Assessment of ground-level ozone pollution with monitoring and modelling approaches in Makkah, Saudi Arabia

Abstract

Long-term exposure to the high levels of ground-level ozone (O3) adversely affects human health, vegetation, and building materials. Atmospheric concentrations of O3 are dependent on the levels of its precursors and meteorological parameters. In this paper, the data of several air pollutants as nitrogen dioxide (NO2), nitric oxide (NO), carbon monoxide (CO), and sulfur dioxide (SO2) and meteorological parameters such as temperature, relative humidity, wind speed and direction, and atmospheric pressure are used to model O3 at five (5) air quality monitoring stations in the Holy City of Makkah, the Kingdom of Saudi Arabia, for the year 2019. The performance of the multiple linear regression model (MLRM) and the generalized additive model (GAM) was compared for modelling the concentrations of O3. It was found that NO, NO2, and relative humidity had a negative association, whereas other covariates had a positive association with O3. To quantify the association between modelled and observed O3 concentrations, several metrics including correlation coefficients and root mean square error (RMSE) were calculated. GAM outperformed MLRM at all five sites and was found to successfully capture nonlinearity in the association between O3 and predictor variables. Correlation coefficients for the cross-validated model using hold-out testing dataset for GAM were 0.82, 0.78, 0.86, 0.79, and 0.76 and for MLRM were 0.69, 0.61, 0.79, 0.64, and 0.63 for Haram, Aziziah, Umrah, Atebiah, and Shawqiah sites, respectively. To conclude, GAM being an advanced nonlinear model performed better than MLRM and therefore is preferred for modelling O3 in Makkah and other similar arid regions.