Pattern scaling constructs future climate change scenarios using the normalized change patterns of GCMs, offers the possibility of representing the whole range of uncertainties involved in future climate change projection. This paper investigates the applicability and uncertainty associated with the pattern scaling method in constructing the changes of future precipitation intensity indices at regional scale, using a two-step ensemble approach. In the first step, the linearity accuracy and GCM internal variability were examined explicitly. The inter-model variability of the GCMs and associated confidence intervals were produced in the second step ensemble. Australia and its 7 administrative regions was selected as the study area and three precipitation intensity indices, including two precipitation extreme indices, were used for the examination: i.e., the 99th percentile daily precipitation intensity (<i>P</i>₉₉), the 20-yr-return extreme precipitation intensity (RP₂₀), and the mean precipitation intensity (precipitation amount per wet day) (RPD). A total of 12 IPCC AR4 GCMs with 6 simulation samples were used for the ensemble. For the 3 precipitation intensity indices, good linear relationships between precipitation intensity indices change and global mean temperature change at the national level were found for most GCMs, however, the linear relationship weakened when the analysis was applied to the administrative regions. In addition, the GCM internal signal-to-noise ratios for each GCM tended to decrease at the regional and grid cell levels, along with the reduction in spatial scale. Both GCM-internal and inter-model variability was significant, and the inter-model variability was larger than GCM-internal variability. The final result of the inter-model ensemble median results show that for Australia, in general, all three indices will increase under global warming, with the change rates being 3.56, 7.62 and 2.26 % K^&minus;1 for <i>P</i>₉₉, RP₂₀ and RPD respectively at the national level.