Pre-study on AIS ‘big data’ analyses for identification of tank cleaning operations at sea

Report, 2023

Following an increase in transport of chemical substances in liquid bulk, e.g. of biofuels, the Swedish Coast Guard are detecting an increasing number of chemical pollution events in Swedish waters and the Baltic Sea. The regulation of tank cleaning is complex and leaves room for different interpretation, yet also legal tank cleaning operations may pose a threat to the marine environment. To carry out environmental risk assessment there is a need to know when and where tank cleaning operations are carried out (in addition to what type of substance, cleaning agents and volume discharged). Currently, chemical tankers are not obliged to report such information. One possible way to get information about when and where tank cleaning operations are carried out is then to analyze the ship activities through AIS data.

This pre-study mainly addresses three aspects of data analysis using AIS information from chemical tankers sailing around Swedish waters, 1) detecting tank cleaning based on the pollution points of the Swedish Coast Guard’s detection records, 2) investigating if there are abnormal/potentially intentional shutdown of a ship’s AIS signals, 3) calculating and detecting abnormal ship sailing trajectories (e.g. sharp unnecessary turns), in order to identify potential tank cleaning operations. The implemented data mining methods are expected to be able to provide data support for maritime regulatory authorities.

AIS big data analysis shows that missing AIS data messages are more frequent for the investigated chemical tankers compared to cargo ships. Even though it is not possible to judge how many of those AIS signals missing scenarios are related to tank cleaning, serious AIS signal loss, i.e., more than 30-50 minutes data while the ship is sailing faster than 5 knots, may be interpreted as suspicion of tank cleaning by looking at their trajectories. For chemical tankers granted prewash exemption, much fewer AIS missing data were observed (no serious AIS signal missing was observed).

Analysis of AIS data around marine chemical pollution spots observed by Swedish Coast Guard, revealed certain sailing patterns (trajectories/speed), which can be used to design machine learning algorithms to identify potential tank cleaning operations. For example, sharp turning/maneuvering, unnecessarily sailing to open sea (outside of Swedish Economic Zone), ballast loading conditions, etc. Further analysis should be able to cluster more patterns to design machine learning algorithms to automatically searching potential tank cleaning operations.