¥ Only the upper tropical ocean influences the
El Niño-Southern Oscillation phenomenon
(ENSO).
¥ All changes in the circulation have deterministic causes.
¥ The ocean is in equilibrium with the atmosphere (the present-day ocean state is caused by
interaction with the present-day atmosphere).
Few meteorologists would argue that large regions
of the atmosphere can be omitted from their models without degrading them: there is no argument
known to me that any part of the atmosphere is
decoupled, laterally, from the rest on time scales
exceeding a day or two. Similarly, it seems
unlikely that any part of the ocean can safely be
assumed to be unchanging, and passive, in attempts
to model the system over periods from a few to
thousands of years. Thus ocean modellers have
gravitated towards making their models truly
global. But the more sophisticated a model is, the
higher is the data quality required to test it (a
crude model only requires crude tests). Such models have to be tested and calibrated everywhere –
an error in high latitudes will eventually ‘bite’ one
in the tropics after the fluid has carried that erroneous information through the system. That transmission may take a long time, but eventually, on
some climate time scale, it will probably happen.
One must have observations everywhere capable
of delineating all space and time scales.
2.1.4.2 Too short records
We know from the geological record that climate
has been changing ever since the time when it
makes sense to speak of such a phenomenon, i.e.
since about 3 billion years ago. Within this enormous span of time, changes have been taking place
on every time scale definable as ‘climate’, i.e. from
a year or two, to millions and billions of years.
There is an enormous temptation to interpret
the record of change as seen in one’s lifetime, or
even in human history, as being representative of
the climate system as a whole. Such a temptation
is fraught with grave dangers. Systems with memory, and the climate system, particularly the ocean
(and cryosphere), have long-lived and multiple
time-scale memories, undergo often highly unintuitive random walks. This type of behaviour is too
readily converted into interesting deterministic
stories. The tendency to long random walks is the
central element of Hasselmann’s (1976) stochastic
theory of climate. It is easy to give examples where
one’s intuition can fail very badly in these circumstances, and I have written about this phenomenon
elsewhere (Wunsch, 1992, 1999a). Consider by way
of example, the ‘temperature’ record in Fig. 2.1.4a
(Wunsch, 1992). Taken by itself, it resembles many
oceanographic time series. One might be tempted
to think of the maximum near time 2500 as representing a climate extreme. But this record was
generated simply as the accumulating sum of that
shown in Fig. 2.1.4b – which is pure white noise.
The existence of a maximum near 2500, a strong
minimum near 1800, and the apparent long-term
trend in between, has no ‘cause’ other than the
stochastic accumulation of a few excess positive or
negative values in a long run of a purely random
sequence.
Another example can be seen in Fig. 2.1.5; here
the transport of a western boundary current and
its spectral density are displayed. Again, the variability is not visually very different from that seen
in practice (e.g. Schott et al., 1988). In this particular case, however, the variability was generated
in a theoretical Stommel-gyre by a purely stochastic variability (white noise in both space and time)
in the interior wind forcing.
It is often argued (as in the WOCE design
period), that large-scale observation systems are
not required – that it suffices to instrument
so-called choke points of the system and to monitor the behaviour of the flow through them. The
most commonly described choke points include
the Drake Passage and the Florida Straits, as well
as the Indonesian Passages, and sometimes the
region south of the Cape of Good Hope. But
Fig. 2.1.5, and others like it, show this type of
strategy to be fallacious: choke points are regions
where any stochastic forcing and response for the
entirety of the ocean sum together, and where one
expects to see large-scale positive and negative
excursions like that in the figure. The temptation
to ascribe such fluctuations to large-scale deterministic changes, e.g. in the wind field, has not
often been resisted (see the literature particularly
on the Florida Current). The circulation certainly
contains both stochastic and deterministic elements. Separating them can be quite challenging,
requiring adequate observations to depict the
response of the complete ocean interior. Often, the
random walk explanation will prove the simplest
SECTION 2 OBSERVATIONS AND MODELS
54
El Niño-Southern Oscillation phenomenon
(ENSO).
