Temporally Coherent Modeling of Compound Tropical Cyclone Flooding and Its Role in Extreme Water Level Estimation

Estimates of extreme coastal flooding depend critically on how interacting processes are represented within flood models. For tropical cyclones, storm surge and wave-driven run-up evolve on different time scales and rarely peak simultaneously, yet many statistical and hydrodynamic flood analyses reduce events to static maxima or impose simplified dependence structures. These modeling choices implicitly determine which combinations of processes are treated as plausible extremes, but their influence on inferred flood behavior is rarely examined explicitly.

Here, we investigate how assumptions about temporal structure affect the characterization of compound tropical cyclone flooding by focusing on how alternative representations of storm evolution modify extreme water level estimates and their interpretation. To this end, we employ the Multivariate Spatio-Temporal Maxima with Temporal Exposure (MSTM-TE) framework by Sando et al. (2024) [1] as a diagnostic and generative framework for reconstructing and simulating storm time series under different assumptions about temporal coherence among metocean drivers.

The analysis is based on a 1000-year synthetic tropical cyclone dataset for the Guadeloupe archipelago (French Antilles), which enables direct comparison between modeled extremes and reference behavior across multiple coastal sites with contrasting exposure conditions. To mimic realistic data constraints, only a limited subset of storm events corresponding to a 50-year period is used for statistical calibration, while the remaining events are retained to evaluate the consequences of modeling assumptions for extreme flood characterization. Extreme total water levels are derived from reconstructed storms at multiple coastal sites with differing exposure to tropical cyclones, using analytical wave run-up formulations that allow surge and wave contributions to be examined jointly in time.

Results show that assumptions about temporal structure play a dominant role in determining both the magnitude and variability of estimated extreme water levels. Approaches that neglect temporal coherence tend to promote unrealistically aligned surge–wave combinations, leading to inflated return levels and ambiguous physical interpretation. In contrast, reconstructions that preserve temporal structure yield narrower uncertainty ranges, reduce upward bias in return-level estimates, and reveal distinct site-specific flood-generating mechanisms. At wave-exposed locations, the upper tail of total water levels is associated with short-lived peaks in wave energy, whereas at more sheltered sites, extreme flooding arises from the coincidence of elevated surge with moderate run-up rather than from either component in isolation.

By explicitly linking modeling assumptions to changes in flood extremes, this study highlights temporal structure as a key source of uncertainty in compound flood analysis. The results demonstrate how spatio-temporal reconstruction frameworks can be used not only to estimate extremes, but also to diagnose the physical plausibility of modeled flood scenarios, offering insights that are directly relevant for flood risk assessment in data-limited coastal regions.

 

[1] Sando, K., Wada, R., Rohmer, J., & Jonathan, P. (2024). Multivariate spatial and spatio-temporal models for extreme tropical cyclone seas. Ocean Engineering, 309, 118365. https://doi.org/10.1016/j.oceaneng.2024.118365