Uncertainty in rainfall—runoff model simulations and the implications for predicting the hydrologic effects of land-use change

Abstract

This paper quantifies the levels of uncertainty in rainfall—runoff model predictions due to the errors in hydrological and climatic data, and considers the implications for prediction of the hydrologic effect of land-use changes. In this study, a rainfall—runoff model, the Monash model, was calibrated on one catchment from each of at five experimental areas in Victoria, Australia. The validity of the optimised parameters was first examined by comparing them with independent estimates. Using the posterior distribution of the pan coefficient and model parameter estimates, the uncertainty of model predictions was determined by 90% prediction limits obtained from a Monte-Carlo simulation. The effects of systematic errors in the model parameters and model inputs (rainfall and evaporation) on runoff were also investigated. The analysis showed that systematic errors in rainfall have the most serious effect on predicted flows, but that estimation errors (random) in pan coefficient and model parameters also have significant effects. With this level of background ‘noise’, up to 65% of the catchment forest area would have to be cleared to produce flow increases detectable at the 90% prediction level, depending on the type of errors and the catchment in question.