Multivariate time series analysis for short-term forecasting of ground level ozone (O3) in Malaysia

• Main Author: Raffee, Ahmad Fauzi
• Format: Thesis
• Language: English; English; English
• Published: 2021
• Subjects: HD72-88 Economic growth, development, planning
• Online Access: http://eprints.uthm.edu.my/4002/1/24p%20AHMAD%20FAUZI%20RAFFEE.pdf; http://eprints.uthm.edu.my/4002/2/AHMAD%20FAUZI%20RAFFEE%20COPYRIGHT%20DECLARATION.pdf; http://eprints.uthm.edu.my/4002/3/AHMAD%20FAUZI%20RAFFEE%20WATERMARK.pdf; http://eprints.uthm.edu.my/4002/

The declining of air quality mostly affects the elderly, children, people with asthma, as well as a restriction on outdoor activities. Therefore, there is an importance to provide a statistical modelling to forecast the future values of surface layer ozone (O3) concentration. The objectives of this study are to obtain the best multivariate time series (MTS) model and develop an online air quality forecasting system for O3 concentration in Malaysia. The implementations of MTS model improve the recent statistical model on air quality for short-term prediction. Ten air quality monitoring stations situated at four (4) different types of location were selected in this study. The first type is industrial represent by Pasir Gudang, Perai, and Nilai, second type is urban represent by Kuala Terengganu, Kota Bharu, and Alor Setar. The third is suburban located in Banting, Kangar, and Tanjung Malim, also the only background station at Jerantut. The hourly record data from 2010 to 2017 were used to assess the characteristics and behaviour of O3 concentration. Meanwhile, the monthly record data of O3, particulate matter (PM10), nitrogen dioxide (NO2), sulphur dioxide (SO2), carbon monoxide (CO), temperature (T), wind speed (WS), and relative humidity (RH) were used to examine the best MTS models. Three methods of MTS namely vector autoregressive (VAR), vector moving average (VMA), and vector autoregressive moving average (VARMA), has been applied in this study. Based on the performance error, the most appropriate MTS model located in Pasir Gudang, Kota Bharu and Kangar is VAR(1), Kuala Terengganu and Alor Setar for VAR(2), Perai and Nilai for VAR(3), Tanjung Malim for VAR(4) and Banting for VAR(5). Only Jerantut obtained the VMA(2) as the best model. The lowest root mean square error (RMSE) and normalized absolute error is 0.0053 and <0.0001 which is for MTS model in Perai and Kuala Terengganu, respectively. Meanwhile, for mean absolute error (MAE), the lowest is in Banting and Jerantut at 0.0013. The online air quality forecasting system for O3 was successfully developed based on the best MTS models to represent each monitoring station.