Drought and climate forecasting in an Australian context

Abstract

Seasonal Climate Outlooks have been issued in Australia since the late 1980s. The Commonwealth Bureau of Meteorology, together with the Queensland Department of Primary Industries, and the Queensland Department of Natural Resources and Mines (under the Queensland Centre for Climate Applications (QCCA) project) have been providing climate forecast output since the late 1980s and early 1990s, respectively. Publicly available forecasts from both organisations generally take the form of probabilities that total seasonal (three-month) rainfall will be above the relative climatological median, with the probability of below median rainfall being immediately derivable. A supplementary derived forecast output from The Commonwealth Bureau of Meteorology and QCCA issued on subscription provides the probability of seasonal rainfall falling into the three terciles (obtained by splitting the distribution into three equal categories by frequency). Tercile probabilities are obtained by means of a notional shifted distribution which is not normally issued pUblicly. This notional shifted distribution could, in principle, be broken down into finer levels of discrimination (quartiles, quintiles, deciles) in a way which is statistically consistent with the larger category probabilities, with the calculation of associated hindcast skill estimates giving some insight into the advisability of such a procedure. However, this type of output is not currently being produced or available for users. Interestingly, tests on historical data suggest rather more skill for forecasting the outer terciles (loosely, the tails of the distribution) than for the middle tercile. With this point in mind, the Bureau's current seasonal forecasting model does not show strong probability shifts towards the middle tercile in its three forecast variables. We have recently applied the 'Binomial Test' as a means of assessing the appropriateness of provision of data in the lowest quartile and decile. In this way we suggest use of the lowest tercile, quartile, or decile in climate forecast output may be useful for enhancing drought prediction.