A Bayesian Network risk model for estimating coastal maritime transportation delays following an earthquake in British Columbia
Introduction
Coastal and especially island communities are comparatively more vulnerable to the immediate impacts and ensuing consequences of disasters, because their comparative geographical remoteness leads to isolation from major population centres and logistics hubs [40]. With the increasingly clearly manifesting effects of climate change, many coastal and island communities are particularly endangered through more frequently occurring hurricanes, floods, storm surges, and other forms of catastrophic devastation [47].
Various coastal populations are heavily or exclusively dependent on maritime transportation for the ingress of essential goods, such as food, oil and medicine [41]. Natural disasters can cause significant damage to critical infrastructures and/or lead to cascading technological accidents. When port or other critical maritime transportation infrastructures are damaged, delivery of goods through maritime logistics chains may be severely hampered for extended periods of time. Even with less destructive natural disasters, operational disruptions in the maritime transportation system can lead to delays in deliveries of essential goods such as fuel or medical supplies to coastal areas, see e.g. Tanner et al. [56] and Errett et al. [23].
For effective emergency response, fast deployment of humanitarian relief logistics and distribution of essential supplies to affected populations in coastal communities is essential. This is evident from the 1995 Kobe earthquake, which had devastating impacts on the port infrastructure in the region, and had severe repercussions to all transportation modes in the region [34]. Another example is the Great Tohoku Tsunami in 2011, which damaged many coastal ports in Japan [58], led to significant transportation disruptions over a large area [55], and led to an estimated economic loss of US$200–300 billion [20].
Several studies have already focused on aspects of maritime transportation disruption, focusing on understanding the impacts of contextual factors, or on proposing strategies to minimize the impacts of disruptions. For example, Gurning et al. [31] provide an analysis of causes and mitigation strategies for disruptions to a maritime supply chain linking Australia and Indonesia. Qu and Meng [53] provide an extreme-scenario economic analysis of shipping disruptions in the Malacca and Singapore Straits, whereas Calatayud et al. [10] analyse the vulnerability of international freight flows to disruptions through a network analysis. Hossain et al. [35] developed a model for assessing the resilience of a port to disruption scenarios. This was extended in Hossain et al. [36] to the interdependencies between the port and its surrounding supply chain network.
In disaster risk management contexts, significant advances have been made in developing location and routing models for humanitarian logistics, mostly focusing on road transportation, see Banomyong et al. [5] and Behl and Dutta [6] for recent reviews. In comparison, work focusing on maritime transportation in a disaster logistics context is relatively scarce. Tatham and Kovacs [57] provide a cost analysis of floating warehouses for disaster logistics to coastal communities, while Wilberg and Olafsen [61] propose a simulation model for assessing the effectiveness of disaster logistics using this seabasing concept. Ozkapici et al. [50] develop a multimodal humanitarian logistics model for the Istanbul area, which includes maritime transportation, to minimize the total transportation time of relief goods to affected communities. Costa et al. [15] propose an object-oriented model for the vulnerability of the fuel distribution network in British Columbia, whereas Chang and Dowlatabadi [12] propose a modelling framework for informing decision making for disaster resilience, focused on coastal shipping.
Understanding plausible delays in the delivery of critical goods, such as fuel, medicine, and food to affected communities is an important aspect of natural disaster preparedness and response risk management [1]. In models for road transport, maritime, and multi-modal disaster logistics, such as the models reviewed in Behl and Dutta [6] and Banomyong et al. [5], and the models by Ozkapici et al. [50] and Costa et al. [15] mentioned above, the duration of the transportation operation delay and the delivery time to affected communities are important parameters, but in these existing models, the values for delays are based mostly on assumptions and high-level judgments.
Hence, there are currently no models available which specifically focus on the delays of maritime transportation due to disruptions caused by natural or NaTech disasters. For the island and coastal communities in British Columbia, which are heavily dependent on maritime transportation, the provincial natural disaster preparedness planning highlights a need to gain better understanding of maritime transportation disruptions and delivery delays [52]. This is especially necessary for disaster risk management for earthquakes, as these occur relatively frequently in the region, with possibly devastating effects [37].
Given the geographically divergent impacts of earthquakes and the different shipping operations in the province, it is desirable that such a model is capable of estimating route delays for a range of scenarios to support region-wide disaster preparedness planning. Furthermore, there often are considerable uncertainties involved in anticipating the particular aspects of NaTech disasters and disaster risk management [22, 54]. Therefore, explicitly accounting for uncertainty in making estimates of such maritime transportation delays caused by earthquake impacts is considered important.
