Comparison of street-level atmospheric pollutant concentrations simulated with a subgrid-scale against a street-network model

A large fraction of the European population is exposed to atmospheric pollutant concentration levels above health-related thresholds, leading to excess mortality and morbidity. Densely populated urban areas are generally more concerned than rural regions due to higher emission of atmospheric pollutants especially from the road network. Atmospheric modelling is a necessary tool to assess urban scale air-quality for both research and operational purposes. It provides a spatially and temporally resolved information for several gaseous and particulate species. It is also used for forecast and scenario evaluation for policy making. 

Modelling atmospheric composition at street level is challenging because pollutant concentration within street-canyons depends largely on local emissions but also on the transport of polluted air masses from remote areas. Therefore, regional scale modelling and local applications must be combined to provide accurate simulations of the atmospheric composition at urban scale. In our study we compare two such strategies. In both cases the regional scale chemistry-transport model CHIMERE, fed by WRF meteorological fields, provides urban background concentrations. To simulate the local component of pollutant concentrations over roads we use i) the statistical sub grid-scale approach embedded in the chemistry-transport model and ii) the street-network model MUNICH. Simulation results over the city of Paris from both modelling approaches are compared to in-situ measurements of the local air-quality network for all available traffic monitors.

The major challenge of this inter comparison exercise is to find a consistent configuration setup for both models allowing a one-to-one comparison of the simulations. To do so we had to implement the same chemical and dynamical mechanisms for gases and suspended particles in both models. We also tested several vertical discretizations to obtain a consistent first-layer depth. Different turbulence parametrizations, including or not the urban canopy model (UCM) within WRF, were compared to obtain stable results for concentrations. We show these latter are particularly sensitive to the parametrization of the anthropogenic heat flux. To obtain realistic heat fluxes and satisfactory results for both modelling strategies we have to include in the simulations all three of the following aspects i) a highly resolved land-cover database (CORINE) ; ii) a three urban class distinction in the UCM and iii) the sub-grid scale urban fraction.

Results of two-month wintertime simulations for NOx, NO2, PM2.5 are discussed. The street-network approach provides better results for both gases and particles especially at high-traffic highways. The conclusion is less straightforward at low-traffic roads. Our results highlight the need to develop a consistent coupling of the street-network model MUNICH with the regional scale chemistry transport model CHIMERE to accurately simulate gases and particulate matter concentrations over the street network.