Consistent relationships between model parameters and catchment attributes for multiple water balance models: a regionalization framework

Regionalization remains essential for predicting runoff in ungauged basins, allowing the transfer of hydrologic information from gauged catchments. This study develops a regionalization framework that links catchment attributes to the parameters of XM, GR2M, WM, and DWBM water balance models across diverse climates. Leveraging on the large CAMELS US dataset for model calibration with its extended available attributes, multiple linear regression is applied to derive parameter–attribute relationships. Parameters governing similar hydrological processes tend to correlate with comparable catchment characteristics. Among the evaluated models, GR2M and DWBM show the most robust performance across climate regions, while XM excels in hyper humid basins. Strong attribute–parameter correlations are observed for most processes, except for groundwater flow. Despite weaker groundwater related relationships, GR2M, WM, and DWBM achieved average NSE values from 0.64 to 0.72. Overall, regionalized models produce slightly reduced yet acceptable predictive skills relative to direct calibration, confirming the effectiveness of the proposed approach.