A data-driven realistic parametric tropical-cyclone wind-field model with terrain-locked asymmetry for improved storm surge simulation

Storm surge hazards have intensified under rising mean sea level, increasing the need for realistic atmospheric forcing in coastal surge models. In the Northwestern Pacific, the spatial structure of tropical cyclone winds strongly controls nearshore water level response, particularly in harbors and shallow coastal zones where wind-field heterogeneity can amplify modeling uncertainty. Over regions with steep and complex topography, tropical cyclone circulations often undergo coherent and persistent asymmetric deformation due to topography–cyclone interaction, commonly referred to as the “topographic locking effect”. Such topography-modulated asymmetry is not adequately represented by conventional symmetric parametric wind models, limiting their reliability for nearshore storm surge applications.

Here we develop a realistic parametric wind-field framework (REP) that captures asymmetry associated with topographic locking using patterns extracted from a historical reanalysis library. We constructed a regional wind-field database from ECMWF ERA5 reanalysis, including sea level pressure and 10-m winds, for 282 typhoon events affecting Taiwan during 1980–2023. For a given storm location and scenario, REP quantifies the contribution of each database member through a designed weighting formula and synthesizes a physically self-consistent two-dimensional wind field via weighted blending, without requiring high-resolution dynamical atmospheric downscaling.

We demonstrate the framework using typhoon events traversing the main island of Taiwan. REP-generated nearshore wind structures are compared against ERA5, and REP winds are further coupled to the COMCOT-SS storm surge model to benchmark surge responses against simulations forced by ERA5. Results show improved consistency with ERA5 in nearshore wind patterns and comparable storm surge evolutions relative to ERA5-forced simulations. In addition, we conduct complementary experiments by replacing the underlying wind-field archive with hindcasts from a tropical-cyclone forecasting numerical weather prediction system, providing an initial assessment of REP’s transferability across databases. Overall, REP offers a computationally efficient and transferable approach to generate reanalysis-like asymmetric wind forcing, supporting storm surge modeling and hazard assessment in regions where topography-modulated cyclone structure is important.