A0182
Title: A nonstationary GEV model for extreme wind speed with sinusoidal and exponential covariates
Authors: Jacque Bon-Isaac Aboy - Universiti Putra Malaysia (Malaysia) [presenting]
Muhammad Aslam Mohd Safari - Universiti Putra Malaysia (Malaysia)
Syafrina Abdul Halim - Universiti Putra Malaysia (Malaysia)
Abstract: The aim is to postulate a nonstationary generalized extreme value (nGEV) model that takes into account the non-stationarity of extreme wind speed data, which is focused on seasonality. This seasonality is captured using a sinusoidal function as a covariate for the location parameter. Moreover, an exponential covariate for the scale parameter is included to represent changes in extreme wind speed variability over time. Maximum likelihood parameter estimates are then estimated computationally. It is shown using Monte Carlo simulation experiments how the proposed model performs compared to the traditional benchmark models in terms of model fit through Akaike information criteria (AIC) and accuracy through root mean squared error (RMSE). For real-life application, the model is applied to fit extreme wind speed data on 33 wind stations in the Kagoshima prefecture of Japan. The importance of including nonstationary statistical characteristics specific to a chosen extreme weather data for better statistical modeling is emphasized.
