
2,500 m
3,500 m
1263, 1,500 m
1265
1266, 2,600 m
1267
1262, 3,600 m
2
4
CCD (km)
2
4
CCD (km)
Atlantic
Indian
Pacific
Tethys
Below ~2 km
Below ~2 km
? Site 259
? Site 259
Atlantic
S. Ocean
Indian
Pacific
¬50 0 50 100 150 200
1,000
1,400
1,800
δ
13
C
inp
= ¬50
o
/
oo
δ
13
C (
o
/
oo
)
¬2
0
2
δ
13
C (
o
/
oo
)
CaCO
3
(wt%)
0
50
100
CaCO
3
(wt%)
Time (kyr ± PEB)
CO
2
(ppmv)
a
b
c
d
e
f
g
h
δ
δδ
Input (Pg C y
¬1
)
Figure 1 | PETM model simulations and palaeorecords. a, PETM carbon-release scenario (model input); t = 0 corresponds to the onset of the PETM.
b, Simulated δ
13
C of TCO
2
in the low-latitude surface Atlantic (LA), deep Atlantic, Indian and Pacific oceans (DA, DI, DP) using the carbon release shown
in a, including the continuous release (solid green and red lines). In simulations without the continuous release (dotted green line), the duration of the δ
13
C
excursion was not captured in the model. c, Observed δ
13
C in bulk CaCO
3
, benthic and planktonic foraminifera
15–17
. d, Simulated wt% CaCO
3
at various
depths in the deep Atlantic. e, Observed wt% CaCO
3
at Walvis Ridge, South Atlantic Ocean
10
. f, Simulated CCD in different basins. g, Observed CCD
before and during the PETM main event (see Supplementary Information). h, Simulated atmospheric CO
2
(PEB: Palaeocene/Eocene boundary).
would therefore erroneously increase the available CaCO
3
for
dissolution/erosion and lead to a significant overestimate of the
carbon input
25
. During the event, the Atlantic CCD shoaled
markedly by at least 2 km (Fig. 1e)
10
, whereas the Pacific CCD
shoaled by only a few hundred metres (see refs 11, 13, 14, 18 and
Supplementary Information).
To simulate the profound differences in observed Atlantic and
Pacific CCD changes, we made additional assumptions on the basis
of earlier suggestions (the model sensitivity to these assumptions
is examined in Fig. 2). First, we assumed a partial carbon injection
directly into the deep Atlantic
3
. Second, we assumed a steady
contribution of North Pacific Deep Water (NPDW) formation
during the event
26
, which makes Atlantic deep waters more
corrosive
18
(the Southern Ocean source remains active but is
reduced relative to its pre-event strength). Both processes enhance
carbonate dissolution in the deep Atlantic. Without NPDW, the
Atlantic CCD shoaling in the model is too small relative to
observations, even at a total carbon input of 4,000 Pg C and
2 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO578
LETTERS
0.8
0.8
0.8
1
1
1
1.2
1.2
1.2
1.4
1.4
1.4
1.6
1.6
1
.6
1.
8
1
.8
2
2.2
Deep Atlantic injection (%)
0
5
10
15
20
25
30
35
40
3
00
300
300
400
400
4
00
500
500
50
0
6
00
600
600
70
0
7
0
0
8
00
800
800
900
90
0
90
0
1,0
00
1,
0
00
2.35
2.4
2.4
2.4
2.5
2.5
2.5
2
.75
2.
