Modelling the non-linear association of particulate matter (PM10) with meteorological parameters and other air pollutants—a case study in Makkah

Abstract

Particulate matter with aerodynamic diameter of 10 μm or less (PM10) causes numerous adverse health and environmental impacts; therefore, it is vital to characterise its behaviour in association with the controlling factors. In this paper, the effects of several meteorological parameters and gaseous pollutants on PM10 concentrations (μg/m3) are analysed employing a quantile regression model (QRM). The study uses air quality and meteorology data collected in the arid climatic conditions of Makkah, Saudi Arabia. In this study, it is shown that the effects of covariates vary at different levels of PM10 distributions, which confirms a non-linear association between PM10 and independent variables. Temperature had positive significant effect at the middle quantiles (0.2 to 0.8) of PM10. The effect of atmospheric pressure was significant only at quantile 0.95 (slope = −1.85). Relative humidity had significant effect at quantiles 0.05 to 0.3 and insignificant effect at higher quantiles. Both wind speed and lag_PM10 demonstrated significant positive effect at all quantiles, and the magnitude of slopes gradually increased as PM10 concentration increased. The effect of CO was significant at all quantiles, and the magnitude of slopes ranged from −8 to −47 at quantile 0.05 and 0.95, respectively. The negative effect of SO2 was significant at most of the quantiles, except at quantiles 0.05, 0.8 and 0.9, where the effect was insignificant. NO showed significant positive effect at all quantiles; in contrast, NO2 had positive effect only at quantiles 0.05 to 0.6. The performance of the model was assessed both locally (at each quantile) and globally (amalgamating the effect of all quantiles). QRM provides a new insight into air quality data analysis and outperforms other regression models.