132
ing warm-phase ENSO years, the PDF of the oceanic regime is enhanced
leading to a greater probability of a poor overall monsoon. During coldphase ENSO years the PDF of the continental regime is favoured, leading
to greater probability of a good monsoon. In this picture, the predictability
of seasonal-mean monsoon fluctuations to imposed external SST anomalies is only partial, a result demonstrated in GeM ensemble integrations
(Sperber and Palmer, 1996).
One simple consequence of the picture put forward is that if the circulation patterns are dominated by a few basic regimes, and if the system
basically responds to weak forcing through changes in regime frequency,
then the EOFs of the system will match the patterns of these regimes
no matter how long a timescale they are computed over. Hence, the fact
that the PNA and NAO patterns emerge as dominant EOFs on weekly,
seasonal, and multi-decadal timescales is entirely consistent with the nonlinear paradigm put forward here.
We will return to the paradigm put forward here in section 6 when we
discuss climate change.
5 Predictability of interdecadal fluctuations
5.1 Internal atmospheric variability
A basic theme underlying this paper is the chaotic nature of climate. Much
of this derives from the atmosphere. The ubiquitous growth of atmospheric
perturbations, as revealed by singular vector analysis above, together with
the underlying nonlinear structure of the atmosphere, suggested, for example, by potential vorticity diagnosis, (eg the wave-breaking and wave,
mean-flow interaction processes illustrated in Hoskins et aI, 1985), is itself
supporting evidence of chaotic variability.
Since chaotic processes are inherently aperiodic, a spectral analysis of
a chaotic time series will reveal power over a range of timescales, possibly
strongly removed from the principal timescale of the dominant instability
process (Lyapunov exponent timescale). For the atmosphere, it is possible that chaotic variability associated with the 'fast' baroclinic timescale,
together with the 'medium' timescale processes associated with regime dynamics, may generate a significant component of 'long' timescale, interannual and interdecadal fluctuations. One could define the word 'significant' through an f-test, comparing the fraction of low-frequency variance
ing warm-phase ENSO years, the PDF of the oceanic regime is enhanced
leading to a greater probability of a poor overall monsoon. During coldphase ENSO years the PDF of the continental regime is favoured, leading
to greater probability of a good monsoon. In this picture, the predictability
of seasonal-mean monsoon fluctuations to imposed external SST anomalies is only partial, a result demonstrated in GeM ensemble integrations
(Sperber and Palmer, 1996).
One simple consequence of the picture put forward is that if the circulation patterns are dominated by a few basic regimes, and if the system
basically responds to weak forcing through changes in regime frequency,
then the EOFs of the system will match the patterns of these regimes
no matter how long a timescale they are computed over. Hence, the fact
that the PNA and NAO patterns emerge as dominant EOFs on weekly,
seasonal, and multi-decadal timescales is entirely consistent with the nonlinear paradigm put forward here.
We will return to the paradigm put forward here in section 6 when we
discuss climate change.
5 Predictability of interdecadal fluctuations
5.1 Internal atmospheric variability
A basic theme underlying this paper is the chaotic nature of climate. Much
of this derives from the atmosphere. The ubiquitous growth of atmospheric
perturbations, as revealed by singular vector analysis above, together with
the underlying nonlinear structure of the atmosphere, suggested, for example, by potential vorticity diagnosis, (eg the wave-breaking and wave,
mean-flow interaction processes illustrated in Hoskins et aI, 1985), is itself
supporting evidence of chaotic variability.
Since chaotic processes are inherently aperiodic, a spectral analysis of
a chaotic time series will reveal power over a range of timescales, possibly
strongly removed from the principal timescale of the dominant instability
process (Lyapunov exponent timescale). For the atmosphere, it is possible that chaotic variability associated with the 'fast' baroclinic timescale,
together with the 'medium' timescale processes associated with regime dynamics, may generate a significant component of 'long' timescale, interannual and interdecadal fluctuations. One could define the word 'significant' through an f-test, comparing the fraction of low-frequency variance