Using historical temperature indices to predict future rainfall extremes

A large risk to life and infrastructure exists from the predicted increase in rainfall extremes with climatic change. But quantifying this increase is exceedingly difficult. In climate models there are issues around models not directly simulating all extreme rainfall producing processes. However, extrapolating historical relationships is also fraught due to the statistical artefacts that exist and the interaction of various spatial and temporal scales. Here we investigate the usefulness of temperature trends for rainfall prediction. Using long historical records across Australia we provide evidence that temperature variability is correlated to rainfall variability. Results are consistent across different depths of rainfall and hold regardless of the temporal scale of rainfall considered. A physical basis for these statistical relationships is discussed and validated. We conclude that historical trends in temperature are a useful conditional variable for informing how rainfall, and rainfall extremes may change in the future.