To predict reversibility of soil and water acidification under conditions of decreasing S-input into forested ecosystems, the amount of reversibly bound SO42- in the soil as well as its desorption behavior must be considered. This study investigated variability of sulfate isotherms and their spatial scaling. The Langmuir isotherm parameters in two forested catchments (one deciduous, one coniferous stand) and their spatial distribution at two different scales (site and catchment scale) were investigated. The soil samples (250 cm(3)) were taken in a systematic grid with spatial distances ranging from 20 x 20 m to 300 X 300 m, Isotherm parameters, soil pH, dithionite- and oxalate-extractable iron and aluminum, crystalline iron oxide, and ent were determined. Significant relationships were found between sulfate sorption isotherm parameters and soil chemical properties for each site. It would be useful to be able to replace the experimental determination of isotherm parameters with the less costly measurements of soil chemical properties such as pH, Al- and Fe-oxides, and C-org content of the soils. Regression analysis, however, resulted in different parameters for each site. Thus, at sites where this relation is unknown, isotherms have to be determined via soil extraction procedures and cannot be predicted by soil chemical properties. To determine the average isotherms with an accuracy of +/- 0.2 mu mol g(-1) SO42- (sorb), a sample size of 55 for the site scale and up to 106 for the catchment scale is required. The semivariance of the isotherm parameters revealed no spatial patterns. The influence of isothenll variability on the prediction of SO42--fluxes with seepage was investigated using the chemical equilibrium model MAGIC. Forecast accuracy depends on proper consideration of the variability of sulfate sorption in a catchment.