Quantifying 'realistic' uncertainty bounds as a part of sound hydrological modelling practice in data scarce regions of southern Africa

This study was based on two premises; the ultimate objective of hydrological modelling is a contribution to sustainable water resources management, and the inherent uncertainties in model results should be realistically quantified. The study uses methods of uncertainty analysis that have been previously applied in the southern Africa region which are based on constraining model outputs using the likely ranges of a set of hydrological indices. One objective was to offer suggestions for sound modelling practice and highlight potential problems. The approach is applied to two case studies where there are very limited streamflow observations. The uncertainty ensemble outputs from the hydrological model are input into a water supply allocation model to assess system performance under different abstraction scenarios. The results are compared with the limited available evidence of system performance and the conclusion is that the uncertainty bands are generally acceptable for future decision making.