Do maritime safety actions deliver lasting risk reduction? Evidence from AIS-based near-miss analysis using difference-in-differences

The maritime domain faces a persistent challenge: mitigating collision risk, particularly in areas with mixed commercial and fishing traffic. Maritime administrations frequently deploy short-term, intensive safety actions to address this, but robust empirical evidence of their lasting effectiveness remains elusive. A recent study published in Survey on ship route planning towards autonomous ship era seeks to address this gap, utilizing a novel AIS-based evaluation framework and a difference-in-differences design to assess a specific safety action in Ningbo–Zhoushan waters. The findings, while not demonstrating a clear reduction in near-miss counts, offer a valuable methodological contribution and underscore the complexities of evaluating maritime safety interventions. Considering the recent incident with the U.S. Coast Guard icebreaker Healy, U.S. Coast Guard Icebreaker Healy Suffers Significant Engineering Casualty During Sea Trials, which highlights the inherent risks and potential for unexpected events in maritime operations, this research is timely and relevant. The approach of using near-miss data, while imperfect, represents a significant step toward better understanding and measuring navigational risk.

The core innovation of this study lies in its use of Automatic Identification System (AIS) data to generate a behavior-level proxy for navigational risk – monthly near-miss counts. By applying DCPA and TCPA criteria to extensive AIS trajectory data, the researchers were able to quantify risk indicators in a way that traditional accident-based analysis cannot. The application of a difference-in-differences design allowed for a quasi-experimental comparison of treated and control areas, aiming to isolate the impact of the safety action. While the analysis did not reveal a statistically significant reduction in near-miss counts during the observation window, it is important to acknowledge the inherent challenges in attributing causality in such complex systems. The study rightly points out that near-miss counts are influenced by traffic intensity and composition, and the policy period coincided with a seasonal fishing moratorium, factors that could have obscured the intervention’s effect. This underscores a broader challenge in maritime safety research: disentangling the effects of specific interventions from the myriad of other variables that influence maritime behavior. Examining the broader geopolitical context, as seen in China Says It ‘Will Not Forget’ Indian Military’s Rescue Of 12 Chinese Crew After Cargo Ship Fire, highlights the interconnectedness of maritime operations and the need for collaborative safety measures.

Beyond the specific findings regarding the Ningbo–Zhoushan safety action, this study's methodological contribution is noteworthy. The combination of AIS-derived behavioral indicators with quasi-experimental policy evaluation provides a valuable framework for assessing non-engineering maritime safety interventions. This approach has the potential to be applied to a wider range of interventions and geographic locations, providing a more data-driven basis for maritime safety policy. The utilization of longitudinal data, aggregated at a sea area–month level, allows for a nuanced understanding of risk dynamics over time. Furthermore, the transparency in acknowledging the limitations of the study—specifically the relationship between near-miss counts and traffic variables—enhances its credibility and provides avenues for future research. The validation of this framework through further studies with varying methodologies and contexts is crucial to establishing its robustness.

Ultimately, this research reinforces the importance of rigorous evaluation in maritime safety. While the immediate findings might not demonstrate a clear success story for the assessed safety action, the development of a practical evaluation framework represents a significant advancement. The challenge remains to refine these approaches, accounting for factors such as traffic patterns, environmental conditions, and human behavior. Looking ahead, the integration of machine learning techniques to analyze AIS data and predict near-miss events holds promise for proactive risk mitigation. A key question for future research is whether similar frameworks, incorporating more granular data and advanced analytical methods, can identify subtle but impactful changes in navigational behavior that contribute to long-term risk reduction, even if not immediately apparent in near-miss counts.