 |
| Titel |
Spurious Rayleigh-Bénard effects in under-resolved simulation of atmospheric convection |
| VerfasserIn |
Z. P. Piotrowski, P. K. Smolarkiewicz, S. P. Malinowski, A. A. Wyszogrodzki |
| Konferenz |
EGU General Assembly 2009
|
| Medientyp |
Artikel
|
| Sprache |
Englisch
|
| Digitales Dokument |
PDF |
| Erschienen |
In: GRA - Volume 11 (2009) |
| Datensatznummer |
250029472
|
|
|
|
|
|
| Zusammenfassung |
| Intrigued by the regularity of convective structures observed in simulations of mesoscale flow
past realistic topography, we take a deeper look into computational aspects of a
classical problem of the flow over a heated plane. We found that the numerical
solutions are sensitive to viscosity, either incorporated a priori or effectively realized in
computational models. In particular, anisotropic viscosity can lead to regular convective
structures that mimic naturally occurring Rayleigh-Bénard (RB) cells but that are
spurious for the problem at hand. We have extended the classical linear theory to
anisotropic viscosity at moderately supercritical Rayleigh numbers, realized effectively in
under-resolved convection simulations. It follows that anisotropic viscosity modifies the
range of unstable RB modes, such that for an effective viscosity much larger in the
horizontal than in the vertical unphysically broad RB cells may be observed. The latter
is relevant to “cloud resolving” global models with relatively fine (for numerical
weather prediction) horizontal resolution δx ~O(103) m. At such a resolution the
simulated convection is still under-resolved and strongly influenced by numerical
filtering.
To better assess the impact of an effective model viscosity on the structure of convective
fields, we have conducted a large series of simulations of thermal convection, with various
degrees of idealization, using the computational model EULAG. We performed an extensive
convergence study and documented differences between the well resolved (viz. realistic)
cellular convection and spurious structures. Comparing various means of enhancing the
effective viscosity in the horizontal, we demonstrated that details of filtering are
inessential. The common denominator of the scale selection is consistent with the linear
theory.
On the practical side we found that some numerical approaches may be preferable when
the resolution is inadequate to capture the realism of convective fields. While control of
effective viscosity is certainly the key to the quality results, resorting to non-dissipative
numerics is not a cure. We found that implicit large-eddy simulation (ILES) approach
based on non-oscillatory forward-in-time numerics minimizes numerical viscosity
and its anisotropy, and produces results superior compared to more standard LES
models. |
|
|
|
|
|
|