Multivariate analyses of the simulated atmospheric response to prescribed sen surface temperatures (1961-1994)

The ECHAM-4 atmospheric circulation model has been forced with the observed sea surface temperatures (SST) and sea-ice coverage through the period 196 1-1994 (with three experiments starting from different initial conditions). Multivariate analyses, based on singular value decomposition, allow the la rge-scale pattern to be identified which maximizes the skill relative to th e observations and the reproducibility between the experiments, that is the proximity between the runs, forced by the same boundary conditions but sta rting from different initial atmospheric conditions. This leads to the forc ed component of the atmospheric variability, mainly by SST, which is theore tically predictable from it at seasonal time scales. The spatio-temporal va riability of the first observed mode of sea level pressure (December-March) is well simulated for the Tropics but also for North America. The forcing in both cases comes primarily from the 'Global Tropical' SST mode. The situ ation in Europe and in the northeast Atlantic is less clear. Although the s patial behaviour of the observed pattern is well simulated, the reproducibi lity is weak. ((C) Academie des sciences / Elsevier, Paris.).