Status: this preprint was under review for the journal OS. A revision for further review has not been submitted.
Evaluation of Peaks-Over-Threshold Method
Soheil Saeed Farand Ahmad Khairi Abd. Wahab
Abstract. Two extreme wave analysis models, namely Peaks-Over-Threshold (POT) and Generalized Pareto Distribution (GPD), were developed in order to improve the POT model and highlight merits and limitations of the two models. Studies have shown that the POT model was not equipped with a suitable approach to determine a true threshold value. This paper proposed an approach to specify the most suitable threshold value for the POT model, which is called Hybrid method. In addition, until now the MIR (minimum ratio of residual correlation coefficient) criterion has been used as a goodness-of-fit method in the POT model. However, the examinations on the method represented that MIR is not always a stable approach in determining true distribution function. This paper proposed an alternative approach instead of the MIR criterion method, it is called Norm of Residuals, and its credibility was examined by the Chi-Square test. The results drawn from this study also demonstrated that the Hybrid method completely matched with the POT model, and the threshold obtained by this method is credible, moreover, the Norm of Residuals method is completely stable in determining the best fitting distribution for the POT model.
Received: 16 Jun 2016 – Discussion started: 07 Jul 2016
Publisher's note: Copernicus Publications remains neutral with regard to jurisdictional claims made in the text, published maps, institutional affiliations, or any other geographical representation in this paper. While Copernicus Publications makes every effort to include appropriate place names, the final responsibility lies with the authors. Views expressed in the text are those of the authors and do not necessarily reflect the views of the publisher.
In this paper, two commonly used extreme value analysis models have been developed (POT and GPD models). Both models use threshold values to censor a range of data. The results of this study show the two models are very sensitive to any changes in threshold value. Moreover, the POT model has shown some imperfections in determining true threshold value and a best fitting distribution function. Two methods were proposed by this paper to deal with the limitations in order to improve the model.
In this paper, two commonly used extreme value analysis models have been developed (POT and GPD...
- Printer-friendly version
- Supplement