Abstract
The Indian Ocean Dipole (IOD) and the El Niño Southern Oscillation (ENSO) are key modes of natural climate variability which co-occur and influence regional rainfall in Australia. This study evaluates the skill of 60 CMIP6 global climate models, including 16 pre-selected models for dynamical downscaling, in simulating the characteristics of IOD & ENSO, their inter-relationship, and their combined and independent influences on Australian rainfall. Focusing on the austral winter-spring season (JJASON) during 1950–2014, we use partial correlation & regression techniques to disentangle the influence of ENSO from the IOD-rainfall relationship and vice versa. Compared with observations, most CMIP6 models overestimate IOD & ENSO variability, frequencies of IOD & co-occurring IOD-ENSO events, and the influence of ENSO on IOD. More models reasonably capture the observed independent IOD-rainfall correlation in Southern (SA) and Southeastern (SEA) Australia than observed combined correlation, while in Eastern (EA) and Northern (NA) Australia, more models reasonably reproduce observed combined ENSO-rainfall correlation than observed independent one. A similar result emerges in regression analyses, but it is more pronounced for ENSO in EA and NA than for IOD in SA and SEA. The ranking of CMIP6 models based on four statistical skill metrics shows that each model demonstrates distinct skill variations, highlighting the importance of skill-specific analyses when evaluating relationships between IOD/ENSO and Australian rainfall. Our results emphasise the need to consider both combined and independent effects of IOD/ENSO on Australian rainfall, and our ranking of the pre-selected CMIP6 models for dynamical downscaling will be helpful to better understand biases in these regional projections for Australia.