Calibration is an essential step that is involved in the application of process based distributed hydrological models. The calibration process focuses on improving the performance index values rather than improving the process dynamics. The poor process dynamics in the model reduces its predictive power. The current study proposes a modified calibration approach (MCA) for the Soil and Water Assessment Tool (SWAT) model, which helps to improve the process dynamics in model through constraining the parameters. The modified calibration method uses the association between physically interpretable SWAT model parameters as soft data for calibration. The proposed calibration approach is validated in the Little River Experimental Watershed, USA. The results indicate that the proposed calibration method improves the process representation and spatial prediction in the calibrated model compared to the standard calibration procedure based on streamflow data. © 2021 Elsevier Ltd