A numerical forecast model based on use of a hybrid vertical coordinate which is terrain-following at low levels but which tends continuously to a pressure coordinate at upper levels is introduced. Results from this model are first presented for 10-day forecasts (extending up to 10mb) of the major stratospheric circulation changes which occurred during the winter of 1979. A considerable degree of success is illustrated, not only in predicting a major wavenumber-two warming up to 10 days ahead, but also in the forecasting of a preceding development of wavenumber one and a subsequent decay of the wavenumber two perturbation. Diagnosis using Eliassen-Palm cross-sections confirms previous observational diagnoses and aids understanding of the performance of the forecast model. Computations of the meridional gradient of zonal-mean potential vorticity are also discussed.
The sensitivity of tropospheric forecasts to stratospheric resolution and the choice of vertical coordinate is also examined. These forecasts are in fact characterized by a marked insensitivity to such aspects. Thus lowering the top forecast level from 10mb to 50mb has little effect on objective forecast scores averaged over the extratropical troposphere of the Northern Hemisphere, although planetary-wave structures in high latitudes are affected to a small degree. The hybrid coordinate gives results similar to those obtained using the usual sigma coordinate, but such differences as are found tend to favour the use of the hybrid representation. Idealized analyses of the stability of the semi-implicit time scheme for hybrid coordinates are confirmed in practice.