Australia's Tinderbox Drought: An extreme natural event likely worsened by human-caused climate change.

We examine the characteristics and causes of southeast Australia's Tinderbox Drought (2017 to 2019) that preceded the Black Summer fire disaster. The Tinderbox Drought was characterized by cool season rainfall deficits of around -50% in three consecutive years, which was exceptionally unlikely in the context of natural variability alone. The precipitation deficits were initiated and sustained by an anomalous atmospheric circulation that diverted oceanic moisture away from the region, despite traditional indicators of drought risk in southeast Australia generally being in neutral states. Moisture deficits were intensified by unusually high temperatures, high vapor pressure deficits, and sustained reductions in terrestrial water availability. Anthropogenic forcing intensified the rainfall deficits of the Tinderbox Drought by around 18% with an interquartile range of 34.9 to -13.3% highlighting the considerable uncertainty in attributing droughts of this kind to human activity. Skillful predictability of this drought was possible by incorporating multiple remote and local predictors through machine learning, providing prospects for improving forecasting of droughts.

The role of internal variability and external forcing on southwestern Australian rainfall: prospects for very wet or dry years.

The cool-season (May to October) rainfall decline in southwestern Australia deepened during 2001-2020 to become 20.5% less than the 1901-1960 reference period average, with a complete absence of very wet years (i.e., rainfall > 90th percentile). CMIP5 and CMIP6 climate model simulations suggest that approximately 43% of the observed multi-decadal decline was externally-forced. However, the observed 20-year rainfall anomaly in 2001-2020 is outside the range of both preindustrial control and historical simulations of almost all climate models used in this study. This, and the fact that the models generally appear to simulate realistic levels of decadal variability, suggests that 43% might be an underestimate. A large ensemble from one model exhibits drying similar to the observations in 10% of simulations and suggests that the external forcing contribution is indeed larger (66%). The majority of models project further drying over the twenty-first century, even under strong cuts to greenhouse gas emissions. Under the two warmest scenarios, over 70% of the late twenty-first century years are projected to be drier than the driest year simulated during the 1901-1960 period. Our results suggest that few, if any, very wet years will occur during 2023-2100, even if strong cuts to global emissions are made.