Large-eddy simulation across a grid refinement interface using explicit filtering and reconstruction

Turbulence in the urban atmospheric boundary layer occurs at too large a Reynolds number for direct computation, requiring various modeling techniques to reduce computational cost. Multiscale formulations, including grid nesting from the synoptic scale to the city scale and adaptive mesh refinement (AMR) within the city, concentrate grid points where they are needed. AMR is particularly useful for studying building-scale phenomena, such as urban plume dispersion, in which individual buildings must be resolved. Transferring the solution between grid sizes is problematic, however, requiring averaging or interpolation operations that generate errors at grid interfaces. Further errors are generated when using large-eddy simulation (LES) on non-uniform grids, due to dependence of computed quantities on the filter width and dependence on turbulence closure models at the grid scale. Here, explicit filtering and reconstruction of the velocity field are used to decrease grid refinement interface errors in a simulation of periodic, forced isotropic flow through a series of grid refinement interfaces. Different explicit filter types and levels of reconstruction are tested. Results from this canonical flow will inform the use of large-eddy simulation of boundary layer flow and flows on more complicated AMR grids.