7
5
2.75
2.8
5
2.85
2.85
Initial carbon input (Pg C)
Deep Atlantic injection (%)
0
5
10
15
20
25
30
35
40
Deep Atlantic injection (%)
0
5
10
15
20
25
30
35
40
Deep Atlantic injection (%)
0
5
10
15
20
25
30
35
40
50
50
50
1
00
100
1
00
2
00
200
200
3
00
300
3
0
0
40
0
400
400
500
5
00
500
550
550 550
Initial carbon input (Pg C)
1,500 2,000 2,500 3,000 3,500 4,000 4,500
1,500 2,000 2,500 3,000 3,500 4,000 4,500
1,500 2,000 2,500 3,000 3,500 4,000 4,500
1,500 2,000 2,500 3,000 3,500 4,000 4,500
a
b
c
d
Atlantic CCD shoaling (km)
SODW
+NPDW
SODW
+NPDW
Pacific CCD shoaling (m)
Initial carbon input (Pg C)
Initial carbon input (Pg C)
Figure 2 | Simulated CCD shoaling. CCD shoaling as a function of initial carbon input (see Fig. 1a) and %carbon injected into the deep Atlantic. The green
and red areas indicate data compatibility and incompatibility, respectively. a, Atlantic CCD shoaling (in km) with a global Southern Ocean Deep Water
(SODW) source. b, Pacific CCD shoaling (in m) with SODW alone. c, Atlantic CCD shoaling when a contribution of North Pacific Deep Water (NPDW) is
included. d, Pacific CCD shoaling including NPDW.
40% direct injection into the deep Atlantic (Fig. 2a). To match
observations, total input and/or direct injection into the deep
Atlantic must be greater (green and red areas in Fig. 2 indicate
data compatibility and incompatibility, respectively). However,
such scenarios lead to excessive CCD shoaling in the Pacific
(inconsistent with the data; red area in Fig. 2b). Alternatively,
with NPDW, the simulated CCD shoaling in the Atlantic is
consistent with observations for all input scenarios between 1,500
and 4,500 Pg C (Fig. 2c). The maximum initial input is constrained
to ∼3,000 Pg C by the observed CCD shoaling of less than ∼300 m
in the Pacific (green area in Fig. 2d). This estimate of the maximum
initial carbon input is largely independent of its duration (see
Supplementary Fig. S1).
An additional key model benchmark (in addition to simulating
adequate δ
13
C and CCD changes) is to replicate the reconstructed
deep-sea [CO
2−
3
] gradient between the different ocean basins
18
.
This basin gradient was reversed during the PETM relative to the
modern. That is, during the PETM main phase, the most corrosive
deep water resided in the Atlantic and not in the Pacific as today
(Fig. 3a). Observations indicate that deep [CO
2−
3
] in the Pacific was
about 1.5 times higher than in the South Atlantic. This ratio is
reproduced by our model (Fig. 3b).
In summary, using the rate of carbon input shown in Fig. 1a
and the input location and circulation changes discussed above,
the model captures the essential features of the observed carbon
isotope and deep-sea dissolution records. This constrains the initial
PETM carbon input to less than 3,000 Pg C, as a larger input would
lead to more intense dissolution, particularly in the Pacific, which
is not supported by the data. The magnitude of the CIE then
requires the isotopic composition of the carbon input to be lighter
than −50h, consistent with a highly
13
C-depleted source such as
biogenic methane. (Note that methane would have been oxidized
rapidly to CO
2
in the water column and/or the atmosphere
3
.) The
pattern of the carbon input scenario required by the model to match
observations (Fig. 1a) seems to be consistent with carbon release
from oceanic gas hydrate reservoirs. The pulsed input pattern
could indicate carbon release from different ocean basins or depth
horizons containing gas hydrate
21
. The continued release could
be explained by non-steady-state fluxes from marine gas hydrate
systems following the initial dissociation of gas hydrate
27
or fluxes
from marine/terrestrial sedimentary reservoirs.
As a result of the carbon input, we calculate an increase in
atmospheric CO
2
from a baseline of 1,000 ppmv to ∼1,700 ppmv
during the PETM main phase (Fig. 1h) (a baseline pCO
2
several
times higher than the pre-industrial value is generally accepted as
the PETM is superimposed on a much warmer climate
4,24
). Thus,
if initiated at a baseline CO
2
of 1,000 ppmv, CO
2
increases by a
factor of ∼1.7. We found this factor to be largely independent of the
assumed baseline CO
2
, for instance, at 500, 1,000 or 1,500 ppmv.
At the accepted equilibrium climate sensitivities of 1.5–4.5
◦
C
warming per doubling of CO
2
(ref. 1), our calculated 1.7-fold
increase in CO
2
would at most have caused ∼3.5
◦
C warming
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 3
© 2009 Macmillan Publishers Limited. All rights reserved.
