Determining the main drivers of PM and CO concentration surfaces in the Greater Beirut area using a landuse regression approach.

Abstract

The Eastern Mediterranean city of Beirut, Lebanon, suffers from poor air quality as compared to many European and North American cities. In this study, high resolution PM₂.₅ (fine particulate matter), PM₁₀ (coarse particulate matter), CO (carbon monoxide) and PM2.5-PM₁₀ annual and seasonal pollution maps are generated for the Greater Beirut Area (GBA) using Land Use Regression models (LUR). The LURs were calibrated and validated on monthly data collected from 58 predefined monitoring locations within the GBA between March 2017 and March 2018. The annual mean concentrations of PM₂.₅, PM₁₀ and CO across the monitoring stations were 68.1 ± 15.7 µg-m³, 83.5 ± 19.5 µg-m³ and 2.48 ± 1.12 ppm respectively. The observed spatio-temporal variability in the recorded PM concentrations was found to be larger than those typically reported in European cities. The performance of the developed LUR models was good, with R² ranging from 0.59 to 0.67 for the PM₂.₅ models, 0.49 to 0.63 for the PM₁₀ models, 0.50 to 0.60 for the CO models, and 0.32 to 0.51 for the PM₂.₅-PM₁₀ ratio models. Overall, the predicted pollution surfaces were able to conserve the inter-pollution correlations that were determined from the field monitoring campaign, with the exception of the cold season. Although the model structure of the generated LUR models differed, building area, distance to main roads, and industrial area were all found to be common predictors across the majority of the annual and seasonal models. The predicted PM surfaces suggested that the entire population of the GBA was exposed to annually-averaged concentrations that exceeded the 24-hour WHO (World Health Organization) air quality standards set for PM₂.₅ and PM₁₀. Finally from a methodological point of view, the results of this study show that LURs generated using low cost purpose-designed monitoring campai