620
von Dobeneck and Schmieder
These controversies of 100 kyr cyclicity and
Pleistocene climate transitions essentially address
the issue, whether long-term' c1eoceanographic
variations are related to basic Milankovitch cycles
and their superstructures (e.g., Hopes), or rather
to extraneous (forced by non·Milankovitchmechanisms) or free oscillations. A s~mple, frequently
applied method to analyze proxy signals in this context is multiple bandpass filtering. In the lower half
of Fig. 11 five signal components of particular interest are extracted from the SUSAS record.
The prominent long-term trend obtained from
lowpass filtering (1 st order Butterworth >350 kyr)
is a remarkable feature not found in oxygen isotope signals and therefore not directly related to
global ice volume. It is mirrored by the sedimentation rate stack (Fig. 5) and also by CaC0 3 concentration (P. Muller, unpublished data), hence by carbonate accumulation. Resembling more a double
base line shift than a cycle, this signal marks two
transitions:
- a first from 1.0 to 0.9 Ma corresponds to the age
ofthe mid-Pleistocene (r)evolution,
- a second from 0.65 to 0.5 Ma culminates in the
large-scale sedimentation event outlined in Fig. 5.
Their duration may have been shorter, since lowpass filtering inevitably broadens step functions.
The three age intervals delimited by these two transitions show different spectral characteristics in the
Milankovitch bands. The precession-related signal
(16-26 kyr) is poorly documented due to bioturbation damping with the exception of sections of
higher sedimentation rates and enhanced forcing.
Apart from superimposed amplitude fluctuations,
the influence of obliquity (35-47 kyr) on the SUSAS
stack diminishes continuously with a major decrease near l.2 Ma. The analysis in the 100 kyr
range was subdivided into two bands from 85 kyr
to 110 kyr and from 110 kyr to 135 kyr to distinguish
the 96 kyr and 127 kyr eccentricity cycles. These
periods clearly dominate the signal in late Pleistocene, but also give important contributions to older
sections. As the amplitude variations in the 41 kyr
and 100 kyr bands resemble in great detail the patterns found in 8180 records of Pacific and Atlantic sites (Park and Maasch 1993), we conclude that
the observed features are of a global nature and
closely related to changes in ice volume.
Evolutionary spectral analysis is an excellent
method to visualize the variability of cyclic climate
characteristics in a record through time (e.g.,
Pestiaux and Berger 1984; Bloemendal and
deMenocal1989; Mwenifumbo and Blangy 1991).
Spectra generated within a moving window are
depicted in three-dimensional' spectrograms' combining frequency and time domain (top of Fig. 11).
These calculations were made using the
'SPECGRAM' algorithm embedded in the
'MATLAB Signal Processing Toolbox'. A 500 kyr
frame was advanced at 10 kyr steps and, in order
to minimize cut-off effects, tapered with a Hanning
window, thereby focusing to the central section at
each step. Long-term trends (periods >350 kyr)
were previously removed by highpass filtering
(purple curve) to avoid disturbance by these unresolved signal components. The resulting data matrix was normalized to an average spectral density
of 1 by dividing all values by the total matrix mean.
This procedure merely scales all spectrograms uniformly and enables their comparison, but does not
alter relative variations in the time or frequency domain.
In Figs. 12a-d spectrograms of forcing and
responding variables are related. The July midmonth insolation signal at 65°N (Berger and Loutre
1991) and a normalized ETP curve are shown as
reference. ETP curves are artificial, but often employed target records composed by calibrating and
adding the time series of Eccentricity, Tilt (obliquity) and Precession (Imbrie etal. 1984). Here, the
cumulative spectral intensity in each band of the
ETP spectrogram was calibrated to equal that of
the SUSAS spectrogram. Being based on the same
orbital variations, both reference signals exhibit
identical amplitude modulation patterns, but very
different spectral power in each band. In the eccentricity band this difference amounts to several
orders of magnitude - an expression of the 100 kyr
cycle problem. Characteristic Milankovitch superstructures as the cyclic 400 kyr modulation of the
100 kyr, 54 kyr and 23 kyr cycles and the synchronous decay of the 100 kyr and 19 kyr cycles are
clearly illustrated in Fig. 12b. The faint 29 kyrcycle is a subordinate component ofthe obliquity signal and shows similar modulation as the main periods near 41 kyr.
von Dobeneck and Schmieder
These controversies of 100 kyr cyclicity and
Pleistocene climate transitions essentially address
the issue, whether long-term' c1eoceanographic
variations are related to basic Milankovitch cycles
and their superstructures (e.g., Hopes), or rather
to extraneous (forced by non·Milankovitchmechanisms) or free oscillations. A s~mple, frequently
applied method to analyze proxy signals in this context is multiple bandpass filtering. In the lower half
of Fig. 11 five signal components of particular interest are extracted from the SUSAS record.