Considering the above, this paper aims to develop a model to estimate the maritime transportation delays to coastal communities in British Columbia caused by an earthquake in the region. To account for the uncertainties, a Bayesian Network (BN) modelling approach is selected, which is further contextualized in a strength of evidence and criticality assessment, which provide information to model users about the evidence base of the model when estimating delays for specific routes. To build a knowledge base for the model development, various evidence gathering activities are executed and reported.
The remainder of this article is organized as follows. In Section 2, the research context is described, providing high-level insights in the maritime transportation system in British Columbia, and earthquakes as the most significant class of natural disasters in that region. Section 3 describes the methodology used for developing the proposed model for maritime transportation delays. Specifically, the choice of BN modelling is justified and briefly explained, and the process for establishing the knowledge base, model development, and model validation is described. In Section 4, the proposed BN model is shown, results of illustrative scenarios are given, and a model validation presented. A discussion is given in Section 5, focusing on the model use, limitations of the presented work, and future research directions. Section 6 concludes.
Section snippets
Research context
As a research context in the current work, Vancouver Island is selected, which is the largest and most populated island in British Columbia, Canada. All communities on this island are heavily dependent on the marine transportation system for receiving essential goods such as foods, fuels and medical supplies. The maritime transportation network to coastal communities on Vancouver Island can be characterized as a short sea shipping hub-and-spoke network, where essential goods are delivered from
Adopted risk perspective and model development process
In this section, the process for developing the proposed model is described. In Section 3.1, the choice for Bayesian Networks as the selected modelling approach is justified, and BNs are contextualized in a two-stage risk analysis framework based on an uncertainty-based risk perspective. The model development process is explained in Section 3.2.
Results
A summary of the main aspects of the background knowledge underlying the final developed model is given in Section 4.1. The actual BN model is presented in Section 4.2. For reasons of brevity, focus here is on the qualitative part of the model, i.e. the graphical component representing the variables and their relations. The discretization of the probabilistic component is outlined as well, whereas for detailed information about the variable states and the probability values for CPTs associated
Discussion
Referring to the two-stage risk analysis framework by Goerlandt and Montewka [28] and the acceptable uses of non-predictive models as suggested by Hodges [33] and Goerlandt and Reniers [30], the developed risk model should not be used as a direct estimate of the delay times for specific routes. Rather, it is recommended to use these estimated delays as a basis for reflection, discussion, and final expert judgment for specific maritime routes in the study area. Hence, the BN model should be seen
Conclusions
In this article, a Bayesian Network risk model is developed to estimate delays of coastal maritime transportation in British Columbia under earthquake conditions. The model considers a range of earthquake conditions, routes, and vessel type, providing high-level insights especially in the relative differences between delays under various scenarios. Several validation tests are performed on the developed model, indicating a plausible representation of the model space and adequate changes of the
Author statement
Floris Goerlandt: Conceptualization, Methodology, Software, Validation, Formal analysis, Investigation, Resources, Data curation, Writing – original draft, Writing – review & editing, Visualization, Supervision, Project administration, Funding acquisition. Samsul Islam: Methodology, Software, Formal Analysis, Data curation, Writing – original draft
Declaration of Competing Interest
I confirm that no conflict of interest exits in the submission of this manuscript. I also confirm that the work is original, has not been published previously, and is not under consideration for publication elsewhere, in whole or in part.
Acknowledgements
The work presented in this paper has been performed in context of the project ‘Shipping Resilience: Strategic Planning for Coastal Community Resilience to Marine Transportation Risk (SIREN)’. This project is financially supported by the Marine Observation, Prediction and Response (MEOPAR) Network of Centres of Excellence, and by Province of British Columbia. This financial support is gratefully acknowledged. The authors are also grateful for the kind support by the domain experts throughout the
References (64)
Practical implications of the new risk perspectives
Reliability Engineering & System Safety
(2013)- et al.
Whose uncertainty assessments (probability distributions) does a risk assessment report: the analysts’ or the experts’?
Reliability Engineering & System Safety
(2011) - et al.
Reliability and validity of risk analysis
Reliability Engineering & System Safety
(2009) - et al.
Vulnerability of international freight flows to shipping network disruptions: A multiplex network perspective
Transportation Research Part E: Logistics and Transportation Review
(2017) - et al.
Good practice in Bayesian network modelling
Environmental Modelling & Software
(2012) - et al.
