<p>Soft data, defined as long-term, process-specific information not directly measurable at the watershed scale, can provide valuable constraints for hydrological modeling. This study explores how such data can improve groundwater recharge simulation. Three calibration strategies were compared: hard calibration (HC), which uses only observed time series; multi-objective calibration (MOC), which integrates soft and hard data into a single objective function; and soft-and-hard calibration (SHC), a two-step approach that uses soft data to constrain the parameter space prior to HC. Based on 20 years of hydroclimate, water quality, and chloride input data from two Ohio watersheds (Maumee and Muskingum), watershed-scale mean annual groundwater recharge was estimated via the chloride mass balance (CMB) method. CMB-estimated recharge rates were 348.5 ± 27.4 mm × year⁻¹ for Maumee and 302.4 ± 25.9 mm × year⁻¹ for Muskingum, accounting for 36.5 ± 0.3% and 30.0 ± 0.3% of mean annual precipitation, respectively. These estimates were then incorporated as soft information to calibrate the abcde-snow model. The SHC approach revealed strong parameter interdependencies, with parameter <i>c</i> accurately predicted from <i>a</i> and <i>b</i> (coefficient of determination = 0.99) via support vector regression. Results showed that HC produced the best streamflow and baseflow simulations, followed by MOC, while SHC achieved the highest accuracy and lowest uncertainty in recharge estimates. Seasonal analysis showed spring recharge peaks driven by snowmelt following winter snow accumulation. This study highlights the value of soft data in revealing parameter interdependencies and enhancing the consistency and process fidelity of hydrological model simulations.</p>