Journal of Environmental Management

Journal of Environmental Management 278 (2021) 111520
Contents lists available at ScienceDirect
Journal of Environmental Management

journal homepage: http://www.elsevier.com/locate/jenvman
Implementing Bayesian networks for ISO 31000:2018-based maritime oil

spill risk management: State-of-art, implementation benefits and
challenges, and future research directions
Tuuli Parviainen a, b, c, d, *, Floris Goerlandt e, f, Inari Helle c, g, Päivi Haapasaari a, d,
Sakari Kuikka b, d
a
University of Helsinki, Marine Risk Governance Group, Ecosystems and Environment Research Programme, Faculty of Biological and Environmental Sciences, P.O Box
65, Viikinkaari 1, FI-00014, University of Helsinki, Finland
b
University of Helsinki, Fisheries and Environmental Management Group, Ecosystems and Environment Research Programme, Faculty of Biological and Environmental
Sciences, P.O Box 65, Viikinkaari 1, FI-00014, University of Helsinki, Finland
c
Helsinki Institute of Sustainability Science (HELSUS), Porthania (2nd Floor), Yliopistonkatu 3, FI-00014, University of Helsinki, Finland
d
Kotka Maritime Research Centre, Keskuskatu 7, FI-48100, Kotka, Finland
e
Aalto University, Department of Mechanical Engineering, Marine Technology, P.O. Box 15300, FI-00076, Aalto, Finland
f
Dalhousie University, Department of Industrial Engineering, Halifax, Nova Scotia, B3H 4R2, Canada
g
University of Helsinki, Environmental and Ecological Statistics Group, Organismal and Evolutionary Biology Research Programme, Faculty of Biological and
Environmental Sciences, P.O Box 65, Viikinkaari 1, FI-00014, University of Helsinki, Finland
A R T I C L E I N F O A B S T R A C T
Keywords: The risk of a large-scale oil spill remains significant in marine environments as international maritime transport
Oil spills continues to grow. The environmental as well as the socio-economic impacts of a large-scale oil spill could be
Pollution preparedness and response substantial. Oil spill models and modeling tools for Pollution Preparedness and Response (PPR) can support
Bayesian networks
effective risk management. However, there is a lack of integrated approaches that consider oil spill risks
Uncertainty
comprehensively, learn from all information sources, and treat the system uncertainties in an explicit manner.
Risk management
ISO 31000:2018 Recently, the use of the international ISO 31000:2018 risk management framework has been suggested as a
suitable basis for supporting oil spill PPR risk management. Bayesian networks (BNs) are graphical models that
express uncertainty in a probabilistic form and can thus support decision-making processes when risks are
complex and data are scarce. While BNs have increasingly been used for oil spill risk assessment (OSRA) for PPR,
no link between the BNs literature and the ISO 31000:2018 framework has previously been made. This study
explores how Bayesian risk models can be aligned with the ISO 31000:2018 framework by offering a flexible
approach to integrate various sources of probabilistic knowledge. In order to gain insight in the current utili
zation of BNs for oil spill risk assessment and management (OSRA-BNs) for maritime oil spill preparedness and
response, a literature review was performed. The review focused on articles presenting BN models that analyze
the occurrence of oil spills, consequence mitigation in terms of offshore and shoreline oil spill response, and
impacts of spills on the variables of interest. Based on the results, the study discusses the benefits of applying BNs
to the ISO 31000:2018 framework as well as the challenges and further research needs.
1. Introduction have been made in terms of maritime safety (UNCTAD 2019; Hassler,
2011; Haapasaari and Dahlbo 2014; Hänninen and Rytkönen 2006;
Global trade largely relies on international maritime transport: an Knudsen and Hassler, 2011; Lagring et al., 2012; Mitchell 1994; Ring
estimated 80 per cent of the volume of world trade is seaborne and in bom 2018), the risk of a large-scale oil spill remains significant. It is well
ternational maritime transport has been projected to continue growing known that there are many different definitions of risk (Aven 2012). For
in the coming decades (UNCTAD 2019). While significant improvements example, the ISO 31000:2018 risk management standard defines risk as
* Corresponding author. University of Helsinki, Marine Risk Governance Group, Ecosystems and Environment Research Programme, Faculty of Biological and
Environmental Sciences, P.O Box 65, Viikinkaari 1, FI-00014, University of Helsinki, Finland.
E-mail addresses: [email protected] (T. Parviainen), [email protected] (I. Helle).
https://doi.org/10.1016/j.jenvman.2020.111520
Received 5 March 2020; Received in revised form 15 September 2020; Accepted 13 October 2020
Available online 7 November 2020
0301-4797/© 2020 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
T. Parviainen et al. Journal of Environmental Management 278 (2021) 111520
the “effect of uncertainty on objectives” (ISO 2018), while, e.g. Aven and uncertainty (caused by limited knowledge) related to oil spills is high,
Renn (2009) define the term as “uncertainty about and severity of the since large-scale oil spills are rare and data related to e.g. site specific
consequences (or outcomes) of an activity with respect to something accident statistics and the environmental conditions/factors are often
that humans value”. Here, we do not suggest any specific definition for scarce. Still, deterministic models typically rely on past information
risk. (historical data and statistics), i.e. require large amounts of data for
In this article, we focus on studies that have mainly analyzed oil spill parameterization and the number of parameters is easily high compared
risk in terms of the uncertainty about the occurrence of an oil spill in the to available observations. Similarly, ontological uncertainty, which re
marine environment, and the related environmental, economic, human- fers to the inherent, natural variability of human and/or natural sys
health, and socio-cultural consequences. The severity of consequences is tems, is high in the model input data. However, deterministic models
a highly value-laden concept requiring additional scientific procedures provide only point estimates or expected values, and do not express or
and the topic is out of the scope of this paper. The environmental as well evaluate the uncertainties related to, e.g. assessing the trajectory of the
as the socio-economic impacts could be substantial in the case of a large- oil in operational oil combating decision-making (Li et al., 2016).
scale oil spill from maritime transport, as evidenced by historic accident Further, complex risks are generally rife with ambiguity, but ambiguity
cases such as Prestige (Spain), Erika (France), and Exxon Valdez (United can also occur in cases where risks are “simple” and uncertainty is low
States) (e.g. Garza-Gil et al., 2006; Miraglia 2002; Kontovas et al., 2010; (Aven and Renn 2010). Previous studies propose that uncertainty in oil
Peterson et al., 2003). spill risk analysis needs to be better addressed and clearly presented
Oil spill models and modeling tools help both the researchers and (Mazaheri et al., 2016) and that uncertainty treatment should be applied
end-users (decision-makers, wider audience) to gain a deeper under as a validity criterion for quantitative risk analysis (Aven and Heide,
standing of oil spills, their assessment, and management. Oil spill models 2009; Goerlandt et al., 2017a).
