Assessing the performance of an ensemble approach to rainfall—runoff modelling for prediction of the impact of climate change on streamflow

Abstract

Climate change over the last four decades has resulted in a significant decline in streamflow in southwest Western Australia, and there is concern for the state of these streams under a future climate. In order to estimate future streamflow, a major exercise was undertaken to simulate the flows under a range of future climate projections. A suite of five rainfall-runoff models was calibrated against observed flow in 106 gauged catchments across the region. The results from weighted means of 31 ensembles, made up of all of the linear additive combinations of the five models, were examined for their ability to reproduce the measured streamflows. The mean of the daily runoff from the Sacramento and IHACRES models consistently resulted in a better model fit than any other combination of the individual models and this 'adopted model' was used to model the runoff under the climate projections. This combination produced the best overall objective function, as determined by the model efficiency of Nash and Sutcliffe (1970) and average bias of less than 2% limit.