This paper summarizes the results of Phase 3 of the Arctic Tanker Risk Analysis Project (ATRA) which provided a prototype Tanker Navigation Safety System (TNSS). TNSS is a ship-board risk management system capable of route planning and decision support based upon a knowledge database. The objective of the project was to provide timely risk assessment information to a mariner or decision-maker in a system capable of integration with existing ECDIS or shipboard PC systems. The historical information was to include: accident location, frequency and type, ice, wind, visibility, environmental sensitivity and other factors. The specifications of the TNSS prototype were expanded so that risk could be assessed for each track in a route plan by applying a predictive accident model patterned after the navigation and collision avoidance process.

Marine accident frequencies reported in 20 different sources have been studied and compared. Further than this, comparisons were made with an analysis done by the present authors. There seems to be consistency within the sources in the use of the terms total loss, casualty and accident/incident. The rates were observed to decrease by an order of magnitude going from accident/incident to casualty and likewise from casualty to total loss. The overall frequencies were found to be in the range of 0·0009 to 0·07 total losses per 106 ship miles, 0·03 to 1 casualties per 106 ship miles, and 0·5 to 13 accidents/incidents per 106 ship miles. It was found that the frequency depends on visibility, brightness, geographical environment, age of vessel and size of vessel. Collision and grounding frequencies were found to increase with decreasing visibility, brightness and more restricted waters. Collision frequencies were found to increase with increasing size of vessel. Fire/explosion and structural damage frequencies were found to increase with increasing age, and collisions to decrease with increasing age. No firm trend was found from the effect of flag state or type of vessel.

State-of-the-art avionics achieves very good reliability, and the electronics in the current generation of communication and navigation satellites is sufficiently good to allow a design-life of around 10 years. Nevertheless, failures will arise. The purpose of failure analysis is to identify the consequences if a particular module does encounter a failure, and to ensure that, by system redundancy or through operational procedures, the effect and/or its probability of occurrence will be acceptably safe. This paper is a contribution to the discussion on the strength and weakness of redundancy in satellite-based Communication, Navigation and Surveillance (CNS) systems as envisaged by the International Civil Aviation Organization (ICAO) to be an integral part of the future air navigation system. This particular paper makes a comparison with today's terrestrial-based systems.

The collision between the dredger Bowbelle and the pleasure cruiser Marchioness on the Thames in August 1989 in the vicinity of the centre of London has drawn attention to the need for river mariners to keep a frequent look-out astern as well as ahead. This paper describes calculations of the minimum frequency at which observations astern are required in order to allow sufficient time for a ship about to be overtaken by a faster vessel to be able to take adequate avoidance action should the overtaking craft fail to keep clear.

Meeting encounters, in which it is uncertain at what range the give-way ship will manoeuvre, are analysed using a model based on probability theory. Course alteration ranges of both giveway and (if necessary) stand-on ships are closely fitted by a simple mathematical model, which is also used to interpret the efficiency of the Collision Regulations.