LETTERS
NATURE GEOSCIENCE DOI: 10.1038/NGEO578
0.5
1.0
1.5
2.0
PETM
Modern
999
1266
690
1220/1221
1209¬1211
0.5
1.0
1.5
2.0
CRB S.ATL SO IND PAC PAC
CRB S.ATL SO IND PAC PAC
PETM
Modern
t = PEB + 2 kyr
t = PEB + 6 kyr
Deep [CO
2
3
] (modelled, norm.)
¬
Deep [CO
2
3
] (reconstr., norm.)
¬
a
b
Figure 3 | Deep-sea carbonate ion basin gradient during the PETM.
a, Reconstruction based on carbonate records from different sites (colour
symbols, as indicated)
18
. b, Simulation from our carbon cycle model (blue
and red lines) at 2 and 6 kyr after the onset of the PETM (t = 0). The black
bars indicate the range of the simulated gradient at different times. The
open diamonds in a,b show the corresponding gradient based on modern
observations. Deep-sea [CO
2−
3
] concentrations are normalized to the
concentration in the South Atlantic Ocean. CRB: Caribbean, S.ATL: South
Atlantic Ocean, SO: Southern Ocean, IND: Indian Ocean, PAC: Pacific
Ocean, PEB: Palaeocene/Eocene boundary.
1,500 2,000 2,500 3,000
1,350
1,400
1,450
1,500
1,550
1,600
1,650
1,700
1,750
Initial carbon input (Pg C)
0.5
1.0
1.5
2.0
2.5
3.0
3.5
CO
2
Main¬phase atmospheric CO
2
(ppmv)
ΔT, Global (°C)
T
2x
= 3.0 °C
T
2x
= 1.5 °C
T
2x
= 4.5 °C
Figure 4 | Simulated atmospheric CO
2
during the main event. Average
pCO
2
over 70 kyr (see Fig. 1h) as a function of initial carbon input (blue, left
axis); pre-PETM CO
2
is 1,000 ppmv. The red graphs (right axis) show the
calculated global temperature increase based on different climate
sensitivities. Note that whereas peak atmospheric CO
2
is a nonlinear
function of the carbon input, the ‘70 kyr-average’ increase is nearly linear
over the range shown. The percentage of deep Atlantic injection has little
effect on the ‘70 kyr-average’.
during the PETM main phase (Fig. 4). This constitutes an enigma
because proxy records globally indicate surface warming by
5–9
◦
C (refs 5–9). If the temperature reconstructions are correct,
then feedbacks and/or forcings other than atmospheric CO
2
caused
a major portion of the PETM warming. The origin of this
additional warming is unknown at present. Possible causes of
the excess warming include increased production and levels of
trace greenhouse gases as a consequence of the climatic warming
(such as CH
4
; ref. 28). Regardless, this mismatch poses a challenge
for our understanding of past episodes of strong and rapid
global warming. Undoubtedly, the climatic boundary conditions
before the PETM were different from today’s—including different
continental configuration, absence of continental ice and a different
base climate, which limits the PETM’s suitability as the perfect
future analogue. Nevertheless, our results imply a fundamental gap
in our understanding of the amplitude of global warming associated
with large and abrupt climate perturbations. This gap needs to be
filled to confidently predict future climate change.
Methods
The PETM carbon release rate was estimated using our initial carbon input of
3,000 Pg C and an input timescale of the order of 5,000 years (ref. 29), giving a
rate of ∼0.6 Pg C y
−1
. The average carbon release rate from fossil-fuel burning and
cement manufacturing from 1954–2004 is ∼5 Pg C y
−1
(ref. 30).