for pollution preparedness and response (PPR) planning and decision- Further, there is a lack of common methodology for oil spill risk
making (i.e. models addressing the accident occurrence, the response assessments (Sepp Neves et al., 2015): commonly accepted methods are
effectiveness, and the ecological, economic, health, and socio-cultural important since oil spill risks are often a transnational problem and
impacts of oil spills) can provide response operators and risk managers require international cooperation in terms of response operations and
with valuable information to support 1) effective response operations contingency planning (Laine et al., 2018; Sepp Neves et al., 2015).
and tactical planning, and 2) strategic (long-term) planning for response Recently, comprehensive risk management approaches for planning and
concerning, e.g. the dimensions of response resources, guiding the preparing for oil spill incidents have increasingly been advocated (IMO
mobilization of resources, and prioritizing protection areas and 2010; Laine et al., 2018; Sepp Neves et al., 2015). The ISO 31000:2018
strategies. International Standard on Risk Management (ISO 2018) provides
Oil spill models and approaches for PPR can generally be categorized guidelines for integrated risk management for all types of organizations
as ex post studies, which are oil spill specific studies conducted after a and is therefore in essential role in communication of academia and
spill has already taken place, and ex ante approaches that are based on industry. The use of the ISO 31000:2018 standard has also been sug
modeled simulations to estimate the possible oil trajectories and the gested as a suitable basis for the evaluation of Pollution Preparedness
final impacts (Nelson and Grubesic 2018). For reviews on oil spill and Response (PPR) risk management (Laine et al., 2018) and for
models and their use in supporting response globally, see e.g. Spaulding dealing with uncertainties when assessing oil spill risks (Sepp Neves
(1988), Spaulding (2017), ASCE (1996), Reed et al. (1999), and Afenyo et al., 2015) in industry activities. As the main focus has been on in
et al. (2015). Widely used models for predicting fate and trajectory of dustry activities, there is a need to improve the link of the academic
spilled oil include, e.g. OSCAR (Reed et al., 1995), SIMAP (French- scientific work to the ISO 31000:2018 standard.
McCay 2004; French-McCay et al., 2004), and the General National In recent years, Bayesian networks (BNs, also known as belief net
Oceanic and Atmospheric Administration Operational Modeling Envi works, Jensen 1996; Pearl 1988) have been increasingly applied in
ronment (GNOME) (Beegle-Krause et al., 2001). In addition, several OSRA (e.g. Helle 2015; Hänninen 2015; Lehikoinen 2014; Mazaheri
models have been developed to assess the environmental (e.g. COWI 2017; Nevalainen 2019a; Valdez Banda 2017; Venesjärvi 2015; and
2007; French-McCay et al., 2004; Kingston, 2002; Olita et al., 2012) and references listed in section 4.1.). Bayesian networks offer a flexible
socio-economic impacts (e.g. Wirtz and Liu, 2006; Wirtz et al., 2007; modeling approach for risk management as outlined by the ISO
Nelson et al., 2015), as well as oil spill response effectiveness (e.g. Fingas 31000:2018 standard/or OSRA. Bayesian networks are well suited for
2011; Li et al., 2014; Ventikos et al., 2004). evaluating complex systems where uncertainty is high as they express
However, limited attention has been paid to uncertainty in maritime uncertainty in terms of probability distributions. Despite the increasing
transport risk analysis (Merrick and van Dorp 2006; Goerlandt and interest in BNs for OSRA, there has previously been no analysis on how
Montewka 2015a; Kelangath et al., 2012) and in other maritime such models can be used to implement the ISO 31000: 2018 standard.
modeling contexts (Li et al., 2016; Sepp Neves et al., 2015; Sebastião and We provide a systemic analysis of the use of BN-based oil spill risk
Guedes Soares 2007; Spaulding 2017). In general, the existing models assessment (OSRA-BN) models for pollution preparedness and response
are deterministic and designed for operational planning or for testing (PPR) planning and decision-making, i.e. models addressing the acci
different scenarios and they rarely assess risks per se, i.e. they do not dent occurrence, the response effectiveness, and the ecological, eco
provide information about the uncertainties (probabilities) and conse nomic, health, and socio-cultural impacts of oil spills. The aim of this
quences of particular scenarios. Therefore, models do not describe how paper is, therefore, to analyze how the existing OSRA-BNs models can be
well we know what may happen or what were the consequences, and aligned with the ISO 31000:2018 framework by offering a flexible
neither do they offer basis to judge which parameters should be known approach to integrate various sources of probabilistic knowledge. This
more precisely by additional scientific effort. paper, however, does not aim to explicitly evaluate the benefits and
Yet, oil spill risks can be considered as systemic and complex risks challenges of the ISO 31000: 2018 standard for oil spill risk assessment
characterized by high levels of uncertainty as well as ambiguity, i.e. the and management.
differing understandings and perceptions of risks, as well as societal The paper is structured as follows. First, the need for a comprehen
values (Aven and Renn 2010). While various definitions exist for un sive oil spill risk management framework is outlined in Section 2, and
certainty in decision-making contexts (Döll and Romero-Lankao 2017; the ISO 31000:2018 risk management standard and the Bayesian
Kwakkel et al., 2016; van Asselt and Rotmans, 2002; Walker et al., 2003) approach to treating uncertainty are introduced. Section 3 describes the
and in modeling (Hamilton et al., 2019; Refsgaard et al., 2007; Regan methods used in the study. We performed a literature review of BN
et al. 2002, 2003), uncertainty is often classified as epistemic and sto models for OSRA for maritime transportation activities, with relevance
chastic or ontological uncertainty (Walker et al., 2003). Epistemic for pollution preparedness and response (PPR) planning and decision-
2
making. Section 4 then presents the results of the literature review and integrated ways to assess and manage oil spill risks is increasingly
provides an overview of how Bayesian OSRA models can be applied to highlighted due to the systemic and complex nature of the risks (IMO
the ISO 31000:2018 risk management framework. This is followed by a 2010; Davies and Hope 2015; Laine et al., 2018; Sepp Neves et al.,
discussion (Section 5) on the potential and challenges related to the 2015). Fig. 1 (based on Chang et al., 2014) presents a framework for
approach, as well as a summary of further research needs. Section 6 is classifying oil spill response and the short and long-term environmental,
left for conclusions. economic, human health, and socio-cultural impacts. Fig. 1 illustrates
the complexities of oil spill risks as well as the interactions between the
2. The need for a comprehensive risk management framework elements. When moving from short term operational policy support to
long term strategic and directive decisions, the role and methodological
2.1. The ISO 3100:2108 framework for oil pollution preparedness and basis of probabilistic advice changes (Barton et al., 2012).
