PhD thesis,
Seasonal predictability of wintertime windstorm climate over the North Atlantic and Europe
Freie Universität Berlin, (May 12, 2011)
Abstract
Strong extra-tropical cyclones are a characteristic feature of mid-latitudinal weather. In extreme cases, such cyclones are accompanied by severe wind speeds. These windstorms are the natural hazard with the highest loss potential in the North Atlantic and European region. Thus, prediction of windstorm climate on seasonal time scales would be beneficial for society and economy. However, the skill of seasonal predictions of wintertime
windstorm occurrence and its sources have not been extensively analyzed. Therefore, this thesis addresses the following issues: a) How can windstorms be objectively identified? b) Which factors are related to wintertime windstorm occurrence on seasonal time scales in observational data, and may therefore be considered as potential sources of seasonal predictability? c) Do state-of-the-art dynamical seasonal prediction models successfully forecast windstorm climate? d) Which factors influence the occurrence of windstorm in such prediction models, i.e., what are the sources of predictive skill?
To this end, an impact-based, objective windstorm identification scheme is developed based on surface wind speeds exceeding the local 98th percentile in consecutive timesteps. The scheme can be applied to both reanalysis products and climate (prediction) model data. It is shown to successfully identify individual historical events. Furthermore, both the spatial and temporal climatologies of the identified events are in good agreement with
observations.
Windstorms identified in reanalysis data are used to analyze their relationship to observed anomalies of hemispheric-scale factors such as North Atlantic sea surface temperature and sea ice, continental snow cover extent, and the North Atlantic Oscillation. The correlation between anomalies of these factors in summer and autumn and the windstorm occurrence in subsequent winter (i.e., with a lead time of four to six months) is statistically significant with coefficients of up to 0.43. Hence, hemispheric-scale factors explain at least 20\% of the inter-annual variability of windstorm climate. A composite study reveals that an anomalous state of the North Atlantic Oscillation in summer supports the generation of specific temperature anomalies in the North Atlantic ocean in autumn, namely the North Atlantic Horseshoe pattern. These anomalies persist until winter and induce anomalous growth conditions for extra-tropical cyclones over the North Atlantic (e.g., increased baroclinicity in areas of increased meridional temperature gradients, or increased latent heat over areas of positive sea surface temperature anomalies). Such conditions are favorable for the development of strong cyclones, and therefore potentially a higher number or stronger wintertime windstorms in the North Atlantic and European region. Such physically motivated links between anomalies in the North Atlantic Ocean, their persistence and their impacts on growth conditions of cyclones are potential sources of windstorm predictability on the seasonal time scale.
The windstorm identification scheme is also applied to data of the DEMETER and ENSEMBLES projects. These data consist of seasonal hindcasts produced by several coupled ocean-atmosphere climate models. December–February windstorm frequency in the 1980–2001 period is statistically significantly predicted by the multi-model ensembles and several single-model ensembles with a lead time of one to three months, with skill ranging between 0.10–0.40. The 1980–2001 period is shown to be generally better predictable than the 1960–1980 period. Additionally, the winters with high windstorm frequency are better predicted than winters with average storm counts.
The sources of predictive skill in these model ensembles are investigated by quantifying the relation between hemispheric-scale factors, growth factors of cyclones, and windstorm occurrence in the prediction models. The relation between hemispheric-scale factors and wintertime windstorms is weaker than in observations and insignificant on the seasonal time scale. Still, composites of hindcast runs with strong North Atlantic Horseshoe anomalies reveal similar relations to growth factors as found in observations. However, the persistence of the oceanic anomalies is very different in the individual models, and generally weaker than in observations, probably related to deficiencies in the coupling between ocean and atmosphere in the prediction models. The preservation of oceanic anomalies in the models' ocean control run from summer until the onset of the hindcasts started in November is shown to be a crucial factor for predictive skill. Models able to retain such anomalies in closer agreement with observations develop stronger anomalies in sea surface temperature, growth factors of cyclones, and windstorms, and show higher skill scores. Therefore, it is argued that the North Atlantic is one source of predictive skill for seasonal windstorm predictions, and that improvements of the ocean-atmosphere coupling in prediction models could enhance skill.
