Bayesian network for the analysis of significant storms in the southern Baltic Sea
Oceanologia · 2026 · 68 (4)
DOI: 10.5697/wslv6710 ↗Kurz gefasst
Fachartikel in der Zeitschrift Oceanologia, 2026. Verfasst von Konrad Furmańczyk und 2 weiteren Autor:innen.
Abstract (Originalsprache)
Coastal emergency managers increasingly require reliable tools to enhance preparedness for extreme coastal storms and their impacts. With the growing availability of methods and frameworks for the development of Early Warning Systems (EWSs) for coastal storm hazards, accurate prediction of dune erosion remains a fundamental challenge. In this context, probabilistic modeling approaches, such as Gaussian Bayesian networks, provide a flexible framework for integrating multiple storm parameters and quantifying erosion forecasts. In this study, the interdependence of the parameters of significant storms is investigated via an expert Gaussian Bayesian network. The network was trained using data from 1978-2009, during which 32 significant storms were recorded. The model’s predictive ability was evaluated for eight additional storm events from 2010-2017, which were not considered during model estimation. In a second validation approach, the leave-one-out (LOO) cross-validation method was applied to the entire dataset of 40 storm events to evaluate the predictive ability and stability of the Bayesian network model parameters. Based on the proposed expert Gaussian Bayesian network, satisfactory forecasts of dune erosion magnitude, expressed as log(D), were obtained. Storm energy was modeled as an event-scale descriptor of storm intensity and included in the Gaussian Bayesian network framework for dune erosion prediction. Model estimation for log(D) yielded satisfactory results, and the coefficient of determination (R²) indicated a relatively good agreement between the modeled and observed values.
Zitieren
APA
Furmańczyk, K., Bugajny, N., & Furmańczyk, K. (2026). Bayesian network for the analysis of significant storms in the southern Baltic Sea. Oceanologia, 68(4). https://doi.org/10.5697/wslv6710
Vancouver
Furmańczyk K, Bugajny N, Furmańczyk K. Bayesian network for the analysis of significant storms in the southern Baltic Sea. Oceanologia. 2026;68(4). doi:10.5697/wslv6710
BibTeX
@article{Furmaczyk2026,
author = {Furmańczyk, Konrad and Bugajny, Natalia and Furmańczyk, Kazimierz},
title = {Bayesian network for the analysis of significant storms in the southern Baltic Sea},
journal = {Oceanologia},
year = {2026},
volume = {68},
number = {4},
doi = {10.5697/wslv6710}
}
EUVIMED