The reduction of nitrogen fluxes from land to sea is an important task in areas with estuarine or marine eutrophication. Wetland creation has been proposed as one method to reduce nitrogen from streams draining agricultural areas. In this study, a scenario of nitrogen removal in created wetlands was evaluated by mathematical modelling of nitrogen fluxes in a catchment (224 km2) in southern Sweden. The scenario was based on topographically realistic siting of 40 potential wetlands with a total area of 0.92 km2 (0.4% of the catchment area). Nitrogen removal in the wetlands was described with a simple and robust first-order model, which was modified and evaluated against data from eight monitored surface-flow wetlands. However, the modifications gave no substantial support for changing the basic model. For catchment-scale modelling this wetland model was incorporated into a dynamic process-based catchment model (HBV-N). The catchment was then divided to several coupled subbasins, so that the wetland influence on nitrogen load could be estimated separately for each potential wetland. The modelling showed that the 40 potential wetlands would reduce the nitrogen transport to the coast with approximately 6%. Specific removal rates ranged between 57 and 466 kg ha-1 yr-1 for the different wetlands, depending on residence time (size and hydraulic loading) and nitrogen concentration in inflow. Due to temperature dependence and seasonal variation in water discharge, significant decrease in nitrogen concentrations mainly occurred during summer periods with low loading. The study illustrates that catchment modelling is a useful method for analysing wetland creation plans, and that wetland creation must cover fairly large areas and be combined with other measures in order to achieve substantial reduction of nitrogen fluxes to coastal waters. Further monitoring of existing wetlands will improve the removal expression and decrease uncertainty. For instance, at present it could not be deducted whether wetlands with low average residence times (< 2 days) have net removal or net resuspension on an annual basis. © 2002 Elsevier Science B.V. All rights reserved.