Developing Probabilistic Flood Inundation Forecasts using SWMM

During hurricane season, major storms have led to significant flooding causing property loss and damage in many urban areas of the United States. There is an urgent need to assess and forecast flood inundation in detail on areas that have been severely affected. We utilize Storm Water Management Model (SWMM) to develop flood inundation forecasts and to provide a quantitative flood impact assessment. SWMM is forced with observed precipitation and 3-day ahead precipitation forecasts to develop inundation forecasts during three major hurricanes (Matthew 2016; Florence 2018; Dorian 2019) for the City of Wilmington NC. The SWMM, consisting of the city drainage network including stormwater pipelines, channels, and structures, was set up for all the sub-watersheds within the city limits to obtain flood inundation using a LiDAR-derived 3 m resolution digital elevation model (DEM). The model parameters were derived based on the properties of the sub-catchment and the stormwater conduits. Both simulated and forecasted flood inundation corroborates observed flooding patterns with maximum inundation impacts occurring near the UNC-Wilmington campus along Wrightsville Beach during Hurricane Florence. Among all storm events simulations, the eastern side near Greenfield Park had been consistently inundated, which suggested there is a need to improve the drainage system in the area. SWMM was forced with probabilistic 3-day ahead forecasts developed using three different statistical methods to develop quantitative probabilistic inundation forecasts for the three hurricanes. The reliability of these probabilistic inundation forecasts is also assessed by comparing with the simulated inundation obtained using observed precipitation. Combining fine-resolution precipitation forecasts with the urban flood model, SWMM, provides opportunities for utilizing the inundation for emergency planning that include storm network maintenance and for developing community evacuation strategies.