Calculation of joint PDFs for climate change with properties matching recent Australian projections

Abstract

The 2007 report Climate Change in Australia presented single-variable probability density functions for climatological change driven by global warming. However, their use is limited to climate impact applications involving one variable or if uncertainty in pairs of variables can be assumed to be independent. As is shown here, local changes in mean rainfall and temperature from 23 individual climate models are often strongly anticorrelated, particularly for summer and annual cases in inland Australia. Relatively large warming tends to coincide with declines in rainfall. A simple iterative approach is developed that produces a joint density function for the pair that includes this anticorrelation and has marginal distributions matching the single-variable ones. An extension of the approach to three variables was also successful. These joint functions can be used in applications where two or more related variables are important. The approach is illustrated using results for Dubbo, New South Wales, in 2070 under the A1B forcing scenario. A brief comparison of the method to an alternative of using a Gaussian copula is made.