Enhancing global maritime traffic network forecasting with gravity-inspired deep learning models.

Area of Science:

• Ecology and Conservation
• Maritime Logistics
• Data Science

Background:

• Aquatic non-indigenous species (NIS) threaten biodiversity and cause economic damage.
• Global trade and transportation facilitate the unintentional spread of NIS.

Purpose of the Study:

• To develop a physics-informed model for forecasting maritime shipping traffic.
• To enhance the risk assessment of NIS spread via transportation networks.

Main Methods:

• Utilized a physics-informed model inspired by the gravity model of international trade.
• Incorporated shipping flux density, port distances, trade flow, and hub centrality.
• Applied transformers to gravity models to capture short- and long-term dependencies.

Main Results:

• Achieved 89% binary accuracy for trajectory prediction and 84.8% accuracy for vessel flow volume.
• Demonstrated over 10% improvement compared to traditional deep-gravity models.
• Provided a versatile framework for assessing international vessel traffic flow.

Conclusions:

• The model enhances understanding of NIS risk assessment.
• Enables policymakers to prioritize management actions by identifying high-risk invasion pathways.
• Offers a adaptable solution for evolving global vessel traffic patterns.