2
how various techniques might be used, we will attempt to identify some
of the opportunities and unresolved issues in climate research that invite
further analysis of observed and model generated datasets. Rather than
focusing on the solutions we will explore the problems. The choice of topics
to be included was largely dictated by the author's preconceived notions
concerning the nature of climate variability, which include the following:
1. The statistically robust, richly textured response to the annual cycle
in sun-earth geometry is still capable of yielding new insights into the
inner workings of the climate system (§2.1)
2. Quasi-periodic phenomena, even if they exist, are of only academic
interest from the standpoint of short term climate variability, because
they cannot possibly account more than a minute fraction of the total variance of the climatic variables of the greatest interest from the
point of view of prediction. By far the most prominent quasi-periodic
climate 'signal' is the equatorial stratospheric quasi-biennial oscillation (§2.2, 3), but this phenomenon exhibits only a tenuous connection
with tropospheric climate variability. It will be argued that the dominant interannual signal in the climate system: the El Nino Southern
Oscillation (ENSO) phenomenon (§2.2) occupies too broad a band of
frequencies to be regarded as quasi-periodic.
3. As demonstrated clearly by recent numerical modeling results of Manabe and Stouffer (1996), not all interdecadal variability in climatic
time series is caused by processes operative on the interdecadal time
scale. Much of it is merely a reflection of inherently unpredictable
sampling fluctuations due to the presence of variability on the intraseasonal and interannual time scales (§2.3-4, and 8).
4. Well defined interdecadal to century scale variability is clearly evident
in certain climatic time series such as hemispheric- or global-mean
surface air temperature, even in their raw, unfiltered form (§2.4).
5. Because of the inherent lack of stationarity of many climatic time series, in the absence of a priori predictions it is extremely difficult to
demonstrate the statistical significance of 'regime shifts' or unprecedented events (§2.5).
6. With a few notable exceptions involving quasi-periodic phenomena,
lead/lag relationships observed in association with interannual to in-
how various techniques might be used, we will attempt to identify some
of the opportunities and unresolved issues in climate research that invite
further analysis of observed and model generated datasets. Rather than
focusing on the solutions we will explore the problems. The choice of topics
to be included was largely dictated by the author's preconceived notions
concerning the nature of climate variability, which include the following:
1. The statistically robust, richly textured response to the annual cycle
in sun-earth geometry is still capable of yielding new insights into the
inner workings of the climate system (§2.1)
2. Quasi-periodic phenomena, even if they exist, are of only academic
interest from the standpoint of short term climate variability, because
they cannot possibly account more than a minute fraction of the total variance of the climatic variables of the greatest interest from the
point of view of prediction. By far the most prominent quasi-periodic
climate 'signal' is the equatorial stratospheric quasi-biennial oscillation (§2.2, 3), but this phenomenon exhibits only a tenuous connection
with tropospheric climate variability. It will be argued that the dominant interannual signal in the climate system: the El Nino Southern
Oscillation (ENSO) phenomenon (§2.2) occupies too broad a band of
frequencies to be regarded as quasi-periodic.
3. As demonstrated clearly by recent numerical modeling results of Manabe and Stouffer (1996), not all interdecadal variability in climatic
time series is caused by processes operative on the interdecadal time
scale. Much of it is merely a reflection of inherently unpredictable
sampling fluctuations due to the presence of variability on the intraseasonal and interannual time scales (§2.3-4, and 8).
4. Well defined interdecadal to century scale variability is clearly evident
in certain climatic time series such as hemispheric- or global-mean
surface air temperature, even in their raw, unfiltered form (§2.4).
5. Because of the inherent lack of stationarity of many climatic time series, in the absence of a priori predictions it is extremely difficult to
demonstrate the statistical significance of 'regime shifts' or unprecedented events (§2.5).
6. With a few notable exceptions involving quasi-periodic phenomena,
lead/lag relationships observed in association with interannual to in-