320 Martin Fischer
16.3.2 Physical ENSO prediction models
Since ENSO is a phenomenon of the coupled ocean-atmosphere system, coupled
ocean-atmosphere models are required for its prediction. Such models have been
developed over the last 10 to 15 years by different groups and applied to ENSO
forecasts (e.g. Cane et al. 1986, Bamett et al. 1988, Leetmaa and Ji 1989, Latif et
al. 1994). The coupled models can be categorized into three classes of models:
Intermediate models, hybrid coupled models (HCM) and coupled general circulation models (CGCM).
Intermediate models
The development of the Lamont-model (Zebiak and Cane 1987) was a milestone
in ENSO forecasting. It is a non linear anomaly model of intermediate complexity,
and has been used operationally for ENSO forecasts since 1986. It is a limited
domain model covering the tropical Pacific. Thus there is an inherent assumption
that the cause of ENSO lies entirely in this region. The model simulates realistically the interannual variability in tropical Pacific SST. The mechanism for the
interannual variability is closely related to the subsurface memory of the system
and is consistent with the delayed action oscillator scenario. This model is one of
the most successful ENSO forecast models. The correlation skill of this model for a
large number of predictions, initialized monthly during the period 1972 to 1995 is
very encouraging (Fig. 16.16). For most ofthe period a useful correlation skill of
0.5 is still found after one year. The prediction skill of this model will be discussed
in more detail in section 16.4.
Hybrid coupled models
In section 16.2.1 it was pointed out that ENSO is a coupled ocean-atmosphere
phenomenon but that the time evolution of the system is mainly determined by the
ocean. At low frequencies the atmosphere over the tropical Pacific can be regarded
as a strongly forced quasi equilibrium system governed by the state of the tropical
Pacific SST. Theoretical investigations, observations, and numerical studies with
coupled ocean atmosphere models support this view. Thus it is possible to express
the atmosphere's response to variations in the oceanic boundary conditions by a
statistical (empirical) model inferred from data. In the following the statistic al
atmosphere model used at the Max-Planck-Institut in Hamburg is described as an
example (LatifandFliigel1991, FliigeI1994).
We as sume that the memory of the coupled ocean-atmosphere system resides
entirely in the tropical Pacific. It is therefore possible to use a minimum atmospheric model with no internal dynamics which passively responds to the anomalous boundary conditions provided by the ocean. Further, we as sume a linear
relation between the dominant anomaly patterns of sea surf ace temperature and
wind stress. In the following, we explain how the atmosphere model was constructed (see also Bamett et al. 1993). We decompose observed sea surf ace temperature T '(x,t) and wind stress anomalies 't'(x,t) into empirical orthogonal functions
(EOF):
16.3.2 Physical ENSO prediction models
Since ENSO is a phenomenon of the coupled ocean-atmosphere system, coupled
ocean-atmosphere models are required for its prediction. Such models have been
developed over the last 10 to 15 years by different groups and applied to ENSO
forecasts (e.g. Cane et al. 1986, Bamett et al. 1988, Leetmaa and Ji 1989, Latif et
al. 1994). The coupled models can be categorized into three classes of models:
Intermediate models, hybrid coupled models (HCM) and coupled general circulation models (CGCM).
Intermediate models
The development of the Lamont-model (Zebiak and Cane 1987) was a milestone
in ENSO forecasting. It is a non linear anomaly model of intermediate complexity,
and has been used operationally for ENSO forecasts since 1986. It is a limited
domain model covering the tropical Pacific. Thus there is an inherent assumption
that the cause of ENSO lies entirely in this region. The model simulates realistically the interannual variability in tropical Pacific SST. The mechanism for the
interannual variability is closely related to the subsurface memory of the system
and is consistent with the delayed action oscillator scenario. This model is one of
the most successful ENSO forecast models. The correlation skill of this model for a
large number of predictions, initialized monthly during the period 1972 to 1995 is
very encouraging (Fig. 16.16). For most ofthe period a useful correlation skill of
0.5 is still found after one year. The prediction skill of this model will be discussed
in more detail in section 16.4.
Hybrid coupled models
In section 16.2.1 it was pointed out that ENSO is a coupled ocean-atmosphere
phenomenon but that the time evolution of the system is mainly determined by the
ocean. At low frequencies the atmosphere over the tropical Pacific can be regarded
as a strongly forced quasi equilibrium system governed by the state of the tropical
Pacific SST. Theoretical investigations, observations, and numerical studies with
coupled ocean atmosphere models support this view. Thus it is possible to express
the atmosphere's response to variations in the oceanic boundary conditions by a
statistical (empirical) model inferred from data. In the following the statistic al
atmosphere model used at the Max-Planck-Institut in Hamburg is described as an
example (LatifandFliigel1991, FliigeI1994).
We as sume that the memory of the coupled ocean-atmosphere system resides
entirely in the tropical Pacific. It is therefore possible to use a minimum atmospheric model with no internal dynamics which passively responds to the anomalous boundary conditions provided by the ocean. Further, we as sume a linear
relation between the dominant anomaly patterns of sea surf ace temperature and
wind stress. In the following, we explain how the atmosphere model was constructed (see also Bamett et al. 1993). We decompose observed sea surf ace temperature T '(x,t) and wind stress anomalies 't'(x,t) into empirical orthogonal functions
(EOF):