1
SALINITY-OXYGEN 18 RELATIONSHIP SIMULATED BY AN
OCEANIC GENERAL CIRCULATION MODEL
Gilles DELAYGUE 1,2 ,
phone (+33) 1 69 08 45 44 fax (+33) 1 69 08 77 16
[email protected] *
Jean JOUZEL1 ,
phone (+33) 1 69 08 77 13 fax (+33) 1 69 08 77 16
[email protected]
Jean-Claude DUTAY1
phone (+33) 1 69 08 31 12 fax (+33) 1 69 08 77 16
1
[email protected]
Laboratoire des Sciences du Climat et de l’Environnement (UMR N°1572), CEA
SACLAY , Orme des Merisiers , 91191 GIF Cedex, France.
2
also at Centre Européen de Recherche et d’Enseignement en Géosciences de
l’Environnement (UMR N°6635) , Europôle de l’Arbois BP 80 , 13545 AIX-EN-PROVENCE
Cedex 04, France.
(4200 words)
Abstract.
The distribution of H218O in the ocean is simulated using an oceanic
general circulation model (GCM). Surface isotopic fluxes prescribed to this model are
derived from an atmospheric GCM, which ensures their consistency. The oceanic GCM
is initialized with an homogeneous composition, and mixes its isotopic gradients so as to
equilibrate with the atmosphere. A good agreement is found between observations and
simulation concerning i) the interbasin H218O gradients, especially the Atlantic being more
enriched than the Pacific, as well as the meridional gradients; and ii) the H218O - salinity
relationship.
* to whom correspondence should be addressed
2
1. Introduction
Information inferred from distributions of stable water isotopes (HDO and/or H218O)
are extensively used by paleoclimatologists. This includes both scientists interested in
reconstructing past climates in polar regions and over the continents, as well as
paleoceanographers aiming to reconstruct past ice volume changes, paleotemperatures and
paleosalinities. In this latter approach, most of the relevant paleoceanographic data come
from the measurement of the oxygen 18 content of the fossil carbonate in foraminifera
(benthic and planktic) and in corals.
The carbonate isotopic content depends on three factors: the temperature -the watercarbonate isotopic fractionation being temperature dependent-, the seawater isotopic
content, and -secondarily- the specy. Once accounted for any specy offset, the change in
carbonate oxygen 18 gives access to either the temperature, assuming that the seawater
isotopic change is known from the benthic foraminiferas, or to the seawater isotopic
variation providing temperature change is independently derived. In both cases, the
seawater isotopic change comprises a global part due to the excess ice stored on the
continent, and a local part due to the evaporation minus precipitation balance -sometimes
called the "salinity effect' because salinity changes in the same way- as well as to the
oceanic circulation. Thus in order to assess either temperature, ice volume or salinity
variations from carbonate oxygen 18, some assumptions about the isotopic variations of
seawater are necessary: usually that the ‘global’ variation due to ice volume is
homogeneously distributed in the oceans, or that the gradients (horizontal and vertical) of
the oxygen 18 in water have not changed [1].
Largely thanks to the pioneering work of Craig and Gordon [2] and to the GEOSECS
program [3] both the water oxygen 18 distribution and the oxygen 18-salinity relationship
are documented at a global scale for present-day ocean. The driving factors are also
relatively well understood: the fractionation at the surface due to atmospheric exchanges
and the mixing by the ocean mentioned above, but also river runoff in coastal areas and
ice processes (sea ice formation and iceberg discharge) in polar regions. This basic
3
understanding is however not sufficient for an entirely correct interpretation of paleodata.
For example [4], the difference of salinity (and of isotope) between surface and deep
waters has, for a given ocean, a different value for Present-day and Last Glacial
Maximum (21000 years Before Present) climates. Recently, Rohling and Bigg [5]
critically assessed the reconstruction of paleosalinity based on the assumption of a
constant and linear oxygen 18-salinity relationship. Instead, they concluded that both the
spatial and temporal variabilities of this relationship need to be constrained.