Given a Palaeocene/Eocene bathymetry
26
, the Pacific seafloor area between
3.5 and 4.5 km depth can be estimated as A
1z
= A
oc
· a
P
· a
1z
' 5.5×10
13
m
2
, where
A
oc
' 3.5 × 10
14
m
2
is the area of the ocean, a
P
' 0.52 is the Palaeocene/Eocene
Pacific fraction and a
1z
' 0.30 is the Pacific area fraction between 3.5 and 4.5 km
depth. The calcite inventory in the top sediment layer of thickness h over this area is
given by M
cal
= A
1z
hρf
c
(1−φ), where ρ = 2,500 kg m
−3
is the sediment density, f
c
is
the calcite dry weight fraction and φ is the porosity. Using f
c
= 0.9 and φ = 0.7 and
converting to carbon units, we have M
C
= M
cal
× 12/100 ' 440 Pg C. The erodible
calcite is larger than the surface inventory by a factor [1 + r
φ
f
c
/(1 − f
c
)], where
r
φ
' 0.4 (ref. 18). Thus, the erodible CaCO
3
inventory in the late Palaeocene Pacific
in the depth range 3.5–4.5 km would have been about 2,000 Pg C.
Restoring the carbonate ion concentration (CO
2
neutralization) by deep-sea
CaCO
3
dissolution in response to CO
2
acidification requires about 0.9 mol
CaCO
3
dissolved per mol CO
2
added. For example, adding 100 µmol kg
−1
CO
2
to a seawater sample at [TCO
2
,TA] = [2.3,2.4] mmol kg
−1
(T,S,P = 10
◦
C, 35,
350 bar), reduces [CO
2−
3
] from 84 to 42 µmol kg
−1
(TCO
2
: total dissolved inorganic
carbon, TA: total alkalinity, T: temperature, S: salinity, P: pressure). Dissolution of
92 µmol kg
−1
CaCO
3
restores [CO
2−
3
] back to 84 µmol kg
−1
.
Received 16 April 2009; accepted 13 June 2009; published online
13 July 2009
References
1. IPCC. Climate Change 2007: The Physical Science Basis (eds Solomon, S., et al.)
(Cambridge Univ. Press, 2007).
2. Zachos, J. C., Dickens, G. R. & Zeebe, R. E. An early Cenozoic perspective on
greenhouse warming and carbon-cycle dynamics. Nature 451, 279–283 (2008).
3. Dickens, G. R. Methane oxidation during the late Palaeocene Thermal
Maximum. Bull. Soc. Geol. Fr. 171, 37–49 (2000).
4. Pagani, M., Calderia, K., Archer, D. & Zachos, J. C. An ancient carbon mystery.
Science 314, 1556–1557 (2006).
5. Kennett, J. P. & Stott, L. D. Abrupt deep-sea warming, palaeoceanographic
changes and benthic extinctions at the end of the Palaeocene. Nature 353,
225–229 (1991).
6. Zachos, J. C. et al. A transient rise in tropical sea surface temperature during
the Paleocene–Eocene Thermal Maximum. Science 302, 1551–1554 (2003).
7. Tripati, A. & Elderfield, H. Deep-sea temperature and circulation changes at
the Paleocene–Eocene Thermal Maximum. Science 308, 1894–1898 (2005).
8. Zachos, J. C. et al. Extreme warming of mid-latitude coastal ocean during the
Paleocene–Eocene Thermal Maximum: Inferences from TEX
86
and isotope
data. Geology 34, 737–740 (2006).
9. Sluijs, A. et al. Subtropical Arctic Ocean temperatures during the
Palaeocene/Eocene Thermal Maximum. Nature 441, 610–613 (2006).
10. Zachos, J. C. et al. Rapid acidification of the ocean during the Paleocene–Eocene
Thermal Maximum. Science 308, 1611–1615 (2005).
11. Lyle, M., Wilson, P. A. & Janecek, T. R. Leg 199 summary. Proc. ODP Init. Rep.
199, 1–87 (2002).
12. Farley, K. A. & Eltgroth, S. F. An alternative age model for the Paleocene–Eocene
Thermal Maximum using extraterrestrial
3
He. Earth Planet Sci. Lett. 208,
135–148 (2003).
13. Colosimo, A. B., Bralower, T. J. & Zachos, J. C. in Proc. ODP, Sci. Results
Vol. 198 (eds Bralower, T. J., Premoli Silva, I. & Malone, M. J.) (Texas A&M
Univ., 2005).
