Clues to Ocean History: a Brief Overview of Proxies
35
poral variations in 015N of the downward
particulate flux appear to concur with the measured and modelled efflux of CO 2 , Np and CH 4
during monsoonal upwelling. As a result, the analysis of olSN continues to provide new possibilities
for constraining the carbon cycle in ways not provided by any other tracer, and to test, e.g. the
denitrification hypothesis put forward by Berger and
Keir (1984) and the iron fertilization hypothesis (increased nutrient utilization in Polar oceans) of
Martin and Fitzwater (1988).
Toward Reconstruction of Surface Water
PCO z (and Atmospheric pCOJ
A number of reconstructions of the partial pressure
of CO in the ocean and in the atmosphere, have
been ;~empted using sediment properties which
were thought to be proxies for carbon dioxide. One
of the earliest attempts (Shackleton and Pisias
1985) relied on ~Ol3C between planktic and benthic
foraminifers in the eastern equatorial Pacific, in
combination with the conceptual two-layer ocean
model of Broecker (1982). The argument strongly
depends on the assumption that ~Ol3C represents
productivity and nutrient content in the ocean, directly correlated to pC0 2 through the Broecker
model. Thus, the soundness ofthe reconstruction
is dependent on how well a core in a particular
(productive) region represents global conditions,
and whether the Broecker model provides a good
theoretical basis. Both of these postulates are subject to considerable doubt (Berger et al. 1996). The
success of the reconstruction can be verified by
tuning with the ice-core CO 2 from Antarctica
(Bamolaetal. 1987; Jouzel etal. 1993)(excluding
poor fits resulting from incorrect time scales). The
result of this exercise yields a correlation coefficient of 0.7 (r2=0.5), for the regression of ~Ol3C
onpC0 2 •
A model-free attempt to link sea-level fluctuation to atmospheric CO 2 by multiple regression of
standard ocean 0 18 0 values (Berger et al. 1996),
yielded a correlation coefficient of 0.9 (r=0.8). The
relationship is of the form:
(7)
It says that the atmospheric content of carbon
dioxide is especially sensitive to sea-level change
whenever sea level is high, and that any sudden
change of sea level provokes a rise in CO 2 , Within
this formulation (Berger et al. 1996),0 18 0 is used
as a proxy for atmospheric CO 2 , but it is only the
global 0 18 0 which is of interest, as a proxy for sea
level and overall temperature.
It has been found that the ol3C-values within
phytoplankton (013C O ,g)' as well as suspended and
sedimentary organic carbon vary with the ambient
concentration of CO 2 in surface waters (Rau et al.
1989; Rau 1994; Jasper and Hayes 1990;
Freemann and Hayes 1992; Bentaleb et al. 1996;
Fischer et al. 1997, 1998) (Fig. 20). The
fractionation, which results in an enrichment of the
12C isotope during carbon fixation, proceeds from
a pool of dissolved carbon in contact with the organism. It was observed that at high CO 2 concentrations, the fractionation Ep (i.e. the difference
between ol3C O ,g and the source=ol3C C02 ) is large
and vice versa. To calculate the Ol3C of dissolved
CO 2 , the Ol3CL C02 has to be determined, for example using foraminifersl-8 l3 C values (vital effects
have to be known) as well as the sea surface temperature. Applying relationships between isotope
fractionation and [C0 2 (aq)] to sediment cores,
several studies aimed to reconstruct past PC0 2
levels in the surface waters (e.g. MUller et al. 1994;
Bentaleb et al. 1996) (Fig. 21). It must be remembered that much of these reconstructions depend
on the paleotemperature record (e.g. derived from
alkenone data) rather than on the Ol3C variation.
Based on mechanistic considerations, a two-step
fractionation model and the assumption of exclusive CO 2 (aq) utilization, Farquhar et al. (1982) argued that a hyperbolic function relates the carbon
isotopic composition to CO 2 (aq). Others used a
relationship that fits best to the observed data sets.
The empirically based calibrations were linear (e.g.
