Multi-scale chemistry-transport modelling of the 2022 extreme Sahara dustevent over Paris

Mineral dust aerosol, predominantly soil-based particles, is a significant and ubiquitous component in the atmosphere, influencing both air quality and regional to global radiative balance. Accurately representing its impact at an urban scale, particularly with resolutions down to a few meters, remains challenging. To enhance our understanding of the global dust cycle, including transport, deposition, and the life cycle of dust aerosol, we adopt a multi-scale modelling approach. This study focuses on simulating the movement of Sahara dust across different scales: from the regional (Europe) to the urban (Paris).

For the regional scale analysis, we utilize the WRF-CHEM model, offering detailed insights into the movement of Sahara dust towards Europe. This model diverges from the conventional practice of using climatology data from inventories. Instead, we initialize and set boundary conditions for gas species and aerosols using CAMS reanalysis data, while ERA-5 reanalysis data provide the meteorological input. WRF-CHEM, an atmospheric chemistry model, incorporates the physics, chemistry, and morphology of dust aerosol, achieving a high resolution of 1km around Paris through a four-tiered grid nesting system (27 km², 9 km², 3 km², and 1 km²). The WRF model is refined hourly with ERA-5 data and every three hours with CAMS data at the boundary of the largest domain. Although the nested grids derive their initial conditions from their parent domain, they do not undergo further nudging.

On the urban scale, we employ code_saturne, a general-purpose CFD open-source solver developed by EDF R&D, to assess air flow and pollutant dispersion around buildings in central Paris. Outputs from the regional scale WRF-CHEM model serve as the initial and boundary conditions for these local scale simulations. A notable challenge in urban modelling is accessing complete and open building geometry data. OpenStreetMaps stands out as a comprehensive source for such geometrical data. To incorporate geometry from various sources (e.g., LiDAR measurements or shape files from the Institut Géographique National of France), we convert these into a point cloud format. This data then feeds into an internal mesh generator, based on the API of code_saturne, allowing us to create meshes with a spatial resolution of 2 meters near buildings and approximately 60 meters at the calculation domain's border. Comparisons with ground-based measurements show a qualitative alignment.