response In terms of oil spills, the use of the ISO 31000:2018 risk management
framework has recently been suggested as a suitable basis for the
Even though there are various interests in using the risk assessment comprehensive evaluation of PPR risk management (Laine et al., 2018;
tools, risk is commonly defined as a combination of the probability of an Sepp Neves et al., 2015). The ISO 31000 Risk Management standard
event and the related negative consequences (Burgman 2005; Jardine (ISO 2009) was first published in 2009, and updated in 2018 (ISO 2018).
et al., 2003; UNISDR 2009; USEPA 1998). Risk assessment is a vital part Experts from different backgrounds contributed to developing the
of the risk management process. Various guidelines and frameworks standard (ISO 2009). The standard is mainly followed by industry
exist, e.g. for environmental risk assessment (EFSA 2020; OECD 2013; players, but it is flexible and is not industry or sector specific.
USEPA 1998; IRGC 2017) as well as for assessing risks related to ship The standard defines risk management as “coordinated activities to
ping (such as the Formal Safety Assessment (FSA) developed by the direct and control an organization with regard to risk”, where risk is
International Maritime Organization (IMO)). While the terminology, defined as the “effect of uncertainty on objectives” (ISO 2018). The ISO
scope, and elements vary among the different operational guidelines, the 31000:2018 definition differs from other risk definitions in traditional
assessment and management process is typically described as an itera risk assessments (Aven 2012; Burgman 2005), as risk is not only defined
tive one that aims to identify the optimal management action to reach in terms of the probabilities of negative or undesirable outcomes, but the
justified balance between expected utilities (e.g. production and car focus is on uncertainty management. The focus of this paper, however, is
rying oil) and potential negative impacts (e.g. oil spills). not on the various risk definitions, rather, we explore the flexibility of
However, the conventional risk management approaches, such that Bayesian networks as a tool to provide relevant knowledge for oil spill
predominantly rely on the traditional, two-dimensional, risk definition, risk assessment and management.
may not be the most suitable for assessing systemic risks characterized The standard comprises the risk management principles, framework,
by complexity, high levels of uncertainty, and ambiguity (Aven and and process (ISO, 2018). The framework provides guidance on how risk
Renn 2010; ; Döll and Romero-Lankao 2017; Sperotto et al., 2017). management processes can be integrated in a specific organizational
Therefore, comprehensive risk management approaches based on setting, i.e. into the activities and functions of an organization. The ISO
stakeholder participation are needed, where the various sources of risk, 31000:2018 risk management process is considered as an iterative
the system interactions, as well as uncertainties related to the system, process including the following steps: 1) defining the scope, context,
are considered in a systematic and iterative manner (Assmuth and Hil criteria, 2) risk assessment (including risk identification, risk analysis,
den 2008; Pollino and Hart 2008). Likewise, the need for new and and risk evaluation), 3) risk treatment, 4) recording and reporting, 5)
Fig. 1. Framework for classifying oil spill response

and short and long-term consequences in PPR
context. Grey boxes (1 A, 1 B) indicate oil spill
occurrence and fate; purple boxes (2 A, 2 B) oil spill
response effectiveness; green boxes (3 A, 3 B),
ecosystem impacts; orange boxes (4 A, 4 B) economic
impacts; yellow boxes (5 A, 5 B) health and socio-
cultural impacts. Adapted from Chang et al. (2014).
(For interpretation of the references to colour in this
figure legend, the reader is referred to the Web
version of this article.)
3
Table 1 Table 1 (continued )

Risk management framework and examples of OSRA-BNs and their applicability Definition in the context of Examples of OSRA-BNs and
(Adapted from ISO 2018). oil spills their applicability
Definition in the context of Examples of OSRA-BNs and how well the models study by Lecklin et al.
oil spills their applicability represent the system in the (2011)).
Scope, context, Defining the objectives and Context specific risk models light of new information or
criteria decisions that need to be for the Baltic Sea (e.g. Lecklin changes in the system.
made, specifying the time et al., 2011). Communication Communication and Visual risk diagrams and
frame and the geographical and consultation consultation refer to causal networks, such as BNs,
scope of the study, the communicating results of can support risk
sources of risks, the nature risk assessments to communication; BNs can be
and type of potential stakeholders such as the combined, e.g. with GIS (
consequences, as well as the industry, insurance Jolma et al., 2014); BNs can
type of the assessment companies, or national and facilitate stakeholder
method to be used. international response deliberation (Goerlandt and
Risk assessment Risk identification refers Identifying the ecological authorities. Reiniers 2017; Parviainen
to identifying and defining impacts of dispersants to et al., 2019).
the assessment endpoints (e. seabirds in the North Sea,
g. impacts on ecosystems, German Bight (Liu and
habitats or species), the Callies 2019); assessing the monitoring and review, and 6) communication and consultation
potential sources of oil economic impacts of oil spill (Table 1).
pollution (causes of events), on commercial fisheries in The eight risk management principles refer to the underlying values
the variables contributing to the Great Australian Bight (
oil spill occurrence and/or Pascoe 2018).
and considerations that are commonly seen as the best practices in risk
the impacts of a spill (after it Quantifying oil outflows from management: e.g. risk management should be structured and compre
has occurred), and the collision (Goerlandt and hensive, inclusive, dynamic, iterative, and based on the best available
dependencies between those Montewka 2014b; Montewka information (ISO, 2018). Hence, risk management needs to be based on
variables. et al., 2015); quantifying the
information that is factual, timely, relevant, accurate and understand
Risk analysis includes ecological impacts of oil in
quantifying the system and the Baltic Sea (Helle et al., able. Risk assessments should utilize both historical and current infor
the dependencies between 2011, 2016; Lecklin et al., mation, and future scenarios should also be included. (ISO, 2018). These
the system variables in order 2011) principles should be integrated in the organization’s risk management
to, e.g. estimate impacts in Evaluating response framework and process (ISO, 2018).
