Spatial filtering framework for scale-aware turbulence modeling on the ICON unstructured grid.

Dynamic sub grid-scale turbulence closures require explicit spatial filtering to separate resolved and sub filter-scale contributions. In unstructured-grid atmospheric models such as ICON, constructing consistent filtering operators is nontrivial due to the triangular mesh and the staggered placement of prognostic variables on cells and edges. This work presents the implementation of a spatial filtering framework for the ICON nonhydrostatic dynamical core, designed as methodological infrastructure for scale-aware turbulence modeling.

A coarse-graining filter based on neighbor averaging has been developed on the ICON triangular grid. Cell-centered variables are filtered using edge-connected neighboring cells, while edge-centered variables are treated consistently using the adjacent cell-edge connectivity. The filter may be applied iteratively to achieve a prescribed effective filter width and is compatible with ICON’s block-based data layout on an unstructured mesh.

The filtering operators are integrated into the diffusion module as a diagnostic operation applied after explicit diffusion and halo synchronization, ensuring consistency across MPI subdomain boundaries. Ongoing work focuses on extending this framework toward a dynamic Smagorinsky-type closure using a test-filter formulation.