Reducing uncertainty in selecting climate models for hydrological impact assessments

Deciding which climate models to use to assess the impact of climate change on water resources is particularly difficult in environments where precipitation dominates resource vulnerability. We show that assessing climate models based on their simulation of mean precipitation provides little guide to a model's ability to simulate the more extreme events that affect hydrological systems. In contrast, a probability density function based assessment using daily climate model data provides a good basis for confidence in a model's ability to simulate the 95th rainfall percentile. We demonstrate that climate models have useful skill in simulating observed probability density functions over two regions of Australia, although the well-known bias of excess rainfall at low rates remains common. We conclude by identifying those climate models that produce the best basis for hydrological impacts assessment over two regions of Australia.