Stationary and non-stationary models of extreme ground-level ozone in Peninsular Malaysia

• Main Authors: Amin, N.A.M (Author), Hamidin, N. (Author), Radi, N.F.A (Author), Zakaria, S.A (Author)
• Format: Article
• Language: English
• Published: Horizon Research Publishing 2021
• Series: Mathematics and Statistics
• Subjects: Akaike’s Information Criterion; Generalized Extreme Value Distribution; Ground-Level Ozone; Likelihood Ratio Test; Maximum Likelihood Estimation
• Online Access: View Fulltext in Publisher; View in Scopus

 High ground-level ozone (GLO) concentrations will adversely affect human health, vegetations as well as the ecosystem. Therefore, continuous monitoring for GLO trends is a good practice to address issues related to air quality based on high concentrations of GLO. The purpose of this study is to introduce stationary and non-stationary model of extreme GLO. The method is applied to 25 selected stations in Peninsular Malaysia. The maximum daily GLO concentration data over 8 hours from year 2000 to 2016 are used. The factors of this distribution are anticipated using maximum likelihood estimation. A comparison between stationary (constant model) and non-stationary (linear and cyclic model) is performed using the likelihood ratio test (LRT). The LRT is based on the larger value of deviance statistics compared to a chi-square distribution providing the significance evidence to non-stationary model either there is linear trend or cyclic trend. The best fit model between selected models is tested by Akaike’s Information Criterion. The results show that 25 stations conform to the non-stationary model either linear or cyclic model, with 14 stations showing significant improvement over the linear model in location parameter while 11 stations follow the cyclic model. This study is important to identify the trends of ozone phenomenon for better quality risk management. © 2021 by authors, all rights reserved.