Abstract
There is likely to be an increase in the area of the globe affected by drought under enhanced greenhouse gas conditions. Therefore water management and drought policy may need to be modified accordingly. Rainfall and potential evapotranspiration (PET) are the key factors defining meteorological drought, and the development of drought projections is facilitated by global climate model (GCM) simulations. This paper assesses how well a set of GCMs can reproduce observed characteristics of historical rainfall and PET on a regional basis and explores the implications for regional drought projections if the poorer performing GCMs are omitted. Fourteen of the GCMs used in the IPCC's 4th Assesment Report are considered and their results compared with 1951–2006 observed rainfall and PET over Australia. The results indicate that some GCMs can reproduce the observed spatial patterns of both the means and variability (represented as the coefficient of variation), but most GCMs fail to reproduce the linear long‐term trends. There is less clear difference between the better and poorer GCMs at a national level, but there is a clearer distinction at the regional level. The omission of the poorer GCMs leads to a clearer sign of the likely change (either increase or decrease) in future drought intensity in some regions. It also results in a decreased range of model‐to‐model uncertainty in some regions. It is hoped such uncertainty reduction can be useful to end users, particularly for those dealing with water management. Copyright © 2010 Royal Meteorological Society