Wednesday, 31 January 2024: 5:45 PM
343 (The Baltimore Convention Center)
Brian O. Blanton, Renaissance Computing Institute, Chapel Hill, NC; University of North Carolina at Chapel Hill, Chapel Hill, NC; and R. Luettich, S. Bunya, and Z. Cobell
Coastal storm surge predictions are needed to inform decision making about resource positioning, road closures, and related emergency management issues, particularly during tropical cyclone events. The ADCIRC Prediction System (APS) is routinely used to make these predictions, usually in its 2-D barotropic mode. The first-order drivers of the coastal water levels are typically astronomical tides and the synoptical and cyclone-driven meteorology, but other physical factors (e.g., offshore baroclinic processes, steric variations, large-scale meteorological systems) can affect water levels are not included in the 2-d barotropic model forcing. Impacts of these lower-frequency drivers, relative to tide and storm surge time scales of 6-24 hours, manifest themselves as biases in the errors and with significant spatial variability.
The information in the biases can be assimilated into ADCIRC to dynamically correct for the unmodeled lower-frequency drivers and compute an improved water level prediction, based on a data assimilation recently developed for ADCIRC. Observed water levels at NOAA gages are compared to prior predictions from a sequence of nowcasts, with the averaged errors then fitted to a surface defined on the ADCIRC grid. The surface is subsequently used in the forecast cycle to adjust the solution in a dynamically consistent way. We describe the real-time implementation of the assimilation approach with examples of performance and skill from recent North Atlantic hurricane events. Results demonstrate that the data assimilation method improves forecast error skill without substantially burdening the nowcast / forecast computational sequence.

- Indicates paper has been withdrawn from meeting

- Indicates an Award Winner