Comparison of a reduced-complexity model to a full model for storm surge predictions

The ability to anticipate the changes in water levels, waves and current velocities associated with storms is critical to determining the storm damages including morphological changes and coastal structure interaction. Typically, storm surge forecasts are generated using complex modeling systems such as ADCIRC, COAWST or Delft3D which solve the Navier-Stokes equations driven by the wind and wave forcing. These models often take time and effort to set up and needs significant computational resource to produce results at the required resolutions. The advantages are obvious – all the relevant physics are represented with a high degree of accuracy in these models. However, the accuracy of the models is often overshadowed by the uncertainties in the forecast of the storm itself. To capture the effects of these uncertainties, we need to resort to ensemble simulations, which brings us to the main disadvantage of these systems – they require significant computational effort to execute even one of the scenarios. Thus, to get timely information about the surge, it is necessary to either reduce the number of members in the ensemble or reduce the resolution at which the model simulates the event, thereby reducing the confidence in the model results. Here we investigate an alternative approach to storm surge predictions – use a reduced complexity model to compute the surge and compare the results to a full model as well as to data to assess the effectiveness of the models. As a case study, we will compare the two approaches using the forecasts from Hurricane Ida (2021) which impacted Louisiana. We will use the Delft3D FM system as representative of the full physics model and compare the results to that produced by SFINCS, which is a reduced complexity model. Comparisons of water levels at available NOAA tide stations are used to validate the model and quantify errors in the system. Wave statistics are compared against available buoys.