Impact of initial conditions and atmospheric model resolution in predicting "Climate Hiatuses"

The results from a set of multi-year hindcasts carried out with the ECMWF coupled system at two different resolutions of the atmospheric component will be presented. The first experiment consists of a control series of ensemble hindcasts with the atmospheric model integrated at T255 with 91 levels in vertical (this is the current resolution of the ECMWF System4). The ocean resolution is the standard NEMO-ORCA1. In the sensitivity experiment the atmospheric resolution has increased to T511 (the ocean resolution is the same). By comparing the control and the sensitivity experiment we estimate the impact of the increased atmospheric resolution on forecast quality. The impact of initial conditions relative to external forcings in multi-year integrations is further assessed using specifically designed sensitivity experiments. They consist, for each atmospheric resolution, of three sets of ensemble hindcasts for three initial dates in 1988, 1994 and 2002 using either the external forcings from the 'correct' decade or swapping the forcings between the three decades. By comparing the three sets of integrations, the impact of external forcing versus initial conditions on the predictability over multi-annual time scales is estimated. In particular we estimate the sensitivity of the model to initial conditions in predicting the multi-year climate oscillations that modulate the global warming trend, also known as "Climate Hiatuses".