One way to approach these problems is to model the oxygen 18 distribution in the
ocean and its relationship with salinity for different climates. This strategy, parallel to that
followed to interpret polar and continental paleoprecipitation [6], has been suggested by
various authors [7,5] . The most promising results have been obtained by Schmidt [8,9,
hereafter SC98] using an ocean model derived from the GISS (Goddard Institute for
Space Studies) coupled general circulation model (GCM) [10].
Hereafter, we follow a similar modelling approach, with a different oceanic GCM.
Our work is specific in that 1) we use consistent isotopic fluxes predicted by an
atmospheric GCM implemented with water isotopes (whereas SC98 parameterizes these
fluxes with respect to climatic variables), 2) we perform long experiments (> 2000 years)
which allows us to investigate oxygen 18 distributions in surface to deep waters (instead
runs presented in SC98 cover less than 100 years which limits interpretation to surface
characteristics), and 3) the ocean starts with an homogeneous isotopic composition (the
oceanic composition in SC98 is instead initialized to observed values), which is
somewhat an insurance that realistic mechanisms are simulated by the model. In this
article, we discuss the results of a present-day simulation. We show that the main
observed characteristics of the oxygen 18 distribution and of the oxygen 18-salinity
relationship are reasonably captured by our isotopic ocean model.
4
2. The atmospheric fluxes and the oceanic tracers
Atmospheric fluxes (evaporation and precipitation) are taken from the isotopic
version of the NASA/GISS atmospheric GCM (8°x10°) developed by Jouzel et al. [11].
This ensures the consistency between water (evaporation and precipitation) and isotopic
fluxes: this is of primary importance here since our tracers originate in the balance of these
fluxes, as discussed in Juillet-Leclerc et al. [7]. Continental runoff (including iceberg
discharge) is taken from a work of Russell and Miller [12] who modelled the continental
drainage using the GISS GCM. The annual isotopic content of the river discharges is
calculated by combining the GISS isotopic outputs with the drainage map of Russell and
Miller. Hydrological fluxes due to sea ice formation and melting are not taken into account
because of their complex seasonal variations. Although they play an important role in
forming bottom waters, they may have a limited effect on both salinity and oxygen 18
characteristics of these waters in the real world [13,14].
The oxygen 18 flux to the oceanic surface is expressed as :
F
18O
=
E. (RS - RE) - P. (RS - RP) - R . (RS - RR)
where RS is the isotopic ratio of the oceanic Surface, RE , RP and RR are the isotopic
ratios of Evaporation (E), Precipitation (P) and Runoff (R).
These fluxes are prescribed monthly to a tracer version of the OPA oceanic GCM of
the Laboratoire d’Océanographie Dynamique et de Climatologie (LODyC, Paris-France),
a primitive equation model described by Madec et al [15]. The degraded tracer version
used here has a medium horizontal 92x77 resolution. Atmosphere-ocean interaction is
described in detail by a fine subsurface resolution (10 levels in the first 100m) and a
turbulent kinetic energy parameterization of the vertical diffusion. This version benefits
from its parent (180x150) dynamics averaged on its own resolution, and reasonably
simulates distribution of geochemical tracers like 14C [15].
Since an important issue of this work is the oxygen 18 - salinity relationship, it is
necessary to simulate a salinity tracer which is consistent with the oxygen 18 for what
concerns their atmospheric fluxes and oceanic mixing. Therefore, we define another tracer
5
by its atmospheric fluxes similar to
F 18O above, but with Rs being the surface salinity
and others R equal to zero. This passive tracer, called ‘salinity’ in the following, has no
dynamical effect, but is transported by the model in the same way than the dynamical
salinity. The basic understanding of these atmospheric fluxes is that evaporation tends to
enrich the surface in salt as well as in oxygen 18 wrt oxygen 16, contrary to precipitation
and runoff.