Rau 1994) or log-linear (e.g. Freeman and Hayes
1992). However, more recent laboratory experiments (Laws et al. 1995; Bidigare et al. 1997; Popp
et al. 1998) show that for instance growth rate, cell
geometry and membrane diffusivity can play an important role confounding a simple use of the
paleobarometer to reconstruct the surface water
PC0 2 •
35
poral variations in 015N of the downward
particulate flux appear to concur with the measured and modelled efflux of CO 2 , Np and CH 4
during monsoonal upwelling. As a result, the analysis of olSN continues to provide new possibilities
for constraining the carbon cycle in ways not provided by any other tracer, and to test, e.g. the
denitrification hypothesis put forward by Berger and
Keir (1984) and the iron fertilization hypothesis (increased nutrient utilization in Polar oceans) of
Martin and Fitzwater (1988).
Toward Reconstruction of Surface Water
PCO z (and Atmospheric pCOJ
A number of reconstructions of the partial pressure
of CO in the ocean and in the atmosphere, have
been ;~empted using sediment properties which
were thought to be proxies for carbon dioxide. One
of the earliest attempts (Shackleton and Pisias
1985) relied on ~Ol3C between planktic and benthic
foraminifers in the eastern equatorial Pacific, in
combination with the conceptual two-layer ocean
model of Broecker (1982). The argument strongly
depends on the assumption that ~Ol3C represents
productivity and nutrient content in the ocean, directly correlated to pC0 2 through the Broecker
model. Thus, the soundness ofthe reconstruction
is dependent on how well a core in a particular
(productive) region represents global conditions,
and whether the Broecker model provides a good
theoretical basis. Both of these postulates are subject to considerable doubt (Berger et al. 1996). The
success of the reconstruction can be verified by
tuning with the ice-core CO 2 from Antarctica
(Bamolaetal. 1987; Jouzel etal. 1993)(excluding
poor fits resulting from incorrect time scales). The
result of this exercise yields a correlation coefficient of 0.7 (r2=0.5), for the regression of ~Ol3C
onpC0 2 •
A model-free attempt to link sea-level fluctuation to atmospheric CO 2 by multiple regression of
standard ocean 0 18 0 values (Berger et al. 1996),
yielded a correlation coefficient of 0.9 (r=0.8). The
relationship is of the form:
(7)
It says that the atmospheric content of carbon
dioxide is especially sensitive to sea-level change
whenever sea level is high, and that any sudden
change of sea level provokes a rise in CO 2 , Within
this formulation (Berger et al. 1996),0 18 0 is used
as a proxy for atmospheric CO 2 , but it is only the
global 0 18 0 which is of interest, as a proxy for sea
level and overall temperature.
It has been found that the ol3C-values within
phytoplankton (013C O ,g)' as well as suspended and
sedimentary organic carbon vary with the ambient
concentration of CO 2 in surface waters (Rau et al.
1989; Rau 1994; Jasper and Hayes 1990;
Freemann and Hayes 1992; Bentaleb et al. 1996;
Fischer et al. 1997, 1998) (Fig. 20). The
fractionation, which results in an enrichment of the
12C isotope during carbon fixation, proceeds from
a pool of dissolved carbon in contact with the organism. It was observed that at high CO 2 concentrations, the fractionation Ep (i.e. the difference
between ol3C O ,g and the source=ol3C C02 ) is large
and vice versa. To calculate the Ol3C of dissolved
CO 2 , the Ol3CL C02 has to be determined, for example using foraminifersl-8 l3 C values (vital effects
have to be known) as well as the sea surface temperature. Applying relationships between isotope
fractionation and [C0 2 (aq)] to sediment cores,
several studies aimed to reconstruct past PC0 2
levels in the surface waters (e.g. MUller et al. 1994;
Bentaleb et al. 1996) (Fig. 21). It must be remembered that much of these reconstructions depend
on the paleotemperature record (e.g. derived from
alkenone data) rather than on the Ol3C variation.
Based on mechanistic considerations, a two-step
fractionation model and the assumption of exclusive CO 2 (aq) utilization, Farquhar et al. (1982) argued that a hyperbolic function relates the carbon
isotopic composition to CO 2 (aq). Others used a
relationship that fits best to the observed data sets.
The empirically based calibrations were linear (e.g.
Rau 1994) or log-linear (e.g. Freeman and Hayes
1992). However, more recent laboratory experiments (Laws et al. 1995; Bidigare et al. 1997; Popp
et al. 1998) show that for instance growth rate, cell
geometry and membrane diffusivity can play an important role confounding a simple use of the
paleobarometer to reconstruct the surface water
PC0 2 •