Evaluation of a modelling system for predicting the concentrations of PM2.5 in an urban area

We present a modelling system that contains a treatment of the emissions and atmospheric dispersion of fine particulate matter (PM2.5) on an urban scale, combined with a statistical model for estimating the contribution of long-range transported aerosols. The model of PM2.5 emissions includes exhaust emissions, cold starts and driving, as well as, non-exhaust emissions originated from urban vehicular traffic. The influence of primary vehicular emissions from the road and street network was evaluated using a roadside emission and dispersion model, CAR-FMI, in combination with a meteorological pre-processing model, MPP-FMI. We have computed hourly sequential time series of the PM2.5 concentrations in 2002 in a numerical grid in the Helsinki Metropolitan Area. The predicted results were compared against measured data at two locations in central Helsinki: urban roadside station of Vallila and urban background station of Kallio. The predicted daily average PM2.5 concentrations agreed well with the measured values; e.g., the index of agreement values were 0.83 and 0.86 at Vallila and Kallio, respectively, and the absolute values of fractional bias 0.13. As expected, the scatter of data points is substantially wider for the hourly concentration values; e.g., the index of agreement values were 0.69 and 0.74. We also computed the spatial concentration distributions of PM2.5. The predicted contribution from long-range transport to the street level PM2.5 varied spatially from 40% in the most trafficked areas to nearly 100% in the outskirts of the area. The emissions originated from cold starts and driving were responsible for <4% of the annual average concentrations at the roadside and urban stations. The model can potentially be used as a practical tool of assessment of urban PM2.5 contributions in various European regions.