Granger causal predictors for maximum rainfall in Australia

Abstract

We describe a novel approach for investigating changes in regional rainfall patterns from short records and for identifying potential causes of these changes. We focus on the predictive potential of past values of the Indian Ocean Dipole (IOD), the Southern Oscillation Index (SOI), and their interactions on rainfall in South Australia and Western Australia. A measure of annual rainfall anomaly based on a signed version of the Hotelling statistic, is proposed. This modified statistic is calculated for rainfall anomaly at 7 sites in South Australia over a period of 59 years. Various plots of rainfall anomaly provide evidence of outlying values and increasing variability over the period. There is no evidence that the rainfall anomaly is associated with past values of temperature or the Southern Oscillation Index or their interaction. In contrast, there is evidence that past values of the IOD have Granger predictive value for rainfall and this is evidence for a Granger causal relationship. A physical explanation for the influence of the IOD on rainfall in South and West Australia is presented.