Such products include forecasts of the principal meteorological elements out to around 10 days though with increasing forecast time the products become, of course, less reliable. However, such forecast data fields (eg Figure 17) are of high quality to around 4 to 5 days and are used by National Meteorological Services in all countries for the provision of services to the public. A review of the current Australian global assimilation and prediction system (GASP) is given by Bourke et al. (1995) and Seaman et al. (1995).

Figure 17 Numerical MSL forecasts for the Southern Hemisphere for 24, 48, 72 and 96 hours based on the numerical MSL analysis for 1200 UTC on 28 Jan 1997 shown in Figure 8 (source National Meteorological Centre Melbourne).

Each of the modelling centres has its own suite of models and comparisons of the results and "skill scores" for particular periods are made leading to improvements to individual model characteristics. In general, however, the finer the grid scale or the higher the wave number which a particular model is able to achieve within the computing capacity available to it and the more realistically it is able to parameterise the effects of sub grid scale phenomena such as convection, surface friction, cloud characteristics etc the better will be its predictions.

The progressive improvements in the quality of forecasts with the introduction of more refined models having improved physical parameterisations may be shown in Figure 18 for global forecasts made by the Australian and principal overseas analysis centres. Here the skill score sequence over some 15 years is depicted. The so called S1 skill score is derived from comparison of forecast against observed pressure gradients for specific lead times. The lower the score the better the forecast and, as indicated, there has been a significant improvement during the period of record. A recent assessment has been made by Adams (1997) of numerical weather prediction skills over the specific region of the Southern Ocean and coastal East Antarctica extending from 50°S to 70°S and from 80°E to 160°E for the summers of 1993–94 and 1994–95. This indicated that the accuracy of the two models considered was less over this region than for the same models over the Australian region and also that Antarctic forecasters were able to produce minor improvements over the numerical guidance.

Figure 18 Time variation in global forecast skill as measured by the S1 skill score (This score essentially provides a measure of the degree of error in the depiction of a forecast pressure field; the lower the score the better the forecast). The upper panel indicates the performance of the Australian operational forecast model against an independent analysis (METANAL); recent models produce their own analysis and can be compared against these (SELF).
The lower panel compares the Australian (GASP) model against those of the equivalent analysis centres, the National Meteorological Centre Washington (US), the British Meteorological Office, Bracknell (UK) and the European Centre for Medium Range Weather Forecasts, Reading UK (ECMWF) (source—Bureau of Meteorology National Meteorological Centre).

Bureau of Meteorology