14. Murphy, B., Lyle, M. & Lyle, A. O. in Proc. ODP, Sci. Res. (eds Wilson, P. A.,
Lyle, M. & Firth, J. V.) 1–12 (Texas A&M Univ., 2006).
4 NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience
© 2009 Macmillan Publishers Limited. All rights reserved.
NATURE GEOSCIENCE DOI: 10.1038/NGEO578
LETTERS
15. Thomas, D. J. et al. Warming the fuel for the fire: Evidence for the thermal
dissociation of methane hydrate during the Paleocene–Eocene Thermal
Maximum. Geology 30, 1067–1070 (2002).
16. Nunes, F. & Norris, R. D. Abrupt reversal in ocean overturning during the
Palaeocene/Eocene warm period. Nature 439, 60–63 (2006).
17. Röhl, U., Westerhold, T., Bralower, T. J. & Zachos, J. C. On the duration of the
Paleocene–Eocene Thermal Maximum (PETM). Geochem. Geophys. Geosyst. 8,
Q12002 (2007).
18. Zeebe, R. E. & Zachos, J. C. Reversed deep-sea carbonate ion basin-gradient
during Paleocene–Eocene Thermal Maximum. Paleoceanography 22,
PA3201 (2007).
19. Dickens, G. R., O’Neil, J. R., Rea, D. K. & Owen, R. M. Dissociation of oceanic
methane hydrate as a cause of the carbon isotope excursion at the end of the
Paleocene. Paleoceanography 10, 965–971 (1995).
20. Zeebe, R. E., Zachos, J. C., Caldeira, K. & Tyrrell, T. Oceans: Carbon emissions
and acidification. Science 321, 51–52 (2008).
21. Dickens, G. R. Carbon addition and removal during the Late Palaeocene
Thermal Maximum: Basic theory with a preliminary treatment of the isotope
record at ODP Site 1051, Blake Nose. Geol. Soc. Lond. Spec. Publ. 183,
293–305 (2001).
22. Bowen, G. J. et al. A humid climate state during the Paleocene–Eocene Thermal
Maximum. Nature 432, 495–499 (2004).
23. van Andel, T. H. Mesozoic/Cenozoic calcite compensation depth and the
global distribution of calcareous sediments. Earth Planet. Sci. Lett. 26,
187–194 (1975).
24. Tyrrell, T. & Zeebe, R. E. History of carbonate ion concentration over the last
100 million years. Geochim. Cosmochim. Acta 68, 3521–3530 (2004).
25. Panchuk, K., Ridgwell, A. & Kump, L. R. Sedimentary response to
Paleocene–Eocene Thermal Maximum carbon release: A model-data
comparison. Geology 36, 315–318 (2008).
26. Bice, K. L. & Marotzke, J. Could changing ocean circulation have destabilized
methane hydrate at the Paleocene/Eocene boundary? Paleoceanography 17,
1018 (2002).
27. Dickens, G. R. Rethinking the global carbon cycle with a large, dynamic
and microbially mediated gas hydrate capacitor. Earth Planet. Sci. Lett. 213,
169–183 (2003).
28. Sloan, L. C. et al. Possible methane-induced polar warming in the early Eocene.
Nature 357, 320–322 (1992).
29. Röhl, U., Bralower, T. J., Norris, R. D. & Wefer, G. New chronology for the late
Paleocene thermal maximum and its environmental implications. Geology 28,
927–930 (2000).
30. Marland, G., Boden, T. A. & Andres, R. J. Global, Regional, and National CO
2
Emissions (CDIAC, ORNL, US DOE, 2007).
Acknowledgements
The research was supported by NSF grant EAR06-28719 to J.C.Z. and
EAR06-28394 to R.E.Z.
Additional information
Supplementary information accompanies this paper on www.nature.com/naturegeoscience.
Reprints and permissions information is available online at http://npg.nature.com/
reprintsandpermissions. Correspondence and requests for materials should be
addressed to R.E.Z.
NATURE GEOSCIENCE | ADVANCE ONLINE PUBLICATION | www.nature.com/naturegeoscience 5
© 2009 Macmillan Publishers Limited. All rights reserved.
