AIRCITY: A Very High Resolution Atmospheric Dispersion Modeling System for Paris

The AIRCITY project, partly funded by the European Union, is now successfully achieved. It aimed at developing a 4D innovative numerical simulation tool dedicated to the dispersion of traffic-induced air pollution at local scale on the whole urban area of PARIS. AIRCITY modeling system is based on PMSS (Parallel-Micro-SWIFT-SPRAY) software, which has been developed by ARIA Technologies in close collaboration with CEA and MOKILI. PMSS is a simplified CFD solution which is an alternative to micro-scale simulations usually carried out with full-CFD. Yet, AIRCITY challenge was to model the flow and pollutant dispersion with a 3 m resolution over the whole city of Paris covering a 14 km × 11,5 km domain. Thus, the choice was to run a mass-consistent diagnostic flow model (SWIFT) associated with a Lagrangian Particle Dispersion Model (SPRAY) on a massively parallel architecture. With a 3 m resolution on this huge domain, parallelization was applied to the computation of both the flow (by domain splitting) and the Lagrangian dispersion (management of particles is split over several processors). This MPI parallelization is more complex but gives a large flexibility to optimize the number of CPU, the available RAM and the CPU time. So, it makes possible to process arbitrarily large domains (only limited by the memory of the available nodes). As CEA operates the largest computing center in Europe, with parallel machines ranging from a few hundred to several thousand cores, the modeling system was tested on huge parallel clusters. More usual and affordable computers with a few tens of cores were also utilized during the project by ARIA Technologies and by AIRPARIF, the Regional Air Quality Management Board of Paris region, whose role was also to build the end-users requirements. These computations were performed on a simulation domain restricted to the hypercenter of Paris with dimensions around 2 km × 2 km (at the same resolution of 3 m).The focus was on the improvements needed to adapt simulation codes initially designed for emergency response to urban air quality applications:• Coupling with the MM5 / CHIMERE operational photochemical model at AIRPARIF (as the forecast background),• Turbulence generated by traffic / coupling with traffic model,• Inclusion of chemical reactions / Interaction with background substances (especially NO / NO2).Finally, in-depth validation of the modeling system was undertaken using both the routine air quality measurements in Paris (at four stations influenced by the road traffic) and a field experiment specially arranged for the project, with LIDARs provided by LEOSPHERE Inc. Comparison of PMSS and measurements gave excellent results concerning NO / NO2 and PM10 hourly concentrations for a monthly period of time while the LIDAR campaign results were also promising. In the paper, more details are given regarding the modeling system principles and developments and its validation.Perspectives of the project will also be discussed as AIRCITY system. The TRL must now be elevated from a demonstration to a robust and systematically validated modeling tool that could be used to predict routinely the air quality in Paris and in other large cities around the world.Copyright © 2014 by ASME