In shipping, collision risk is a serious safety threat. Risk probability estimations used for policymaking are derived from traffic density statistics, largely ignoring the decision-making process on board. Conflict detection and resolution on board is done using rather rudimentary but effective mental-model-based techniques. In this article, the authors analyse traffic using the concept of complexity. The actual geometry of the ships involved in the conflict defines how well the crews on board can resolve the conflict. This geometry is transformed into a complexity value. A reliable detection and resolution of conflicts by human operators decreases in certain situations. A previous study has shown that when complexity reaches a threshold, the risk of a near miss increases significantly. In this study, three actual collisions at open sea are analysed. It will be shown that situations of high complexity, which decreases human reliability, can be predicted well in advance, allowing for a safe resolution. The technique also allows for alerting and a decision support for the crew.

Situation awareness (SA) is generally seen as a mental representation of the system state, an objective measure of the ‘situation out there’. In this article, the authors make an argument that SA can only have a meaning in relation to the task of the user and characteristics of the system. This will be argued with the help of a specific environment: vessel traffic monitoring. The long-time constants and the complex constraints imposed on the ship require that the operator monitoring the traffic has a good SA: the operator must make long-term predictions about possible traffic developments. For this, being in control and having SA are inseparable characteristics of the same process.

A ship’s safety domain is a principal element in conflict detection and avoidance manoeuvres and consequently collision risk. Conflict detection depends on the ability to predict the intrusion of another vessel into the safety domain of the own-ship. To resolve a conflict the own-ship must manoeuvre so that the intruder remains outside this domain. In this study the shape, size and realisation of a safety domain on the North Sea are examined using Automatic Identification System (AIS) data collected from the Dutch sector of the North Sea (March–May 2014). In this study, encounters between ships over 45 m in length and travelling at 5–25 knot were analysed. Five hypotheses were tested: (1) the safety domain is elliptical rather than circular; (2) the domain size is a function of the ship speed; and (3) ship size has no effect on domain size when distances are measured hull-to-hull; (4) safety domains are created using minimal course changes; (5) those course changes are based on CPA calculations. The empirical results confirm that domains can be described by ellipses. The major-axis length increases with speed while the minor-axis is independent of speed, and ship size does not influence domain size as defined significantly. Analysis of manoeuvring behaviour shows that bridge teams seem to rely on ARPA-derived CPA and TCPA calculations, initiating avoidance manoeuvres typically between 600 and 1800 s before the predicted closest point of approach. Most course changes are modest and aim to maintain a CPA that corresponds to the empirically derived elliptical domain. The study suggests that an elliptical safety domain can be defined for vessels operating in deep-sea conditions. Such a domain would provide a practical basis for improving conflict-detection algorithms, decision-support tools for bridge teams, and autonomous-ship navigation systems while remaining consistent with the COLREGs.