The simulation results from physically-based hydrology and erosion models contain uncertainty. This is largely the result of uncertainty over the value of the model's input parameters. If such uncertainties are in the form of probability distributions of model output, it becomes apparent that they are not inconsiderable. Such variation in model output may reduce the utility of the simulations and it is therefore desirable for it to be reduced. Two methods are considered: one which relies on improved parameter set selection, and a second which derives physical parameters from observed hydrographs. Both methods are compared with a blind application of the model and observed data. The results suggest that, while both methods may be successful in reducing the variation in model output, the coincidence with the observed data deteriorates.