 |
| Titel |
High dimensional decision dilemmas in climate models |
| VerfasserIn |
A. Bracco, J. D. Neelin, H. Luo, J. C. McWilliams, J. E. Meyerson |
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| ISSN |
1991-959X
|
| Digitales Dokument |
URL |
| Erschienen |
In: Geoscientific Model Development ; 6, no. 5 ; Nr. 6, no. 5 (2013-10-15), S.1673-1687 |
| Datensatznummer |
250085000
|
| Publikation (Nr.) |
 copernicus.org/gmd-6-1673-2013.pdf |
|
|
|
|
|
| Zusammenfassung |
| An important source of uncertainty in climate models is linked to the
calibration of model parameters. Interest in systematic and automated
parameter optimization procedures stems from the desire to improve the model
climatology and to quantify the average sensitivity associated with potential
changes in the climate system. Building upon on the smoothness of the response
of an atmospheric circulation model (AGCM) to changes of four adjustable parameters,
Neelin et al. (2010) used a quadratic
metamodel to objectively calibrate the AGCM.
The metamodel accurately estimates global
spatial averages of common fields of climatic interest, from precipitation,
to low and high level winds, from temperature at various levels to sea level
pressure and geopotential height, while providing a computationally cheap
strategy to explore the influence of parameter settings. Here, guided by the
metamodel, the ambiguities or dilemmas related to the decision making process
in relation to model sensitivity and optimization are examined. Simulations
of current climate are subject to considerable regional-scale biases. Those
biases may vary substantially depending on the climate variable considered,
and/or on the performance metric adopted. Common dilemmas are associated with
model revisions yielding improvement in one field or regional pattern or
season, but degradation in another, or improvement in the model climatology
but degradation in the interannual variability representation. Challenges are
posed to the modeler by the high dimensionality of the model output fields
and by the large number of adjustable parameters. The use of the metamodel in
the optimization strategy helps visualize trade-offs at a regional level,
e.g., how mismatches between sensitivity and error spatial fields yield
regional errors under minimization of global objective functions. |
| |
|
| Teil von |
|
|
|
|
|
|
|
|