Sensitivity of extreme storm surge estimates induced by tropical cyclones to different widely used approaches

Tropical cyclones represent one of the major natural hazards for coastal regions worldwide, primarily due to the extreme storm surges they generate. Reliable estimates of storm surge associated with return periods are essential for coastal risk assessment, infrastructure design, and climate adaptation planning. These estimates, however, are highly sensitive to the characteristics of the accuracy of the input forcing and the underlying datasets used for the extreme value analysis, including their temporal length, time resolution and the representation of rare but high-impact events.

The goal of this study is to analyze this sensitivity. To this end, several tropical cyclone–induced storm surge datasets are considered for the Caribbean Sea and Gulf of Mexico region, differing in duration, structure, input forcing, and underlying assumptions. The datasets are derived using commonly adopted approaches, including a 32-year (1993–2024) continuous storm surge hindcast forced by ERA5 reanalysis wind and pressure fields, as well as event-based storm surge simulations generated using a parametric Holland wind model for both historical and synthetic tropical cyclones. The historical hurricane dataset is analysed for the full period of available records (1851–2024) and separately for the period overlapping with the ERA5 hindcast (1993–2024), enabling a consistent comparison across datasets. Return level curves obtained from these datasets are compared to evaluate the sensitivity of extreme storm surge estimates to dataset length, input forcing, and the inclusion of synthetic events. The results provide valuable insights into the uncertainties affecting storm surge return period estimates and emphasize the importance of carefully selecting datasets when assessing tropical cyclone–induced coastal hazards.