Rainfall–runoff modelling in northern Australia: A guide to modelling strategies in the tropics

Abstract

A prolonged drought across southern Australia has led to renewed interest in water resource development of northern Australia, and to increased demand for runoff predictions from ungauged catchments in Australia’s tropics. In contrast to more temperate settings where there is a plethora of rainfall–runoff modelling studies, the world’s tropical regions, including those of Australia, have received little attention and thus the predictive skill of various rainfall–runoff models and methods in tropical basins is less known. Using data from 105 catchments in tropical Australia, five daily rainfall–runoff models and three methods of regionalising model parameters were compared. When locally calibrated, the more complex rainfall–runoff models performed best. However, when the models were used to predict streamflow in ‘ungauged’ catchments the differences in model performance was negligible. The adoption of multiple criteria to select an optimal parameter set resulted in an improved ability to simulate low-flows with no loss in predictive capacity for higher flows. An ‘informed’ transposition of parameter sets from gauged to ungauged catchments was better than random assignment of intact parameter sets for medium to high-flows, but not for low-flows. Assigning model parameters on the basis of spatial proximity outperformed physical similarity methods, particular with respect to model bias. The use of spatially distributed rainfall data did not improve model performance over the use of catchment average rainfall data. When models were locally calibrated there was a weak inverse correlation between catchment area and model performance. However, constraining donor–target parameter allocation by similar catchment area did not improve predictive capability in ungauged catchments. Although model performance was not as good as that reported for southern Australia and other temperate regions of the world, this study confirmed that modelling strategies similar to those claimed as successful elsewhere have application in tropical savanna environments.