The simulation begins with an homogeneous oceanic composition of 34.6 permil for
the salinity and V-SMOW (Vienna-Standard Mean Ocean Water with an H218O/H216O
ratio, RSMOW, of 2005.2*10-6) for the oxygen isotopic ratio. The model is first run, for a
spin-up period of 2000 years, with a lower resolution version (47x39) but with the same
dynamics degraded to this resolution. Outputs from this simulation serve as initial
conditions for the 92x77 version run, which lasts 200 years (this acceleration procedure is
described in Aumont et al.[16]). At this time, because the atmospheric fluxes are not
corrected and thus not exactly balanced, a slight drift of the salt and isotope masses
remains, but it is considered as negligeable for our first order study: over the last 100
years of simulation, both salinity and oxygen isotopic ratio increase by only 0.004 permil
in the first layer. Averaged over the whole oceanic surface, the precipitation flux is about
1.22 m/yr, evaporation 1.28 m/yr and runoff 0.09 m/yr, so that the global water flux
imbalance is +.03 m/yr. The time step for the tracer transport is one day.
To illustrate how the atmospheric fluxes (evaporation, precipitation and runoff) force
the oceanic surface composition, isotopic content and salinity are estimated without
oceanic horizontal advection (Figures 1a and 1b with the isotopic content expressed as
the deviation δ wrt the V-SMOW, in permil : δ = (R/RSMOW - 1)*1000). This estimation
is established from the conceptual 2-box model described by Craig and Gordon [2]:
variations of the surface box composition are forced by atmospheric fluxes and buffered
by the exchange with the other box considered as a reservoir. The model parameters are
derived from an empirical correlation between the GISS atmospheric fluxes and
GEOSECS observed δ18O, computed by Juillet-Leclerc et al. [7]. They infer a vertical
6
mixing flux of 4.8 10-4 m2/s, and a deep water δ18O of 0.22 permil. To compute the
salinity, we use a deep value of 34.6 permil. This box model is expected to represent an
ideal ocean where the prescribed atmospheric fluxes are balanced only by vertical mixing,
so that the simulated tracer fields are representative of the atmospheric forcing. The main
enrichment zones are centered above subtropics, along both basin sides in the Pacific,
along the south-east side and north-west side (Gulf of Mexico) in the Atlantic. The Indian
and western Pacific equatorial zones show up a strong depletion due to precipitation, not
seen in the Atlantic. In subglacial latitudes, the limit between enrichment and depletion lies
around 40°. In higher latitudes, fluxes are markedly lower because the sea ice cover
artificially prevents them in the model, except where runoff from rivers (northern
hemisphere) or ice shelf (southern hemisphere) is prescribed.
3. Isotopic field and relation to salinity :
The seasonal variability of salinity and oxygen 18 will not be described here since
we are interested only in their average characteristics. An important point is that the
simulated oxygen 18 - salinity relationship, on a basin scale, are rather stable over the
seasons. Overall the salinity variability seems underestimated, probably because the
coarse resolution of the atmospheric GCM somehow smooths the variability of the
prescribed fluxes.
Figures 1c and 1d show the surface (first model level) salinity and oxygen 18 fields
simulated by the GCM after 2200 years. Characteristics similar to those of the
atmospheric fluxes (Figures 1a and 1b) are obvious: a strong latitudinal gradient results
from enriched subtropics and depleted higher latitudes, with intermediate values in the
intertropical zone except in the Atlantic which is much more enriched than the two others
basins. These features are summarized in Figure 2 which compares the latitudinal
distribution of the oxygen 18 GEOSECS surface data [3] with the modelled ones.
Although these data are not likely to represent the whole ocean, they give an idea of the
7
spatial variability of its isotopic content. There is a quite good agreement between
observed and modelled values, especially in the Atlantic, although the latter are too high in
the subtropics. In the Pacific, the very low modelled values in the intertropical zone
correspond to the monsoonal south-eastern Asia characterised by a strong P-E budget (PE> 1.3 m/year). The large number of points there is due to the highest spatial resolution
of the model in the lowest latitudes. The Atlantic is clearly more enriched, by more than
0.5 ‰, compared to the Pacific. This difference corresponds to a net imbalance of the
basin moisture exchanges, as simulated by the GISS AGCM [17].
