The simulation of weathering, solute distribution or acidi®cation at the catchment scale is predominantly done with either a mass balance or process level model. The former redistributes total elemental concentrations between known points with measured total concentration, but does not explicitly include catchment hydrology. The latter includes compartmental hydrological models and detailed descriptions of spatially averaged chemical reactions. Interestingly, the model applications tend towards hydrologically dierent watershed structures: mass balance modelling favours bare rock watersheds similar to the Apache Leap Research Site, whereas process level models are applied most often to forested watersheds, among which the Hubbard Brook Experimental Forest is an example. Although constrained either by mineral or water compositions on the one hand, or calibrated against stream or lake water chemistry on the other, both approaches basically ®t parameters to the geochemical circumstances of the speci®c watershed of interest. Limited success is attained if the hydrological conditions remain within the circumstances of the parameter ®tting. The principal dierences in model formulation and approach to mass balance modelling are discussed. Without advancements in model calibration and rigorous model testing, and the development of methods for optimizing the important reactions and pathways, the transportability of models between watersheds or the simulation of extreme events will continue to be inadequate.