A comparison between parameter regionalization and model calibration with flow duration curves for prediction in ungauged catchments

Abstract. Streamflow prediction in ungauged catchments is essential for hydrological applications because most catchments in the world are ungauged. The flow duration curve (FDC) has received increasing attention as a criterion for calibration of runoff models in ungauged catchments; however, performance of the FDC-based model calibration was barely evaluated in comparison with parameter regionalization. In this study, we conducted a comparative assessment between the conventional proximity-based parameter regionalization and the model calibration with regional FDCs. Leave-one-out evaluations of the two methods showed that the proximity-based parameter regionalization outperformed the model calibration with regionalized FDCs in hydrograph prediction. No information of flow timing in FDCs seems to cause high uncertainty in flow predictions. The relative merits of the model calibration with regional FDCs were low bias in predicted flows and high reproducibility of the baseflow index. From the evaluation of flow signature reproducibility, it was indicated that the rising limb density is likely to be orthogonal to the FDCs, and thus additional conditioning with the rising limb density is expected to improve the model calibration with regional FDCs.