Identificación y separación de eventos medioambientales en el análisis de valores extremos para la proyección de periodos de retorno

Abstract

The present study aims at identifying and separating stochastic events according to their origin and nature in order to more rigorously fulfill the requirements of extreme value theory (EVA) and thus to obtain more robust statistical parameters. For this, wave events from the Eastern Equatorial Pacific (3 ° N, 278 ° W) were used with ERA-INTERIM database models that cover the period from 1979 to 2015. The spectral partition technique was applied to obtain independent time series. In the study area four wave systems were identified: (1) the long-distance waves of the Southern Hemisphere Storm Belt (WS1), (2) the waves from the North Pacific Ocean (WS2), (3) one more system generated from the southwest Pacific (WS3) and (4) waves produced locally due to the wind that crosses the isthmus of Panama (WS4). Subsequently to this methods for estimation of extreme values were applied: (1) initial distribution, (2) peak over threshold and (3) annual maximums, obtaining periods of return. When performing the statistical analysis of extreme values for the integrated series and individually for each of the partitioned series it is observed in this particular case that the SW1 series predominates in the extreme values of the integrated series, with the other series having less relevance. The application of EVA to independent series is more consistent, projects robust results and allows a deeper physical vision in the stochastic analysis.