A Dynamical Systems Approach to Studying Mid-Latitude Weather Extremes

International audienceExtreme weather occurrences carry enormous social and economic costs and routinely garner widespread scientific and media coverage. The ability to predict these events is therefore a topic of crucial importance. Here we propose a novel predictability pathway for extreme events, by building upon recent advances in dynamical systems theory. We show that simple dynamical systems metrics can be used to identify sets of large-scale atmospheric flow patterns with similar spatial structure and temporal evolution on timescales of several days to a week. In regions where these patterns favour extreme weather, they afford a particularly good predictability of the extremes. We specifically test this technique on the atmospheric circulation in the North Atlantic region, where it provides predictability of large-scale wintertime surface temperature extremes in Europe up to one week in advance. Significance Statement Extreme weather events carry major social and economic costs; improving their predictability is therefore of crucial importance. Forecasting the occurrence of a given extreme event can be more or less difficult depending on the state of the atmosphere from which the forecast is initialised. In this study we apply diagnostics from the field of dynamical systems analysis to identify the atmospheric states providing the best predictability and investigate their link to wintertime temperature extremes in Europe. We find that these states of " maximum predictability " correspond to significant changes in the frequency of very warm or cold spells, and are often followed by large-scale extreme temperature events. These findings can provide a useful complement to existing operational forecast tools