an area, the likelihood of effectiveness (Helle et al.,
actually polluting 2011; Lehikoinen et al.,
vulnerable areas, or the 2013b) and cost-benefit ( 2.2. Bayesian networks (BNs) as a tool for evaluating uncertainty and
effectiveness of oil spill Helle et al., 2015b) or
response measures. cost-effectiveness of response
identifying decision options
Risk evaluation includes measures (Montewka et al.,
the consideration, 2013). Bayesian networks (Jensen, 1996; Pearl 1988) are increasingly used
comparison, and to evaluate uncertain and complex systems. They build on the subjec
prioritization of reduction
tivist Bayesian approach, which, in comparison to the frequentist
alternatives based on the
results of risk analysis. approach in statistics, expresses the personalist view of probability
Risk treatment The formulating and BNs inform which treatment (Aven and Reniers 2013; Gelman et al., 2013) and allows the use of
selecting of risk treatment to implement as they allow probability distributions, which is not possible in classical statistics.
options, planning and for assessing risks, e.g. the Whereas the frequentist approach is based on induction, the Bayes’
implementing risk effectiveness of response
treatment, assessing the measures with the use of
theorem enables two-way reasoning, i.e. from cause to effect and vice
effectiveness of the decision nodes (e.g. Helle versa (Jensen, 1996; Pearl, 1998; Pearl and MacKenzie 2018). The
treatment, deciding whether et al., 2015; Lehikoinen et al., diagnostic use, from the effect back to potential causes, enables learning
the remaining risk is 2013a). from uncertain evidence.
acceptable, and if not
Bayesian networks help to conceptualize complex systems and
acceptable, taking further
actions (e.g. reconsider describe uncertainty in probabilistic terms. In BNs, different types of
objectives and/or carry variables can be included in the same network, e.g. sources of risks, the
further analysis to gain a uncertainties related to the risks, and the system endpoints, as well as
better understanding of the the interaction between the variables. Bayesian networks are often
risk).
Recording and Recording and reporting is The visual nature of BNs
depicted as directed acyclic graphs (DAGs) with nodes and arcs, where
reporting necessary in order to serves as a way of structuring the nodes represent random variables and the arcs the probabilistic
communicate risk information in a systemic dependencies between the variables (Fig. 2). A detailed description of
management activities and manner, i.e. it records BNs can be found, e.g. from Jensen (1996) or Jensen and Nielsen (2007).
outcomes, provide information from various
Conditional probability tables (CPTs) are used to quantify BNs. The
information for decision- sources; Strength of Evidence
making and wider policy (SoE) can further support quantification requires the defining of possible values to each variable
activities, improve risk reporting and recording (e.g. and assigning conditional probabilities for the strength of the relation
management activities, and Lu et al., 2019). ships, i.e. the dependencies, between the variables. CPTs contain in
support interaction with formation of the probabilities of a given state of a child node given the
stakeholders.
Monitoring and Monitoring and review BNs can be easily updated as
state of its parent nodes and, therefore, they also describe the quality of
review refers to assessing the need new evidence becomes knowledge, i.e. uncertainties. In addition, BNs can be applied even when
for modifying risk treatment available; BNs can build on data are scarce as the dependencies can be quantified using different
options. In terms of risk previous models (e.g. the types of data, information, and knowledge (i.e. new experimental data,
assessment, monitoring and work by Lehikoinen et al.,
empirical or theory-based models, existing publications, statistics),
review of risk models is 2013b builds on a previous
necessary for examining including expert elicitation, which may be crucial when considering
actions that have not yet been tested in practice.
Finally, BNs can be used as decision support tools by including
4
Fig. 2. A simplified example of a directed acyclic graph in the context of an oil spill. The nodes represent variables and the arrows linking the variables indicate that
the state of the receiving ‘child’ node (acute impact on Baltic herring) is conditionally dependent on the state of the originating ‘parent’ node (oil concentration in
water). The graph is quantified with the use of conditional probability tables.
distinctive decision and/or utility nodes in the network and effectively 4. Results
embedding the models in a decision analysis context. Bayesian networks
with decision and/or utility nodes are generally referred to as influence 4.1. Review results
diagrams (IDs). Decision and utility nodes can also be associated with
cost-benefit analysis, which enables comparing the actions also in eco The review shows an increased use of BNs for PPR: Figure S1 (Ap
nomic terms. Further, IDs can be used to assess, for instance, the value- pendix 1) shows a temporal evolution of the number of publications (by
of-information, i.e. whether knowing the state of a certain variable country) using BN for PPR. The first article on using BN modeling as an
would change the expected value of a given decision (Raiffa and approach for quantifying oil spill risks, response, and ecological impacts,
Schlaifer, 2000; Helle et al., 2015). was published in 2005 b y Juntunen et al. (2005). Thereafter, there was
an increased interest in the approach from 2009 onwards, with a slowly
3. Methods increasing trend thereafter. It is seen that by far most contributions
originate from Finland (31 publications) (Fig. S1). There was an early
We performed a literature review in order to gain insight in the interest in Estonia (3 publications), and more recent interest in the
utilization of OSRA-BN models for pollution preparedness and response approach in Canada (4 publications) (Fig. S1). Other countries where
(PPR) planning and decision-making, and to assess the suitability of the BNs have been used for oil spill risk related modeling for pollution
models to the implementation of the ISO 31000:2018 framework. The preparedness and response include Australia, the United States of
review focused on BN models for oil spill risks from maritime trans America, Norway, Germany, Poland, and Sweden. In this analysis, full
portation activities, with relevance for pollution preparedness and counting is applied, i.e. where an article is co-authored by authors
response (PPR) planning and decision-making. In particular, we focused affiliated with institutions from different countries, each country is
on articles presenting BN models that analyze the occurrence of oil counted for that publication.
spills, consequence mitigation in terms of offshore and shoreline oil spill BNs have been applied to gain understanding of various aspects of oil
response, and the impacts of spills on ecosystem and other variables of spill response and impacts as defined in the framework by Chang et al.
