Maritime navigation accidents and risk indicators: An exploratory statistical analysis using AIS data and accident reports

Abstract

This paper presents the results of statistical analyses of maritime accidents data and AIS data from Norwegian waters, to identify conditions that are associated with navigation related accidents (groundings and collisions) and could be used as risk indicators. Vessels involved in accidents reported in the accident database of the Norwegian Maritime Directorate (NMA), have been traced in historical AIS records, and data related to each ship have been transformed into variables. These variables are related to the behavior of the ship in front of the accident (e.g. nautical miles sailed, hours in operations, number of port calls etc.), technical and organizational conditions (ship categories, flag state, age, gross tonnage, Paris MoU ratings etc.) and the area where the accident occurred (number of vessels in the area, port calls in the area, nautical miles in the area etc.). Both the AIS data and the data from the NMA accident database have first been analyzed using correspondence analysis (categorical variables), F-tests (continuous variables), and then combined in a multivariate logistic regression model with “navigation accidents” and “other accidents” as dependent variables. The model is a strong predictor for whether the accident is navigation-related or not. Specifically, some vessel types, less vessel length, poor visibility condition, and a flag of convenience increased this probability.