Parameter estimation and threshold selection uncertainty in extreme wind speed distribution – A frequentist approach with generalized Pareto distribution using automatic threshold selection

Kenéz, Ágnes and Joó, Attila László (2020) Parameter estimation and threshold selection uncertainty in extreme wind speed distribution – A frequentist approach with generalized Pareto distribution using automatic threshold selection. IDŐJÁRÁS / QUARTERLY JOURNAL OF THE HUNGARIAN METEOROLOGICAL SERVICE, 124 (3). pp. 311-330. ISSN 0324-6329

Preview

Text 346d73f37e061b012d3c37f31d411eff-124-3-1-kenez.pdf Download (3MB) | Preview

Abstract

Nowadays, the use of probabilistic modeling for the design of engineering structures is becoming more and more widespread due to the advances in computer technology. In order to have a comprehensive picture about a meteorological phenomenon, e.g., wind actions on structures, uncertainties must be taken into account. From structural engineering and practical points of view, the effect of the length of short time series available for the analysis on the final results can be interesting to define a minimum observation-length. In this way, the real condition at the site can be utilized to assess wind loading effects on the structure. This paper deals with the effect of uncertainties associated with the parameter estimation and threshold selection. A four-year record of wind speed data of Sződliget is analyzed, and these results are compared with the results of neighboring sites, Penc, and Budapest. The peak over threshold (POT) method with maximum likelihood estimation are selected to obtain the basic wind velocity. The suitable threshold is chosen using an automatic threshold selection approach. According to our results, the applied automated threshold selection method provide reliable results, and it is simple and computationally inexpensive. This method may reduce associate errors of threshold selection in the future. It was found that at least approximately 100 realizations should exceed the specified threshold to earn reliable results. It means that 1–1.5-year and 4-year records of wind speeds are necessary for statistical inference in case of weakly dependent observations and for statistically independent events, respectively.

Actions (login required)

Edit Item