The problems involved in formulating and evaluating categorical rainfall predictions are discussed. Equitable skill scores, based on predicted/observed contingency tables, provide a relatively simple means of assessing performance and effectively distinguish between true and artificial skill. These have been used to assess the . seasonal rainfall predictions issued by two agencies over recent years - the National Climate Centre (NCC), 1989–1992, and Austweather Pty Ltd, 1984–1992. Although the NCC predictions performed well in 1991, the predictions have not yet demonstrated any overall significant skill. The Austweather predictions, although confined to Western Australia, exhibit significant overall skill. This is particularly noteworthy since these predictions are issued with lead times up to eight months.
Ian Smith (Thu,) studied this question.