Deeper in the ocean, waters are generally too enriched (by .05-.1 ‰) due to some
bias in either the model circulation or the atmospheric fluxes. Figures 3 and 4 compares
the GEOSECS compositions of deep waters with simulated ones (nearest grid point of the
model). In the Atlantic basin (Figure 3b), the model clearly forms two deep water masses
comparable to the North Atlantic Deep water and the Antarctic Bottom Water apparent in
the GEOSECS profiles (Figure 3a), but the AABW-like water is too enriched by .3 ‰ in
the model. This is possibly due to deficiency of the oceanic model in ventilating the deep
waters in high latitudes. Note that the Craig and Gordon [2] mean values for deep waters
correspond to the GEOSECS measurements for the NADW (around 0.2 ‰ wrt VSMOW), but not for the AABW (estimated at -0.35‰ wrt V-SMOW by Craig and
Gordon from measurements in the Weddell Sea, and slightly higher - by .1‰ - from the
GEOSECS campaign). The simulated surface values are also consistent with the observed
ones: in northern high latitudes, surface waters are enriched compared to deeper ones
because of the north Atlantic drift; conversely to southern high latitudes where surface
waters are depleted by the Antarctic runoff and the precipitation minus evaporation
imbalance. In the Pacific ocean, the few GEOSECS vertical profiles (5 deep profiles,
located in high south latitudes) do not allow a precise comparison, we just note that the
simulated deep water compositions are quite close to these observed ones (Figures 4a and
4b).
The relationship between oxygen 18 and salinity in surface waters has been
extensively studied by Craig and Gordon [2] with the currently available data. Since then,
8
GEOSECS expeditions and limited other ones have brought new data, without any global
analysis of the isotopic field. It is far beyond the scope of this paper to do such an
analysis, but a striking feature coming out of these data is that a slope characteristic of
each basin does not appear so clearly as in Craig and Gordon study. For instance a slope
increase in high latitudes has already been shown by Fairbanks et al. [1]. Thus
comparison with previously published slopes must be considered with care. Also, a
correction of the δ18O values published by Craig and Gordon in 1965 (of -0.1 ‰ wrt the
V-SMOW as explained by H. Craig in Östlund et al., [3], p.7) must be recalled. The
relationship between surface δ18O and salinity is represented in Figure 3 for the Atlantic
and 4 for the Pacific. The agreement with the regressions proposed by Craig and Gordon
is quite good in each basin, but not as good compare to the GEOSECS values. Especially,
the model predicts a higher slope close to 0.7 in latitudes south of 40°S which does not
clearly appear in this set of data.
4. Atmospheric vs oceanic control of the surface :
Following Juillet-Leclerc et al. [7], we try to separate the effects of atmospheric
fluxes and oceanic circulation on determining both the surface fields and their
relationship. We base our reasoning on the salinity, which can be compared to observed
values over the whole ocean, by contrast to the less documented oxygen 18. Both fields
respond to similar forcings and are qualitatively very close. Figure 1a shows the influence
of atmospheric fluxes alone on salinity, 1c includes the full oceanic circulation, and 1e
represents climatological observations of Levitus [18]. Although the box model (Figure
1a) captures some basic features of the salinity field, ie the zonal distribution and the
Atlantic enrichment, the GCM-simulated field is much more closer to the observations.
The ocean dynamics readily explains the difference. The tropical gyres smooth and
displace the maximal subtropical values by transporting depleted water i) from the
equatorial band along the western part of the basins and ii) from the high latitudes along
9
the eastern parts. Upwelling of deep waters characterised by mean values contributes to
strongly smooth surface waters, either depleted as in the equatorial Pacific or enriched as
along the western coasts of Africa and South America. In the north Atlantic basin, the net
transport of surface waters from the south increases the salinity, and the north Atlantic
drift brings enriched waters northern than 60°N. The same differences exist for the
oxygen 18 between figures 1b and 1d, which is a good argument for the validity of the
simulated oxygen 18.
