In this study we analyzed the large scale spatial patterns of river pH, alkalinity, and CO2 partial pressure (PCO2) in North America and their relation to river catchment properties. The goal was to set up empirical equations which can predict these hydrochemical properties for non-monitored river stretches from geodata of e.g. terrain attributes, lithology, soils, land cover and climate.For an extensive dataset of 1120 river water sampling locations average values of river water pH, alkalinity and PCO2 were calculated. The catchment boundaries and catchment properties were calculated using GIS and different sets of geodata. The correlations between the hydrochemical properties and the catchment properties were explored using simple and multiple linear regression analysis.For each of the considered hydrochemical parameters, a multiple regression equation was fitted: for pH with the predictor's mean annual precipitation and areal proportions of carbonate rocks (r2=0.60); for alkalinity, in addition to these two predictors, with subsoil pH and areal proportions agricultural lands (r2=0.66); and for pPCO2 (i.e. the negative logarithm of PCO2) with mean air temperature, mean catchment slope gradient, and mean annual precipitation (r2=0.43). Based on these results, we argue that spatial patterns in river water pH and alkalinity are governed by catchment processes related to chemical rock weathering. For the PCO2, on the other hand, the spatial patterns are governed by in-river processes on which catchment properties can have an indirect effect. We conclude that our approach can be used to predict averages of these parameters for non-monitored river stretches, which in-turn allows for a better spatially explicit representation of the rivers' carbonate system at the regional to global scale, which will be needed for a refined analysis of rivers in the global carbon cycle. © 2012 Elsevier B.V. ; SCOPUS: ar.j ; info:eu-repo/semantics/published