Unit:
Department of PhysicsLibrary of the School of Science
Dissertation committee:
Γεώργιος Κάλλος, Ομότιμος Καθηγητής, Τμήμα Φυσικής, ΕΚΠΑ
Σαράντης Σοφιανός, Επίκουρος Καθηγητής,Τμήμα Φυσικής, ΕΚΠΑ
Γεώργιος Γαλάνης, Καθηγητής, Σχολή Ναυτικών Δοκίμων
Έλενα Φλόκα, Καθηγήτρια, Τμήμα Φυσικής, ΕΚΠΑ
Εμμανουήλ Αναγνώστου, Καθηγητής, University of Coinnecticut, USA
Παναγιώτης Νάστος, Καθηγητής, Τμήμα Γεωλογίας & Γεωπεριβάλλοντος, ΕΚΠΑ
Ιωάννης Πυθαρούλης, Αναπληρωτής Καθηγητής, Τμήμα Γεωλογίας, ΑΠΘ
Original Title:
Extreme value analysis of environmental parameters and quantification of the associated uncertainty
Translated title:
Extreme value analysis of environmental parameters and quantification of the associated uncertainty
Summary:
Extreme weather is a term generally used for describing weather patterns with a low frequency of occurrence, from the strong cyclonic presence to extreme sea state or extensive heat waves. Such events can be associated with severe social and economic consequences. Therefore, their analysis and in-depth understanding is more than critical not only for the pure scientific interest but also for its impact in the society and economy. In this context, the potential risk from extreme weather can be expressed through the concept of return periods. These are based on Extreme Value Theory and they are practically a statistical estimate for the recurrence of extreme phenomena based on limited data. The scope of this dissertation, therefore, is to quantify risk associated to atmospheric and wave parameters in terms of return periods employing different approaches, extreme value methodologies and tools.
The overall analysis is performed through three proposed approaches, focusing on grid-points (single locations), employing a characteristic value of an entire region and studying the phenomena themselves. Beginning with the first approach, wind speed probability distribution is examined focusing on both its higher and lower values alongside the duration of the event. The main purpose is to determine the probability of occurrence of extreme events by combining their intensity and duration, adopting the concept of return periods, and to quantify the associated uncertainty. Among the highlights in the study of low wind events was that the Maximum Likelihood method for the parameter-fitting was found to be suitable in the Intensity Given Duration approach. In the Duration Given Intensity approach the Rayleigh distribution outperformed other theoretical distributions in the application of AM methodology. Regarding the wind speed probability distribution upper tail, the intensity, duration and frequency of the events were found to be highly affected by the topography.
There are several sectors that apart from the traditional approaches that are applied in single locations they also require additional information on a regional basis. Therefore, in the present work, an effort towards the characterization of wide areas according to their extremes is made. To achieve this, several regional-scale summary measures are proposed. These summarize the performance of the region into a singular value and can help identify selected cases and support an overall risk assessment from particular scenarios. It was found that the spatially maximum values or high spatial quantiles should not be selected as extreme indices for large areas as their performance deviates. Also, in general, the events are characterized by lower return periods in terms of significant wave height. Through this process a transition is made from the grid-oriented approach to one characterizing a region. However, an additional finding was that the extremity of a variable under study may have different impact depending on the weather pattern associated to.
This issue was addressed emphasizing also in weather phenomena with distinct characteristics through an object-oriented approach. Thus, the probability of occurrence of an event is estimated based on different environmental parameters, an element necessary in the modeling of damages. For the application of the proposed methodology, the Mediterranean cyclones with tropical characteristics were selected as a phenomenon to be investigated. The aim was to identify the areas at risk and estimate the extremity of such cyclones. The most affected regions were found to be mainly in the central and the western Mediterranean both regarding extreme winds and waves. In the estimation of their return periods, a similar behavior among the methods applied was met.
The employment of different meteorological parameters and methodologies to estimate the above can be valuable from a climatic point of view and help towards the implementation of more targeted measures to deal with potential damages. This can be of great assistance to many sectors and in particular to decision makers and stakeholders.
Main subject category:
Science
Keywords:
Return periods, Extreme value theory, Impact of extreme weather events, Atmospheric modeling systems
Number of references:
212
