 |
| Titel |
Frequentist and Bayesian analyses of the uncertainty associated with regional wind estimations |
| VerfasserIn |
E. García-Bustamante, J. F. González-Rouco, P. A. Jiménez, J. Navarro, J. P. Montávez |
| Konferenz |
EGU General Assembly 2010
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 12 (2010) |
| Datensatznummer |
250043379
|
|
|
|
|
|
| Zusammenfassung |
| A statistical downscaling technique is used to analyze the surface wind variability in a region
of complex terrain in the northeast of the Iberian Peninsula. This type of procedure provides
an added value with respect to the global model simulations that show limitations in
reproducing the regional spatial scales. However, the downscaling strategy involves a
source of uncertainty that adds to the cascade of uncertainties associated with the
estimations.
The uncertainties in the downscaling step are analyzed on the basis of the methodological
sensitivity to changes in some options of the statistical model. 14 years of observations were
used to understand the relation of the regional wind over the northeastern Iberian Peninsula
and the large scale circulation over the North Atlantic area. The sensitivity of the downscaling
technique (Canonical Correlation Analysis) is explored by sampling all parameters that are
important for the model configuration. The relative importance of each parameter together
with the spatial and temporal variability of the uncertainty in the wind estimates is
explored.
This can be considered as a classical (frequentist) approach to the treatment of the
uncertainties. From this point of view, a certain degree of subjectivity is involved in
the selection of possible values for the parameters of the model. Alternatively, a
more objective assessment of this type of methodological uncertainty is possible
applying a Bayesian analysis. The procedure implies that prior knowledge in the
model parameters can be updated by using the available observations during the
calibration period. Thus, the optimal parameters and the uncertainty associated
with the method are estimated based on the possible constraints imposed by the
observations.
A robust assessment of the uncertainties derived from the downscaling step has many
implications for the understanding of past and future estimations of the wind field. It will be
illustrated why single estimations should be managed with care and it also will be show how
a Bayesian treatment can assign weights (probabilities) to each possible combination of
parameters discriminating which cases are more realistic according to the information
provided by the observed fields. |
|
|
|
|
|
|