interest. Articles that discuss the use of BN models for this purpose, e.g. (2014), see Fig. 1. Table 2 gives an overview of the articles in the
presenting a conceptual approach, or discussing the challenges in database. It is seen that by far most models explicitly address the oil spill
developing those, were included as well. occurrence itself. Most models consider variables considering the size
We applied a rigorous and traceable process to identify the relevant and type of oil (1 A) only as starting points to understand response or
literature, based on recommendations by Van Wee and Banister (2016). impacts, which are the main focus of the models. However, some models
We used the following steps to identify the relevant literature. First, a are dedicated specifically to provide accurate probabilistic estimates of
search was made in the Web of Science Core Collections (WoSCC) ab oil outflow sizes in collision (Goerlandt and Montewka 2014b) or
stract database. All combinations from the following three sets of key grounding (Montewka et al., 2015) accidents, conditional to traffic and
words were used: {“marine” OR “maritime” OR “shipping”} AND impact conditions. A significant number of models also includes vari
{“model” OR “approach”} AND {“oil”} AND {“Bayesian”}. Second, we ables related to weathering and transport of oil (1 B), to explicitly ac
inspected the title and abstract of the results to screen relevant articles, i. count for the extent of the area affected by oil, given the sea and
e. that the articles focused on maritime transport oil spill risks and that meteorological conditions.
the information was related to oil spill response. The selected articles Often, the models quantify the probabilities through expert judg
were used to construct an initial dataset. Finally, we used backward and ment (Juntunen et al., 2005; Lecklin et al., 2011), but some models
forward snowballing to identify further articles, i.e. the reference lists in (Helle et al., 2011; Lu et al., 2019) use oil drift models as a basis for
the articles in the dataset were inspected, and Web of Science was used quantification in one part of the model.
to search for articles citing these articles. The search was performed in The assessment of offshore response effectiveness (2 A) has been an
June 2019, leading to a final dataset of 39 articles. important use for the developed BN models, e.g. Aps et al. (2010), Helle
et al. (2011), Lehikoinen et al. (2013b), Liu and Callies (2019), and Lu
et al. (2019). However, shoreline response (2 B) has been included in
5
Table 2
Analysis of OSRA-BN models for oil spill PPR according to framework of Fig. 1. Color coding: Grey (1 A, 1 B): oil spill occurrence and fate; Purple (2 A, 2 B): oil spill
response effectiveness; Green (3 A, 3 B): ecosystem impacts; Orange (4 A, 4 B): economic impacts; Yellow (5 A, 5 B): health and socio-cultural impacts. CA =
presentation of conceptual approach |D = discussion on BNs for OSRA for oil spill PPR.
6
only a handful of models (Helle et al. 2011, 2015; Montewka et al., are possible, often due to differing views about the key causalities. BNs
2013). can be used for exploring different scenarios as well as the multiple
In terms of impacts, most models focus on assessing the acute im perspectives and dimensions (e.g. economic, social, and environmental)
pacts (3 A) of oil to specific marine or coastal species, e.g. Aps et al. of risk, which can then be either integrated within the same network or
(2009a), Aps et al. (2009b), Helle et al. (2011) and Rahikainen et al. assessed separately. For example, Parviainen et al. (2019) provided a
(2017), while some models also address long-term impacts (3 B) to conceptual approach for building qualitative influence diagrams to
certain species (Lecklin et al., 2011; Venesjärvi 2012; Pascoe 2018). examine how different stakeholders perceive and identify oil spill risks
Economic impacts, especially acute ones (4 A) such as the direct costs caused by offshore oil industry operations in the Norwegian Barents Sea.
related to offshore oil combating and shoreline clean-up (Helle et al., BNs require the quantification of the system, and the influence diagrams
2015b; Montewka et al., 2013), have been considered. Long-term eco built in the study could be further quantified. Fully quantified BNs can
nomic impacts (4 B) have been evaluated significantly less often, and if be used to promote informed discussion on tradeoffs or to calculate the
so in a more exploratory sense (Afenyo et al., 2019; Pascoe 2018). optimal tradeoffs between different objectives (see section 4.3.3).
Human health impacts (5 A) and socio-cultural impacts (5 B), have been
included in exploratory and more conceptual models (Afenyo et al., 4.3. Risk assessment
2019; Parviainen et al., 2019), but no advanced quantifications have
been performed. 4.3.1. Risk identification
A significant number of articles provides a conceptual approach Risk identification includes deciding on the system end points, i.e.
(CA), i.e. presenting the idea of using BN models for oil spill impact and what is at risk. Once risks are identified, BNs can help to conceptualize
response management. These may involve the use of BNs as real-time complex systems as well as describe the different possible interactions
environmental decision support systems during an oil spill response between the variables and the system endpoints in a probabilistic
(Davies and Hope 2015), using BNs alongside other modeling ap manner (section 2.2). In terms of oil spill risks, with the use of BNs, Liu
proaches (Klemola et al., 2009), presenting the idea as a new solution for and Callies (2019) studied the use of dispersants and the ecological
a specific geographical area (Khan et al., 2014), or presenting a new use impacts of dispersants to seabird distributions in the North Sea, German
case for BNs in a pollution preparedness and response context. An Bight. The system included several variables related to e.g. the oil type,
example of the latter is the model by Parviainen et al. (2019), which uses season, drift paths to target areas, currents, and the ecological impacts.
influence diagrams based on the logic of BNs to study how different Pascoe (2018) applied BNs to explore the economic impacts of a po
stakeholders frame oil spill risks. In this work, however, the strength of tential oil spill on commercial fisheries in the Great Australian Bight.
the probabilistic dependencies between variables was not quantified. The model captured several key biophysical and economic factors to
Also, Nevalainen et al. (2017) lays out a conceptual approach to analyze identify the system and the potential economic consequences for
the ecological impacts of oil spills in the Arctic marine environment different fisheries (both wild-catch fisheries and aquaculture).
without quantifying the model parameters. Studies by Lecklin et al. (2011) and Helle et al. (2011) were the first
A final cluster of articles provides discussions on the development to identify risks to assess the ecological impacts of a potential oil spill in
and use of BNs for oil spill occurrence, response, and impacts. For the Baltic Sea using BNs. Lecklin et al. (2011) identified risks to assess
instance, Kuikka et al. (2011) describe challenges in developing BNs, the biological acute and long-term impacts to Baltic Sea marine and
particularly in interdisciplinary research teams. Kuikka et al. (2013) coastal species using BNs. The study also ranked species sensitivity to
discuss the use of BNs as learning systems, both in terms of using the oil. Helle et al. (2011) identified the risks to six different species in order
models as a tool to understand the value of information about particular to assess the impacts of an oil spill to the different parts of the ecosystem
system aspects which are uncertain at the time of the model develop in the Gulf of Finland.
