Uncertainties in climate change and precipitation patterns reduce the predictability of groundwater and surface water management. In the process of constructing groundwater flow models for existing groundwater management, large-scale watersheds are typically the primary focus. However, in watershed environments with complex boundary conditions where coastal lagoons and rivers coexist, it is necessary to establish conditions distinct from those of large-scale watersheds. This study, targeting the Songji lagoon watershed on the east coast, determined that defining boundary conditions in small-scale basins can significantly influence the predictive stability of the model and the results of sensitivity analysis. It quantitatively investigated the impact of boundary condition settings on the calibration and predictive accuracy of groundwater flow models. To this end, nine scenarios were constructed by stepwise combination of hydraulic conductivity, river, lake, and drainage conditions. Steady-state simulations using MODFLOW and parameter calibration based on PEST were performed to simulate groundwater flow for each scenario. Analysis results indicated that the scenario employing a stepwise calibration method, where river and lake water levels (stage) and conductance were set first, yielded the most effective water level prediction accuracy (R² = 0.998 and RMSE = 0.138). This demonstrates that, in small coastal basins, a boundary-focused calibration strategy is an effective approach that reduces spatial uncertainty and enhances model reliability compared to traditional parameter calibration.

Keywords: boundary-focused calibration, coastal lagoon, small watershed, groundwater flow, MODFLOW