Advanced risk-based event attribution for heavy regional rainfall events

Risk-based event attribution (EA) science involves probabilistically estimating alterations of the likelihoods of particular weather events, such as heat waves and heavy rainfall, owing to global warming, and has been considered as an effective approach with regard to climate change adaptation. However, risk-based EA for heavy rain events remains challenging because, unlike extreme temperature events, which often have a scale of thousands of kilometres, heavy rainfall occurrences depend on mesoscale rainfall systems and regional geographies that cannot be resolved using general circulation models (GCMs) that are currently employed for risk-based EA. Herein, we use GCM large-ensemble simulations and high-resolution downscaled products with a 20-km non-hydrostatic regional climate model (RCM), whose boundary conditions are obtained from all available GCM ensemble simulations, to show that anthropogenic warming increased the risk of two record-breaking regional heavy rainfall events in 2017 and 2018 over western Japan. The events are examined from the perspective of rainfall statistics simulated by the RCM and from the perspective of background large-scale circulation fields simulated by the GCM. In the 2017 case, precipitous terrain and a static pressure pattern in the synoptic field helped reduce uncertainty in the dynamical components, whereas in the 2018 case, a static pressure pattern in the synoptic field provided favourable conditions for event occurrence through a moisture increase under warmer climate. These findings show that successful risk-based EA for regional extreme rainfall relies on the degree to which uncertainty induced by the dynamic components is reduced by background conditioning.