Seasonal and Station Effects Modelling to Extreme Temperature Data in South Africa

Abstract

AbstractExtreme value theory has been widely applied to weather variables, and rigorous approaches have also been employed to investigate the seasonality and dependencies to extreme values of weather variables. To investigate the seasonal and station effects of daily maximum and minimum temperatures data, station and season specific effects model have been introduced in the parameters of general Pareto distribution. Then, the seasonality and station variations that are inherent in the data under consideration were assessed applying mainly the Bayesian approach. Non-informative and informative priors were used for estimation of the parameters. The seasonal and station effects parameters of the general Pareto distribution were estimated through the introduced models, allowing the sharing of information between stations and seasons. Simulation study was also carried out to investigate the precision of estimators for the GPD parameters with and without the effects, station and seasonal, to simulated data. The models employed improved precision of the station and seasonal effects parameter estimators at individual stations and in individual seasons. The study also depicted the significance of introducing seasonal and station variabilities when modelling extreme values using univariate method, which allows information to be pooled across stations and seasons. Results obtained in this study have essential scientific and practical applications. In an extreme temperature setting, designing a level without taking the station and seasonal effects into account could lead to significant under-protection. Hence, it is important to consider what is expected to be colder or warmer than usual by identifying the effects of stations and seasons in the analysis. This would benefit greatly local governments, researchers and farmers, which they can use to suggest adaptation and mitigation steps to improve resilience.