A new parametrization is presented for the representation in large-scale models of turbulent form drag due to subgrid orography with horizontal scales smaller than around 5000 m. The scheme is based on earlier work in which surface stress and its vertical distribution is formulated in terms of a slope parameter for sinusoidal hills. The new elements of the current parametrization are: (i) the spectrum of the orography is represented by a piecewise empirical power law, (ii) the turbulent form drag scheme integrates over the spectral orography to represent all the relevant scales, and (iii) the wind forcing level of the drag scheme depends on the horizontal scale of the orography, which implies that the smaller scales respond to lighter winds than do the larger scales. The scheme is controlled by a single parameter, namely the spectral density of the orography at scales of a few kilometres, as can be accurately obtained from a global I km data set. The empirical orographic power-law spectrum is derived by analysing fine resolution data as available for the USA. In order to be applicable in a largescale model, simplifications are applied. Single-column simulations are performed to assess the consequences of these simplifications. Comparisons of single-column simulations with output from fine-scale simulations for sinusoidal hills show that the approximations are within the uncertainty of the fine-scale models. Furthermore, it is felt that the scheme is sufficiently accurate given the uncertainty in the characterization of the orography with current global datasets.
