Modelling Australian annual mean rainfall data: a new approach based on fractional integration

Abstract

This paper analyses Australian annual rainfall data from 1900 to 2006. We use new statistical techniques based on the concept of fractional integration, which allow us to examine the dynamic behaviour of the series in a much richer way than the classic approaches based on stationary or nonstationary processes. Moreover, the possibility of a structural break at an unknown date is also taken into account. The results indicate that a break takes place at 1973, the time trend coefficients being insignificant in the two subsamples, and the degree of dependence between the observations is higher after the break. Regional data are also explored and the results show significant differences in the degree of persistence across the regions, especially after the breaks.