Collision Risk Assessment in a Narrow Channel with Gini Coefficient Approach

Abstract Tanjung Perak Port in Surabaya, Indonesia, serves as a primary hub for trade and economic activities in the eastern region. It makes this port has a great traffic density and prone to a collision accident. Gini coefficient was employed to utilize the Automatic Identification System (AIS) data for ship collision frequency analysis in the port area. AIS data contained voyage related data, one of which is Course Over Ground (COG), which were then subjected to Gini coefficient calculations, measuring the dispersion of COG values. Hierarchical clustering grouped COG values into four clusters: the first comprised of Leg 1, 2, 3, 4, and 6, the second included Leg 8 and 7, the third involved Leg 5 and 10, and the fourth comprised of Leg 9. Cluster proximity validation, indicated by a cophenetic correlation of 0.90484, signified excellent relationships between clusters. Subsequent risk characterization classified clusters 1, 2, and 4 as “extremely remote” risk levels, while cluster 3 was deemed “remote,” based on IMO frequency index assessments. IWRAP software facilitated accident frequency calculations, revealing discrepancies in risk levels across different areas in the port. Notably, while IWRAP assigned a “remote” risk level to leg 4, the Gini coefficient characterization indicated “remote” risk levels for legs 10 and 5. Recommendations for reducing ship collision frequency through variable geometry and causal probability values in the Western Shipping Lane of Surabaya should be provided if the ship collision frequency exceeds ALARP level according to the UK HSE standards.