Assessing historical and future runoff modelled using rainfall from the analogue downscaling method

In this study, two time slices of GCM daily rainfall - one historical (1961-2000) and one future (2046-2065) (IPCC AR4 A2 scenario), are downscaled from 11 GCMs to 0.05° (~5km) grid cells covering southeast Australia. The downscaled gridded rainfall is used to drive a widely used hydrological model and results are compared with the runoff, modelled using observed rainfall and rainfall scaled using a perturbation method (daily scaling) to reflect climate change. The results show that the analogue method underestimates rainfall and subsequently the modelled runoff. Therefore, it is necessary to use an inflation factor to scale the daily rainfalls to match the observed 1961-2000 seasonal means. Nevertheless, differences remain between daily analogues and observed rainfall distributions, and these translate to errors in the modelled daily and mean runoffs. The percentage change in future runoff (for the period 2046-2065 relative to 1960-2000) modelled using rainfall from the analogue method and the daily scaling method are largely similar with the large majority of the results indicating a decrease in runoff particularly in the southern parts of the region. There are differences in some of the GCMs and the cause of the differences needs to be further investigated. The range in the modelled change in future runoff using the analogue method informed by the 11 GCMs is smaller than the range when using the daily scaling method. The analogue method is a simple downscaling method and can therefore be relatively easily used with many GCMs to represent the range of uncertainty in the climate change projections. It has the potential to capture a fuller range of potential changes in future rainfall characteristics compared to simple perturbation or scaling methods. Like the scaling methods, analogues also ensure that the daily spatial correlation over large regions is preserved. However, the use of analogues based solely on observed historical data is a limitation. The modelling results here indicate that the analogue method can be useful for hydrological impact studies over large regions. More research is nevertheless required to continue to improve the analogue method (including posterior corrections of GCM predictor variables used to define the weather states and bias-adjusting the daily rainfall outputs from the analogue method) in order to produce daily rainfall that are sufficiently similar to the observed daily rainfall for direct use in hydrological models.