Zur Seitenansicht


Return Periods of Extratropical Storms in Europe : A Spatial Prediction Approach
VerfasserBrunhuber, Stefan
GutachterMayr, Georg ; Hanley, John
HochschulschriftInnsbruck, Univ., Masterarb., 2017
Arbeit an der Bibliothek noch nicht eingelangt - Daten nicht geprüft
Datum der AbgabeJuni 2017
Schlagwörter (DE)Extratropische Stürme / Europa / Windböen / Extremwertstatistik / Wiederkehrperiode / räumliche Vorhersage / generalisiertes additives Modell
Schlagwörter (EN)extratropical storms / Europe / wind gusts / extreme value theory / return period / spatial prediction / generalized additive model
URNurn:nbn:at:at-ubi:1-6861 Persistent Identifier (URN)
 Das Werk ist frei verfügbar
Return Periods of Extratropical Storms in Europe [15.37 mb]
Zusammenfassung (Englisch)

The objective of this work is to produce accurate estimates of return periods and corresponding return levels of severe surface wind gusts associated with extratropical storms during winter half-year in Europe. A spatial generalized additive model (GAM) is used in combination with extreme value theory. To overcome the spatial limitations of coarse surface observations a GAM can help to regionalize wind statistics where no observational data is available. Gridded return levels of atmospheric model analysis of the mean wind speed are included as spatial background information. Additional explanatory variables like topography and distance from sea are considered. We model return periods and corresponding return levels of wind speed events over an automatically selected threshold with the generalized Pareto distribution (GPD). This is done for i.i.d. daily maxima of the 3-sec gust from surface observations of 1500 weather stations (with at least a record period of 15 years) and for i.i.d. daily maxima of the mean wind speed from ERA-Interim reanalysis data (with a record period of 37 years). Reliable and plausible predictions are obtained for ERA-Interim due to high temporal homogeneity in the data. Bootstrapping of the GPD predictions often leads to high uncertainties for the surface observational data which is far less temporally homogeneous. On average plausible estimates of return levels occur up to the 50 year return period. However, we observe unrealistically high estimates starting at the high end of the possible range of the 25 year return period obtained from bootstrapping. For that reason we use the GAM to only model the return levels of the 5 and 25 year return period on the elevation of Earths terrain in the UK, Germany, Italy, and whole Europe. Model validated is carried out with accuracy measures like RMSE, MAE, BIAS, and AIC. Additionally cross-validation is conducted. Models containing all three covariates exhibit best scores (full sample and cross-validated), except for the Italian model. For the 5 and 25 year return period highest return levels of the surface gusts occur in a sector spanning from the British Isles into the Northern parts of Western and Central Europe. This pattern becomes more pronounced with increasing return period. On the other hand, severe extratropical storms affect Northern and Southern Europe less frequently. The results show that return periods of extratropical storms in Europe have characteristic regional patterns which have been established by storm tracks and intensity.