Abstract
A forecast intercomparison study using the same six initial states as were used for a study by Baumhefner and Downey (1978) was carried out using the United Kingdom (UK) Meteorological Office operational 10-level model, and two 11-level general circulation models, one a 2° finite-difference model and the other a T42 spectral model. The same diagnostics were produced as by Baumhefner and Downey, including standard deviations, Hoymbder diagrams and error fields. The conclusions show that there are large errors common to all the models, similar to those in Baumhefner and Downey, despite the use of higher resolutions in this experiment. In general, the amplitudes of the transient waves are increased by the higher resolution but phase errors are, not reduced. Errors in the planetary waves also are not reduced. The overall skill scores for these models therefore are no better than those given by Baumhefner and Downey. However, synoptic assessments indicate that there is useful extra detail in these models. The spectral and finite difference general circulation models give similar results in many respects. The results are consistent with the conjecture that the planetary wave forecasts depend on the forecast area. Implications for design of forecast models are discussed.
