2
H.-J. Bolle
thus results in four different descriptions of nature which are neither identical nor can
they be identified with nature: Proxy data, ground based measurements, observations
from space, and model simulations.
One interesting argument articulated at the workshop in Casablanca was that if
various models come to nearly the same conclusion about the reaction of the climate
system to external forcing, then the probability is high that they simulate the trend of
the natural system correctly. This behaviour of models can, however, also be
interpreted in a different way. It is possible that the closer model results come to each
other the more likely it is that the physics implemented in the models equal each
other, whether this reflects the response of the real nature or not. Of course modellers
try hard to make the models a mirror of nature by continuous improvement of the
physics implemented into the algorithms. This can only be done, if there is a standard
against which the models can be tested. The only "standard" we have are measured
and estimated data sets of a few state variables which are another abstraction of the
real behaviour of nature. With the aid of these selected measured data models are
initialized and these data can also be used to validate the output of the models.
Numerical weather forecast models are moored to measured data every six hours, for
climate models this procedure is not so easy, especially if they are running into the
future.
For the initialization and validation procedures the quality of the measured data
is essential. Since they are not more than random samples which are taken, maybe,
at nearly the same time but at widely spaced locations, it is essential to check,
whether they provide a consistent picture, to prove, how good is their precision, and
to estimate, how large their error with respect to the natural state variables may be.
This last estimate is difficult to perform because no absolute standard exists. As an
example, if one measures precipitation at a few positions and calculates the
precipitation of an area which can be compared to the grid size of a model, then there
is no way to prove how close this calculation comes to reality.
Measured data are inevitable for the improvement of process descriptions in
models as well for their initialization and validation. They provide a kind of irregular
honeycomb like structure to stabilize model performance. But beyond their use with
models there is a strong motivation to use measured data as an diagnostic too\. Even
if they do not precisely reflect the natural processes and area averages, a series of data
measured at the same location under identical environmental conditions provides
information about changes and variability. Long term data series give more precise
and reliable information about local changes than models can provide. The analysis
of measured data allows to study, as an example, changes in the occurrence of
extreme events which cannot be resolved in climate models. There are periods in the
data series that represent such extreme climatic periods or climate "excursions".
Studies of the connections of such events with large scale climate phenomena such
as the North Atlantic Oscillation and monsoons may improve the understanding how
the Mediterraneran area responds to changes of the global general circulation system.
Such studies, to which it is believed the data presented here will make an important
contribution, may then lead to an answer of the key question how Mediterraneran
climate and with it Mediterranean ecosystems may react to global climate change.
The following contributions deal with various aspects of mostly empirical data
analysis with the goal to provide a consistent picture of the Mediterranean climate
H.-J. Bolle
thus results in four different descriptions of nature which are neither identical nor can
they be identified with nature: Proxy data, ground based measurements, observations
from space, and model simulations.
One interesting argument articulated at the workshop in Casablanca was that if
various models come to nearly the same conclusion about the reaction of the climate
system to external forcing, then the probability is high that they simulate the trend of
the natural system correctly. This behaviour of models can, however, also be
interpreted in a different way. It is possible that the closer model results come to each
other the more likely it is that the physics implemented in the models equal each
other, whether this reflects the response of the real nature or not. Of course modellers
try hard to make the models a mirror of nature by continuous improvement of the
physics implemented into the algorithms. This can only be done, if there is a standard
against which the models can be tested. The only "standard" we have are measured
and estimated data sets of a few state variables which are another abstraction of the
real behaviour of nature. With the aid of these selected measured data models are
initialized and these data can also be used to validate the output of the models.
Numerical weather forecast models are moored to measured data every six hours, for
climate models this procedure is not so easy, especially if they are running into the
future.
For the initialization and validation procedures the quality of the measured data
is essential. Since they are not more than random samples which are taken, maybe,
at nearly the same time but at widely spaced locations, it is essential to check,
whether they provide a consistent picture, to prove, how good is their precision, and
to estimate, how large their error with respect to the natural state variables may be.
This last estimate is difficult to perform because no absolute standard exists. As an
example, if one measures precipitation at a few positions and calculates the
precipitation of an area which can be compared to the grid size of a model, then there
is no way to prove how close this calculation comes to reality.
Measured data are inevitable for the improvement of process descriptions in
models as well for their initialization and validation. They provide a kind of irregular
honeycomb like structure to stabilize model performance. But beyond their use with
models there is a strong motivation to use measured data as an diagnostic too\. Even
if they do not precisely reflect the natural processes and area averages, a series of data
measured at the same location under identical environmental conditions provides
information about changes and variability. Long term data series give more precise
and reliable information about local changes than models can provide. The analysis
of measured data allows to study, as an example, changes in the occurrence of
extreme events which cannot be resolved in climate models. There are periods in the
data series that represent such extreme climatic periods or climate "excursions".
Studies of the connections of such events with large scale climate phenomena such
as the North Atlantic Oscillation and monsoons may improve the understanding how
the Mediterraneran area responds to changes of the global general circulation system.
Such studies, to which it is believed the data presented here will make an important
contribution, may then lead to an answer of the key question how Mediterraneran
climate and with it Mediterranean ecosystems may react to global climate change.
The following contributions deal with various aspects of mostly empirical data
analysis with the goal to provide a consistent picture of the Mediterranean climate