Informed climate change adaption relies on hydrological models to transform the projections of climate forcing into key societal impacts, such as flood and drought risk. The development of hydrological models that can provide credible projections of catchment runoff under future climate forcings is regarded as one of the most challenging aspects of hydrological science. Particularly as the models are being extrapolated outside the domain over which there model has been calibrated. To improve the reliability of runoff projections from hydrological models requires systematic scrutiny of the assumptions of the hydrological modelling process. A key issue is the impact of hydrological non-stationarity on runoff projections. A strategy will be presented for diagnosing and interpreting hydrological non-stationarity using a case study from South Australia’s major water supply catchment. This case study will demonstrate how incorporating hydrological non-stationarity can improve runoff projections under a drier changing climate. The broader implications of this strategy for providing robust hydrological predictions for climate change impact assessment will be outlined.