Dynamic modelling of hydrochemistry is a valuable tool to study and predict the recovery of surface waters from acidification, and to assess the effects of confounding factors (such as delayed soil response and changing climate) that cause hysteresis during reversal from acidification. The availability of soil data is often a limitation for the regional application of dynamic models. Here we present a method to upscale site-specific soil properties to a regional scale in order to circumvent that problem. The method proposed for upscaling relied on multiple regression models between soil properties and a suite of environmental variables used as predictors. Soil measurements were made during a field survey in 13 catchments in the Pyrenees (NW Spain). The environmental variables were derived from mapped or remotely sensed topographic, lithological, land-cover, and climatic information. Regression models were then used to model soil parameters, which were supplied as input for the biogeochemical model MAGIC (Model for Acidification of Groundwater In Catchments) in order to reconstruct the history of acidification in Pyrenean lakes and forecast the recovery under a scenario of reduced acid deposition. The resulting simulations were then compared with model runs using field measurements as input parameters. These comparisons showed that regional averages for the key water and soil chemistry variables were suitably reproduced when using the modelled parameters. Simulations of water chemistry at the catchment scale also showed good results, whereas simulated soil parameters reflected uncertainty in the initial modelled estimates.