This study reassesses the application of the geochemical model MAGIC in the prediction of long-term changes of water quality in response to changes in atmospheric deposition. It does so within theMonte Carlo based GLUE methodology in which it is possible to evaluate the performance of sets of model parameters in predicting the available observations as a means of constraining the uncertainty in current and future predictions. This work was prompted by previous work which showed that, for a typical upland site in Wales, MAGIC predictions were dominated by the depositional scenario used. Uncertainties in the depositional scenario are taken into account by using estimates of uncertainty for the different depositional sources including European anthropogenic sources as produced by the HARM model. The results show almost no change in predictive uncertainty bounds, in the form of 5th and 95th percentiles of the likelihood-weighted distributions, owing to tight observational data constraints. The implications of this lack of change with respect to predictive capability and possible over-constraint by observed data are discussed.