A fully distributed coupled hydrological-sediment yield model was developed. An assessment was made of the predictive uncertainty in the individual model predictions, as well as the uncertainty propagated from the primary hydrological model to the secondary sediment yield model, using the Generalised Likelihood Uncertainty Estimation (GLUE) methodology. The value of additional data, in the form of additional periods of flow data, as well as deterministic (based on landuse and soil type) and random spatial parameterisation of hydrological parameters in restricting model uncertainty of the spatially lumped model parameterisation were examined, using Bayesian updating. The results revealed significant model uncertainty in both the hydrological and sediment yield models, with uncertainty bounds widest at peak flow and sediment flux, and predictive failure in recession flows, similar to other applications of GLUE methodology. Uncertainty in the sediment yield model was found to be due to uncertainty inherited from the hydrological model, as well as simplifying assumptions made about sediment removal and transport, and resulted in lower model efficiencies and generally poorer qualitative sedigraph fit. The model validation exercise revealed that the calibrated 'optimum' parameter set was not 'optimum' for all validation periods and resulted in inaccurate spatial and temporal hydrological response predictions for the validation periods. This suggested that traditional split-sample model calibration methods may not be effective in capturing the true spatial and temporal variability of the system. Successive periods of flow data were effective in reducing the calibration period uncertainty bounds. Similarly, the use of sediment yield predictions to update hydrological model uncertainty resulted in a reduction in hydrological model uncertainty. Spatially distributed parameterisation was found to also improve model predictions, resulting in a reduction in uncertainty bounds, particularly for soil-distributed parameterisation. However, stochastic parameterisation of spatially variable hydrological parameters provided equally acceptable predictions for both models, suggesting that a deterministic approach might not be required to capture the spatial variability in hydrological and sedimentological response in the study catchment, and that a stochastic approach may be adequate.