DescriptionStorm surge represents a major threat for coastal communities in the United States, accounting for 50% of fatalities due to tropical cyclones (TCs) and causing significant economic losses. Cyclones along the Northeast United States have been some of the most destructive, partly due to their effect in regions with high population density. Hurricane Sandy was a high-impact event producing record-breaking storm surges around the Mid-Atlantic Bight region and causing billions of dollars in damages. Much of the impact from Hurricane Sandy is attributed to its atypical near-perpendicular angle of landfall. This event prompted the need to study a wide range of possible TC scenarios and to understand the role of atmospheric forcing in modulating storm surge. Motivated by the damages from TC-induced storm surge events, we seek to determine the sensitivity of storm surge to atmospheric forcing, in our attempt to contribute towards improved predictions and mitigation of storm surge impacts. Improvement of storm surge predictions can be accomplished by advancing and developing modeling systems, and by understanding the relation between storm surge and TC physical parameters. The work in this dissertation seeks to determine the influence of different wind models on storm surge forecasts and to assess the sensitivity of storm surge to cyclone landfall angle.
To address these goals, we perform simulations of TCs, and their associated storm surge, by coupling state-of-the-art atmospheric and hydrodynamic models, namely the Weather Research and Forecasting model and the Advanced Circulation Model. The modeling framework facilitates the use of different wind models and the creation of synthetic cyclones that provide the desired spread in TC characteristics, particularly the angle of landfall. The coupled simulations are also used to inform an artificial neural network (ANN) model on the relationship between various TC parameters and storm surge, in our attempt to make accurate storm surge predictions at various station locations around the Mid-Atlantic Bight. We show that a higher resolution atmospheric simulation is not necessary to accurately depict the storm surge magnitude and spatial extent. While the sensitivity of storm surge and inundation to the TC impact angle varies along the coast, cyclones perpendicular to the coast generally produce the largest impacts. Results also emphasize the dependency of the storm surge impact to cyclone landfall location. We successfully train the ANN model to formulate timely storm surge predictions with a mean squared error of 0.08 m, demonstrating the potential of ANNs as forecasting tools.
We develop a modeling framework that can be employed to study the fundamental mechanisms modulating storm surge. Our results have important implications in how storm surge modeling can be improved, informing us on the current limitations in storm surge assessment and on alternative methods for improved forecasts that will ultimately lead to a reduction of impacts from TC-induced storm surge.