Groundwater numerical modeling is a crucial scientific tool for understanding groundwater circulation and supporting regional water resource planning and management. The effectiveness of these models depends largely on the accuracy of hydrogeological parameters within aquifers, which are often spatially heterogeneous and randomly distributed due to complex geological and tectonic factors. Traditional modeling approaches frequently overlook this randomness, compromising the precision and resolution of groundwater simulations. This study focuses on a section of the Qingshui River in the Huaihe River Basin. Using field and laboratory data, probability distribution functions for key parameters like hydraulic conductivity, specific yield, and specific storage were developed. These functions were integrated into the groundwater model to reflect the inherent stochastic nature of aquifer properties. This integration significantly enhanced model accuracy, reducing the root mean square error of simulated water levels from 0.47–1.43 m to 0.13–0.16 m and improving the Nash-Sutcliffe efficiency coefficients (NSE) from −2.96–0.73 to 0.94–0.98. Additionally, the model facilitated analysis of the interactions between river and groundwater, particularly in the hyporheic zone, under various scenarios. It identified spatial and temporal variations in groundwater recharge dynamics and delay effects at different distances from the river channel. For instance, recharge rates at 50 m and 150 m from the river were 0.295 m/day and 0.015 m/day, respectively, indicating stronger recharge closer to the river. The study also assessed the impact of varying river flows, riverbed permeability, and irrigation practices on water exchanges between the river and groundwater. These factors were found to significantly influence the intensity of water exchange, seepage, and groundwater reserves. This research provides valuable insights for managing river-groundwater interactions and analyzing the ecological environment of surrounding groundwater systems, underscoring the importance of incorporating stochastic characteristics into groundwater modeling.