High resolution simulations using ICON-LEM to study convective storms in an urban environment around Paris

The population in cities is steadily increasing, requiring more sustainable and secure urban environments. This includes the adaptability and resilience towards natural disasters. Floods and storms can have a large negative impact on livelihoods and infrastructure. At the same time, we are increasing our ability to predict such events. Numerical weather prediction models nowadays can perform on large regions with convection resolving grid resolution. 

 

A current research question is to what extent forecasting improvements can be obtained by running locally nested large-eddy simulations in advance of convective storms. To investigate this, we use a setup with the atmospheric model ICON in the numerical weather prediction mode (ICON-NWP, 2.5 km resolution) as well as in the large eddy simulation mode (ICON-LEM). The study is embedded into the WMO-WWRP endorsed Paris Research Demonstration Project (RDP) focussing on the Olympic Games of Paris in 2024 with the objective to advance meteorological research for future weather forecasting systems in urban areas.

 

The ICON-LEM is set-up and run for Paris and its urban surroundings using a circular domain applying different horizontal resolutions (100 - 600m). We use a two-moment cloud microphysics parameterization scheme and the Smagorinsky scheme for the subgrid scale turbulence. For the evaluation on the observational side, we use the super site SIRTA. There, for instance, we can use the ground-based remote sensing profilers (e.g. microwave radiometer, Doppler lidar) as well as radiosondes and surface measurements. We ran the model for single cases with convective storms and high temperatures.

 

We analyzed the boundary layer growth in the simulations using ICON-NWP during the day, which showed good agreement between model and observations. The vertical profiles of the radiosondes showed that the height of the inversions was not accurately met and there is less humidity in the model. We plan to further extend our analysis to evaluate the precipitation and establish if and what refinements to certain parameters such as the roughness length can improve the simulations.