10
70 1980
10
70
1981
1985
1986
10
1982
1983
1984
Figure 7: Time-height section of 5-month zonal wind anomalies over the equator. Contour
interval 10 m/s. Westerlies are shaded. The ordinate ranges from the 70-mb pressure
level near the tropical tropopause at 17 km to the 10-mb level near 31 km and the scale is
logarithmic in pressure and approximately linear in height. Labels on the abscissa denote
the midpoint of the calendar year. From Wallace et al. (1993)
tions of the total variance of the time series, is usually more informative
to regard the time series as aperiodic (i.e., without periodicities or quasiperiodicities) and to focus on the gross features of the spectra and the
relationships between variables that prevail across wide ranges of frequencies and account for substantial fractions of the total variance.
2.3 Aperiodic and random variability
Whether the variability associated with a particular aperiodic phenomenon
should be viewed as deterministic or random depends upon the range of
frequencies under consideration. For example, baroclinic waves and their
attendant cyclones and fronts are deterministic and quite predictable out
to several days, yet they have to be treated stochastically in seasonalto-interannual climate prediction. The "stormtracks" (the envelopes in
which the most vigorous baroclinic wave activity resides) might evolve
in a deterministic manner on the seasonal to interannual time scale, but
not the individual cyclones within them. In a similar manner, ENSO is
deterministic and to some extent predictable on the seasonal to interannual
time scale, but it would have to be treated stochastically in dealing with
decade-to-century scale variations.
70 1980
10
70
1981
1985
1986
10
1982
1983
1984
Figure 7: Time-height section of 5-month zonal wind anomalies over the equator. Contour
interval 10 m/s. Westerlies are shaded. The ordinate ranges from the 70-mb pressure
level near the tropical tropopause at 17 km to the 10-mb level near 31 km and the scale is
logarithmic in pressure and approximately linear in height. Labels on the abscissa denote
the midpoint of the calendar year. From Wallace et al. (1993)
tions of the total variance of the time series, is usually more informative
to regard the time series as aperiodic (i.e., without periodicities or quasiperiodicities) and to focus on the gross features of the spectra and the
relationships between variables that prevail across wide ranges of frequencies and account for substantial fractions of the total variance.
2.3 Aperiodic and random variability
Whether the variability associated with a particular aperiodic phenomenon
should be viewed as deterministic or random depends upon the range of
frequencies under consideration. For example, baroclinic waves and their
attendant cyclones and fronts are deterministic and quite predictable out
to several days, yet they have to be treated stochastically in seasonalto-interannual climate prediction. The "stormtracks" (the envelopes in
which the most vigorous baroclinic wave activity resides) might evolve
in a deterministic manner on the seasonal to interannual time scale, but
not the individual cyclones within them. In a similar manner, ENSO is
deterministic and to some extent predictable on the seasonal to interannual
time scale, but it would have to be treated stochastically in dealing with
decade-to-century scale variations.