Regional Application of the Pitman Monthly Rainfall-Runoff Model : Incorporating uncertainty into hydrologic modelling in Southern Africa

Effective and sustainable management of water resources demand reliable quantifications of water amount, distribution and quality. However, the available data are frequently insufficient for the practical application. Hydrologic models provide the necessary information on water availability and quality. It is impossible to accurately represent all hydrological processes in a model and the information available to establish a model for any specified basin is typically less than perfect. This inevitably results in predictions that are imperfect and that could span a range of equally plausible simulations. Uncertainty is thus unavoidable in any hydrologic modelling undertaking. This manuscript explores the combined use of two approaches for the generation of ensemble predictions for basins in South Africa. A priori parameter estimates, with associated prior probability distributions, are generated directly from basin physical attributes data and by simple Monte Carlo sampling from the feasible parameter space multiple model outputs are generated. Unacceptable models are rejected based on indices of characteristics of catchment hydrological behaviour, such as runoff coefficient.