Higher-order patterns of aquatic species spread through the global shipping network

Mandana Saebi, Jian Xu, Erin K. Grey, David M. Lodge, Nitesh V. Chawla

TL;DR

This paper introduces a novel "Species Flow Higher-Order Network" (SF-HON) model to better predict the spread of aquatic invasive species via global shipping. By accounting for complex, indirect pathways beyond simple port-to-port connections, SF-HON significantly improves risk prediction accuracy compared to traditional models, offering crucial insights for targeted prevention strategies.

What Problem Does This Paper Address?

Non-indigenous species (NIS) introduced through global shipping pose a massive threat to marine ecosystems and economies, costing trillions of dollars annually. While ballast water has received some regulatory attention, biofouling remains a significant and less-managed risk. Effectively preventing these introductions requires accurate models of how species spread. Existing models often assume a "first-order" spread, meaning species only move directly between two ports. However, ships often visit multiple ports in sequence, creating indirect pathways for species transfer that are missed by simpler models. This paper addresses the limitation of first-order models by developing and validating a "higher-order" network approach that captures these complex, multi-port transfer patterns.

Methods

The core innovation is the development of Species Flow Higher-Order Networks (SF-HON). Unlike traditional "first-order" networks (SF-FON) that model direct port-to-port connections, SF-HONs explicitly model higher-order dependencies. This means they consider the sequence of ports visited by a ship, not just individual port pairs.

The process involves:

  1. Data Integration: Combining shipping trajectory data, environmental data (temperature, salinity), and biogeographical data (ecoregions).
  2. First-Order Risk Calculation: Estimating the probability of NIS introduction, establishment, and being non-indigenous for each direct port-to-port link, separately for ballast water and biofouling.
  3. Rule Extraction: Identifying significant higher-order patterns in ship movements. This involves checking if adding a previous port to a ship's trajectory significantly alters the predicted NIS spread risk.
  4. Network Wiring: Constructing the SF-HON by representing these higher-order rules as nodes and weighted edges. A node in an SF-HON can represent a sequence of ports, allowing for a more nuanced representation of species flow.
  5. Network Analysis: Employing clustering techniques (Infomap) to identify groups of tightly connected ports.

Datasets

The study utilizes several large-scale datasets:

  • Vessel Movement Data: Lloyd's Maritime Intelligence Unit (LMIU) database, containing millions of ship voyages and port information.
  • Ballast Discharge Data: National Ballast Information Clearinghouse (NBIC) for US ports.
  • Environmental Data: Global Ports Database and World Ocean Atlas for port temperature and salinity.
  • Biogeographical Data: Marine Ecoregion of the World (MEOW) for ecoregion information.
  • NIS Introduction Data (for validation):
    • AquaNIS: Non-indigenous species data for European countries.
  • USGS: Non-indigenous aquatic species database for the United States.
  • NEMESIS: National Exotic Marine and Estuarine Species Information System for North America.

These datasets cover a 15-year period (1997-2012) and are used to validate the SF-HON model against first-order models and an existing higher-order approach.

Key Insights and Ideas

  • Higher-Order Pathways are Crucial: The assumption of simple, direct port-to-port spread is a significant oversimplification. Indirect pathways, dictated by ship trajectories, are critical for accurately modeling NIS spread.
  • Ballast vs. Biofouling are Fundamentally Different: The mechanisms and patterns of NIS spread via ballast water and biofouling are distinct. Ballast water transport creates more interconnected, higher-order networks, while biofouling tends to result in more localized, isolated clusters.
  • Global Recession Impacted Spread Differently: The 2007-2009 global recession led to reduced shipping activity, which disproportionately affected ballast water NIS spread (decreasing it) while biofouling NIS spread increased, likely due to longer ship stays in ports.
  • Vessel Type Matters: Different vessel types contribute to NIS spread differently. Large vessels (Bulk, Auto carriers) are more significant for ballast water introductions, while smaller vessels with longer port stays are more relevant for biofouling.

Key Results

  • Improved Accuracy: SF-HON models consistently outperformed first-order models (SF-FON) and a "full-trajectory" baseline model in predicting NIS introductions across US and European datasets. The SF-HON models achieved lower Mean Squared Error (MSE).
  • Ballast SF-HON Dominance in US: For US datasets (USGS, NEMESIS), the ballast SF-HON provided the most accurate predictions.
  • Biofouling SF-HON Dominance in Europe: For the European AquaNIS dataset, the biofouling SF-HON performed best. This discrepancy is attributed to the prevalence of freshwater NIS in US data (more susceptible to ballast water) versus marine NIS in Europe where biofouling might be more prevalent due to shorter voyages.
  • Network Structure Differences: Ballast SF-HONs showed a dramatic increase in nodes and edges compared to SF-FONs (37.56x nodes, 5.08x edges), indicating a much richer higher-order network structure. Biofouling SF-HONs also expanded, but to a lesser extent (9.39x nodes, 3.13x edges).
  • Clustering Insights: Ballast SF-HONs exhibited small, overlapping, and globally connected clusters, suggesting a need for both local and global prevention strategies. Biofouling SF-HONs showed larger, more isolated clusters, indicating that targeting inter-cluster connections would be a more effective strategy.

Limitations and Future Work

  • Data Completeness: The study acknowledges that some NIS records might represent secondary spread rather than initial introductions, and the exclusion of multiple records for the same species might lead to an incomplete dataset, potentially reducing the model's power.
  • Model Refinement: The models for ballast and biofouling introduction are based on available data at a global scale and could be improved with more accurate and detailed vessel and voyage-specific introduction risk data.
  • Ballast Water Management Practices: The models did not explicitly account for ballast water exchange or other management practices that have been implemented in some jurisdictions over time.
  • Species-Specific Analysis: Future work could expand by breaking down risk analysis by specific temperature and salinity tolerance groups to identify pathways for particular species.
  • Further Validation: The authors suggest that the models can be further improved with more accurate and detailed data about vessels and voyage-related introduction risks.

Summary updated

Abstract

Abstract The introduction and establishment of non-indigenous species (NIS) through global ship movements is a significant threat to marine ecosystems and economies. While ballast-vectored invasions have been partly addressed by some national policies and an international agreement regulating the concentrations of organisms in ballast water, biofouling-vectored invasions remain a large risk. Development of additional realistic and cost-effective ship-borne NIS policies requires an accurate estimation of NIS spread risk from both ballast water and biofouling. In this paper, we demonstrate that first-order Markov assumptions limit accurate modeling of NIS spread risks through the global shipping network. In contrast, we show that higher-order patterns overcome this limitation by revealing indirect pathways of NIS transfer. We accomplish this by developing Species Flow Higher-Order Networks (SF-HON), which we developed independently for ballast and biofouling, for comparison with first-order Markovian models of ballast and biofouling. We evaluated SF-HON predictions using the largest available datasets of invasive species for Europe and the United States. We show that not only does SF-HON yield more accurate NIS spread risk predictions than first-order models and existing higher-order models, but also that there are important differences in NIS spread via the ballast and biofouling vectors. Our work provides information that policymakers can use to develop more efficient and targeted prevention strategies for ship-borne NIS spread management, especially as management of biofouling is of increasing concern.
BibTeX
@inproceedings{saebi2020higher,
  author = {Mandana Saebi and Jian Xu and Erin K. Grey and David M. Lodge and Nitesh V. Chawla},
  title = {{Higher-order patterns of aquatic species spread through the global shipping network}},
  booktitle = {PLoS},
  year = {2020},
  doi = {10.1371/journal.pone.0220353},
}

Loading related content...

Figure