Recent performance of a method fof forecasting Australian seasonal tropical cyclone activity

The relationship between Australian region seasonal tropical cyclone numbers and prior values of the Southern Oscillation Index (SOI) is examined, with particular emphasis on the period after 1979. The first study indicating that the Southern Oscillation could be used to predict Australian tropical cyclone numbers was published in 1979, so later data provide an independent test of the relationship. Interannual variations of cyclone numbers have continued to be related with prior values of the SOI throughout the post-1979 period, and showed skill relative to persistence or climatological 'predictions'. However, the relationship has been confounded by secular changes in the numbers of cyclones, including a drop in cyclone numbers after 1986 that was not associated with a corresponding drop in the SOI. This would have led to substantial over-prediction of seasonal cyclone numbers in the seasons 1986/87-1990/91, although the changes in cyclone numbers from year to year would still have been predicted quite accurately. Use of the SOI to predict the change in cyclone numbers from last season to the coming season, rather than predicting the expected numbers directly from the SOI, could reduce the confounding effect of possible secular changes in cyclone numbers, the SOI, or of relationships between them.