|
Titel |
What is really a drift in a coupled climate model used for climate forecasts ? |
VerfasserIn |
Emilia Sanchez-Gomez, Christophe Cassou, Elodie Fernandez |
Konferenz |
EGU General Assembly 2013
|
Medientyp |
Artikel
|
Sprache |
Englisch
|
Digitales Dokument |
PDF |
Erschienen |
In: GRA - Volume 15 (2013) |
Datensatznummer |
250081426
|
|
|
|
Zusammenfassung |
The climate research community has been facing a new scientific challenge with the
evaluation and understanding of the predictability at interannual to decadal (I2D hereinafter)
timescales. A considerable international effort has been devoted to the production of
near term climate predictions in a set of I2D coordinated experiments within the
CMIP5 (Coupled Model Inter-comparison Project Phase 5) framework, where models
components are initialized from observations. Preliminary results show that initialized
simulations increases the forecast skill comparing with non-initialized experiments
for leadtimes ranging from 2-3 years for the Pacific and 6-8 years for the North
Atlantic.
In despite of these encouraging results, the science of near-term climate prediction is in
its early stages. Due to the imperfect climate simulated by coupled models, when initialized
from observations, they are affected by important drifts at the beginning of the forecast
experiment that may alter their performance in terms of skill. The climate forecast
community has therefore to face with fundamental scientific and technical questions, as to the
initialization strategies, the minimization of the drift its understanding and a posteriori
correction.
Most of the I2D forecast studies are focused on skill scores on a particular variable on a
given region, and the model drift is a posteriori removed by averaging all the forecasts as a
function of leadtime. However, the model initial shock has not been carefully analyzed and
documented for the most of forecasts systems. In this work we present a detailed analysis of
the drift of decadal forecasts performed with the CNRM-CERFACS coupled model
(CNRM-CM5) when initialized from NEMOVAR-COMBINE ocean reanalysis.
CNRM-CM5 produces a strong and quick initial shock over the Tropical and also the
Austral Oceans. We show, based on EOFs analysis, that the model drift projects
on the main climate internal modes over the Tropical Pacific and Atlantic during
the first 4 years of integration. In particular, over the Tropical Pacific, the model
artificially creates a sequence of El Niño/La Niña episodes during the first 4 years of
integration, while in the Atlantic; the so-called meridional mode is excited with a 2
years swing between the two hemispheres. The spurious ENSO teleconnection
due to the drift, perturbs the atmosphere over the Northern Hemisphere. At longer
timescales, the Atlantic Meridional Variability (AMV) pattern projects onto the model
drift.
The present analysis highlights the fact that, to derive to its own climate, the model
precisely uses the internal modes of variability that we seek to predict, putting some shade
and uncertainties on traditional skill score. |
|
|
|
|
|