To evaluate the effects of nitrogen (N) emission policies, reliable information on nitrate concentrations and leaching fluxes from forest ecosystems is necessary. Insight into the regional variability of nitrate concentrations, to support local policy on emission abatement strategies is especially desirable. In this paper, three methods for the calculation of a spatial distribution of soil nitrate concentrations in Dutch forest ecosystems are compared. These are (i) a regression model based on observed nitrate concentrations and additional data on explanatory variables such as soil type, tree species and nitrogen deposition (ii) a semi-empirical dynamic model WANDA, and (iii) a process-oriented dynamic model SMART2. These two dynamic models are frequently used to evaluate the effects of reductions in nitrogen deposition at scales ranging from regional to countrywide. The results of the regression model evaluated the performance of the two dynamic models. Furthermore, the results of the three methods are compared with the steady-state approach currently used for the derivation of nitrogen critical loads. Both dynamic models, in the form of cumulative distribution functions, give similar results on a national scale. Regional variability is predicted differently by both models. Discrepancies are caused mainly by a difference in handling forest filtering and denitrification. All three methods show that, despite the high nitrogen inputs, Dutch forests still accumulate more N than they release. This implies that, in respect of groundwater quality, presently acceptable nitrogen deposition is higher than the (long-term) critical loads. However, in areas with high atmospheric nitrogen input, all three methods indicate that the EU standard for nitrate in groundwater (50 mg NO 3 l –1 ) is exceeded. Steady-state with nitrogen deposition seems to have been reached in about 10% of the forested area, with a nitrate concentration greater than 50 mg NO 3 –1 . Keywords: soil modelling, up-scaling, model validation, critical load