AbstractModel calibration allows reducing parameter uncertainty and, therefore, uncertainty in simulation results. In the present study, automatic model calibration with the shuffled complex evolution algorithm is demonstrated using the example of the distributed conceptual model SWAT‐G, a derivative of the Soil and Water Assessment Tool (SWAT). The special challenge in calibrating SWAT is that it describes many physical processes and therefore is highly parameterized. Moreover, as SWAT or SWAT‐G, respectively, is a distributed model, the parameters may take on different values in different spatial subunits of a catchment model. A strategy of imposing constraints on the parameters to limit the number of independently calibrated values is outlined and applied to the model of a mesoscale low mountain range catchment. Copyright © 2005 John Wiley & Sons, Ltd.