Sea-Surface Temperature Estimations Using a Modem Analog Technique
73
Methodology
Faunal Analysis
The core samples were washed at the laboratory
at Bremen University, then sieved through a 150mm screen to obtain foraminifera for faunal analysis. Because the primary purpose ofthis study is
to compute temperatures using the SIMMAX
Modem Analog Technique (MAT), this size fraction was chosen to be consistent with the reference
(modem analog) data set. The tropical cores were
examined at 10-cm intervals, whereas the Rio
Grande Rise cores were examined at 5-cm intervals because of the lower sedimentation rates in
that area. To facilitate examination and counting of
foraminiferal species abundances, each sample was
sieved into sub-samples offour size fractions: 150212mm, 212-315mm, 315-500mm and >500mm.
Each size fraction was subsequently divided with
a microsplitter until about 100 (but never fewer than
85) specimens remained for counting. This resulted
in counts of about 450 to 600 specimens per sample, with the average count >500 individuals per
sample. The counts for the four different size fractions were recalculated to real percentages based
on the number of splits for each size range. The
resulting species percentages were employed to
estimate past sea-surface temperatures using the
SIMMAX Modem Analog Technique. Taxonomic
concepts follow those of Be (1967), Be and
Tolderlund (1971), and Parker (1962).
SIMMAXMAT
This technique for estimating paleotemperatures,
developed and described in detail by Ptlaumann et
al. (1996), is a variation of the standard modem
analog technique. It uses a similarity index based
on the scalar product of the normalized faunal percentages to compare fossil samples with the modem reference database, and is an advancement over
earlier methods in that it employs a weighting procedure based on the inverse geographical distance
of the most similar modem analog samples from
the subject sample. That is, after a predetermined
number (ten were used in this study) of most similar modem samples are chosen, the sea-surface
temperatures associated with the geographically
closest samples are given greater weight in the
averaging procedure used in assigning a value to
the fossil sample. The suitability of applying this
weighting procedure has been questioned in that it
implies a degree of stability in regional environments through the past (see Wolff et al. this volume). On the other hand, the weighting procedure
results in excellent correlation coefficients when
applied to the core-top data set (0.994 for caloric
winter and 0.993 for caloric summer), and standard deviations of less than 1°C. The SIMMAX
program can be run without applying the distanceweighting option, but in the present study it is employed. An additional improvement of the
SIMMAX technique is the use of an expanded
core-top database containing 738 samples in the
Atlantic Ocean, including published CLIMAP data
(ptlaumann et al. 1996). The use of the scalar products similarity index is also unique to the SIMMAX
technique; previous MAT methods commonly used
a square chord dissimilarity coefficient (e.g. Prell
1985) which tends to weight toward rare species
(Ptlaumann et al. 1996). The scalar product index
used by SIMMAX results in optimal correlation coefficients. The 39 species or forms distinguished
in the original examinations were combined to the
extent that the following 26 species/formae were
employed in the temperature calculations in order
to be consistent with the core-top reference data
set:
Globigerinella aequilateralis
Globigerina bulloides
Globigerina calida
Globigerinoides conglobatus
Globorotalia (Turborotalia) crassaformis
Sphaeroidinella dehiscens
Globigerinella (Beella) digitata
Neogloboquadrina dutertrei
Globigerina falconensis
Globigerinita glutinata
Globorotalia hirsuta
Globorotalia inflata
Globorotalia menardii (including G. cultrata
and G. tumida)
Pulleniatina obliquiloculata
Neogloboquadrina pachyderma (sinistral)
73
Methodology
Faunal Analysis
The core samples were washed at the laboratory
at Bremen University, then sieved through a 150mm screen to obtain foraminifera for faunal analysis. Because the primary purpose ofthis study is
to compute temperatures using the SIMMAX
Modem Analog Technique (MAT), this size fraction was chosen to be consistent with the reference
(modem analog) data set. The tropical cores were
examined at 10-cm intervals, whereas the Rio
Grande Rise cores were examined at 5-cm intervals because of the lower sedimentation rates in
that area. To facilitate examination and counting of
foraminiferal species abundances, each sample was
sieved into sub-samples offour size fractions: 150212mm, 212-315mm, 315-500mm and >500mm.
Each size fraction was subsequently divided with
a microsplitter until about 100 (but never fewer than
85) specimens remained for counting. This resulted
in counts of about 450 to 600 specimens per sample, with the average count >500 individuals per
sample. The counts for the four different size fractions were recalculated to real percentages based
on the number of splits for each size range. The
resulting species percentages were employed to
estimate past sea-surface temperatures using the
SIMMAX Modem Analog Technique. Taxonomic
concepts follow those of Be (1967), Be and
Tolderlund (1971), and Parker (1962).
SIMMAXMAT
This technique for estimating paleotemperatures,
developed and described in detail by Ptlaumann et
al. (1996), is a variation of the standard modem
analog technique. It uses a similarity index based
on the scalar product of the normalized faunal percentages to compare fossil samples with the modem reference database, and is an advancement over
earlier methods in that it employs a weighting procedure based on the inverse geographical distance
of the most similar modem analog samples from
the subject sample. That is, after a predetermined
number (ten were used in this study) of most similar modem samples are chosen, the sea-surface
temperatures associated with the geographically
closest samples are given greater weight in the
averaging procedure used in assigning a value to
the fossil sample. The suitability of applying this
weighting procedure has been questioned in that it
implies a degree of stability in regional environments through the past (see Wolff et al. this volume). On the other hand, the weighting procedure
results in excellent correlation coefficients when
applied to the core-top data set (0.994 for caloric
winter and 0.993 for caloric summer), and standard deviations of less than 1°C. The SIMMAX
program can be run without applying the distanceweighting option, but in the present study it is employed. An additional improvement of the
SIMMAX technique is the use of an expanded
core-top database containing 738 samples in the
Atlantic Ocean, including published CLIMAP data
(ptlaumann et al. 1996). The use of the scalar products similarity index is also unique to the SIMMAX
technique; previous MAT methods commonly used
a square chord dissimilarity coefficient (e.g. Prell
1985) which tends to weight toward rare species
(Ptlaumann et al. 1996). The scalar product index
used by SIMMAX results in optimal correlation coefficients. The 39 species or forms distinguished
in the original examinations were combined to the
extent that the following 26 species/formae were
employed in the temperature calculations in order
to be consistent with the core-top reference data
set:
Globigerinella aequilateralis
Globigerina bulloides
Globigerina calida
Globigerinoides conglobatus
Globorotalia (Turborotalia) crassaformis
Sphaeroidinella dehiscens
Globigerinella (Beella) digitata
Neogloboquadrina dutertrei
Globigerina falconensis
Globigerinita glutinata
Globorotalia hirsuta
Globorotalia inflata
Globorotalia menardii (including G. cultrata
and G. tumida)
Pulleniatina obliquiloculata
Neogloboquadrina pachyderma (sinistral)