Climate impact, adaptation and vulnerability research underpins strategies for adaption to climate change and help to conceptualise what life may look like in decades to come. This work draws on climate projections, which in turn use outputs from global climate models and other modelling. The process of using available climate modelling into adaptation research involves numerous steps, including the selection of a sub-set of global model outputs, then regionalising these outputs to make them locally-relevant. There are several key decisions through the production process, where information may be selectively highlighted through poor sub-sampling, become skewed through the use of a particular regionalisation or scaling technique, or lose particular features of the climate change signal such as the change to extremes.

Here we consider the chain of processes leading to an application-ready data set, where each step may have a significant impact on the climate change signal. We present worked examples set in an Australian context in human health, natural systems and water availability. Our examples demonstrate that choices impact on the final results differently depending on various factors such as the application needs, the range of uncertainty of the projected variable, the amplitude of natural variability and the size of study region. Final results can differ in the direction of change, or by a factor of two or more in magnitude, depending on the choices made. The work illustrates the potential pitfalls when using unwise, non-representative datasets when conducting impact, adaptation and vulnerability research.