The Synthesis Report of the IPCC Third Assessment in 2001 concluded that one of the key uncertainties regarding attribution of climate change is "relating regional trends to anthropogenic climate change". Most studies that have successfully detected an anthropogenic influence on recent climate change have considered global scale variations. It is harder to attribute the effects of anthropogenic climate change on regional scales than on global scales because of the greater magnitude of the natural climate variability relative to forced climate change signals.

The results from a regional climate change detection and attribution study for Australia are described. There has been a significant increase in observed temperatures and a reduction in diurnal temperature range averaged across Australia over the last 50 years. There is a large out-of-phase relationship between rainfall variations and temperature variations in Australia on interannual timescales, with wet years associated with cooler temperatures and reduced diurnal temperature range. These rainfall-associated temperature variations contribute a large part of the interannual variability of Australian-mean temperatures. Much of this interannual variability of Australian temperatures and rainfall is associated with the impact of El Nino-Southern Oscillation on Australia. However, on century timescales, the warming over Australia has not been associated with decreased rainfall; rather there has been an observed increase in Australian-mean rainfall.

In this study, the natural variability of temperatures associated with rainfall variations is removed and we consider the residual temperature variations. This reduces the "noise" in the temperature variations, enhancing the signal-to-noise ratio for anthropogenic climate change detection. The same approach is applied to both observed and model-simulated Australian-mean temperatures and rainfall. We show that control climate model simulations represent the observed relationship between Australian-mean temperatures and rainfall variations.

Comparison of anthropogenically-forced model simulations with observed Australian-mean temperatures shows that most of the observed warming is likely to be due to human activity. Removing the rainfall-induced temperature variations increases the confidence in this conclusion and allows attribution statements to be made about the observed warming over even smaller regions.