Temporal analysis and gravity-informed marine traffic forecasting for non-indigenous species risk assessment through ballast water

Song, Ruixin (2024) Temporal analysis and gravity-informed marine traffic forecasting for non-indigenous species risk assessment through ballast water. Masters thesis, Memorial University of Newfoundland.

[English] PDF - Accepted Version Available under License - The author retains copyright ownership and moral rights in this thesis. Neither the thesis nor substantial extracts from it may be printed or otherwise reproduced without the author's permission. Download (10MB)

Abstract

Non-indigenous species (NIS) spreading through ballast water and establishing themselves in the new environment threaten biodiversity and marine ecosystems. Ballast water risk assessment (BWRA) models estimate the risk for NIS introduction by ballast water, and the environmental similarity between water source and destination locations is important in these models. Previous BWRA models rely on annual-scale environmental data and potentially neglect seasonal variability in the environmental factors. This research investigates the impact of incorporating monthly-scale environmental data on the evaluation of environmental similarity between source and recipient locations. The statistical comparison reveals that using monthly-scale data generally results in smaller environmental distances across all regions, indicating a higher risk of NIS invasions into Canadian waters than previously estimated with annual data. In addition, this work introduces a novel physics-inspired framework to forecast maritime shipping traffic, enhancing the assessment of NIS spread through global transportation networks. Integrating graph analysis, the gravity model, and the self-attention mechanism from the Transformers, this framework outperforms existing methods, achieving an 89% accuracy for discriminating existing and non-existing shipping trajectories and an 84.8% accuracy in estimating the number of vessels owing between port areas. This represents more than 10% improvement over the traditional deep-gravity model and nearly 50% improvement over the machine learning regressional models, offering a more accurate tool to identify high-risk invasion pathways and prioritize ballast water management in the future.

Actions (login required)

View Item