¥ All changes in the circulation have deterministic causes.
¥ The ocean is in equilibrium with the atmosphere (the present-day ocean state is caused by
interaction with the present-day atmosphere).
Few meteorologists would argue that large regions
of the atmosphere can be omitted from their models without degrading them: there is no argument
known to me that any part of the atmosphere is
decoupled, laterally, from the rest on time scales
exceeding a day or two. Similarly, it seems
unlikely that any part of the ocean can safely be
assumed to be unchanging, and passive, in attempts
to model the system over periods from a few to
thousands of years. Thus ocean modellers have
gravitated towards making their models truly
global. But the more sophisticated a model is, the
higher is the data quality required to test it (a
crude model only requires crude tests). Such models have to be tested and calibrated everywhere –
an error in high latitudes will eventually ‘bite’ one
in the tropics after the fluid has carried that erroneous information through the system. That transmission may take a long time, but eventually, on
some climate time scale, it will probably happen.
One must have observations everywhere capable
of delineating all space and time scales.
2.1.4.2 Too short records
We know from the geological record that climate
has been changing ever since the time when it
makes sense to speak of such a phenomenon, i.e.
since about 3 billion years ago. Within this enormous span of time, changes have been taking place
on every time scale definable as ‘climate’, i.e. from
a year or two, to millions and billions of years.
There is an enormous temptation to interpret
the record of change as seen in one’s lifetime, or
even in human history, as being representative of
the climate system as a whole. Such a temptation
is fraught with grave dangers. Systems with memory, and the climate system, particularly the ocean
(and cryosphere), have long-lived and multiple
time-scale memories, undergo often highly unintuitive random walks. This type of behaviour is too
readily converted into interesting deterministic
stories. The tendency to long random walks is the
central element of Hasselmann’s (1976) stochastic
theory of climate. It is easy to give examples where
one’s intuition can fail very badly in these circumstances, and I have written about this phenomenon
elsewhere (Wunsch, 1992, 1999a). Consider by way
of example, the ‘temperature’ record in Fig. 2.1.4a
(Wunsch, 1992). Taken by itself, it resembles many
oceanographic time series. One might be tempted
to think of the maximum near time 2500 as representing a climate extreme. But this record was
generated simply as the accumulating sum of that
shown in Fig. 2.1.4b – which is pure white noise.
The existence of a maximum near 2500, a strong
minimum near 1800, and the apparent long-term
trend in between, has no ‘cause’ other than the
stochastic accumulation of a few excess positive or
negative values in a long run of a purely random
sequence.
Another example can be seen in Fig. 2.1.5; here
the transport of a western boundary current and
its spectral density are displayed. Again, the variability is not visually very different from that seen
in practice (e.g. Schott et al., 1988). In this particular case, however, the variability was generated
in a theoretical Stommel-gyre by a purely stochastic variability (white noise in both space and time)
in the interior wind forcing.
It is often argued (as in the WOCE design
period), that large-scale observation systems are
not required – that it suffices to instrument
so-called choke points of the system and to monitor the behaviour of the flow through them. The
most commonly described choke points include
the Drake Passage and the Florida Straits, as well
as the Indonesian Passages, and sometimes the
region south of the Cape of Good Hope. But
Fig. 2.1.5, and others like it, show this type of
strategy to be fallacious: choke points are regions
where any stochastic forcing and response for the
entirety of the ocean sum together, and where one
expects to see large-scale positive and negative
excursions like that in the figure. The temptation
to ascribe such fluctuations to large-scale deterministic changes, e.g. in the wind field, has not
often been resisted (see the literature particularly
on the Florida Current). The circulation certainly
contains both stochastic and deterministic elements. Separating them can be quite challenging,
requiring adequate observations to depict the
response of the complete ocean interior. Often, the
random walk explanation will prove the simplest
SECTION 2 OBSERVATIONS AND MODELS
54