Extreme heavy rainfall events often result in flooding and have severe impacts with associated costs. Indices representing the frequency and intensity of rainfall extremes have been shown to be associated with climate modes, such as the El Niño-Southern Oscillation and the Indian Ocean Dipole, suggesting there may be predictability for rainfall extremes on seasonal timescales. Accurate seasonal prediction of rainfall extreme indices would aid stakeholders in their efforts to prepare for, and mitigate the impacts of, rainfall extremes.

In this study we use observed rainfall data, reanalyses and the state-of-the-art ACCESS-S seasonal prediction system. We first examine the potential predictability for a set of extreme rainfall indices, including the most intense multi-day rain events in a season and the number of days above specific thresholds and percentiles, based on statistical relationships with climate modes and sea surface temperature anomalies at various lead-times. We then use reanalysis data to investigate the processes behind the predictability of rainfall extremes. We examine whether ACCESS-S hindcast simulations have predictive skill for the extreme rainfall indices using a range of verification statistics. We also investigate whether the processes that result in predictability of rainfall extremes are captured in the ACCESS-S simulations.

By demonstrating predictability of rainfall extremes and predictive skill in ACCESS-S it is hoped that Bureau of Meteorology seasonal outlooks may incorporate a set of indices for rainfall extremes in the future.