The use of data from large-scale forest inventories for the development and application of explanatory ecosystem models is often hampered by a lack of quantitative site data. To circumvent such limitations, we describe a method to estimate soil water-holding capacity (SWC), a key parameter in many ecosystem simulation models, for the sample plots of the Austrian Forest Inventory (AFI). In accordance with the limited soil data which are customarily recorded by forest inventories, a functional model for the calculation of SWC from soil depth, coarse fractions in two soil layers, water-holding capacity of the fine fraction and the organic carbon content of the mineral soil is developed. Bayesian probability theory is employed to establish relationships between measured site attributes of the AFI and the soil parameters required for calculating SWC. More detailed site and soil data of 514 sample plots of the Austrian Forest Soil Survey (AFSS) which can be considered a subsample of the AFI are used for model calibration. As explanatory variables orientation, slope, elevation, topography, soil-type and humus-type are included in the predictive models for the estimation of the required soil parameters. Model performance in reproducing the calibration dataset was evaluated by means of intraclass correlation coefficients. A predictability index from the field of information theory is used to evaluate the contribution of the independent variables in explaining the variability (entropy) of the dependent variables. A preliminary model validation with a subset of AFSS data indicated good agreement of observed and predicted values for the individual soil parameters as well as for the composite parameter SWC.