Estimating future rainfall distributions in a changing climate for water resource planning: Victoria, Australia

Abstract

AbstractHere we use observations and simulations from 40 global climate models from phase 5 of the Coupled Model Intercomparison Project (CMIP5), under preindustrial, historical, and a high emission scenario (RCP8.5) to provide estimates of Victorian cool season (April–October) rainfall for the coming century. This includes a new method which exploits recent research that estimated the relative contribution of external forcing and natural variability to the observed multidecadal decline in cool season rainfall in Victoria from 1997. The new method is aimed at removing the influence of external forcing on Victoria’s cool-season rainfall, effectively rendering a stationary time-series. The resulting historical record is then modified by scaling derived from the mean projected change evident in climate models out to 2100. The results suggest that the median value of the All-Victoria rainfall PDF will decrease monotonically over the remainder of the twenty-first century under RCP8.5. The likelihood that All-Victoria rainfall in any given year from 2025 onward will be below the observed 5th percentile of the observations (291 mm) increases monotonically, becoming three times larger by the end of the century. The new method is assessed using cross-validation and its ability to hindcast observed multidecadal rainfall change. The latter indicates that CMIP5 models poorly replicate recent interdecadal rainfall change. So, while we have more confidence in the new method because it accounts for the non-stationarity in the observed climate, limitations in the CMIP5 models results in us having low confidence in the reliability of the estimated future rainfall distributions.