A Bayesian Approach to Evaluating Environmental Risks from Maritime Accidents

Maritime accidents represent a significant source of disturbance to marine ecosystems, particularly in the context of the increasing transport of hazardous cargo. Despite advances in maritime safety, analytical tools capable of quantifying the relationship between accident-generating factors and ecological consequences within a coherent probabilistic framework remain limited. This study develops a Bayesian network model to assess the environmental risk associated with maritime accidents and to support decision-making in maritime safety and marine ecosystem protection.
A systematic literature review conducted using the PRISMA protocol informed the construction of a multi-nodal Bayesian network integrating key variables such as ship type, accident type, weather conditions, hazardous cargo, spill volume, environmental sensitivity, and response time. The model was implemented in GeNIe and evaluated through four simulated accident scenarios representing different operational and environmental configurations.
The results indicate that high-risk scenarios, such as night-time tanker collisions under severe weather conditions, can generate a 78% probability of large spill volumes and major ecological consequences. In contrast, favourable operational conditions involving small vessels, calm weather, and rapid response measures resulted in an 85% probability of low or moderate environmental impact. Sensitivity analysis identified hazardous cargo, spill volume, response time, and environmental sensitivity as the most influential variables affecting ecological outcomes.
The proposed Bayesian framework provides a transparent and flexible probabilistic tool for ecological risk assessment in maritime transport. By integrating operational, environmental, and response-related factors, the model supports predictive analysis, risk prioritisation, and the development of preventive strategies aimed at strengthening maritime safety and sustainable marine ecosystem management.