ment, as well as in terms of using the outputs of a model as inputs for Further, Nevalainen et al. (2017; 2019) proposed the use of BNs to
further models with different aims. Further, Nevalainen et al. (2019b) identify risks in order to assess the biological impacts of oil spills on
apply BNs in sensitivity analysis related to the index approach devel different functional groups in Arctic waters. Previous research on oil
oped to estimate the vulnerability of Arctic biota to oil spills. spill risks in the Arctic also investigated the biophysical and
socio-economic impacts with the use of BNs (Afenyo et al., 2019) and
4.2. Defining scope, context, criteria qualitative influence diagrams (Parviainen et al., 2019). Afenyo et al.
(2019) developed a comprehensive probabilistic model, named as the
As the review results demonstrate (section 4.1), OSRA-BNs have been Socio-Economic Impact Model for the Arctic (SEMA), which was used to
increasingly applied for PPR depicting complex systems and covering identify risks to assess the social, economic, and biophysical impacts (in
the uncertainties related to the occurrence of oil spills, their magnitude, terms of costs) of an oil spill due to shipping activity in the Arctic.
the effectiveness of various response measures, as well as possible
ecological, economic, health, and socio-cultural impacts. 4.3.2. Risk analysis
Context-specific oil spill risk assessments are important (ISO, 2018). Following the ISO 31000:2018 standard, risk identification is fol
Risk models need to consider the specific context of the risks: the nature lowed by risk analysis, where, in a BN application context, the aim is to
of spill risks varies by region due to differences in the type of shipping, quantify the system and the relationships between the system variables
the physical features of the marine environment, regulations and policy using conditional probabilities. BNs are useful for risk analysis. Oil spill
measures, and varying economic, social, and cultural context. For risk analysis is often challenging due to lack of observational data. As
example, models assessing the oil-induced impacts to ecosystems in noted earlier (section 2.2), BNs allow for the integration of different
other marine environments may not be applicable in the low-saline, types of knowledge sources. In addition, expert elicitation is often used
brackish, water conditions of the Baltic Sea (Lecklin et al., 2011). As for constructing BNs in the cases where statistical data, publications for
noted in Section 4.1, so far, the OSRA-BN studies have mostly focused on meta-analysis or simulation models are not available.
the Baltic Sea. For example, in their study on the biological impacts of a potential oil
Further, the ISO 31000:2018 framework emphasizes stakeholder spill in the Baltic Sea, Lecklin et al. (2011) used already published
participation in defining the scope, context, and criteria of risk assess studies and expert elicitation in order to quantify the biological impacts
ments (ISO, 2018). Especially in the case of complex risks, where as well as the recovery potential of a selected group of organisms. In
defining the system variables is difficult and the causal interactions addition, BNs can be combined with deterministic simulations models,
between the variables are highly uncertain, reaching a consensus on the e.g. Helle et al. (2016) studied the effects of an oil spill caused by
model structure might be challenging and alternative model structures different tanker accident scenarios to threatened species and habitats in
7
the Northern Baltic Sea using BNs. The study (ibid.) combined a BN Lehikoinen et al. (2013b) found that environmental variables, e.g. wave
describing tanker accidents and uncertainties related to them, proba height and wind speed, make the effectiveness of response operations
bilistic maps showing the movement of oil, and a database of threatened highly uncertain.
species and habitats. Winter period and ice cover would significantly reduce response
Risk analysis includes a comprehensive overview of uncertainties effectiveness. Most of the existing BN response models, however, have
(ISO 2018). BNs are designed to handle and communicate uncertainties excluded winter season with ice, as the conditions differ significantly
explicitly (section 2.2). Bayesian models estimating oil outflows from from open sea. Recent research on mechanical recovery in sea ice con
collision (Goerlandt and Montewka 2014b) and grounding (Montewka ditions in the Baltic Sea by Lu et al. (2019) provided a comprehensive
et al., 2015) accidents have provided a way for assessing the un model on oil spill recovery under different sea ice and atmospheric
certainties related to the size of outflows when only limited data conditions, for different oil types, accident locations, spill sizes and port
regarding e.g. ship design or the specific accident scenario is available. locations.
In the study by Valdez Banda et al. (2016), BNs have been quantified for Considering the high levels of uncertainty related to response
sea ice conditions. The study (ibid.) primarily focused on accident pre effectiveness, the role of preventive actions has been highlighted: pre
vention measures for ship collision occurrence, but also provided esti ventive actions can also be more cost-efficient than response measures
mates of oil spill sizes conditional to navigation conditions, i.e. gave (Helle et al. 2011, 2015; Hänninen 2014; Haapasaari and Dahlbo 2014).
estimates of the likely oil spill sizes in different accident scenarios,
which is useful in a PPR context. 4.4. Risk treatment
Further, Helle et al. (2011) applied BNs to describe uncertainties
related to the effectiveness of oil combating from an ecological Risk treatment refers to actions and the implementation of measures
perspective in the Baltic Sea. Helle et al. (2016) further illustrated the to reduce the probability of an oil spill and to mitigate the consequences
uncertainties related to the environmental impacts of oil spills in their based on risk assessment (section 2.1). BNs do not have a direct role in
study on the effects of an oil spill caused by different tanker accident risk treatment, i.e. in the implementation of treatment measures in
scenarios to threatened species and habitats in the Northern Baltic Sea. practice. The OSRA-BNs, however, demonstrate the need for action: BNs
inform risk treatment as they allow for assessing risks and the effec
4.3.3. Risk evaluation tiveness of response measures with the use of decision nodes (e.g. Helle
BNs with decision and/or utility nodes can be used for comparing the et al., 2015; Lehikoinen et al., 2013a).
effectiveness of different measures and decision alternatives (section OSRA-BNs also imply that risk treatment measures need to pay
2.2). When evaluating oil spill risks, BNs have been used to compare the careful attention to uncertainties: as shown in the previous sections (e.g.
effectiveness of different oil combating methods on the risk level and to sections 4.3- 4.3.3), the existing BN models for Baltic Sea demonstrate
evaluate what is the potential of a response system to reduce the amount that the uncertainties related to oil spill occurrence, response effec
of oil in the water. For example, Liu and Callies (2019) have evaluated tiveness and the environmental impacts are high. BNs also support
the benefits of using dispersants in oil combating in the German Bight. iterative and participatory risk treatment (see e.g. sections 4.2, 4.6, and
Most of the response models have focused on the Baltic Sea (Helle 5.2).
et al., 2013; Juntunen et al., 2005; Lehikoinen et al., 2013b; Lu et a.
