Incorporating rainfall uncertainty into catchment modelling

A framework for incorporating areal rainfall uncertainties into catchment modelling is presented and demonstrated through the daily streamflow simulation of the Mooi River catchment using the Australian Water Balance Model (AWBM). The framework is an extension of the typical hybrid manual-automatic model calibration-validation in which uncertainties are imposed as perturbations (disturbances) on the rainfall time series. The differences in areal rainfall obtained from different rain-gauge densities are used to generate the perturbations, and their variability is found to reduce as areal rainfall magnitude increases. The applied probability distributions of perturbations are therefore obtained for specified ranges of rainfall magnitude. The effect of incorporating uncertainties is assessed by finding out the impact of imposing perturbations on the validation performance of the rainfall-runoff modelling. This is done by carrying out 100 rainfall-runoff calibration-validation runs with perturbed and with unperturbed rainfall and comparing the 100 generated validation period runoff time series with the observed (historical) runoff series. With perturbations applied, the 5-95 percentile bounds from the resulting 100 streamflow ensembles contain 52% of the observed time series for the validation period. Without perturbations, only 25% of the observed flows fall within the bounds. The framework has the potential for practical use, but this would require a more rigorous identification of appropriate distributions of the perturbations.