Development and application of a high resolution system for the evaluation of urban air quality

A hybrid modelling system (HMS) was developed to provide hourly concentrations at the urban local scale. The system is based on the combination of a meteorological model (WRF), a chemical and transport eulerian model (CAMx), which computes concentration levels over the regional domains, and a lagrangian dispersion model (AUSTAL2000), accounting for dispersion phenomena within the urban area due to local emission sources; a source apportionment algorithm is also included in the HMS in order to avoid the double counting of local emissions. The HMS was applied over a set of nested domains, the innermost covering a 1.6 x 1.6 km(2) area in Milan city center with, 20 m grid resolution, for NOx simulation in 2010. For this paper the innermost domain was defined as "local", excluding usual definition of urban areas. WRF model captured the overall evolution of the main meteorological features, except for some very stagnant situations, thus influencing the subsequent performance of regional scale model CAMx. Indeed, CAMx was able to reproduce the spatial and temporal evolution of NOx concentration over the regional domain, except a few episodes, when observed concentrations were higher than 100 ppb. The local scale model AUSTAL2000 provided high-resolution concentration fields that sensibly mirrored the road and traffic pattern in the urban domain. Therefore, the first important outcome of the work is that the application of the hybrid modelling system allowed a thorough and consistent description of urban air quality. This result represents a relevant starting point for future evaluation of pollution exposure within an urban context. However, the overall performance of the HMS did not provide remarkable improvements with respect to stand-alone CAMx at the two only monitoring sites in Milan city center. HMS results were characterized by a smaller average bias, that improved about 6-8 ppb corresponding to 12-13% of the observed concentration, but by a lower correlation, that worsened around 1-3% (e.g. from 0.84 to 0.81 at Senato site), due to the concentration peaks produced by AUSTAL2000 during nighttime stable conditions. Additionally, the HMS results showed that it was unable to correctly take into account some local scale features (e.g. urban canyon effects), pointing out that the emission spatialization and time modulation criteria, especially those from road traffic, need further improvement. Nevertheless, a second important outcome of the work is that some of the most relevant discrepancies between modelled and observed concentrations were not related to the horizontal resolution of the dispersion models but to larger scale meteorological features not captured by the meteorological model, especially during winter period. Finally, the estimated contribution of the local emission sources accounted on the annual average for about 25-30% of the computed concentration levels in the innermost urban domain. This confirmed that the whole Milan urban area as well as the outside background areas, accounting all sources outside the innermost domain, play a key role on air quality. The result suggests that strictly local emission policies could have a limited and indecisive effect on urban air quality, although this finding could be partially biased by model underestimation of the observed concentration. (C) 2016 Elsevier Ltd. All rights reserved.