Coastal risk assessment and hazard forecast analysis via a Bayesian network

Abstract

In this study, a Bayesian network (BN) module is proposed for coastal hazard risk assessment and forecasting based on the statistical prior and conditional probabilities, which capture the mechanisms of transitivity and causality. The BN module is capable of quantifying the probabilities of various risk scenarios, thereby effectively modeling and interpreting natural phenomena. Rather than focusing on isolated variables, this BN-based risk assessment quantifies the causal interrelationships among multiple factors. Using long-term coastal, environmental, and socio-economic data from Taitung city, located along the coast of Taiwan, the causal models and conditional probabilities are applied to forecast hazard risks under a 2 °C global warming scenario. The predictive accuracy of the module is approximately 74%, demonstrating its substantial practical value for coastal planning. Causality verification indicates a strong correlation (R2 ≈ 0.81) between wave energy and the shoreline change rate. Under the 2 °C global warming scenario, the hazard risk grade for wave energy-propagated shoreline change is projected to increase in both the estuary and sediment-supply-terminated sectors; specifically, the maximum probability of a high hazard risk grade increases to 60% in the estuary sector of the Bainen River. Furthermore, the risk grades associated with shoreline change rates become more polarized due to climate change, identifying the estuary and sediment supply termination sectors as erosion hotspots. Ultimately, hazard forecasting via BNs provides an informative and well-founded framework for sustainable coastal management.