Catchments are more spatially variable than catchment models assume them to be. Does spatial heterogeneity undermine the validity of spatially aggregated catchment geochemical models? Here we test whether catchment spatial heterogeneity invalidates a recently proposed technique that uses the bulk chemistry of runoff to predict how key water quality variables (e.g., titratable alkalinity in concentrations of base cations, hydrogen ions, and inorganic aluminum) will respond to changes in stream water in acid anion concentrations. We show that these predictions are valid for hypothetical heterogeneous catchments, created by mathematically ''mixing'' measured runoff chemistry data from chemically diverse streams. The chemical information in the combined runoff yields reliable acidification predictions, without any direct information concerning either the runoff chemistries of the individual source streams for the geochemical properties of their catchments. These results show that catchment acid buffering can be predicted directly from runoff chemistry in spatially heterogeneous catchments, even if the individual source regions cannot be identified and their geochemical characteristics are unknown. ¿ American Geophysical Union 1993