Existing real-time coastal flooding guidance systems in the US tend to underestimate total water level (TWL) predictions in upstream tidal areas of the Chesapeake Bay rivers, impacting flood forecasts for highly vulnerable areas, such as the National Capital Region. These under-predictions are mostly due to missing physical processes, lack of integration between hydrological and hydrodynamic models, and an oversimplification of the model setups used to predict TWL. In this study, an integrated TWL forecast system was introduced, where a high-resolution two-dimensional coastal storm surge model (ADCIRC) was implemented to simulate the combined influence of various flood drivers (storm tide, river flows, urban runoff, and local wind forcing) in the Potomac River. In this framework, the downstream boundaries of storm tide predictions are provided by existing coastal guidance systems, whereas, streamflow forecasts at upstream rivers and local urban runoff are provided by the National Weather Service and the National Water Model. Additionally, high-resolution wind fields from the North American Mesoscale and the National Blend of Models are added to account for local wind effects on TWL. This model setup was successfully validated with a range of historical events and it also demonstrated improved forecast performance against the existing large-scale coastal guidance systems in a reforecast evaluation during 2020. Unlike other studies, we provided a comprehensive evaluation on the influence of individual flood drivers on TWL modeling and clearly demonstrated that the absence of one or more flood drivers in the model framework can underestimate simulated TWL in the National Capital Region.