Climate impacts of the NAO are sensitive to how the NAO is defined

We analyze the sensitivity of the effects the North Atlantic Oscillation (NAO) exerts on surface temperature and precipitation in Europe to the definition of the NAO index. Seven different NAO indices are examined: two based on station sea level pressure (SLP) data, two based on action centers, and three based on correlation/covariance structures described by principal component analysis (PCA). The analysis is based on monthly mean data; winter and summer seasons are analyzed separately. Temporal correlations between indices are weaker in summer than in winter for most pairs of indices. In particular, low correlations are found between station indices on the one hand and PCA-based indices on the other hand. The NAO effects are quantified by correlations between the indices and station data in Europe. Effects of the NAO on precipitation amount and wet day probability are very similar, while NAO effects on maximum temperature are stronger than those on minimum temperature. The sensitivity of the NAO effects on both surface temperature and precipitation to the choice of the NAO index is considerably higher in summer. Correlations differ among the NAO indices not only in their magnitude but in some regions in summer also in their sign. These effects can be explained by a northward shift of the whole NAO pattern and its action centers in summer, away from the sites on which the station indices are based, and by a decoupling of the Azores high and Icelandic low from the centers of high covariability, identified by PCA. Considerable differences in SLP anomaly patterns associated to individual NAO indices also contribute to different responses in temperature and precipitation. Finally, we formulate two recommendations to future analyses of NAO effects on surface climate: use several different NAO indices instead of a single one, and for summer do not use station indices because they do not represent the circulation variability related to the NAO.