Assessing and selecting CMIP6 GCMs ensemble runs based on their ability to represent historical climate and future climate change signal

The latest advances in climate change science were recently summarised in the sixth assessment report of the IPCC, including the contribution of CMIP6 models to understand changes in future climate. The CMIP6 is composed of hundreds of simulations, where the same model can have a few dozen different realizations for a single shared socio-economic pathway (SSP). This wealth of environmental data can be challenging for end-users interested in selecting ensemble runs to perform downscaling and impact assessments. Here, we assess the performance of the CMIP6 historical ensemble runs against observational data (Australian Water Availability Project – AWAP) and reanalysis (ERA-Interim) using a combination of metrics such as the Kling Gupta efficiency (KGE) over Australian regions. The assessment was based on precipitation, minimum and maximum temperature and sea surface temperature for the period 1995-2014, accounting for seasonal, monthly and daily time steps. We also assessed the climate change signal for precipitation and temperature for mid-century and end-of-the-century and developed an algorithm to automatically select the best-ranked ensemble runs and represent the spread in the climate change signal – that is the Skill Spread Selection. The results are presented as a performance score ranging from 0 to 100 which can be used to rank and select ensemble runs with distinct future climate signals. The analysis has great potential to inform scientists and practitioners on the strengths and limitations of individual ensemble runs and offers a robust and practical solution for the selection of CMIP6 realizations.