Coupling of LES models of different spatial resolution performing on massively parallel computers
Responsible: Gerald Steinfeld
Project type: Diplomarbeit (equivalent to master thesis)
Duration: 17/02/2002 - 16/03/2003

Large-eddy-simulations (LES) resolve the large, energy containing elements of a turbulent flow explicitly, while those elements that are smaller than a specific scale, which depends on the chosen spatial filter, have to be parameterized. As the maximum eddy size decreases with declining distance to a given physical boundary, even the dominant flow pattern are subgrid scale and therefore not resolved explicitly any more in the vicinity of the boundary.

The presently known parameterization schemes are not able to describe the turbulence near a physical boundary in a fully satisfactory way. Another possibility to consider the turbulence structure near a physical boundary is to run - in addition to the original LES-model - a nested fine grid model in those areas of the coarse-grid domain which are situated near to the physical boundaries.

Two different kinds of nested grid models are known in literature. On the one hand there exists the concept of one-way-interaction (OWI). Here, information is only transmitted from the coarse to the fine grid, as the boundary conditions for the fine grid model are derived from the coarse grid model. On the other hand, the additional consideration of fine grid data for the evaluation of the coarse-grid data leads to the concept of two-way-interaction (TWI).

Within the scope of this thesis the possibility to run OWI- and TWI-nesting was added to the parallelized LES-model PALM. It was found out that in the case of using a OWI the non-physical boundaries of the fine grid become an artificial source of production of TKE. A simulation with the same resolution in the coarse and the nested grid showed that this behaviour appears to be independent of the concrete method applied to derive boundary conditions from coarse grid data for the nested fine grid.

In a similar simulation using the TWI-scheme a different development between coarse and fine grid fields did not occur, which means that the coarse grid data always contains the necessary information to provide appropriate boundary conditions for the nested grid. Further, "real" nesting simulations, applying the TWI scheme led to an improvement compared with the corresponding OWI-scheme results (e.g. no maximum of the variances of u and v resp. TKE could be observed at the last grid point below the fine grid boundary), but the obtained results were still not totally convincing so that further investigations on these problems exceeding the scope of this thesis would be necessary before nesting can be used as a standard tool in PALM.