A scheme is presented for estimating the ability to forecast a wide range of scales of weather features. First, the predictability curves of Anthes and Baumhefner (1984) are normalized to be independent of the time scale of the atmospheric phenomenon. Under such a normalization it is then evident that both turbulent scales and synoptic scales have nearly the same degree of predictability. In particular, the useful forecast duration of weather patterns from a deterministic “perfect” Eulerian model appears to be no greater than the Lagrangian lifetime of the feature being forecast. If one extends this argument to mesoscales, then small mesoscale features that are internally forced, such as some severe storm cells, are anticipated to have usefully accurate model-produced predictabilities of less than one hour, while larger-scale features would have correspondingly greater predictabilities.
1 Manuscripts for focus on forecasting should be sent directly to Dr. Robert Burpee, editor, focus on forecasting, Hurricane Research Division, AOML, 4301 Rickenbacker Causeway, Miami, FL 33149.