Enhancing the accuracy of tropospheric ozone prediction using probability distribution

Abstract

Tropospheric ozone or ground-level ozone, mainly found near ground level, has adverse effects on human health. Distribution fitting is useful for predicting the probability, or forecasting the frequency of recurrence, of a phenomenon in a specific period of time. This study aimed to find the best fit distribution of ground-level ozone for specific industrial, rural, and suburban areas of monitoring locations in Malaysia, which were Kuala Terengganu, Jerantut, and Banting. Secondary data from 2017 to 2020 used in this study were obtained from the Department of Environment Malaysia (DoE). This study employed eight probability distributions namely Weibull, gamma, lognormal, logistic, log-logistic, Birnbaum–Saunders, Nakagami, and inverse Gaussian. The method of moments was used to estimate the parameters for each distribution and the best distribution can be used for predicting the return period of the concentration. The descriptive statistics analysis showed that ground-level ozone reached the highest peak at 1400 and 1500 hours, due to the UV radiation from sunlight, while the lowest concentration reading was at 0700 hours at all monitoring locations. By comparing the analysis of the eight distributions, Nakagami was found to be the best fit distribution to the actual monitoring data for Kuala Terengganu, Jerantut, and Banting stations from 2017 to 2020. As a result, this study suggests that the Nakagami distribution be used to predict exceedances and return periods, based on the performance indicators. Thus, it can take the place of the typical distributions employed in fitting the distribution of air pollutants, such as the lognormal distribution and the gamma distribution.

Keywords

• air pollution modelling
• tropospheric ozone
• return period
• statistical distribution