Accurate implementation of river interactions with subsurface water is critical in large-scale hydrologic models with a constant horizontal grid resolution when models apply kinematic wave approximation for both hillslope and river channel flow. The size of rivers can vary greatly in the model domain, and the implemented grid resolution is too coarse to accurately account for river interactions. Consequently, the flow velocity is underestimated when the width of the rivers is much narrower than the selected grid size. This leads to inaccuracy and uncertainties in calculations of water quantities. In addition, the rate of exfiltration and infiltration between the river and the subsurface may be overestimated as the modeled area of water exchange between rivers and subsurface is larger than reality. Therefore, the present study tests the approximation of subscale channel flow by a scaled roughness coefficient in the kinematic wave equation. For this purpose, a relationship between grid cell size and river width is used to correct flow velocity, which follows a simplified modification of the Manning-Strickler equation. The rate of exfiltration and infiltration between the subsurface and river is also corrected across riverbeds by a scaled saturated hydraulic conductivity based on the grid resolution even though the grid size is relatively large. The scaling methodology is implemented in a hydrological model coupling ParFlow (PARallel FLOW) v3.5 and the Community Land Model (CLM) v4.5. The model is applied over the Upper Rhine Basin (between France and Germany) for a time period from 2012 to 2014 and at a spatial resolution of 0.055° (∼6 km). The validity of the results is examined with satellite and in situ data through an innovative application of the First Order Reliability Method (FORM). The scaling approach shows that soil moisture estimates have improved, particularly in the summer and autumn seasons when cross-validated with independent soil moisture observations provided by the Climate Change Initiative (CCI). The results underline the use of a simple scaling procedure of the Manning coefficient and saturated hydraulic conductivity to account for the real infiltration/exfiltration rate in large-scale hydrological models with constant horizontal grid resolution. The scaling procedure also shows overall improvements in groundwater level estimation, particularly where the groundwater level is shallow (less than 5 m from the surface). By using the scaling approach, the average bias in soil moisture for the study domain was decreased from 0.17 mm³/mm³ to 0.1 mm³/mm³. The FORM results show that the probability of a substantial divergence between the ParFlow-CLM-S soil moisture results and the CCI-SM observation, which is defined as more than 0.25% of the CCI-SM observation value, is less than 0.05, 0.11, 0.15, and 0.08 for autumn, winter, spring, and summer, respectively.