This presentation describes a new dynamical downscaling experiment to predict regional climate change at 5km resolution for Victoria during 1960-2100 using CSIRO's Conformal Cubic Atmospheric Model (CCAM). The experiment includes both RCP4.5 and RCP8.5 emission scenarios, as well as downscaling 6 CMIP5 Global Climate Models (GCMs) recommended by Climate Change in Australia, ACCESS1-0, CNRM-CM5, GFDL-ESM2M, HadGEM2, MIROC5, NorESM1-M plus ERA-Interim for the present climate. To reduce the impact of GCM biases, the Sea Surface Temperature (SSTs) from the GCMs are bias and variance corrected, based on errors during 1980-2010. The global atmosphere is simulated at 50 km resolution with the corrected SSTs and then downscaled to 5 km resolution over Victoria using a scale-selective spectral filter. CCAM employs the Australian developed CABLE land-surface, UCLEM urban and CSIRO-Mk3.6 prognostic aerosol models to better simulate the Australian climate. The downscaled results are reasonably consistent with the GCM host model in terms of changes in mean temperature and rainfall. However, the representation of regional phenomena such as the urban heat island or extreme rainfall are significantly improved compared to the host GCM. The CCAM simulations predict enhanced drying at higher elevations that was not resolved in the GCMs. Changes in extreme rainfall will also be discussed.