Non-stationarity in extreme rainfalls across Australia

Abstract

Future flooding is likely to exceed current design flood levels which are based on historical extreme rainfall characteristics. The Clausius-Clapeyron relationship explains the intensification of extreme rainfalls as approximately 7% per one degree warming as atmospheric water holding capacity increases with temperature. Therefore, to prepare for a future warmer climate, we need to develop methodologies to project future rainfall intensities across the range of durations and exceedance probabilities used in engineering design. However, the studies that have investigated changes in extreme rainfalls across Australia have had disparate results and are not spatially or temporally comprehensive – hampering our understanding of changes in extreme rainfalls across different durations and exceedance probabilities.

This study investigates the impact of climate change on rainfalls from the annual maximum to the 1 in 100-year rainfall across a range of storm durations for the continent of Australia. We find increases in short duration (<1 h) annual maximum rainfall are greater than increases in long duration (>1 h) annual maxima across Australia from 1967 to 2021. These results are consistent regardless of the data period or data set chosen for analysis. We estimate events rarer than the annual maxima through fitting non-stationary Generalize Extreme Value models. We find that events of rarer severity have increased more than frequent events. Further, we identify the parameterisation of a model with non-stationary location and scale parameters to capture the changes in historic design quantiles that are consistent with our physical understanding of rainfall intensification, empirical quantile changes, and historical trends. We conclude that trends in annual maxima are best represented by non-stationary models that incorporate changes in both location and scale parameters, not by solely varying either location or scale parameters.