10
Ghapter 1: The Development of Glimate Research
application of statistical analysis in climate research is that we have only one
realization of the climate variability. Therefore almost all studies re-cook the
same information (not necessarily the same data, but data from the same
limited piece of the trajectory the climate system has moved along in the
past hundred "instrumental" years). Not surprisingly, every now and then
strange "significant" results are created, which are random events which have
been upgraded to "significance" by a type of data screening.
The development of climate research has shown that a successful understanding of climate dynamics· requires a clever combination of these three
components. Sometimes the dynamics are too complex and the simplification required for the dynamical theory cannot be sensibly done. Then the
numerical models have to demonstrate that our dynamical concepts, which
are encoded in our basic physical equations, really describe the phenomenon
at hand. Furthermore, a statistical analysis of observed data should clarify
that the model results are not artifacts of the model, which might stern from
thermodyanmic processes which are disregarded or inadequately parameterized. Models have advanced so much in the recent past that their bahaviour is
sometimes as difficult to interpret as the real atmosphere. Statistical methods are sometimes the only way to shed light on some intricate processes
generated from a numerical esperiment. In the following we will try to introduce some of the most useful techniques to analyze climate data, either
observed or simulated, with a special emphasis on delicate junctures, where
the risk of backfiring for the carefree apprentice is sometimes very high.
Ghapter 1: The Development of Glimate Research
application of statistical analysis in climate research is that we have only one
realization of the climate variability. Therefore almost all studies re-cook the
same information (not necessarily the same data, but data from the same
limited piece of the trajectory the climate system has moved along in the
past hundred "instrumental" years). Not surprisingly, every now and then
strange "significant" results are created, which are random events which have
been upgraded to "significance" by a type of data screening.
The development of climate research has shown that a successful understanding of climate dynamics· requires a clever combination of these three
components. Sometimes the dynamics are too complex and the simplification required for the dynamical theory cannot be sensibly done. Then the
numerical models have to demonstrate that our dynamical concepts, which
are encoded in our basic physical equations, really describe the phenomenon
at hand. Furthermore, a statistical analysis of observed data should clarify
that the model results are not artifacts of the model, which might stern from
thermodyanmic processes which are disregarded or inadequately parameterized. Models have advanced so much in the recent past that their bahaviour is
sometimes as difficult to interpret as the real atmosphere. Statistical methods are sometimes the only way to shed light on some intricate processes
generated from a numerical esperiment. In the following we will try to introduce some of the most useful techniques to analyze climate data, either
observed or simulated, with a special emphasis on delicate junctures, where
the risk of backfiring for the carefree apprentice is sometimes very high.