Correlating real-world incidents with vessel traffic off the coast of Hawaii, 2017-2020

Abstract

Abstract Objectives: Because of the high-risk nature of emergencies and illegal activities at sea, it is critical that algorithms designed to detect anomalies from maritime traffic data be robust. However, there exist no publicly available maritime traffic data sets with real-world labeled anomalies. As a result, most anomaly detection algorithms for maritime traffic are validated without ground truth. Data description: We introduce the HawaiiCoast_GT data set, the first ever publicly available automatic identification system (AIS) data set with a large corresponding set of true anomalous incidents. This data set—cleaned and curated from raw Bureau of Ocean Energy Management (BOEM) and National Oceanic and Atmospheric Administration (NOAA) automatic identification system (AIS) data—covers Hawaii’s coastal waters for four years (2017-2020) and contains 88,749,176 AIS points for a total of 2,622 unique vessels. This includes 208 labeled tracks corresponding to 154 rigorously documented real-world incidents.