Large eddy simulation of pollutant dispersion in under-resolved urban environments

The ability of large eddy simulation (LES) to capture transient turbulent behaviour can be vital for accurate pollutant dispersion simulation in a dense urban environment, however a significant drawback of LES is its computational expense. Typical guidelines for LES urban flows recommend a resolution of at least 10 cells per significant feature, approximately 1 metre for a typical dense city environment, requiring millions or billions of cells for realistic applications. A method fully explicit in time and space results in a theoretical factor of sixteen speed-up for every factor of two reduction in resolution, hence there is a driving desire for a method that can produce consistent results at a reduced resolution.

An incompressible LES code using the Diffuse Obstacle Boundary (DOB) Immersed Boundary Method (IBM) is integrated into the AMReX framework, providing efficient implementation of adaptive mesh refinement (AMR) capability as well as parallelisation on hybrid CPU/GPU hardware. Applied to urban dispersion scenarios, the combination of the DOB IBM and AMR enables resolution to be concentrated around the source, whilst still accounting for flow channelling due to under-resolved buildings in the far field. The code is tested against the Mock Urban Setting Test (MUST) and Michelstadt wind tunnel experimental data, showing that the method can produce consistent dispersion results even at greatly reduced resolutions.

UK Ministry of Defence © Crown owned copyright 2025/AWE.