2019; Montewka et al., 2013). For example, Juntunen et al. (2005) 4.5. Recording and reporting
compared the effectiveness of different oil combating strategies in order
to assess the effects of oil spills on the ecosystem of the Gulf of Finland. Recording and reporting of oil spill risks need to embody the prin
In the Baltic Sea, the response needs to be rapid and effective. In the case ciples underlying risk management, such as transparency, the use of best
of an accident, oil spill response would rely on mechanical clean-up as available knowledge, a structured approach, and inclusiveness (section
according to the HELCOM agreement, the use of chemical dispersants or 2.1). The systematic and visual nature of BNs serves as a way of struc
in-situ burning is not recommended in the Baltic sea (HELCOM 2001). turing the information, i.e. it records information from various sources.
Collecting the oil before it reaches the shore is considered vital to protect When developing BNs, documenting the information behind the condi
the coastal ecosystems. However, for example in the Gulf of Finland, tional probability tables and the variables needs to be carried out in a
effective oil spill response is challenging since the Gulf is narrow and oil systematic manner. For example, in the study of Valdez Banda et al.
would reach the shores fast, i.e. within some days or even in some hours (2016) and Lu et al. (2019), the information has been provided in the
(e.g. in the case of a grounding). Similarly, oil response in the archi appendix of the journal articles. Strength of Evidence (SoE) assessment,
pelagos of the Baltic Sea would be complicated as, e.g. the navigation of which refers to the qualitative assessments of the quality of the evidence
clean up vessels in the narrow inlets would be difficult (Montewka et al., underlying the BNs, can also be utilized to further support systematic
2013). and transparent recording and reporting. For example, Lu et al. (2019)
The existing Bayesian models highlight the uncertainties that relate included an SoE assessment for the quantified uncertainty measures in a
to the assessment of the response effectiveness (Helle et al., 2011; BN model assessing response effectiveness in winter conditions, and
Lehikoinen et al., 2013b) and the costs and benefits of oil spills (Helle Goerlandt and Montewka (2014b) provided a qualitative uncertainty
et al., 2015b) or the cost-effectiveness of different response measures and bias assessment related to oil outflows.
(Montewka et al., 2013). Using different scenarios, Helle et al. (2011)
demonstrated that the effects of different oil combating strategies are 4.6. Monitoring and review
difficult to predict as their efficiency is strongly dependent on the
environmental conditions, which are stochastic and where impacts are BNs can be easily updated once new information and evidence (e.g.
highly uncertain. The model included three oil combating options observations, field data, model results) becomes available, which facil
(mechanical recovery offshore, dispersants, and oil deflection booms) itates monitoring and review. BNs can also be used as a basis for new
and several variables representing the uncertainty related to the acci models. For example, the oil dispersion model (including probability
dent, behavior of spilled oil, environmental conditions, efficiency of oil estimates of the amount of spilled oil, evaporation of oil, length of oiled
combating, and biological effects. The deployment time of the booms, an coastal water, type of the accident, and leakage stop) developed by
important restriction for the operational oil combating, was not included Juntunen et al. (2005) has later been applied to other models to esti
as a random variable, but it was still taken into account when calculating mate, e.g. the ecological impacts of oil spills (Lecklin et al., 2011; Helle
the proportion of the populations that can be safeguarded by booms. et al., 2011). Similarly, the work of Lecklin et al. (2011) has later been
Similarly, when assessing the optimal placement of combating vessels, used to develop other, more specific, ecological impact models (Helle
8
et al., 2011) as well as response models (Lehikoinen et al., 2013b). probabilistic assessment of risks where uncertainty, e.g. related to
achieving the desired outcome, is expressed in an explicit manner. In
4.7. Communication and consultation sum, BNs can help decision-makers to realistically evaluate the chances
(likelihood) of achieving the desired result/outcome as well as the un
Visual risk diagrams and causal networks, such as BNs, can support certainties associated with it. The effective calculus of discrete proba
risk communication. SoE assessments can further improve communi bilities enables the use of the model to interactively learn from the case.
cation on uncertainties (Lu et al., 2019). The models can also be com Further, OSRA-BNs support iterative and adaptive management. BNs
bined with user-friendly interfaces, e.g. Geographic Information use previous knowledge (e.g. models and expert knowledge) effectively
Systems (GIS) for spatial risk assessments (Jolma et al., 2014; Helle as the modular structure of BNs allows for modifying and updating
et al., 2016), which supports the use of BNs as communication tools. In a previous oil spill models by adding new information/nodes: this sup
broader sense, risk diagrams also make the possible conflicting views ports continuous learning processes. BNs also assist iterative modeling
over risks explicit and facilitate stakeholder deliberation, and finally, as they allow for updating the prior information as new evidence be
can help in selecting appropriate risk management measures as well as comes available. The probability distribution narrows as new knowledge
ranking them (Goerlandt and Reniers 2017; Parviainen et al., 2019). accumulates and uncertainty about the phenomenon and the parameters
Hence, it is recommended to use BNs alongside other visualization tools decreases.
to maximally facilitate communication.
5.2. BNs and OSRA
5. Discussion
Risk management approaches need to take into consideration the
The ISO 31000:2018 standard provides a comprehensive framework complexity of oil spill risks, the high levels of uncertainty, as well as
for contextualizing, assessing, evaluating, and treating risks. In Section ambiguity. The overview in Section 4 indicates that existing OSRA-BNs
4, an overview is given of how the existing probabilistic BNs can support are valuable for understanding systemic risks, such as oil spill risks.
the implementation of ISO 31000:2018-based maritime oil spill risk Notably, OSRA-BNs can improve and support risk communication about
management. The work suggests that BNs have several useful features uncertainties as well as stakeholder participation. The role of stake
for oil spill risk management and can serve as a valuable tool for holder participation throughout the modeling process is increasingly
implementing the ISO 31000:2018 principles and process. emphasized in the management of complex risks (ISO 2018; Aven and
The benefits of OSRA-BNs over alternative modeling approaches Renn 2010): stakeholder participation can enhance the legitimacy of
include features such as the possibility to comprehensively represent decision-making, support effective implementation of models, and pro
complex systems in a graphical fashion and the flexibility in using mote social learning (Aven and Renn 2010; Shrader-Frechette 1991;
various knowledge sources (both quantitative and qualitative informa Stern and Fineberg 1996).
