Due to the unique geographic location, complex navigation environment and intense vessel traffic, a considerable number of maritime accidents occurred in estuarine waters during recent years which caused serious loss of life, property and environmental contamination. Based on the historical data of maritime accidents from 2003 to 2012, which is collected from Shenzhen Maritime Safety Administration, this paper conducted a risk analysis of maritime accidents by applying Bayesian network and fault tree analysis. First a Bayesian network model was introduced to describe the consequence of accidents based on the accident investigation report. Then fault tree analysis was applied to estimate the probability on the basis of accident statistics and ship traffic flow. Finally the risk of maritime accidents in Shenzhen Waters was depicted through the consequence multiplied by the probability of an accident.

Oil spills from maritime activities can lead to very extensive damage to the marine environment and disrupt maritime ecosystem services. Shipping is an important activity in the Northern Baltic Sea, and with the complex and dynamic ice conditions present in this sea area, navigational accidents occur rather frequently. Recent risk analysis results indicate those oil spills are particularly likely in the event of collisions. In Finnish sea areas, the current wintertime response preparedness is designed to a level of 5000 tonnes of oil, whereas a state-of-the-art risk analysis conservatively estimates that spills up to 15000 tonnes are possible. Hence, there is a need to more accurately estimate oil spill scenarios in the Northern Baltic Sea, to assist the relevant authorities in planning the response fleet organization and its operations. An issue that has not received prior consideration in maritime waterway oil spill analysis is the dynamics of the oil outflow, i.e. how the oil outflow extent depends on time. Hence, this paper focuses on time-dependent oil spill scenarios from collision accidents possibly occurring to tankers operating in the Northern Baltic Sea. To estimate these, a Bayesian Network model is developed, integrating information about designs of typical tankers operating in this area, information about possible damage scenarios in collision accidents, and a state-of-the-art time-domain oil outflow model. The resulting model efficiently provides information about the possible amounts of oil spilled in the sea in different periods of time, thus contributing to enhanced oil spill risk assessment and response preparedness planning.