The Application Of Generalized Pareto Distribution In Waves Extreme Parameter Estimation

Wave is one of the most important parameter in the ocean, it is necessary to dowave research for coast protection, preventing soil erosion, shipping and offshoreoil-gas facilities protection. In the engineering design process, extreme value of varioushydrometeorological parameters are often needed. As an important hydrologicalparameters, wave parameters of extreme value is necessary for some ocean engineering.Wave observation data in the ocean are limited, they often cannot meet the needsof ocean engineering. With the development of the wave numerical model, it isbecoming more and more important to get wave parameters of extreme value throughwave numerical simulation method. Based on the SWAN, the third generation wavenumerical model, this paper using using CCMP wind field data (from January1988toDecember2010) as input data, to simulate the wave field of China sea. Comparing withthe buoy data of No.22001and No.21004in east China sea, the sumulation results aregood enough to be used as samples to calculate wave parameters of extreme value.Generalized extreme value distribution(GEV) is widely used to calculate waveparameters of extreme value. This paper introduces the generalized pareto distribution(GDP), using partial duration series method to obtain sample sequence. Comparedwith GEV model, based on the partial duration series method, GDP model can fully usethe raw data. GDP model has higher fitting accuracy and stability, including in the highvalue area, its more practical and more persuasive. The rectangular region nearNo.22001buoy are discussed, including scale parameters and the return period ofextreme value distribution, both are crucial regional marine environmental conditions.With the progress of science-technology and the further development of the designconcept, in order to optimize the ocean engineering design scheme, directional waveparameters of extreme value is often taken into account. Using GEV model to calculatedirectional return period of extreme wave value, this is a problem, the return period ofextreme value in the strong wave direction is greater than the omnidirectional return period of extreme value. Analyzing the causes, and is likely to be GEV model sampleobtain methods. Using GPD model to calculate directional return period of extremewave value, this problem did not appear again. After limiting threshold, it is possible ina given year wave of a certain direction has no sample lager than threshold, and ithighlights the annual maximum statistical method’s stability in obtaining sample. Interms of the results in this paper, using the GDP model to calculate the return period ofextreme value is slightly greater than the GEV model calculate the return period ofextreme value.