Masan Bay is a semi-enclosed coastal embayment where external riverine loads and internal sediment loads interact, leading to recurring water quality problems each summer. In this study, (1) a hydrodynam-ic model was applied to reproduce the ocean cir...
Masan Bay is a semi-enclosed coastal embayment where external riverine loads and internal sediment loads interact, leading to recurring water quality problems each summer. In this study, (1) a hydrodynam-ic model was applied to reproduce the ocean circulation and temperature of Masan Bay, (2) a phospho-rus circulation box model was used to simulate the spatiotemporal variation of phosphorus, and (3) a machine learning model was employed to predict and reproduce COD concentrations. The model was developed and validated for 2022, the driest year (1,050 mm of rainfall), and 2023, the wettest year (2,161 mm) during the past seven years (2018–2024). Comparison with observations demonstrated that the model successfully captured the seasonal variability of total phosphorus (TP) and COD. Scenario simulations showed that TP largely satisfied the target concentration (0.031 mg/L), whereas COD failed to meet the target (2.10 mg/L) under realistic conditions. In particular, exceedance of COD was pronounced during August–September due to high rainfall and internal processes, indicating the necessity of differen-tiated seasonal targets. Furthermore, reducing wastewater treatment plant loads alone had limited effects on improving water quality in the inner bay, while sediment load reduction within the inner bay was most effective in lowering TP and COD. This study highlights the importance of simultaneously managing both external and internal loads and supports the feasibility of season-specific targets, providing a scientific basis for designing more effective water quality management strategies in Masan Bay.