The box model approach also allows to precise the origin of the oxygen 18 - salinity
relationship: since horizontal advection does not mix the different boxes, a linear
relationship arises only if it does exist within each box from the mixing between
atmospheric fluxes and the deep water reservoir. Actually the same slopes are simulated
by the box model (Figures 3b and 4b, small points), but with two differences. First the
scatter of grid points is higher. Secondly the slopes for highest salinities are lower,
decreasing the overall slope lower. These differences can be conceptually explained by the
oceanic mixing limiting the range of values created by the different atmospheric forcings.
This helps explain the low slopes for the highest salinity, a characteristics of tropical
surface waters which disappears by mixing with equatorial and temperate latitudes. Thus
the oxygen 18 - salinity relationship can be considered as a mixing line, between a ‘pure’
oceanic source (close to the average deep ocean) and a ‘pure’ freshwater source consisting
in the atmospheric isotopic fluxes. As explained by Craig and Gordon [2], the continental
runoff can represent an important contribution to this freshwater source, and play a role in
setting the oxygen 18 - salinity slope, only in coastal region and not in the open ocean.
This is the case for instance in the Bay of Bengal where observed and simulated
relationships are different (see Delaygue et al., submitted to JGR Ocean).
Modelling the oceanic distribution of oxygen 18 also allows to separate the different terms
of its transport, between diffusion, advection and atmospheric fluxes. Although the
advective fluxes penetrating the sides of each surface box are individually three orders of
magnitude higher than the other fluxes, they compensate each other. It happens that the
net advective flux is of the same order of magnitude, but generally lower, than the
10
diffusive (mainly vertical) and atmospheric fluxes which thus primarily control the
isotopic composition of the surface, as in the box model. This advective term is higher
than the atmospheric one in area where the simulated salinity and oxygen 18 present
strong horizontal gradients (Figure 1b) : this happens at temperate and equatorial latitudes
where the atmospheric flux sign changes from positive (tropical zone) to negative
(high/equatorial latitudes) values.
Conclusion :
We have implemented the oxygen 18 water isotope cycle in a tracer version of an
oceanic GCM in order to simulate the oceanic distribution of this water isotope and its
relationship with salinity for present-day conditions. Using as boundary conditions the
outputs of an atmospheric GCM implemented with water isotopes, we have performed a
long simulation (2200 years) which allows us to examine properties of both surface and
deep oceanic waters. The global oxygen 18 distribution in surface waters is satisfyingly
simulated both for its latitudinal distribution and interbasin characteristics. The oxygen 18
- salinity relationship compares well with available data although noticeable deficiencies
are noted for high latitudes where the slope is too high. Comparison of surface properties
(oxygen 18 and salinity) as simulated by the Craig and Gordon [2] two-box model and by
our isotopic OGCM illustrates the relative influence of surface fluxes and of the oceanic
circulation on these properties. Vertical profiles also favorably compare with limited
available observations (although with a slight bias). Overall, our simulation confirms that
global ocean modelling is a promising way to study the oceanic oxygen 18 distribution
and its relation to salinity.
Various directions can be envisaged for future research in this field. First, it will be
useful, as a sensitivity test, to perform experiments with fluxes (evaporation, precipitation
and runoff) produced by a different atmospheric GCM (for example by the isotopic
version of the ECHAM model, [19]). Second, the development of an isotopic version of
11
a coupled atmospheric-oceanic GCM will certainly represent a stepforward with respect
to our current approach which consists in using uncoupled atmospheric and oceanic
isotopic GCMs. However, the results obtained with our current approach are already
quite satisfying and well adapted to simulate a glacial state using appropriate oceanic and
atmospheric fluxes. This is important to test the hypothesis used to reconstruct past SSTs,
i.e. that the deep to surface isotopic difference holds in time. Also investigating the
difference between spatial and temporal oxygen 18 salinity relationships is critical in order
to reconstruct past salinity. Our model is also well adapted to explore how isotopic fields
behave when a meltwater pulse disrupts the isotopic balance at the oceanic surface.
