| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 4.22 MB | Adobe PDF |
Orientador(es)
Resumo(s)
Low coastal areas are increasingly exposed to extreme meteo-oceanographic events that generate large storm surges and energetic waves, leading to wave overtopping and coastal flooding. Early Warning Systems (EWSs) can enhance short-term preparedness and mitigate the impacts of these natural hazards; however, their implementation is often constrained by the high computational cost of process-based models and by the limited availability of site-specific data required for reliable impact predictions (e.g., model devel¬opment, calibration and validation). Among the different tools available to make overtop-ping predictions, in this study, we developed a hybrid approach that combined flexible pro¬cess-based numerical modeling (XBeach) with a probabilistic approach based on Bayesian Networks. The framework was implemented at two contrasting beach profiles (strongly urbanized and semi-natural) along an embayed beach in Santa Catarina, Brazil, allowing explicit assessment of spatial variability in storm response. The Bayesian Networks acted as computationally efficient surrogates of the numerically intensive XBeach simulations, providing probabilistic estimates of storm-induced impacts as a function of offshore oce¬anic conditions at a fraction of the computational cost. To ensure robust predictive skill, the Bayesian Networks were trained using several thousand real storm scenarios and their associated hazards and impacts simulated with XBeach. Accurate model simulations of wave overtopping required quantitative information on bathymetry and overtopping dis¬charges, which were not readily available in the study area, as is often the case in data-scarce regions worldwide. This limitation was addressed by constructing an empirical profile formulation (e.g., Dean’s equilibrium profile) for beach profile reconstruction and by using non-professional imagery to estimate damage levels and infer associated over¬topping discharges. By coupling physically based modeling with data-driven probabilistic inference, the proposed framework enabled rapid, robust, and uncertainty-aware predic¬tions of storm impacts suitable for operational EWS applications. Moreover, the extensive training dataset provides meaningful insights into the variability of hazard intensity and impact severity along embayed beaches and allows identification of critical storm condi¬tions driving extreme responses. Overall, this work advances coastal hazard prediction by providing a transferable and computationally efficient methodology that explicitly ad¬ dresses data limitations, thereby supporting more effective coastal risk management and EWS development in vulnerable and data-limited coastal settings.
Descrição
Palavras-chave
Coastal storm Wave overtopping XBeach non-hydrostatic Bayesian network Santa Catarina
Contexto Educativo
Citação
Editora
Springer Nature
Licença CC
Sem licença CC