12
Weferetal.
face. Thus, in regions where evaporation is greater
than precipitation, surface waters are enriched with
18 0. The reverse is true for regions of excess precipitation. The effect is weak in the tropics but
strong in high latitudes (Craig and Gordon 1965;
Fairbanks et al. 1992). The error introduced by
neglecting this effect is on the order of 1°C, but
depends strongly on regional circumstances, being
greater near the boundaries of evaporation- versus precipitation-dominated regimes.
The use of oxygen isotopes in estimating paleotemperature may be illustrated by addressing the
problem of tropical glacial temperatures, mentioned
above. Glacial-interglacial differences in8 18 0 values of Globigerinoides sacculifer have been
mapped in the western and central tropical Atlantic using data from surface and core sediments. A
temperature change in the northwestern equatorial
Atlantic of2-3 cC, as depicted by Wolff et al. (I 998),
represents an intermediate estimate between the
results of CLIMAP (~I DC) and tropical SST estimates from corals (Guilderson et al. 1994) (~5°C)
and from the South American Continental record
(e.g. Stute et al. 1995; Thompson et al. 1995)
(5-6°C).
Using laboratory cultures, Spero et al. (1997)
recently demonstrated a strong correlation between
the isotopic composition (both carbon and oxygen)
of planktic foraminiferal shells and the carbonate
chemistry of the culture water (Bijma et al. this
volume). This fractionation effect is not entirely
new: McCrea (1950) demonstrated that the
18 0/ 16 0 ratio of synthetic calcite decreases as the
carbonate ion concentration increases. However,
the possible implications for paleoceanographic
and -climatic reconstructions have been neglected
for nearly 50 years. In view of the factthatthe oceanic carbonate chemistry changes significantly on
a glacial/interglacial time-scale (e.g. Archer et al.
1989; Broecker and Peng 1994; Sanyal et al. 1995),
this isotopic fractionation effect has considerable
effect on temporal isotopic variability in the fossil
record and thus potentially confounds direct interpretation offoraminiferal stable isotope data. The
influence of the oceanic carbonate chemistry on
8 13 C recorded by planktic foraminifers provides a
novel explanation for the large, negative swings in
planktic 8 13 C observed in the glacial sections of
Southern Ocean cores (Lea et al. this volume). It
also presents an alternative hypothesis to the terrestrial biosphere-to-ocean transfer of carbon that
is generally accepted as an explanation for lower
glacial she1l813C values (Shackleton 1977). Correcting for the influence of the carbonate ion effect on shell 8 18 0 lowers glacial tropical SST estimates by up to 1°C, which brings oxygen isotope
paleotemperatures closer to those marine SST
proxies (e.g. coral Sr/Ca ratios) and those terrestrial indicators (e.g. snow-line and ice core ( 18 0)
that suggest more intense tropical cooling at the last
glacial maximum (Guilderson et al. 1994;
Thompson et al. 1995).
The exact meaning of an estimate such as the
one just made is obscure. First, we do not know to
what extent the water composition really changed;
the uncertainty regarding ice-mass effect and
evaporation-precipitation effect adds up to about
1°C. Second, we cannot be sure that the species
chosen (G. ruber) records temperature in the same
fashion during interglacials and glacials. It may
change for example its seasonal preference of
depth habitat. Also, it may change possible "vital
effects", that is, idiosyncratic ways of fractionating
isotopes within its body, depending on vital activities, including interactions with symbionts. The
uncertainty from this source is near 0.5°C. Finally,
it is not clear what such an estimate means relative to "average annual temperature", or even
whether "average annual temperature" is a useful
parameter with respect to climatology and biogeography, e.g. G. ruber has maxima in abundance
and flux during warm seasons in todays oceans (e.g.
Bijma et al. 1990). The seasonal temperature range
is necessary in order to model the system correctly.
From the nature of the temperature gradient
records, it is clear that the range of8 18 0 in planktic
foraminifer species at a given location contains
information about the thickness of the (warm)
mixed layer, and therefore about the depth of the
thermocline. Mulitza et al. (1997) compiled 8 18 0
data from various planktic foraminifers with different depth habitats from the South Atlantic (Fig. 5).
