MODELLING TIME SERIES TROPOSPHERIC OZONE DATA AND THE PRECURSORS TO STUDY EFFECT AND RELATIONSHIP IN PETALING JAYA AND SHAH ALAM MALAYSIA

Abstract

Understanding the role of pollutant precursors is very important to provide meaningful insights for planning and designing localized air pollution control strategies. This study is conducted with the aim of describing the relationship and assessing the effect of several precursors (OX, NO, NO2, SO2, CO) towards O3 using Multiple Linear Regression (MLR) at an industrial (Petaling Jaya) and an urban (Shah Alam) location in Selangor, Peninsular Malaysia. Statistical modelling and analysis were conducted based on secondary data of time-series observations of multi-variables from the period of the year 2015-2017 obtained from the Department of Environment, Malaysia. Two models were developed, evaluated and compared involving models without (default) and with a proper statistical procedure to highlight the importance of data randomization and outlier treatment in the modelling of MLR for time series air quality data set. The results have shown that the model without using a good quality data set has resulted in a false estimated MLR model. Based on the p-value, NOx and CO are the two most significant O3 precursors in Petaling Jaya and Shah Alam indicating emission sources from vehicles contributed to O3 changes at the sites. Stronger effect of SO2 on O3 with a positive relationship in the industrial site, Petaling Jaya station compared to an urban residential area, Shah Alam whereby the relationship of SO2 with O3 is negative. The lag (AR1) variable is also shown significant. The study also concluded that the MLR model for time series observations requires good data quality (normality and non-autocorrelated) as the approach is critical to ensure a correct MLR model that satisfies the model assumptions with accurate parameter estimates. The study results contribute critical methodological knowledge to future researchers and the findings are fruitful for environmental bodies to help in managing O3 pollution at the study locations.

Keywords

• air pollution
• modelling
• multiple linear regression
• ozone
• precursors