NSW Department of Primary Industries | CSU Homepage
30 May 2006
Analysis by Graham Centre researcher and colleagues questions usefulness of seasonal rainfall forecasts
A research paper published last year in the peer-reviewed journal Meteorological Applications by Graham Centre researcher David Buckley and his colleagues concluded that the seasonal rainfall forecasts issued by the Australian Bureau of Meteorology have been of limited benefit to industries reliant on rainfall predictions for their decision making.
Since June 1997 the Australian Bureau of Meteorology has issued three-month seasonal rainfall forecasts in two formats: (1) a map of Australia depicting the probability of exceeding median rainfall or the probability of rainfall in each of the three terciles (< 33 %, 33 - 67 % or > 67 % of the rainfall observations on record), and (2) the probability of 'dry' (< 33 %) and 'wet' (> 67 %) seasons for 262 townships throughout Australia. The purpose of the forecasts is to provide information for decision makers whose operations benefit from knowledge of rainfall patterns over the coming season such as farmers and futures traders.
The authors, Vizard, Anderson and Buckley, compared the township-specific 'wet' and 'dry' forecasts for the period up to May 2005 to the rainfall observations of these towns. A variety of statistical analyses were conducted on the data set to assess the predictive value of the forecasts. Unfortunately, all of the statistical measures indicated that the seasonal forecast system was of little value. Even after changes to the method used to generate the forecasts in March 2000 the forecasts improved somewhat but continued to be of little predictive value.
The researchers came to the conclusion that some users would have been better off ignoring the forecast, especially those who relied upon large shifts in seasonal patterns for their decision making. If the seasonal forecasting model is to deliver uniform and widespread value to end-users then new rainfall indicators with greater predictive power must be developed.
For further information contact David Buckley on 69334145.