Acknowledgments. The authors are grateful to O. Aumont and J. Orr for the model use,
to G. Schmidt for fruitful discussion, to R. Koster for providing GISS outputs and to E.
Bard for stimulation. Simulations were run at CEA Grenoble on CRAY C90.
12
LIST OF FIGURES
FIGURE 1 :
Results from the 2200 year simulation with OPA oceanic GCM and from a 2-box model
(where a deep homogeneous reservoir balances the atmospheric fluxes). All fields
concern the surface (i.e. the first 10 m level) and are annual averages. Salinity is
expressed in psu; oxygen 18 in deviation to V-SMOW (in permil).
a. salinity simulated with a box model
b. oxygen 18 simulated with a box model
c. salinity simulated with the oceanic GCM
d. oxygen 18 simulated with the oceanic GCM
e. salinity compilation from observations [18]
FIGURE 2 :
Latitudinal gradients of zonally averaged surface oxygen 18, from GEOSECS [3] and our
GCM simulation.
FIGURE 3 :
Oxygen 18 - salinity relationship in the Atlantic surface, from a) data [2,3] and b)
simulations (OPA; box model). Data from the first 200 m for GEOSECS, the first level
(10 m) for OPA.
Values of deep waters (depth< 3000 m) are also indicated. Model deep values correspond
to the nearest grid location.
FIGURE 4 :
Same as Figure 3 but for the Pacific ocean.
13
REFERENCES
[1] R.G. Fairbanks, The origin of continental shelf and slope water in the New York
Bight and Gulf of Maine: evidence from H218O/H216O ratio measurements, J. Geophys.
Res. 87 (1982) 5796-5808.
[2] H. Craig, L.I. Gordon, Deuterium and Oxygen 18 variations in the Ocean and the
Marine Atmosphere, in: E. Tongiorgi (Ed.), Stables Isotopes in Oceanographic Studies
and Paleotemperatures, Consiglio Nazionale delle Ricerche, Laboratorio di Geologia
Nucleare, Pisa, Spoleto, Italy, 1965, pp. 9-130.
[3] H. Östlund, H. Craig, W.S. Broecker, D. Spenser, GEOSECS Atlantic, Pacific and
Indian ocean expeditions, IDOE NSF, Washington D. C., 1987.
[4] W.S. Broecker, Oxygen isotope constraints on surface ocean temperatures, Quater.
Res. 26 (1986) 122-134.
[5] E.J. Rohling, G.R. Bigg, Paleosalinity and δ18O : a critical assessment, J. Geophys.
Res. 103 (1998) 1307-1318.
[6] J. Jouzel, G. Hoffmann, R. Koster, V. Masson, Water isotopes in precipitation :
Data/model comparison for present-day and past climates, Quat. Sci. Rev. in press
(1999).
[7] A. Juillet-Leclerc, J. Jouzel, L. Labeyrie, S. Joussaume, Modern and last glacial
maximum sea surface δ18O derived from an Atmospheric General Circulation Model,
Earth Planet. Sci. Lett. 146 (1997) 591-605.
14
[8] G.A. Schmidt, Oxygen-18 variations in a global ocean model, Geophys. Res. Lett.
25 (1998) 1201-1204.
[9] G.A. Schmidt, Oxygen-18 tracers in an ocean GCM: implications for interpreting
proxy data, in: 6th International Conference on Paleoceanography, Lisbon, 1998.
[10] G.L. Russell, J.R. Miller, D. Rind, A coupled atmosphere-ocean model for transient
climate change, Atmosphere-Ocean 33 (1995) 683-730.
[11] J. Jouzel, G.L. Russell, R.J. Suozzo, R.D. Koster, J.W.C. White, W.S. Broecker,
Simulations of the HDO and H218O atmospheric cycles using the NASA GISS general
circulation model : the seasonal cycle for present-day conditions, J. Geophys. Res. 92
(1987) 14739-14760.
[12] G.L. Russell, J.R. Miller, Global river runoff calculated from a global atmospheric
circulation model, J. Hydro. 117 (1990) 241-254.
