Using Rock Magnetic Proxy Records for Orbital Tuning and Extended Time Series Analyses
621
The numerical difference between the spectrograms ofthe SUSAS record and the ETP curve is
displayed in Fig. 12d. For the two precessional
bands at 19 kyr and 23 kyr, the assumption of a
constant proxy response appears justified, as both
records largely compensate. In contrast, the residues in the 41 kyr band show a variable response
to obliquity forcing. Two marked spectral intensity
declines at 1.2-1.1 and 0.7-0.6 Ma (Fig. 12c) cannot be explained as a result of steady response to
orbital forcing (Fig. 12b) and suggest that the potential of obliquity to drive the observed carbonate
dissolution cycles decreased stepwise during Pleistocene times. This finding is in accordance with
analyses made for 8 18 0 records (e.g., Ruddiman
et al. 1989; Joyce et al. 1990) and implies that the
climatic system was less sensitive to obliquity forcing during the late Pleistocene. The coincidence of
target and proxy records in the 100 kyr band is
even lower, as indicated by larger residues and
steeper gradients. While the ETP model looses
100 kyr power (Fig. 12b), the SUSAS stack, like
most climate records, documents an intensification
of 100 kyr oscillations in late Pleistocene.
This trend is interrupted by an early Pleistocene
100 kyr maximum appearing between 1.2 and
1.1 Ma. This 'premature 100 kyr incident', also
observed in 8 18 0 records (e.g., Mudelsee and Stattegger 1997), merges with a phase of strong 70 kyr
cyclicity (Fig. 12c) which has also been found in
other paleoceanographic records (e.g., Ruddiman
et al. 1989; Robinson 1990; Bassinot et al. 1994;
Bolton et al. 1995). Two low-frequency positive
maxima centered at 0.95 and 0.6 Ma in the SUSAS
and the residual spectrogram (Fig. l2c, d) result
from remainders ofthe earlier discussed base line
transitions of the long-term susceptibility trend
(Fig. 11 bottom).
The most striking result ofthe residual spectrogram is the reciprocity of spectral intensities in the
41 and 100 kyr bands (dashed outlines in Fig. 12d).
In relation to the stationary ETP model, the SUSAS
100 kyr cyclicity develops over-proportionally in
sections with reduced response to obliquity (blue
outline) and retreats, where the response to obliquity is strong (yellow outline). These findings imply an exchange of spectral energy between obliquity and eccentricity in overall accordance with
insolation threshold models (e.g., Saltzman and
Verbitsky 1993; Raymo 1997). Intensified northern
hemisphere summer insolation capable of triggering major deglaciations hence should result from an
interference of obliquity maxima with precession
index minima. At times, when the insolation threshold is relatively low, obliquity maxima alone will
trigger the withdrawal of continental ice shields and
41 kyr cyclicity prevails. When the insolation
threshold is higher, the required peak insolation is
only reached by optimum interaction of obliquity and
precession. As eccentricity modulates the precession amplitude, sufficiently large precessional peaks
occur only during 100 kyr eccentricity maxima
causing the observed 100 kyr climate cycles.
Raymo (1997) assumes a linear, pC0 2 controlled rise of the insolation threshold over the past
1.2 m.y. and obtained a single transition from 41 kyr
to 100 kyr predominance at about 0.7 Ma. Our evolutionary spectral analyses (Figs. 12c, d) imply, that
this turnover was predated by a 100 kyr episode
around 1.2 Ma.
Sub-Milankovitch signal variations of the
CEARIS stack
A most interesting feature of all CEARlS records
are consistent high-frequency variations in the subMilankovitch to millennial range. As shown in
Fig. 1 Oc, the < 15 kyr signal components extracted
by high-pass filtering can be matched in all detail.
Again, spectral analysis is a promising strategy to
detect the environmental factors causing this unusually regular pattern. Stochastic processes would
yield smeared 'colored noise' spectra with poorly
defined maxima, while periodic signals are represented by much sharper peaks appearing at characteristic frequencies - under the premise that a
record is undisturbed, accurately dated, and covering enough time to provide ample spectral resolution. Because these conditions are never fully met
in practice, elaborate statistical techniques have
been developed to distinguish harmonic and random
signal oscillations (e.g., Thomson 1982; for applications see Yiou et al. 1991; Nobes et al. 1991;
Cortijo et al. 1995).
