Runoff Model Calibration and Regionalization to a Disturbed Basin Under Various Potential Evapotranspiration Estimation Methods

Droughts adversely affect human life, economic activities, and ecosystems, For this reason, considerable efforts have been made to prepare for and mitigate their impacts. This study examines the natural runoff estimation in a disturbed basin by regionalizing a runoff model calibrated for an undisturbed basin. Four potential evapotranspiration (PET) estimation methods—Blaney-Criddle, Hargreaves, FAO Penman-Monteith, and Priestley-Taylor—were compared. Model performance was evaluated under two types of parameter range settings in automatic calibration: theoretically possible ranges (TPR) and empirically applicable ranges (EAR). The Seomijingang River basin was selected as the study area. The grid based rainfall-runoff model (GRM) model was calibrated using daily inflow data to the Seomjingang dam and subsequently regionalized to the downstream basin to estimate its natural runoff while considering water use circulation. Although the optimized parameter settings differed according to the calibration conditions, model performance during calibration was generally consistent across cases. Differences among the four PET methods in terms of the estimated natural runoff through regionalization were minimal. However, EAR-derived parameters performed slightly better in the simulation results. These findings suggest that any of the four PET estimation methods can be applied for model calibration using optimization techniques. However, regionalizing the optimized model by employing EAR, which considers the potential for model overfitting, provides more reliable runoff simulations. Therefore, the procedures and methods proposed in this study can be utilized when regionalizing a rainfall-runoff model calibrated using an automatic optimization method.