Predictions of future climate and climate change are available from a variety of global climate models. There is no widely adopted protocol, however, how we – as a community – synthesise  regional climate projections from the available climate model outputs and observation data. Consequently, various approaches to form probabilistic regional climate change projections have been proposed and used. Here we compare different methods  to derive climate change projections for near-surface temperature and precipitation in Australia from simulations submitted to the World Climate Research Programme’s (WCRP) Coupled Model Intercomparison Project phase 5 (CMIP5) archive. The methods range from reporting empirical percentiles of the available simulations to fully probabilistic models of the uncertainties in CMIP5. All but the simplest approaches include a weighting of climate models based on evaluation against the observed climate. We explicitly compare the effect of weighting based on skill in reproducing recent mean climate against weighting based on recent trends. Preliminary results suggest that the choice of method generally affects the sharpness of the projection (here the difference between the projected 10^th and 90^th percentile) and thus the uncertainty in projected changes much more than the general tendency for a drying/wetting or the magnitude of warming. Also, both observed mean climate and recent changes provide little constraint for future projected changes in Australia.