Francisco J. Doblas-Reyes
ECMWF, Shinfield Park
RG2 9AX, Reading, UK
Valentina Pavan
Atmospheric Dynamics Group
Dept. of Physics, University of Bologna
Via Magnanelli 6/3, 40033 Casalecchio di Reno, Bologna, Italy
David B. Stephenson
Dept. of Meteorology, Univ. of Reading
RG6 6BB, Reading, UK
An evaluation of the skill of North Atlantic Oscillation (NAO) seasonal hindcasts in a multi-model framework
has been carried out. The multi-model approach consists in merging a set of four atmospheric model ensemble
hindcasts to create a super-ensemble system. Deterministic (ensemble mean) and probabilistic (categorical)
NAO hindcasts have been considered. The super-ensemble mean hindcasts present statistically significant
skill. As for the probabilistic hindcasts, evidences have been found of a consistent positive skill, with a
high agreement among different scoring measures. The scarcity of statistically significant results indicates
that the predictable NAO signal should be small. Two different definitions of the NAO index have been used. A
first set of NAO hindcasts has been obtained as the projection of model anomalies onto a predefined NAO
pattern. The second obtains the index as the first principal component of the individual model anomalies over
the Euro-Atlantic region. An overall degradation of the skill has been found with the first definition. This
might be a consequence of the projection method being unsuitable because of the systematic spatial shifts of
the simulated NAO patterns.