[13] J.R. Toggweiler, B. Samuels, Effect of sea ice on the salinity of Antarctic Bottom
Waters, J. Phys. Oceano. 25 (1995) 1980-1996.
[14] A.P. Worby, N.L. Bindoff, V.I. Lytle, I. Allison, R.A. Massom, Winter ocean/sea
ice interactions studied in the East Antarctic, EOS Transactions 77 (1996).
[15] G. Madec, P. Delecluse, M. Imbard, C. Lévy, OPA 8.1 ocean general circulation
model reference manual, Notes du Pôle de modélisation, IPSL (PDF-version available at:
http://www.ipsl.jussieu.fr/modelisation/notes/OPA8.1-Total.pdf), 1998.
15
[16] O. Aumont, J. Orr, D. Jamous, P. Monfray, G. Madec, O. Marti, A degradation
approach to accelerate simulations to steady-state in a 3-D Tracer Transport model of the
Global Ocean, Clim. Dyn. 14 (1998) 101-116.
[17] F. Zaucker, W.S. Broecker, The influence of atmospheric moisture transport on the
fresh water balance of the Atlantic drainage basin : general circulation model simulations
and observations, J. Geophys. Res. 97 (1992) 2765-2773.
[18] S. Levitus, Climatological atlas of the world ocean, technical report NOAA prof.
paper 13, NOAA, Washington D. C., 1982, 173 pp.
[19] G. Hoffmann, M. Werner, M. Heimann, Water isotope module of the ECHAM
atmospheric General Circulation Model: A study on timescales from days to several years,
J. Geophys. Res. 103 (1998) 16871-16896.
2
2
0
ATLANTIC
-1
-2
1
0
Figure 2
1
δ18O (permil/ SMOW)
δ18O (permil/ SMOW)
PACIFIC
-1
-2
OPA
GEOSECS (depth<200m)
-3
-80 -60 -40 -20 0
20
latitude
OPA
GEOSECS (depth<200m)
40
60
80
-3
-80 -60 -40 -20
0
20
latitude
40
60
80
δ18O (‰)
-2
33
34
Figure 3a
35
36
37
-2
-1.5
Regression of surface data
N-Atlantic: s=0.59 (R=.96)
S-Atlantic: s=0.52 (R=.94)
-0.5
0
-1.5
deep south waters
deep north waters
0.5
1
-1
32
- ATLANTIC OBSERVATIONS GEOSECS:
surface data
deep waters
CRAIG & GORDON
North Atlantic
1.5
-1
-0.5
0
0.5
1
1.5
δ18O (‰)
-2.0
33
34
Figure 3b
35
Salinity (‰)
36
37
-2.0
-1.5
-1.5
Regression for OPA surface points:
N-Atlantic: s=0.55 (R=.98)
S-Atlantic: s=0.70 (R=.97)
-1.0
.0
-1.0
south deep waters
north deep waters
.5
1.0
1.5
-.5
32
- ATLANTIC SIMULATIONS OPA surface points
OPA deep waters
box model surface points
-.5
.0
.5
1.0
1.5
δ18O (‰)
-2
33
34
Figure 4a
35
Salinity (‰)
36
37
-2
-1.5
-1.5
Regression of surface data
N-Pacific: s=0.49 (R=.96)
S-Pacific: s=0.47 (R=.93)
-1
-1
south deep waters
0
0.5
1
-0.5
32
- PACIFIC OBSERVATIONS GEOSECS:
surface data
deep waters
CRAIG & GORDON
North Pacific
South Pacific
1.5
-0.5
0
0.5
1
1.5
δ18O (‰)
-2.0
-1.5
-1.0
-.5
.0
.5
1.0
1.5
32
33
Figure 4b
south deep waters
36
Regression for OPA surface points:
N-Pacific: s=0.59 (R=.97)
S-Pacific: s=0.66 (R=.98)
34
35
Salinity (‰)
- PACIFIC SIMULATIONS OPA surface points
OPA deep points
box model surface points
37