Wintertime variability in the Euro-Atlantic region in observations and in ECMWF seasonal ensemble experiments

An empirical orthogonal function (EOF) decomposition of monthly mean analys es of 500 hPa height (1949-94) is used to describe the interannual variabil ity of the large-scale flow in the Euro-Atlantic region during winter. The first four EOFs resemble low-frequency variability patterns identified in p revious studies, such as the North Atlantic Oscillation,and the eastern Atl antic and Eurasian teleconnection patterns. The second EOF and fourth EOF ( EOF4) are associated with the occurrence of El Nino-like sea surface temper ature (SST) anomalies according to an observational analysis; for EOF4, cor relations with SST anomalies in the west Pacific are also important. Indica tions of an influence of tropical SST anomalies on the blocking-like third EOF, emerging from the European Centre for Medium-Range Weather Forecasts ( ECMWF) re-analysis (ERA) data, are not confirmed by a longer SST record. A negligible correlation exists between the first EOF and El Nino indices dur ing the ERA period, although a link may be detected in earlier decades. In the second part of the paper, a 14-winter set of seasonal ensemble simul ations, performed with the ECMWF atmospheric model forced by observed SST, is validated by comparing the projection of ERA data onto the four leading Euro-Atlantic EOFs with those of ensemble experiments, and computing compos ite anomalies of analyses and model fields. The performance of the model is uneven and depends on the large-scale pattern considered. Model biases and flow-dependent errors affect the simulations of some of the EOFs: In parti cular the model has strong problems reproducing the occurrence of European blocks. The best correlation (69%) between the time series of analysis and ensemble-mean principal components is obtained for EOF4, which seems to be forced by SST anomalies in the west Pacific associated with El Nino/Souther n Oscillation (ENSO) events. The effects of this forcing are felt on the zo nal wind structure over Eurasia, and are reasonably reproduced by the ECMWF model. Conversely, the propagation into the Atlantic region of planetary w aves originating in the east Pacific is poorly simulated, thus degrading th e model performance in predicting Euro-Atlantic anomalies during some stron g ENSO events.