Relative accuracy of HWRF reanalysis and a parametric wind model during the landfall of Hurricane Florence and the impacts on storm surge simulations

Abstract

Prediction and reanalysis of storm surge rely on wind and pressure fields from either parametric tropical cyclone wind models or numerical weather model reanalysis, and both are subject to large errors during landfall. This study assesses two sets of wind/pressure fields for Hurricane Florence that made landfall along the Carolinas in September 2018, and appraises the impacts of differential structural errors in the two suites of modeled wind fields on the predictive accuracy of storm surge driven thereby. The first set was produced using Holland 2010 (H10), and the second set is the Hurricane Weather Research and Forecasting (HWRF) reanalysis created by the NWS National Centers for Environmental Prediction (NCEP). Each is validated using a large surface data set collected at public and commercial platforms, and then is used as input forcing to a 2-D coastal hydrodynamic model (Delft3D Flexible Mesh) to produce storm surge along the Carolina coasts and major sounds. Major findings include the following. First, wind fields from HWRF are overall more accurate than those based on H10 for the periphery of the storm, though they exhibit limitations in resolving high wind speeds near the center. Second, applying H10 to the best track data for Florence yields an erroneously spike in wind speed on September 15 th when the storm reduced to a tropical depression. Third, HWRF wind fields exhibit a progressively negative bias after landfall, likely due to deficiencies of the model in representing boundary layer processes, and to the lack of assimilation of surface product after landfall for compensating for these deficiencies.Fourth, using HWRF reanalysis as the forcings to Delft3D yields more accurate peak surges simulations, though there is severe underestimation of surge along the shoreline close to the track center. The peak surge simulations by Delft3D are biased low when driven by H10, even though over several locations the H10 model clearly overpredicts surface wind speeds. This contrast highlights the importance of resolving wind fields further away from the center in order to accurately reproduce storm surge and associated coastal flooding.

Determining the Relative Contributions of Runoff and Coastal Processes to Flood Exposure across the Carolinas during Hurricane Florence

“…Third, we did not investigate the uncertainty in the model forcings but used the best available products. The meteorological data (e.g., wind, rainfall) can contribute to model errors when hindcasting hydrodynamic processes resulting from TCs (Rahman et al, 2022;Ratcliff, 2022). For example, the temporal and spatial resolution of the data applied can impact the timing and volume of runoff computed by hydrologic models (Quintero et al, 2022).…”

A novel deep learning approach for typhoon-induced storm surge modeling through efficient emulation of wind and pressure fields

“…Parametric models have been reported to work well for storm surge simulations 15 , 16 . Nevertheless, poor predictive skills are typically exhibited in areas far from the typhoon center 8 , 17 or during the typhoon transitioning stage into an extratropical cyclone and landfall due to topographic effects 18 . Similarly, standard mesoscale NWP models also exhibit some drawbacks partly caused by the insufficient grid resolutions for resolving a typhoon’s intensity 19 .…”

“…Recent atmospheric and computer science developments have led to substantial improvements in the NWP models using a finer NWP model grid resolution 20 or a data assimilation scheme 21 to improve the typhoon structure resolvability. Therefore, many recent studies advocated using NWP models for storm surge simulations 8 , 9 , 17 , 22 .…”