Predictability of German Bight Storm Activity

Abstract

Extreme events like storms and storm surges are an everlasting challenge for populated coastal areas such as the German Bight. These events pose a threat to criticalinfrastructure and property, and require constant planning, adaptation, and precautionary measures, which oftentimes take multiple years to be implemented. Therefore, coastal protection and regional stakeholders in the German Bight may greatly benefit from seasonal-to-decadal predictions of coastal hazards like storm and surge activity. Historical records of German Bight storm activity show a pronounced multidecadal variability, but no significant trend. The historical evolution of storm surges at the coast is mainly characterized by changes in the mean sea level and coastal engineering measures like damming and dredging. Apart from the further projected rise in sea level, however, current climate projections suggest low confidence in the response of regional storm and storm surge activity to global warming. Hence, there appears to be potential in initialized decadal prediction systems to provide forecasts of the local storm and surge climate on a seasonal-to-decadal scale. In this thesis, I thus investigate the capabilities of a large-ensemble decadal prediction system based on the MPI-ESM-LR climate model to predict these climate extremes. In the first part, I evaluate the skill of the model for German Bight storm activity and winter mean sea-level pressure anomalies and find that the model is most skillful in predicting long averaging periods of more than five years. For shorter periods, such as the upcoming year, the model shows little to no forecast skill. Subsequently, motivated by the lack of skill for shorter forecast lead times, I draw on physical predictors of winter storm activity and an already established ensemble subselection technique to prove that seasonal predictions of German Bight storm activity can be significantly improved. I also illustrate how this skill improvement is associated with a better representation of the large-scale circulation in the model. Lastly, I build on the findings of the first part and show to what extent the skill for storm activity is utilizable for surge predictions. I introduce two approaches to derive surge statistics from storm-related parameters, since the model does not explicitly predict water levels. I demonstrate that these approaches provide a fair conversion from storms to surges, but the resulting prediction skill of the decadal hindcast system for surges is remarkably lower than for storm activity. Overall, I provide in this thesis an overview of the limits and capabilities of a large-ensemble decadal prediction system in predicting the German Bight storm and storm surge climate on timescales ranging from several months up to ten years.