472
the latter gives rise, via an oscillatory instability, to a limit cycle that
eventually becomes destabilized in turn.
The signal-to-noise (S/N) enhancement described in Sec. 2 led in fact,
in the case of seasonal-to-interannual variability taken as the Ariadne's
thread of methodological exposition, to the robust identification of separate 4/1-, 4/2- and 4/3-year peaks. This separation of the peaks and
description of the associated spatio-temporal variability, in data (Rasmusson et al. 1990, Jiang et al. 1995) and coupled general circulation models
(Robertson et al. 1995a, b) has helped bolster the case for the Devil's
staircase scenario provided by simple and intermediate models of the EI
Nino/Southern Oscillation (ENSO: Jin et al. 1994, 1996; Tziperman et
al. 1994). This scenario for the interaction between the intrinsic ENSO
cycle of 2-3 years (depending on models and parameters) and the seasonal
cycle could provide the cause for both the warm events occurring in boreal
winter and the irregularity of their spacing from year to year.
A similar dialog between spectral studies of (inter )decadal variability
and dynamieal studies of oscillatory model behavior is taking shape. Interdecadal spectral peaks at 10-11, 14-15 and 25-27 years were identified
by combining the application of the various spectral methods described
in Sec. 3, in both global (Ghil and Vautard 1991, Vautard et al. 1992)
and regional (Plaut et al. 1995) time series. Using (M-)SSA-related but
distinct methods, Allen and Smith (1994) described the spatial patterns
associated with the peakks at 10-11 and 25-27 years, while the peak at
14-15 years was confirmed by SSA or coral-reef proxy records (Quinn et
al. 1993) and its associated spatial pattern described using MTM-related
methods of intrumental temperature data (Mann and Park 1995). On
the other hand, a plethora of oscillatory mechanisms is emerging in two(Quon and Ghil 1995) and three-dimensional (Weaver et al. 1993, Chen
and Ghil1995, Rahmstorf 1995) models of the ocean's thermohaline circulation [(THC); see summary table, with further references, in Ghil (1994)],
as well as in coupled ocean-atmosphere models (Delworth et al. 1993, Chen
and Ghil1996). As in the case of EN SO variability, a full hierarchy of models will probably be needed to clarify the connections between the (dimly)
observed and the (imperfectly) modeled oscillations.
On the methodological side, there is much to be done in terms of refining statistical significance tests for each one of the methods presented
here, as well as extending further their multi-channel applications. Beyond this, we showed that SSA can follow well, via RCs, variations in
the latter gives rise, via an oscillatory instability, to a limit cycle that
eventually becomes destabilized in turn.
The signal-to-noise (S/N) enhancement described in Sec. 2 led in fact,
in the case of seasonal-to-interannual variability taken as the Ariadne's
thread of methodological exposition, to the robust identification of separate 4/1-, 4/2- and 4/3-year peaks. This separation of the peaks and
description of the associated spatio-temporal variability, in data (Rasmusson et al. 1990, Jiang et al. 1995) and coupled general circulation models
(Robertson et al. 1995a, b) has helped bolster the case for the Devil's
staircase scenario provided by simple and intermediate models of the EI
Nino/Southern Oscillation (ENSO: Jin et al. 1994, 1996; Tziperman et
al. 1994). This scenario for the interaction between the intrinsic ENSO
cycle of 2-3 years (depending on models and parameters) and the seasonal
cycle could provide the cause for both the warm events occurring in boreal
winter and the irregularity of their spacing from year to year.
A similar dialog between spectral studies of (inter )decadal variability
and dynamieal studies of oscillatory model behavior is taking shape. Interdecadal spectral peaks at 10-11, 14-15 and 25-27 years were identified
by combining the application of the various spectral methods described
in Sec. 3, in both global (Ghil and Vautard 1991, Vautard et al. 1992)
and regional (Plaut et al. 1995) time series. Using (M-)SSA-related but
distinct methods, Allen and Smith (1994) described the spatial patterns
associated with the peakks at 10-11 and 25-27 years, while the peak at
14-15 years was confirmed by SSA or coral-reef proxy records (Quinn et
al. 1993) and its associated spatial pattern described using MTM-related
methods of intrumental temperature data (Mann and Park 1995). On
the other hand, a plethora of oscillatory mechanisms is emerging in two(Quon and Ghil 1995) and three-dimensional (Weaver et al. 1993, Chen
and Ghil1995, Rahmstorf 1995) models of the ocean's thermohaline circulation [(THC); see summary table, with further references, in Ghil (1994)],
as well as in coupled ocean-atmosphere models (Delworth et al. 1993, Chen
and Ghil1996). As in the case of EN SO variability, a full hierarchy of models will probably be needed to clarify the connections between the (dimly)
observed and the (imperfectly) modeled oscillations.
On the methodological side, there is much to be done in terms of refining statistical significance tests for each one of the methods presented
here, as well as extending further their multi-channel applications. Beyond this, we showed that SSA can follow well, via RCs, variations in