Author(s)

G.A. TULARAM

Abstract

This paper explores the relationship between El Niño Southern Oscillation Index (SOI) and rainfall in South East Queensland. Several statistical models of the SOI and rainfall data are analysed including interactions. The results of the analyses show that although the SOI and rainfall can be predicted for short terms using time series analysis methods, mean squared error (MSE) and Theil’s U showed that in comparison to persistence model forecasts may not be significantly better. Although the short term predictions may be worse, this does not mean that these cannot be used; rather, it shows that for different fitting and forecast periods, models can fail in predicting ahead. In fact weather forecasts can predict the short term rainfall well – but cannot predict monthly precipitation for two or three or more months ahead. In climate predictions the limit of predictability is reached when the standardised MSE reaches the climate variance (=1). All the rainfall models used in this study show an MSE well below 1, i.e. the models did not reach the limit of predictability.

Keywords

El Niño, predicting rainfall, rainfall modelling, SOI, time series