Sea-Surface Temperature Estimations Using a Modern Analog Technique
77
in a fairly tight temperature range between 27° and
28° C, and another with a wider range between 19°
and 24°, representing the tropical cores and Rio
Grande Rise cores, respectively. The few points
present between 24° and 27°C represent the samples in the cool excursion associated with isotope
stage 6 in the tropical cores.
The most prominent warm-water indicators are
Globigerinoides ruber, G. sacculifer, and
Globigerinella aequilateralis, all of which are
present in all the tropical samples and have significantly lower average abundance values in the Rio
Grande Rise samples. Globigerinita glutinata also
belongs to this group and is present in all the tropical samples, but the abundance difference between
warm and cool samples is not as pronounced.
Trends of higher abundance with increasing temperature can be seen within the Rio Grande Rise
samples for G. glutinata and G. ruber.
Pulleniatina
obliquiloculata,
Neogloboquadrina dutertrei, Globorotalia menardii, and
C. nitida are also important in the "warm" group
because of their near absence in the Rio Grande
Rise cores, although they are also missing in some
samples in the tropical cores.
The most important of the cool-water species
are Globorotalia inflata, which occurs consistently in very large abundances in the Rio Grande
Rise cores and rarely in the tropical cores, and the
dextral variety of N. pachyderma, which attains
higher abundances with cooler temperatures even
within the temperature range of the Rio Grande
Rise cores. Other members ofthe "cool" group are
Globorotalia hirsuta. Globigerina buZZoides,
Globigerina falconensis, Globigerina calida,
and Globorotalia scitula. The species G.
truncatulinoides, as a group, also shows slightly
higher abundances in the cooler samples. Fig. 4,
where the two coiling varieties are distinguished by
different symbols, reveals that the left-coiling form
(sinistral) would have been an excellent indicator
of cooler surface temperatures, whereas the dextral forms show no pattern of preference within the
entire temperature range. Unfortunately, in the
SIMMAX MAT calculations it was necessary to
group the two forms for consistency with the reference data set.
Cluster Analyses
Another tool that can provide some insight into the
workings of the SIMMAX temperature calculations is cluster analysis. Because the foraminiferal
faunal content of each sample was compared with
a data set of modem samples to identify those most
similar and from there to determine
paleotemperature based on the most similar modem samples, it will be enlightening to group the fossil
samples according to their similarity with each
other and discover what temperature relationships
are exhibited among and within the clusters.
Because hierarchical clustering techniques begin by combining the two most similar observations
(samples), then successively connecting the next
two most similar samples (treating groups as single observations after their combination), we would
expect the initial level of combinations to involve
samples with very close or even identical temperature estimations. Dendrograms illustrating the eight
clusters obtained for each core are presented in
Figs. 5 (a) through 5 (d). The grouping patterns and
general sizes of the clusters can be seen in these
diagrams, in which the samples are represented
along the x-axes and labeled with their estimated
temperatures. The most similar samples can be
picked out by the elevation on the y-axis (Squared
Euclidian Distance) of the line connecting them,
which indicates how early in the clustering procedure they were combined. For example, in Fig. 5
(a), the dendrogram for Core GeoB 1312, the 15th
and 16th samples counting from the left along the
x-axis (each with a value of22.l 0), are connected
by a line lying almost on top of the x-axis, which
means they are very similar and were combined
early in the clustering procedure. In fact, this was
the third pair in this core to be combined. The first
pair of samples to be combined, each a with temperature of21.2°, can be seen farther to the right
along the x-axis where the connecting bar is lying
practically on top ofthe x-axis. An interesting point
to consider while following the succession of increasingly less similar samples being connected is
the degree to which the clusters being formed remain unique in terms of the temperature ranges
they contain. Will there be any overlap of tempera-
77
in a fairly tight temperature range between 27° and
28° C, and another with a wider range between 19°
and 24°, representing the tropical cores and Rio
Grande Rise cores, respectively. The few points
present between 24° and 27°C represent the samples in the cool excursion associated with isotope
stage 6 in the tropical cores.
The most prominent warm-water indicators are
Globigerinoides ruber, G. sacculifer, and
Globigerinella aequilateralis, all of which are
present in all the tropical samples and have significantly lower average abundance values in the Rio
Grande Rise samples. Globigerinita glutinata also
belongs to this group and is present in all the tropical samples, but the abundance difference between
warm and cool samples is not as pronounced.
Trends of higher abundance with increasing temperature can be seen within the Rio Grande Rise
samples for G. glutinata and G. ruber.
Pulleniatina
obliquiloculata,
Neogloboquadrina dutertrei, Globorotalia menardii, and
C. nitida are also important in the "warm" group
because of their near absence in the Rio Grande
Rise cores, although they are also missing in some
samples in the tropical cores.
The most important of the cool-water species
are Globorotalia inflata, which occurs consistently in very large abundances in the Rio Grande
Rise cores and rarely in the tropical cores, and the
dextral variety of N. pachyderma, which attains
higher abundances with cooler temperatures even
within the temperature range of the Rio Grande
Rise cores. Other members ofthe "cool" group are
Globorotalia hirsuta. Globigerina buZZoides,
Globigerina falconensis, Globigerina calida,
and Globorotalia scitula. The species G.
truncatulinoides, as a group, also shows slightly
higher abundances in the cooler samples. Fig. 4,
where the two coiling varieties are distinguished by
different symbols, reveals that the left-coiling form
(sinistral) would have been an excellent indicator
of cooler surface temperatures, whereas the dextral forms show no pattern of preference within the
entire temperature range. Unfortunately, in the
SIMMAX MAT calculations it was necessary to
group the two forms for consistency with the reference data set.
Cluster Analyses
Another tool that can provide some insight into the
workings of the SIMMAX temperature calculations is cluster analysis. Because the foraminiferal
faunal content of each sample was compared with
a data set of modem samples to identify those most
similar and from there to determine
paleotemperature based on the most similar modem samples, it will be enlightening to group the fossil
samples according to their similarity with each
other and discover what temperature relationships
are exhibited among and within the clusters.
Because hierarchical clustering techniques begin by combining the two most similar observations
(samples), then successively connecting the next
two most similar samples (treating groups as single observations after their combination), we would
expect the initial level of combinations to involve
samples with very close or even identical temperature estimations. Dendrograms illustrating the eight
clusters obtained for each core are presented in
Figs. 5 (a) through 5 (d). The grouping patterns and
general sizes of the clusters can be seen in these
diagrams, in which the samples are represented
along the x-axes and labeled with their estimated
temperatures. The most similar samples can be
picked out by the elevation on the y-axis (Squared
Euclidian Distance) of the line connecting them,
which indicates how early in the clustering procedure they were combined. For example, in Fig. 5
(a), the dendrogram for Core GeoB 1312, the 15th
and 16th samples counting from the left along the
x-axis (each with a value of22.l 0), are connected
by a line lying almost on top of the x-axis, which
means they are very similar and were combined
early in the clustering procedure. In fact, this was
the third pair in this core to be combined. The first
pair of samples to be combined, each a with temperature of21.2°, can be seen farther to the right
along the x-axis where the connecting bar is lying
practically on top ofthe x-axis. An interesting point
to consider while following the succession of increasingly less similar samples being connected is
the degree to which the clusters being formed remain unique in terms of the temperature ranges
they contain. Will there be any overlap of tempera-