Typically, the SSO specification involves some measure of the variability and some measure(s) of the shape of the real orography in a model grid-box. The shape of the SSO is important because when the orography is ridge-like the drag should be proportional to that component of the flow perpendicular to the major axis of the orography.
A major weakness common to most if not all measures currently used to define the SSO shape is that they do not converge as the resolution of the orography dataset increases. The shape of the SSO therefore depends very strongly on the resolution of the orography dataset and hence the tuning of the SSO parametrization also depends strongly on the resolution of the orography dataset. This is clearly unphysical and, until recently, led to the unsatisfactory situation of there being a factor of 100 difference in the SSO parametrization coefficient used in the operational mesoscale and operational global forecast versions of the Unified Model (UM).
In this talk, a simple method is described which makes the SSO characterisation much more robust. By filtering the smallest scales from a high-resolution dataset, this method ensures the SSO characteristics converge and hence minimises the need to retune the parametrization if either the orography dataset resolution or the model resolution are changed. The impact of the new method on the SSO parametrization used in the UM operational forecast suite will be described and illustrated.