Investigations into seasonal predictability of North Atlantic winter climate
Skilful seasonal predictability of European climate would bring widespread socio economic benefits. However, little useful skill has been identified to date. This work extends prior research to show evidence for significant skill in predicting the winter North Atlantic Oscillation (NAO). The research divides into two topics. First, the work clarifies what is the best lagged predictor for the winter NAO by standardising the assessment of four previously published lagged NAO predictors. A new NAO predictor---the zonal gradient in summer northern hemisphere (NH) subpolar air temperature---is examined. This predictor outperforms other NAO predictors over assessment periods out to 100 years. This finding suggests that it is NH subpolar re gions rather than the midlatitudes or the tropics which provide the best NAO lagged predictability. A physical mechanism linking summer NH subpolar climate and the winter NAO is proposed. Summertime subpolar atmospheric perturbations lead a pattern of North Atlantic sea surface temperature which persists into autumn. This pattern could feed back onto the atmosphere to influence the sign and magnitude of the winter NAO. Second, the study explores further the mechanisms which un derpin the observed NAO predictability---in particular how summer NH snow cover links to the winter NAO---by using a coupled general circulation model (CGCM). A validation of the CGCM snow cover representation 1972-2002 is presented. The CGCM captures well the observed annual cycle and spatial distribution of NH snow cover. However, the CGCM exhibits critical deficiencies in the seasonal interaction between snow cover and the atmosphere. The CGCM exhibits a significant link between summer snow cover and the winter NAO 1972-2002. However, this link is nonstationary during the twentieth century and does not function through the mechanism seen in observations. Winter NAO predictability in a single CGCM integration cannot therefore be distinguished from internal model variability.