Assessment of which scales of orography can be credibly resolved in a numerical model

Idealized simulations are performed in order to examine which scales of orography (relative to the grid scale) can credibly be represented in numerical models. A wide range of Row regimes is considered. These include neutral flow over small hills embedded in the boundary layer and highly stable flow over and around large mountains. In none of these cases do simulations with only two grid points per hill show any appreciable skill. Four and six points per hill give qualitatively reasonable results, but still tend to underestimate the drag relative to that obtained in the best resolved simulations. Implications for the smoothing of the resolved orography fields (and for the inputs to the sub-grid parametrizations) in large-scale models are discussed.