A regional parameter estimation scheme for a pan-European multi-basin model

• Model parameter regionalization for prediction in ungauged basins across Europe.
• Model parameters can be linked to catchment descriptors with good transferability.
• Regionalization within homogeneous catchments improves estimation in a few groups.
• North-south gradient in model prediction skill, decreasing towards south.

Study regionEurope.Study focusA semi-distributed continuous hydrological model, HYPE, was applied to model daily stream flows in more than 35,000 subcatchments across Europe. A stepwise regionalization approach was implemented to estimate different groups of model parameters. HRU based parameters were estimated first for each soil and landuse class, respectively. Lake and reservoir parameters were estimated separately. Catchments were grouped based on similarity of their characteristics and model parameters defined at a catchment scale were then regionalized for each group as functions of the catchment characteristics by simultaneously calibrating the model for a number of catchments to concurrently optimize the overall model performance and the functional relationships between the parameters and the catchment characteristics. Calibration was performed at 115 discharge stations and the approach was validated at 538 independent stations.New hydrological insights for the regionParameters could be linked to catchment descriptors with good transferability, with median NSE of 0.54 and 0.53, and median volume error of −1.6% and 1.3% in the calibration and validation stations, respectively. Although regionalizing parameters for different groups of catchments separately yielded a better performance in some groups, the overall gain in performance against regionalization using a single set of regional relationships across the entire domain was marginal. The benefits of separate regionalization were substantial in catchments with considerable proportion of agricultural landuse and higher mean annual temperature.