Clues to Ocean History: a Brief Overview of Proxies
7
for this purpose: abundances of microfossils living
in surface waters, oxygen isotope composition of
planktic foraminifers, the ratio of magnesium or
strontium to calcium in calcareous shells, ratios of
certain organic molecules (various types of
alkenones produced by coccolithophorids).
That microfossils bear witness to temperature
was recognized early in the history of oceanography. John Murray, naturalist of the
CHALLENGER Expedition (1872-1875) found
that planktic foraminifers provide for much of the
sediment on the sea floor, and that differences in
species composition contain clues to the temperature of the waters in which they lived (Murray
1897). Schott (1935), working on sediments of
the METEOR Expedition (1925-1927) introduced
quantitative counting of species within the fossil assemblages on the sea floor, and noted the changes
in composition downcore. He realized that surfacewater temperature changed as the climate fluctuated between glacial and interglacial conditions.
Parker (1958) and Ruddiman (1971) formalized this
approach by plotting warm- and cold-water faunal
percentages as a function of depth in core. The
simplest possible equation relating percentages of
species abundance to temperature is as follows
(Berger 1971):
T(est) = Sum (Pi X T) / Sum (p)
(4),
where Pi stands for percentage of species i and Ti
is the temperature at which the species is most
abundant.
In a first comprehensive attempt at utilizing
abundance data to estimate paleotemperature,
Imbrie and Kipp (1971) introduced factor regression to paleontology as a basis for developing transform equations. In this method, each assemblage
is expressed as a mixture of factor loadings, which
stand for idealized end-member assemblages.
These loadings are then related to the temperature
field by multiple regression, in the calibration process. Results show that such estimates are useful,
especially in the mid-range of the temperature field
(Fig. 1).
An alternative method, the "Modern Analog
Technique" (MAJ), has been introduced by Hutson
30'
.' :!
::t 25'
:>
....
< £I:
~ 20'
:::e
... ....
£I: 15'
... .... z
~ 10"
e
...
~
:; 5°
i=
en
...
/. ..
5'
10'
15°
20'
25°
30"
MEASURED WINTER TEMPERATURE (C.l
Fig. 1. Scatter diagram showing measured winter sea
surface temperatures versus those estimated from faunal
assemblages in core-top samples using factor analysis
and transfer function methods. From Imbrie and Kipp
(1971).
(1980) and further developed by Prell (1985) and
by Howard and Prell (1992). In this method, there
is no factoring. Instead, SST is directly estimated
from statistical differences of fossil assemblages
from modern assemblages, which are tagged with
the overlying water temperature. The MAT technique may be more robust toward non-analog conditions (that is, conditions in the past which are
outside the present calibration set). On the whole,
however, there seems to be little difference in results between the two methods, as long as the calibration sets are the same (Weferetal. 1996; Ortiz
and Mix 1997) (see Fig. 2).
The statistics for estimating surface temperatures from microfossil assemblages are well established. Both methods are dependent on extensive
reference data sets that must be calibrated with
present-day surface-water temperatures (for example, Levitus 1982). An improvement of the
paleotemperature determinations mainly depends
on the improvement of the reference data set, by
obtaining more real core tops, and by applying uni-
7
for this purpose: abundances of microfossils living
in surface waters, oxygen isotope composition of
planktic foraminifers, the ratio of magnesium or
strontium to calcium in calcareous shells, ratios of
certain organic molecules (various types of
alkenones produced by coccolithophorids).
That microfossils bear witness to temperature
was recognized early in the history of oceanography. John Murray, naturalist of the
CHALLENGER Expedition (1872-1875) found
that planktic foraminifers provide for much of the
sediment on the sea floor, and that differences in
species composition contain clues to the temperature of the waters in which they lived (Murray
1897). Schott (1935), working on sediments of
the METEOR Expedition (1925-1927) introduced
quantitative counting of species within the fossil assemblages on the sea floor, and noted the changes
in composition downcore. He realized that surfacewater temperature changed as the climate fluctuated between glacial and interglacial conditions.
Parker (1958) and Ruddiman (1971) formalized this
approach by plotting warm- and cold-water faunal
percentages as a function of depth in core. The
simplest possible equation relating percentages of
species abundance to temperature is as follows
(Berger 1971):
T(est) = Sum (Pi X T) / Sum (p)
(4),
where Pi stands for percentage of species i and Ti
is the temperature at which the species is most
abundant.
In a first comprehensive attempt at utilizing
abundance data to estimate paleotemperature,
Imbrie and Kipp (1971) introduced factor regression to paleontology as a basis for developing transform equations. In this method, each assemblage
is expressed as a mixture of factor loadings, which
stand for idealized end-member assemblages.
These loadings are then related to the temperature
field by multiple regression, in the calibration process. Results show that such estimates are useful,
especially in the mid-range of the temperature field
(Fig. 1).
An alternative method, the "Modern Analog
Technique" (MAJ), has been introduced by Hutson
30'
.' :!
::t 25'
:>
....
< £I:
~ 20'
:::e
... ....
£I: 15'
... .... z
~ 10"
e
...
~
:; 5°
i=
en
...
/. ..
5'
10'
15°
20'
25°
30"
MEASURED WINTER TEMPERATURE (C.l
Fig. 1. Scatter diagram showing measured winter sea
surface temperatures versus those estimated from faunal
assemblages in core-top samples using factor analysis
and transfer function methods. From Imbrie and Kipp
(1971).
(1980) and further developed by Prell (1985) and
by Howard and Prell (1992). In this method, there
is no factoring. Instead, SST is directly estimated
from statistical differences of fossil assemblages
from modern assemblages, which are tagged with
the overlying water temperature. The MAT technique may be more robust toward non-analog conditions (that is, conditions in the past which are
outside the present calibration set). On the whole,
however, there seems to be little difference in results between the two methods, as long as the calibration sets are the same (Weferetal. 1996; Ortiz
and Mix 1997) (see Fig. 2).
The statistics for estimating surface temperatures from microfossil assemblages are well established. Both methods are dependent on extensive
reference data sets that must be calibrated with
present-day surface-water temperatures (for example, Levitus 1982). An improvement of the
paleotemperature determinations mainly depends
on the improvement of the reference data set, by
obtaining more real core tops, and by applying uni-