An operational lee wave forecasting code is described and tested on a number of actual cases. Although the solution procedure involves a well-known numerical algorithm, its implementation for operational purposes needs to be intelligent and flexible. The technique involves an eigenvalue problem based on the vertical structure equation for stationary, internal gravity waves. Since the orientation of the trapped lee waves is not known a priori, the code solves the eigenvalue problem for a range of wind directions centred on the low-level wind direction and makes a decision as to which direction will have the largest amplitude and most highly trapped lee wave. The amplitude of the lee wave motion depends on the intensity of orographic forcing in a spectral band centred on the resonant wavelength. For complex terrain this is difficult to quantify. Here, it is assumed that the Fourier spectrum of the orography follows a power law in wavenumber and that the wave amplitude is attributed with a ‘wave launching-height’ condition. One of the test cases considered is the notorious ‘Sheffield gale’ of 16 February 1962, and in that case the lee wave code does predict an unusually large-amplitude wave. Copyright © 1997 Royal Meteorological Society.