Assessing the performance of single and multi-criteria calibration approaches for hydrological modelling: a comparative analysis

Accurate hydrological modelling is crucial for understanding natural processes and managing water resources. However, simulation accuracy depends on the availability of field observations for calibration and validation. It is therefore critical to develop effective calibration strategies to reduce prediction uncertainties. This study applies the DREAM model to the experimental basin of Fiumarella of Corleto in Southern Italy to assess the benefits of single and multicriteria calibration approaches. The former uses total runoff; the latter optimizes total runoff, baseflow, and annual water balance. The study also compares uniform or spatially-based parameterization, including correction factors and recession constants. Parameters were optimized through automatic calibration using a genetic algorithm (GA) and the Kling-Gupta efficiency (KGE) as the objective function. Results show that spatially distributed information improves model reliability compared to a uniform parameterization set-up. The multi-objective calibration constrained on baseflow and balance allowed us to optimize the model, reducing variability compared to mono-objective calibration.