Evaluating and Using Seasonal Rainfall Forecasts for Decision-making on Farms

A project undertaken by CSIRO Plant Industry, and supervised by JR Donnelly

Sheep and beef producers in Australia respond to forecasts of poor seasonal weather by taking actions such as reducing stocking rates or purchasing supplementary feed to reduce the impact of the lower-than-expected rainfall. Such actions are expensive, so a favourable outcome depends critically on the accuracy of the forecasts. The Southern Oscillation Index (SOI) is widely promoted as an indicator of future seasonal rainfall but its usefulness needs careful evaluation.

Currently sheep and beef producers may inadvertently make inappropriate decisions based on SOI forecasts, often with substantial loss of profits. The primary purpose of this project is to determine how good the SOI is as a predictor of rainfall and what are the appropriate management responses a grazier should make in response to seasonal rainfall forecasts.


Detailed statistical modelling of monthly rainfall records over 80 years from six locations across the Australian continent showed that the SOI (Southern Oscillation Index) effect was contemporaneous with rainfall and accounted for 10% of rainfall variability. The use of SOI as predictor does not give useful improvement in rainfall forecasts.

Australian agricultural producers face high levels of seasonal uncertainty. Since 1993, the Australian Bureau of Meteorology has issued three-month seasonal rainfall outlooks, in which the probability of dry or wet seasons has been estimated for various districts and townships in Australia. These forecasts have the potential to guide farmers about changes to their management for the upcoming season.

The utility of such forecasts to help rational decision-making in agriculture is highly dependent upon their accuracy and any substantial bias in the forecast probability of 'dry' or 'wet' seasons will markedly erode their value. Our analyses of the actual rainfall outcome compared with these seasonal forecasts take into account the correlated nature of the data and indicate that the forecasts have generally had poor accuracy. Consequently the seasonal outlooks have had limited utility for agricultural purposes. Considerable improvement in accuracy is required if these forecasts are to be useful for improving management decisions on Australian farms.