When a hurricane threatens North Carolina, researchers use computational models to predict how the ocean waters will rise, and what areas will be flooded.  Emergency managers rely on fast and accurate storm surge predictions from these models to make decisions and estimate damages during storm events.  These models use unstructured, finite-element meshes to describe the ocean, continental shelf, coastline, floodplains and channels over a range of spatial scales.  One of the challenges for such models is providing a high level of resolution along the coast without significantly increasing the computational time. In this presentation, we will describe how these models are constructed, discuss their performance during recent storms including Matthew (2016) and Florence (2018), and share new results to improve their efficiency in real-time.