Developing land use regression models to predict nitrogen oxides and ozone concentrations across an urbanizing gradient : the case of Lebanon.

Abstract

Most developing countries suffer from elevated ambient traffic-related air pollution. In the Greater Beirut Area (GBA), the absence of a functioning fixed-station air quality monitoring network with an adequate spatial coverage has curtailed the assessment of personal exposure to air pollutants. The development of Land Use Regression (LUR) models that can predict the intra-urban variability in ambient pollution surfaces as a function of traffic, meteorological and GIS-based explanatory variables have proven effective and efficient. The models have been successfully used and applied across cities in North America, Europe and Asia. In this study, nitrogen oxides (NOx), nitrogen dioxides (NO2) and ozone (O3) concentrations were monitored across the GBA over a year using passive air quality samplers. The annual average concentrations of NOx and NO2 in the study area were 89.7 and 36.0 ppb respectively. These concentrations are higher than levels reported across most European and many Asian cities. On the other hand, O3 concentrations were largely low (GBA wide mean was 26.9 ppb), particularly in the dense and congested urban areas of the GBA. Based on these measurements, annual and seasonal LUR models were developed for the study area. Traffic related predictors were found to have a strong predictive role across all LUR models. Moreover, the role that local point sources had on the ambient levels was also evident in the final model structures. Overall, the performance generated models was good with low biases, a high model robustness, and acceptable R2 that ranged between 0.66 and 0.73 for NO2, 0.56 and 0.60 for NOx, and 0.54 and 0.65 for O3. The developed LUR models were then used to develop the first ambient pollution concentration maps for the study area for NO2, NOx and O3.