Assessment of the uncertainty of estimated extreme quantiles by regional frequency analysis. (15th October 2019)

• Record Type: Journal Article
• Authors: Lucas, C.; Muraleedharan, G.; Guedes Soares, C.
• Is Part Of: Ocean engineering. Volume 190(2019)
• Publication Date: 2019-10-15
• Subjects: Regional frequency analysis -- L-moments -- Growth curve -- Extreme quantiles -- Monte-Carlo simulations -- Error bounds; Ocean engineering -- Periodicals; Ocean engineering; Periodicals; 620.4162
• DOI: 10.1016/j.oceaneng.2019.106347
• Languages: English
• ISSNs: 0029-8018

Abstract

Abstract: The accuracy of the estimated regional extreme quantiles off Portugal determined by the regional frequency analysis is ascertained by constructing confidence intervals. The regional frequency analysis was performed for significant wave height data from 35 sites in a grid of 0.25° x 0.25° off Portugal, extracted from the 44 years HIPOCAS hindcast wave database. The position and site statistics are used to form homogeneous regions or clusters by the regional frequency analysis algorithm . The regions obtained by applying the regional frequency analysis to 35 points are used in this study for Monte Carlo simulations to determine the precision of the regional estimated quantiles from the actual data (i.e. the hindcast data) that have the same number of sites, sample sizes of the sites and the regional averages of the actual data. The between-site differences in population for the simulated region are determined to be less than the sample of the actual data as the samples are more dispersed than the respective population due to sampling variability. Simulated regions also integrate heterogeneity, misidentification of the frequency distribution and cross-correlations of the sites of the actual data. Highlights: The accuracy of the estimated regional extreme quantiles is ascertained by constructing confidence intervals. Regional frequency analysis algorithms are applied to significant wave height data from a 35 sites. The regions obtained by RFA are used for Monte Carlo simulations to determine the precision of the regional quantiles. Simulated regions have heterogeneity, misidentification of the frequency distribution and cross correlations of the sites. The 90% error confidence intervals formed by the simulation based procedure provides the magnitude of the errors.