Since all fundamental Milankovitch periods exceed 18 kyr (e.g., Berger and Loutre 1992;
621
The numerical difference between the spectrograms ofthe SUSAS record and the ETP curve is
displayed in Fig. 12d. For the two precessional
bands at 19 kyr and 23 kyr, the assumption of a
constant proxy response appears justified, as both
records largely compensate. In contrast, the residues in the 41 kyr band show a variable response
to obliquity forcing. Two marked spectral intensity
declines at 1.2-1.1 and 0.7-0.6 Ma (Fig. 12c) cannot be explained as a result of steady response to
orbital forcing (Fig. 12b) and suggest that the potential of obliquity to drive the observed carbonate
dissolution cycles decreased stepwise during Pleistocene times. This finding is in accordance with
analyses made for 8 18 0 records (e.g., Ruddiman
et al. 1989; Joyce et al. 1990) and implies that the
climatic system was less sensitive to obliquity forcing during the late Pleistocene. The coincidence of
target and proxy records in the 100 kyr band is
even lower, as indicated by larger residues and
steeper gradients. While the ETP model looses
100 kyr power (Fig. 12b), the SUSAS stack, like
most climate records, documents an intensification
of 100 kyr oscillations in late Pleistocene.
This trend is interrupted by an early Pleistocene
100 kyr maximum appearing between 1.2 and
1.1 Ma. This 'premature 100 kyr incident', also
observed in 8 18 0 records (e.g., Mudelsee and Stattegger 1997), merges with a phase of strong 70 kyr
cyclicity (Fig. 12c) which has also been found in
other paleoceanographic records (e.g., Ruddiman
et al. 1989; Robinson 1990; Bassinot et al. 1994;
Bolton et al. 1995). Two low-frequency positive
maxima centered at 0.95 and 0.6 Ma in the SUSAS
and the residual spectrogram (Fig. l2c, d) result
from remainders ofthe earlier discussed base line
transitions of the long-term susceptibility trend
(Fig. 11 bottom).
The most striking result ofthe residual spectrogram is the reciprocity of spectral intensities in the
41 and 100 kyr bands (dashed outlines in Fig. 12d).
In relation to the stationary ETP model, the SUSAS
100 kyr cyclicity develops over-proportionally in
sections with reduced response to obliquity (blue
outline) and retreats, where the response to obliquity is strong (yellow outline). These findings imply an exchange of spectral energy between obliquity and eccentricity in overall accordance with
insolation threshold models (e.g., Saltzman and
Verbitsky 1993; Raymo 1997). Intensified northern
hemisphere summer insolation capable of triggering major deglaciations hence should result from an
interference of obliquity maxima with precession
index minima. At times, when the insolation threshold is relatively low, obliquity maxima alone will
trigger the withdrawal of continental ice shields and
41 kyr cyclicity prevails. When the insolation
threshold is higher, the required peak insolation is
only reached by optimum interaction of obliquity and
precession. As eccentricity modulates the precession amplitude, sufficiently large precessional peaks
occur only during 100 kyr eccentricity maxima
causing the observed 100 kyr climate cycles.
Raymo (1997) assumes a linear, pC0 2 controlled rise of the insolation threshold over the past
1.2 m.y. and obtained a single transition from 41 kyr
to 100 kyr predominance at about 0.7 Ma. Our evolutionary spectral analyses (Figs. 12c, d) imply, that
this turnover was predated by a 100 kyr episode
around 1.2 Ma.
Sub-Milankovitch signal variations of the
CEARIS stack
A most interesting feature of all CEARlS records
are consistent high-frequency variations in the subMilankovitch to millennial range. As shown in
Fig. 1 Oc, the < 15 kyr signal components extracted
by high-pass filtering can be matched in all detail.
Again, spectral analysis is a promising strategy to
detect the environmental factors causing this unusually regular pattern. Stochastic processes would
yield smeared 'colored noise' spectra with poorly
defined maxima, while periodic signals are represented by much sharper peaks appearing at characteristic frequencies - under the premise that a
record is undisturbed, accurately dated, and covering enough time to provide ample spectral resolution. Because these conditions are never fully met
in practice, elaborate statistical techniques have
been developed to distinguish harmonic and random
signal oscillations (e.g., Thomson 1982; for applications see Yiou et al. 1991; Nobes et al. 1991;
Cortijo et al. 1995).
Since all fundamental Milankovitch periods exceed 18 kyr (e.g., Berger and Loutre 1992;