ADCIRC Model Verification of Tropical Cyclone Storm Surge Along the Louisiana Gulf Coast

Storm surge is the deadliest and costliest threat associated with hurricanes, therefore, considerable effort has been invested in forecasting coastal storm surge from tropical cyclones using specialized computer models such as ADCIRC (Advanced Circulation) and SLOSH (Sea, Lake, and Overland Surges from Hurricanes). Currently, there is no such thing as a perfect model so forecasting storm surge in advance is a difficult if not impossible task. In an effort to better predict storm surge heights, ADCIRC uses a varied resolution mesh grid to save computing power away from the coast (100 km) while providing high resolution (100 m or less) in coastal regions. ADCIRC is preferred over SLOSH from an accuracy standpoint because ADCIRC is able to use a finer resolution (as fine as 20 m) and accounts for finer more acute details. However, due to the increased accuracy and resolution, ADCIRC inherently takes longer to run and uses far more computing power so the National Hurricane Center uses the SLOSH model for time efficiency. The slowness with ADCIRC is constantly improving and is becoming more efficient or simply runs on less resolved grids. The goal of this study is to evaluate verified storm surge heights from specific tropical cyclones in Louisiana, including Betsy, Andrew, Katrina, Rita, Gustav, Isaac taken from the Surgedat database provided by LSU, to the values generated from the Advanced Circulation (ADCIRC) storm surge model to assess the accuracy. The data for the model is gathered from the UNC RENCI data suppository and converted into files compatible for ADCIRC to run. The model is then run through a remote access application (design-safe). After, an interpolation is applied to the ADCIRC model output through python to display an accurate gradient field of estimated storm surge. The actual high water marks are gathered from the SURGEDAT database and plotted on a map with the ADCIRC gradient display to provide a visual of how the model compares to actual high water marks. Next, a spearman correlation test is used to statistically evaluate the accuracy of the ADCIRC model run. This analysis uses the maximum elevation of each element in the model during the passage of the storm. Descriptive statistics, correlation coefficients, confidence intervals, mean normalized bias, root mean square error, mean error, mean absolute error, mean normalized error, and scatter index among possible other factors are all used to evaluate the models accuracy. These results will display how and where the ADCIRC model can be improved in the future. This knowledge of error will help researchers and storm surge modelers improve the inner workings of the ADCIRC model to assist in the forecasting of storm surge before a hurricane makes landfall. This extra accuracy will allow coastal communities to be more prepared for an impending storm.