Evaluation and projection of extreme rainfall from a large ensemble of high–resolution regional climate models in Australia

Incorporating climate change into intensity−duration−frequency (IDF) curves is broadly conducted using either a climate model simulation–based approach or a covariate–based approach. However, as of now, there has been no research comparing these two approaches in the context of rainfall IDF derivation under climate change. To this end, this study evaluates the 1–hour annual rainfall maxima from an ensemble of 60 CORDEX–CMIP6 simulations using a high–resolution regional climate model, the Conformal Cubic Atmospheric Model, for Australia. We quantify rainfall changes for the near (2041–2070) and far (2071–2100) future compared to a reference period (1961–1990) across various durations and Annual Exceedance Probabilities (AEPs) under three emissions scenarios for 39 locations across Australia. We then compare these projections with covariate–based frequency model projections. The 1–hour extreme (1 in 100 AEP) hindcast event shows a negative bias relative to observations, with a wide degree of variability across the ensemble. Projected changes for a high emissions scenario with a 3 of global temperature increase show a median increase of 33.9% for 1–hour and 18.9% for 1–day extreme events by the end of the century. Additionally, the reference 1 in 100 AEP event is projected to be 2.3 and 1.6 times more frequent for the 1–hour and 1–day durations, respectively. Projections also indicated extreme rainfall increases at the rate of 8.7% which exceeds Clausius–Clapeyron (CC) scaling for 1–hour duration events, and nearly equal CC scaling for longer 1–day durations. Covariate–based projections indicated larger quantile increases for 1–hour events with no change for 1–day events. Regional downscaling provides robust evidence for extreme rainfall changes despite uncertainties. • Dynamically downscaled CMIP6 1 hour 100–year rainfall events are negatively biased • Increases in the 1 hour 100–year rainfall events are projected by 2071–2100 • Projected rainfall increases are larger for shorter durations and rarer events • 1 hour 100–year rainfall is projected to occur twice as often by 2071–2100