North of 15°S , well-separated 8 18 0 values of shallow- and deep-living species denote a deep
thermocline; merging values denote thermocline
shallowing. Therefore, this proxy might only be
Weferetal.
face. Thus, in regions where evaporation is greater
than precipitation, surface waters are enriched with
18 0. The reverse is true for regions of excess precipitation. The effect is weak in the tropics but
strong in high latitudes (Craig and Gordon 1965;
Fairbanks et al. 1992). The error introduced by
neglecting this effect is on the order of 1°C, but
depends strongly on regional circumstances, being
greater near the boundaries of evaporation- versus precipitation-dominated regimes.
The use of oxygen isotopes in estimating paleotemperature may be illustrated by addressing the
problem of tropical glacial temperatures, mentioned
above. Glacial-interglacial differences in8 18 0 values of Globigerinoides sacculifer have been
mapped in the western and central tropical Atlantic using data from surface and core sediments. A
temperature change in the northwestern equatorial
Atlantic of2-3 cC, as depicted by Wolff et al. (I 998),
represents an intermediate estimate between the
results of CLIMAP (~I DC) and tropical SST estimates from corals (Guilderson et al. 1994) (~5°C)
and from the South American Continental record
(e.g. Stute et al. 1995; Thompson et al. 1995)
(5-6°C).
Using laboratory cultures, Spero et al. (1997)
recently demonstrated a strong correlation between
the isotopic composition (both carbon and oxygen)
of planktic foraminiferal shells and the carbonate
chemistry of the culture water (Bijma et al. this
volume). This fractionation effect is not entirely
new: McCrea (1950) demonstrated that the
18 0/ 16 0 ratio of synthetic calcite decreases as the
carbonate ion concentration increases. However,
the possible implications for paleoceanographic
and -climatic reconstructions have been neglected
for nearly 50 years. In view of the factthatthe oceanic carbonate chemistry changes significantly on
a glacial/interglacial time-scale (e.g. Archer et al.
1989; Broecker and Peng 1994; Sanyal et al. 1995),
this isotopic fractionation effect has considerable
effect on temporal isotopic variability in the fossil
record and thus potentially confounds direct interpretation offoraminiferal stable isotope data. The
influence of the oceanic carbonate chemistry on
8 13 C recorded by planktic foraminifers provides a
novel explanation for the large, negative swings in
planktic 8 13 C observed in the glacial sections of
Southern Ocean cores (Lea et al. this volume). It
also presents an alternative hypothesis to the terrestrial biosphere-to-ocean transfer of carbon that
is generally accepted as an explanation for lower
glacial she1l813C values (Shackleton 1977). Correcting for the influence of the carbonate ion effect on shell 8 18 0 lowers glacial tropical SST estimates by up to 1°C, which brings oxygen isotope
paleotemperatures closer to those marine SST
proxies (e.g. coral Sr/Ca ratios) and those terrestrial indicators (e.g. snow-line and ice core ( 18 0)
that suggest more intense tropical cooling at the last
glacial maximum (Guilderson et al. 1994;
Thompson et al. 1995).
The exact meaning of an estimate such as the
one just made is obscure. First, we do not know to
what extent the water composition really changed;
the uncertainty regarding ice-mass effect and
evaporation-precipitation effect adds up to about
1°C. Second, we cannot be sure that the species
chosen (G. ruber) records temperature in the same
fashion during interglacials and glacials. It may
change for example its seasonal preference of
depth habitat. Also, it may change possible "vital
effects", that is, idiosyncratic ways of fractionating
isotopes within its body, depending on vital activities, including interactions with symbionts. The
uncertainty from this source is near 0.5°C. Finally,
it is not clear what such an estimate means relative to "average annual temperature", or even
whether "average annual temperature" is a useful
parameter with respect to climatology and biogeography, e.g. G. ruber has maxima in abundance
and flux during warm seasons in todays oceans (e.g.
Bijma et al. 1990). The seasonal temperature range
is necessary in order to model the system correctly.
From the nature of the temperature gradient
records, it is clear that the range of8 18 0 in planktic
foraminifer species at a given location contains
information about the thickness of the (warm)
mixed layer, and therefore about the depth of the
thermocline. Mulitza et al. (1997) compiled 8 18 0
data from various planktic foraminifers with different depth habitats from the South Atlantic (Fig. 5).
North of 15°S , well-separated 8 18 0 values of shallow- and deep-living species denote a deep
thermocline; merging values denote thermocline
shallowing. Therefore, this proxy might only be