Real-time high-resolution forecasting of the coastal ocean during a hurricane

Alexander Rey, Queen's University, Civil Engineering, Kingston, ON, Canada and Ryan P Mulligan, Queen's University, Kingston, ON, Canada
Abstract:
The coastal ocean experiences dynamic changes during severe weathers event such as hurricanes, and forecasting these conditions is challenging. While existing forecast models offer considerable guidance, they provide relatively low resolution in shallow coastal areas and do not typically forecast wind- and wave-driven currents. This limits their application for nearshore research, such as deploying sensors at specific field sites prior to a storm event. To address these challenges and provide high-resolution (100 m) coastal forecasts for the DUring Nearshore Event eXperiment (DUNEX) research community, a real-time (RT) modelling system was developed. The RT model runs every 6-hours using Delft3D-SWAN to provide a 36-hour forecast of the significant wave height, depth-averaged current velocity, and water levels from combined tide and storm surge for an ideal test site in coastal North Carolina, USA. The domain includes the shelf and coast of the Outer Banks, tidal inlets, and the large back-barrier Albemarle-Pamlico estuarine system. Detailed conditions from large scale forecasts are provided by the National Oceanic and Atmospheric Administration (NOAA) and used as model inputs, including atmospheric conditions (winds, pressure, precipitation) at 3 km resolution, ocean wave boundary conditions, and water level boundary conditions. The real-time model results are communicated via a website with past and present forecasts shown together with observations at 10 sites in the ocean and estuary. This allows the effects of changes to the forecast hurricane track and intensity to be visualized and the model performance to be validated in real-time, thereby communicating the accuracy of the model. The performance of the RT model for Hurricane Dorian in September 2019 is statistically assessed by examining differences between observations, forecasts runs, and a hindcast run with the best available input data.