Storm Surge Prediction Using Subgrid Model

A tropical cyclone is an intense circular storm that originates over the ocean, progresses toward a coastline and is characterized by intense winds and low atmospheric pressure in the center. As a result of tremendous surface winds around the storm and relatively low surface pressure, seawater rises above mean tidal levels, which is called a storm surge. The most destructive component of a tropical cyclone is the storm surge. A single storm can cause billions of dollars in property damage. In 2005, Hurricane Katrina's storm surge caused catastrophic damage to the New Orleans area and its residents, amounting to nearly 125 billion worth of damage. Thus, accurate and efficient storm surge predictions are essential to prevent destruction of life and property along coastlines. Two categories of storm surge models exist: (i) low resolution and (ii) high resolution, and these models are distinguished by run times and accuracy. Low resolution models are much faster but generally have lower accuracy and vice versa for high-resolution models. Using these models is the trade-off between computational cost and accuracy. A promising intermediate path to reach models with both accuracy and low computational cost is through the use of subgrid models. Subgrid models use information from smaller scales to correct the flow variables averaged over the grid-scale. In this dissertation, upscaled forms of the shallow water equations for storm surge modeling are presented which include the subgrid corrections. Results show that, for coarse grids in complex geometries, subgrid corrections significantly improve model accuracy when compared to standard solutions. However, excessively coarse grids can lead to artificial cross flows between hydrologically disconnected areas separated by barriers smaller than the grid size. An approach based on introducing cell and edge clones, consisting of connected groups of pixels in each cell, is able to remove such artificial cross flows. Finally, the subgrid corrections are implemented in a widely used storm surge model. This work highlights how subgrid corrections can improve the accuracy of storm surge models with minimum additional computational cost.