232
Ghapter 13: Spatial Patterns: EOFs and GGA
Figure 13.1: First two characteristic horizontal distributions ff; ofnormalized
temperature in the Mediterranean Sea, as inferred from the output of a 9-year
run with a numerical ocean model. (horn Korre.., and Pinardi).
44..1
57%
~..IL-______________________ ~ ____ ~~~ __ ~ __ ~ __ ~ __ ~
32%
38.1
~..I~----------------------~----~~~--~--~--~--~
.. .3
·1.7
&..
13.4
2o.t
2'"
summer - indicating warmer than average conditions) plus a slight upward
trend (gradual warming).
13.2.3 Specification of Guess Patterns
There are various ways to define the patterns ffk:4
• The very general approach is Hasselmann's "Principal Interaction Patterns" formulation (PIP; Hasselmann, 1988). The patterns are implicitly
defined such that their coefficients Ckk(t) approximate certain dynamical
equations, which feature unknown parameters.
• A simplified version of the PIPs are the "Principal Oscillation Patterns"
(POPs, H. von Storch et al. , 1988, 1993), which model linear dynamics and which have been successfully applied for the analysis of various
processes (see Chapter 15).
• A standard statistical exercise in climate research aims at the identification of expected signals, such as the atmospheric response to enhanced
greenhouse gas concentrations or to anomalous sea-surface temperature
conditions (see Chapter 8). This identification is often facilitated by the
specification of patterns determined in experiments with general circulation models (H. von Storch, 1987; Santer et al., 1993, Heger! et al., 1994).
'See also Section 8.3.2.
Ghapter 13: Spatial Patterns: EOFs and GGA
Figure 13.1: First two characteristic horizontal distributions ff; ofnormalized
temperature in the Mediterranean Sea, as inferred from the output of a 9-year
run with a numerical ocean model. (horn Korre.., and Pinardi).
44..1
57%
~..IL-______________________ ~ ____ ~~~ __ ~ __ ~ __ ~ __ ~
32%
38.1
~..I~----------------------~----~~~--~--~--~--~
.. .3
·1.7
&..
13.4
2o.t
2'"
summer - indicating warmer than average conditions) plus a slight upward
trend (gradual warming).
13.2.3 Specification of Guess Patterns
There are various ways to define the patterns ffk:4
• The very general approach is Hasselmann's "Principal Interaction Patterns" formulation (PIP; Hasselmann, 1988). The patterns are implicitly
defined such that their coefficients Ckk(t) approximate certain dynamical
equations, which feature unknown parameters.
• A simplified version of the PIPs are the "Principal Oscillation Patterns"
(POPs, H. von Storch et al. , 1988, 1993), which model linear dynamics and which have been successfully applied for the analysis of various
processes (see Chapter 15).
• A standard statistical exercise in climate research aims at the identification of expected signals, such as the atmospheric response to enhanced
greenhouse gas concentrations or to anomalous sea-surface temperature
conditions (see Chapter 8). This identification is often facilitated by the
specification of patterns determined in experiments with general circulation models (H. von Storch, 1987; Santer et al., 1993, Heger! et al., 1994).
'See also Section 8.3.2.