![Hier klicken, um den Treffer aus der Auswahl zu entfernen](images/unchecked.gif) |
Titel |
Development of new ensemble methods based on the performace skills of regional climate models over South Korea |
VerfasserIn |
M. S. Suh, S. G. Oh, D. K. Lee, D. H. Cha, S. J. Choi, S. Y. Hong, H. S. Kang |
Konferenz |
EGU General Assembly 2012
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 14 (2012) |
Datensatznummer |
250065010
|
|
|
|
Zusammenfassung |
It is well known that multi-model ensembles can reduce the uncertainties of the
model results and increase the reliability of the model results. In this paper, the
prediction skills for temperature and precipitation of five ensemble methods were
discussed by using the 20 years simulation results (from 1989 to 2008) by four
regional climate models (RCMs : SNURCM, WRF, RegCM4, and RSM) driven
by NCEP-DOE and ERA-interim boundary conditions. The simulation domain is
CORDEX (COordinated Regional climate Downscaling Experiment) East Asia and the
number of grids is 197 x 233 grids with a 50-km horizontal resolution. The new
three ensemble methods, PEA_BRC, PEA_RAC and PEA_ROC, developed in this
study, are performance based ensemble averaging methods by using bias, RMSE
(root mean square errors) and correlation, RMSE and absolute correlation, and
RMSE and original correlation, respectively. The other two ensemble methods are
equal weighted averaging (EWA) and multivariate linear regression (Mul_Reg).
Fifteen years and five years data from 20-year simulation data were used to derive
the weighting coefficients and cross-validate the prediction skills of five ensemble
methods. The total number of training and evaluation is 20 times through a cyclic
method from 20 years data. The Mul_Reg (EWA) method among the five ensemble
methods shows the best (worst) skill without regard to seasons and variables during the
training period. And the PEA_RAC and PEA_ROC show very similar skills with
Mul_Reg for all variables and seasons during training period. However, the skills
and stabilities of Mul_Reg are drastically reduced when it applied to prediction
regardless of variables and seasons. However, the skills and stabilities of PEA_RAC are
slightly reduced. As a result, the PEA_RAC shows the best skill without regard to the
seasons and variables during the prediction period. This result confirms that the
new ensemble methods developed in this study, the PEA_RAC, can be used for
the prediction of regional climate without regard to the variables and averaging
time scale. In addition, the simplicity of deriving process of weighting coefficients
and applications are also the strong points of the ensemble method, PEA_RAC. |
|
|
|
|
|