Object-oriented model of the seismic vulnerability of the fuel distribution network in coastal British Columbia
Reliability Engineering & System Safety
(2019) - et al.
Probabilistic approach for characterising the static risk of ships using Bayesian networks
Reliability Engineering & System Safety
(2020) - et al.
Towards a probabilistic model for predicting ship besetting in ice in Arctic waters
Reliability Engineering & System Safety
(2016) - et al.
Validity and validation of safety-related quantitative risk analysis: A review
Safety Science
(2017) - et al.
A framework for risk analysis of maritime transportation systems: A case study for oil spill from tankers in a ship–ship collision
Safety Science
(2015)
On the assessment of uncertainty in risk diagrams
Safety Science
Prediction in a risk analysis context: Implications for selecting a risk perspective in practical applications
Safety Science
A Bayesian network based approach for modeling and assessing resilience: A case study of a full service deep water port
Reliability Engineering & System Safety
Modeling and assessing interdependencies between critical infrastructures using Bayesian network: A case study of inland waterway port and surrounding supply chain network
Reliability Engineering & System Safety
Bayesian networks for maritime traffic accident prevention: Benefits and challenges
Accident Analysis & Prevention
Probabilistic decision-support framework for community resilience: Incorporating multi-hazards, infrastructure interdependencies, and resilience goals in a Bayesian network
Reliability Engineering & System Safety
Leveraging a Bayesian network approach to model and analyze supplier vulnerability to severe weather risk: A case study of the U.S. pharmaceutical supply chain following Hurricane Maria
International Journal of Disaster Risk Reduction
Bayesian network-based seismic damage estimation for power and potable water supply systems
Reliability Engineering & System Safety
A Bayesian Network risk model for assessing oil spill recovery effectiveness in the ice-covered Northern Baltic Sea
Marine Pollution Bulletin
Critical aspects for collision induced oil spill response and recovery system in ice conditions: A model-based analysis
Journal of Loss Prevention in the Process Industries
Enhancing human performance in ship operations by modifying global design factors at the design stage
Reliability Engineering & System Safety
A proposed validation framework for expert elicited Bayesian Networks
Expert Systems with Applications
The economic importance of the Straits of Malacca and Singapore: An extreme-scenario analysis
Transportation Research Part E: Logistics and Transportation Review
Realising advanced risk assessment of vessel traffic flows near offshore wind farms
Reliability Engineering & System Safety
Incorporation of formal safety assessment and Bayesian network in navigational risk estimation of the Yangtze River
Reliability Engineering & System Safety
Natural Disasters and Risk Management in Canada: An Introduction
A systematic review of humanitarian operations, humanitarian logistics and humanitarian supply chain performance literature 2005 to 2016
Annals of Operations Research
Humanitarian supply chain management: a thematic literature review and future directions of research
Annals of Operations Research
Modified Mercalli intensities for nine earthquakes in Central and Western Washington between 1989 and 1999
The fuel transportation system in coastal British ColumbiaAttributes and vulnerabilities
Modeling the dynamics of community resilience to coastal hazards using a Bayesian network
Annals of the American Association of Geographers
Canada's earthquakes: ‘The Good, the Bad, and the Ugly
Geoscience Canada
Cited by (24)
Developing a probability-based technique to improve the measurement of landslide vulnerability on regional roads
2024, Reliability Engineering and System SafetyThe YODO algorithm: An efficient computational framework for sensitivity analysis in Bayesian networks
2023, International Journal of Approximate ReasoningShip navigation from a Safety-II perspective: A case study of training-ship operation in coastal area
2023, Reliability Engineering and System SafetyOptimization of maritime support network with relays under uncertainty: A novel matheuristics method
2023, Reliability Engineering and System SafetyCitation Excerpt :Increasing numbers of maritime incidents have caused serious security concerns and negatively affected national interests, and consequently the maritime support network has attracted widespread attention [1,2].
Datasets of disrupted transportation networks on Canada's West Coast in a plausible M9.0 Cascadia Subduction Zone earthquake scenario
2023, Data in BriefCitation Excerpt :The communities’ resilience level, damages to infrastructures (i.e. ports and roads), and delays to ships trajectories are classified as medium. Even though these data are generated using models developed specifically to Vancouver Island in the context of a Cascadia earthquake event, and are published in peer-reviewed journal articles, the models represent simplifications of impacts of the scenario, and involve several assumptions and limitations as recognized in those articles [11,14,19]. Finally, the damages to air infrastructures and roads unblocking times are classified as “low” strength of evidence.