The prominent long-term trend obtained from
lowpass filtering (1 st order Butterworth >350 kyr)
is a remarkable feature not found in oxygen isotope signals and therefore not directly related to
global ice volume. It is mirrored by the sedimentation rate stack (Fig. 5) and also by CaC0 3 concentration (P. Muller, unpublished data), hence by carbonate accumulation. Resembling more a double
base line shift than a cycle, this signal marks two
transitions:
- a first from 1.0 to 0.9 Ma corresponds to the age
ofthe mid-Pleistocene (r)evolution,
- a second from 0.65 to 0.5 Ma culminates in the
large-scale sedimentation event outlined in Fig. 5.
Their duration may have been shorter, since lowpass filtering inevitably broadens step functions.
The three age intervals delimited by these two transitions show different spectral characteristics in the
Milankovitch bands. The precession-related signal
(16-26 kyr) is poorly documented due to bioturbation damping with the exception of sections of
higher sedimentation rates and enhanced forcing.
Apart from superimposed amplitude fluctuations,
the influence of obliquity (35-47 kyr) on the SUSAS
stack diminishes continuously with a major decrease near l.2 Ma. The analysis in the 100 kyr
range was subdivided into two bands from 85 kyr
to 110 kyr and from 110 kyr to 135 kyr to distinguish
the 96 kyr and 127 kyr eccentricity cycles. These
periods clearly dominate the signal in late Pleistocene, but also give important contributions to older
sections. As the amplitude variations in the 41 kyr
and 100 kyr bands resemble in great detail the patterns found in 8180 records of Pacific and Atlantic sites (Park and Maasch 1993), we conclude that
the observed features are of a global nature and
closely related to changes in ice volume.
Evolutionary spectral analysis is an excellent
method to visualize the variability of cyclic climate
characteristics in a record through time (e.g.,
Pestiaux and Berger 1984; Bloemendal and
deMenocal1989; Mwenifumbo and Blangy 1991).
Spectra generated within a moving window are
depicted in three-dimensional' spectrograms' combining frequency and time domain (top of Fig. 11).
These calculations were made using the
'SPECGRAM' algorithm embedded in the
'MATLAB Signal Processing Toolbox'. A 500 kyr
frame was advanced at 10 kyr steps and, in order
to minimize cut-off effects, tapered with a Hanning
window, thereby focusing to the central section at
each step. Long-term trends (periods >350 kyr)
were previously removed by highpass filtering
(purple curve) to avoid disturbance by these unresolved signal components. The resulting data matrix was normalized to an average spectral density
of 1 by dividing all values by the total matrix mean.
This procedure merely scales all spectrograms uniformly and enables their comparison, but does not
alter relative variations in the time or frequency domain.
In Figs. 12a-d spectrograms of forcing and
responding variables are related. The July midmonth insolation signal at 65°N (Berger and Loutre
1991) and a normalized ETP curve are shown as
reference. ETP curves are artificial, but often employed target records composed by calibrating and
adding the time series of Eccentricity, Tilt (obliquity) and Precession (Imbrie etal. 1984). Here, the
cumulative spectral intensity in each band of the
ETP spectrogram was calibrated to equal that of
the SUSAS spectrogram. Being based on the same
orbital variations, both reference signals exhibit
identical amplitude modulation patterns, but very
different spectral power in each band. In the eccentricity band this difference amounts to several
orders of magnitude - an expression of the 100 kyr
cycle problem. Characteristic Milankovitch superstructures as the cyclic 400 kyr modulation of the
100 kyr, 54 kyr and 23 kyr cycles and the synchronous decay of the 100 kyr and 19 kyr cycles are
clearly illustrated in Fig. 12b. The faint 29 kyrcycle is a subordinate component ofthe obliquity signal and shows similar modulation as the main periods near 41 kyr.