Partitioning uncertainty in climate projections for the Murray Darling Basin using large ensembles

Abstract The Murray Darling Basin (MDB) in southeast Australia is a highly managed system, aiming to deliver resilient and sustainable outcomes for communities, industries, and the environment that share the water in the basin. Climate projections from the Coupled Model Intercomparison Project Phase 6 (CMIP6) are informing the upcoming review of the Basin Plan. However, internal variability in future climate projections is often underappreciated in regional impact assessments, such as those undertaken in the MDB. Applying methods that partition uncertainty can be useful to quantify the climate change ‘signal’ to consider in planning. In this study two methods are used to estimate changes in climate in the MDB (1) future time slice approach, and (2) pattern scaling. We estimate the range of projected changes from a CMIP6 multi-model ensemble consisting of one member from all available climate models. Further, we use single model initial condition large ensembles (SMILEs) available from eight models to partition the sources of uncertainty and examine their evolution with increasing global temperatures. Our results show that pattern scaling estimates, by more specifically reflecting the climate change signal in the ensemble, can be a useful tool for visualising and considering forced change and natural variability in a coherent framework. This is particularly relevant for assessments that use change signals from climate models in conjunction with stochastic methods to understand natural variability. Analyses of large ensembles show that estimates with higher confidence in the MDB such as the increases in temperature and PET, and declines in austral winter rainfall compared to summer, are associated with lower internal variability and higher signal to noise ratios. Our results have implications for the use for climate projections for water resources assessments in the basin.