314 Martin Fischer
1988, Storch et al. 1995). A very brief description of the method is given in Appendix B. The POP analysis yields a set of osci11ation patterns. Each of them represents a damped osci11ation in a two dimensional phase space which is represented
by the complex plane. In Figs. 16.3, 16.4, and 16.5 the results of a combined POP
analysis of sea surface temperature, upper ocean heat content, and pseudo wind
stress anomalies are presented. The real part corresponds to the intermediate phase
and the negative of the imaginary part represents the extreme warm phase. The
similarity between the heat content anomaly patterns and the theoretical patterns
that follow from the delayed action oscillator theory is obvious. In Fig. 16.4 the
corresponding pseudo wind stress anomaly patterns are displayed. To each POP
mode two time series can be attributed by projecting the original data sets onto the
POP patterns. In Fig. 16.5 these time series are presented. The quasi periodic character is obvious and a phase shift of about 90° between the two time series indicated the predictability of ENSO. Furthermore, we may infer from the POP
patterns, together with the corresponding time series that the heat content and wind
stress anomalies propagate, whereas the SST anomaly is a standing pattern with
varying amplitude. This also agrees weB with the delayed action osci11ator theory.
These characteristics become even more evident in a Hovm6Ber diagram (Fig.
16.6). Instead of the original data, the time evolution as described by the POP
model is presented. This filters out patterns that are not related to ENSO. Especially in the heat content it is obvious how anomalies in the west Pacific propagate
within one to two years to the east Pacific and finally result in an extreme warm or
cold event. Thus, heat content anomalies in the west Pacific may be used as a precursor for El Nifio/La Nifia predictions.
16.2.2 ENSO predictability
From observational studies as weB as from modeling studies it is evident that
ENSO is predictable on lead times up to a year. The easiest way to prove this is to
look at the time series of upper ocean heat content in the west and in the east
Pacific. In the following, averages over the two index regions NINO-3 (5°N - 5°S,
150 0 E - 150 0 W) and NINO-4 (5°N - 5°S, 150 0 W - 90 0 W) are frequent1y used (Fig.
16.7). Most ofthe scientific groups working on ENSO and ENSO predictions use
averages over these regions to quantify the quality of simulations and predictions.
From Fig. 16.8 one may infer that extreme ENSO events are predictable on lead
times up to about one year. The averaged temperature anomaly down to a depth of
275 meters is used as a measure for upper ocean heat content. The anomalous heat
content in the west Pacific leads the corresponding signal in the east Pacific by
about one year. The maximum correlation of about 0.82 between the two time
series (not shown here) is observed at a time lag of 11 months which is a clear indication for the predictability of ENSO. Even more evidence for the predictability of
ENSO comes from the POP analysis from the previous section. The leading POP
mode explains about 32% of the total variance in the combined data set which
demonstrates that this is the dominant pattern. The rotation period is about four
years and the decay time about 3 years. The two POP coefficient time series vary
1988, Storch et al. 1995). A very brief description of the method is given in Appendix B. The POP analysis yields a set of osci11ation patterns. Each of them represents a damped osci11ation in a two dimensional phase space which is represented
by the complex plane. In Figs. 16.3, 16.4, and 16.5 the results of a combined POP
analysis of sea surface temperature, upper ocean heat content, and pseudo wind
stress anomalies are presented. The real part corresponds to the intermediate phase
and the negative of the imaginary part represents the extreme warm phase. The
similarity between the heat content anomaly patterns and the theoretical patterns
that follow from the delayed action oscillator theory is obvious. In Fig. 16.4 the
corresponding pseudo wind stress anomaly patterns are displayed. To each POP
mode two time series can be attributed by projecting the original data sets onto the
POP patterns. In Fig. 16.5 these time series are presented. The quasi periodic character is obvious and a phase shift of about 90° between the two time series indicated the predictability of ENSO. Furthermore, we may infer from the POP
patterns, together with the corresponding time series that the heat content and wind
stress anomalies propagate, whereas the SST anomaly is a standing pattern with
varying amplitude. This also agrees weB with the delayed action osci11ator theory.
These characteristics become even more evident in a Hovm6Ber diagram (Fig.
16.6). Instead of the original data, the time evolution as described by the POP
model is presented. This filters out patterns that are not related to ENSO. Especially in the heat content it is obvious how anomalies in the west Pacific propagate
within one to two years to the east Pacific and finally result in an extreme warm or
cold event. Thus, heat content anomalies in the west Pacific may be used as a precursor for El Nifio/La Nifia predictions.
16.2.2 ENSO predictability
From observational studies as weB as from modeling studies it is evident that
ENSO is predictable on lead times up to a year. The easiest way to prove this is to
look at the time series of upper ocean heat content in the west and in the east
Pacific. In the following, averages over the two index regions NINO-3 (5°N - 5°S,
150 0 E - 150 0 W) and NINO-4 (5°N - 5°S, 150 0 W - 90 0 W) are frequent1y used (Fig.
16.7). Most ofthe scientific groups working on ENSO and ENSO predictions use
averages over these regions to quantify the quality of simulations and predictions.
From Fig. 16.8 one may infer that extreme ENSO events are predictable on lead
times up to about one year. The averaged temperature anomaly down to a depth of
275 meters is used as a measure for upper ocean heat content. The anomalous heat
content in the west Pacific leads the corresponding signal in the east Pacific by
about one year. The maximum correlation of about 0.82 between the two time
series (not shown here) is observed at a time lag of 11 months which is a clear indication for the predictability of ENSO. Even more evidence for the predictability of
ENSO comes from the POP analysis from the previous section. The leading POP
mode explains about 32% of the total variance in the combined data set which
demonstrates that this is the dominant pattern. The rotation period is about four
years and the decay time about 3 years. The two POP coefficient time series vary