tion). Further, BNs are especially valuable for treating uncertainty as The graphical form of BNs allows the communication of uncertainty
they explicitly express uncertainty between system variables in the form and risks in an understandable way to decision makers and other
of probability distributions. Overall, BNs for OSRA can support and stakeholders. The users can easily identify the main sources of uncer
improve iterative and participatory management approaches that are tainty affecting in the model as well as where uncertainty can be
based on continuous learning processes. reduced. Communication about uncertainties is especially important
when the results are used to inform public policy, as decision-makers
5.1. BNs for ISO 31000:2018 risk management and managers need to be informed about the possible outcomes and
base their decisions on robust quantitative estimates (Sperotto et al.,
OSRA-BNs can help the implementation of the ISO 31000:2018 risk 2017). However, the end-users might not be aware of the uncertainties
management principles in practice. OSRA-BNs effectively facilitate involved in modeling results, or the risk assessment results expressed as
conceptualizing complex systems. In BNs, different types of variables point estimates may seem definitive if the limitations in data and the
can be included in the same network, e.g. sources of risks, the un inadequacies of the models are not communicated appropriately (Mer
certainties related to the risks, and the system endpoints. The causal and rick and Van Dorp 2006; Goerlandt and Montewka 2015a). Therefore,
visual structure allows for evaluating the interactions in a structured and uncertainties need to be understood also by the non-scientific commu
systematic manner. BBN can be used also in a diagnostic manner, i.e. nity in order to avoid misjudged information as well as to prevent
updating the probabilities of causes by observing the state of impacts. overconfidence in management responses (Uusitalo 2007). Accounting
This may be very useful in case specific analysis, i.e. after an accident. for the uncertainties include in models is important as it can strengthen
The use and integration of both qualitative and quantitative infor the acceptance of model outputs as well as the implementation of risk
mation from different knowledge domains ensures that BNs are based on treatment options based on the outputs. In sum, if uncertainties are left
best evidence that is factual, timely, and relevant. The use of different untreated and are poorly communicated, the models may, contrary to
knowledge forms, such as expert elicitation, to identify system variables their purpose, hamper effective risk management (Uusitalo et al., 2015).
and to quantify the cause-effect relationships between the system vari Participatory approaches based on visual risk diagrams, such as BNs,
ables allows for risk analysis even when quantitative data are lacking or can provide better understanding of how risks and risk control options
scarce, which makes the approach especially suitable for context- are defined and prioritized by different stakeholders (Haapasaari et al.,
specific oil spill modeling, where region-specific data are often 2015; Parviainen et al., 2019). Involving stakeholders in the modeling
limited. BNs can also be easily customized for specific contexts as, e.g. processes and in identifying, analyzing, and evaluating risks ensures
existing general models can be modified with region specific inclusive risk management that is considerate of human and cultural
information. factors. Participation of stakeholders can be seen as especially important
OSRA-BNs also allow for risk evaluation, i.e. for comparing man during the first phase of the process as setting the scope and context for
agement actions and identifying the optimal measures to minimize risks. risk assessment and management, i.e. framing of the risk, defines the rest
For example, the effectiveness and/or cost-efficiency of alternative of the management process (Brugnach et al., 2011; Döll and
response measures can be compared by adding new decision and utility Romero-Lankao 2017; Parviainen et al., 2019).
nodes to a BN and by assigning fixed distributions to input variables and Regional stakeholder committees can support the participation of
testing how the probability distributions of the output variables change stakeholders in assessing regional oil spill risks (Haapasaari et al., 2015).
as a response to this. The results, i.e. the outcome variables, are Such committees can provide an arena for discussing risk perceptions,
expressed as probability distributions, which provides a basis for a the tolerability of risks and the costs of reducing them, as well as the
9
development and use of decision support tools. The use of BNs can For risk evaluation, most of the models focus on the Baltic Sea. The
provide an effective method to explore how different stakeholders existing BN models for the Baltic Sea often combine some of the aspects
perceive and prioritize risks and risk control options. The use of full of oil spill risk pollution preparedness and response. However, the
probability distributions can support discussions over acceptable risks models are not comprehensively integrated (e.g. combine the estimation
that is informed by science. Indeed, rather than aiming to accurately of oil fate and potential trajectory with the assessment of oil spill
quantify risks and uncertain systems, Rae and Alexander (2017) propose response effectiveness or with potential impacts) and even if they are
that risk assessments should primarily be seen as a means of describing integrated, it is not necessarily optimal as the models are typically
uncertainties. A similar view is expressed by Goerlandt and Reniers developed for specific conditions and/or policy questions. Further, most
(2018), who suggest that risk models and assessments are better un OSRA-BNs exclude winter conditions: the existing models for ice
derstood as non-predictive artefacts, which serve as a basis for reflection covered waters focus on prevention measures for ship collision occur
and discussion, rather than as tools for accurate measurement. In sum, rence and on mechanical recovery.
participatory OSRA-BNs can support knowledge co-production, enable While expert knowledge has been used for developing BN models,
experimentation between a range of stakeholders, and finally, enable the existing models have not included a wider range of societal stake
social learning (Parviainen et al., 2019). holders in constructing and developing the risk models. Most of the
The challenges related to the use of BNs in environmental modeling OSRA-BNs focus on quantifying risks and the system dependencies,
have been extensively discussed by several authors (Aguilera et al., however, BNs can also be applied to explore stakeholder perceptions
2011; Düspohl et al., 2012; Phan et al., 2016; Pollino and Henderson, (Haapasaari et al., 2012; Kuikka and Varis, 1997; Varis and Kuikka
2010; Uusitalo, 2007). The main challenges include, e.g. the high level 1997; Parviainen et al., 2019). More research is needed on how well
of expertise needed for model development, the labor-intensive process stakeholders understand the BNs approach, how to effectively facilitate
of model development, the limited capacity of BNs in dealing with knowledge sharing and social learning, and what factors enable suc
continuous variables, and the growing computational effort when cessful stakeholder deliberation.
modeling complex systems. As with models in general, BNs can also We have provided an overview of existing OSRA-BN models and
oversimplify the system, as it is not always possible to include all their use, but the overall framework provided is only a suggested
(potentially) relevant factors in a model. This may lead to overly opti approach for how the models can be integrated in the ISO 31000:2018
mistic outcomes. The major drawbacks related to the use of BNs in oil structure. Naturally, the ISO 31000: 2018 risk management framework
spill risk assessment and management, however, include their limited and the suggested approach for integrating BNs in the ISO framework is
capacity to represent temporal and spatial dynamics (temporal de open for discussion and requires testing and possible adjustments. For
velopments cannot be easily represented), the limitations of expert example, Aven and Ylönen (2019) argue that the strong focus on in
elicitation (Rae and Alexander 2017), and the difficulty of performing a dustry standards, such as the ISO 31000, in risk management may pose a
quantitative validation of model results (Goerlandt et al., 2017a,b). threat to the development of risk science. Therefore, further research is
To deal with temporal and spatial dynamics, BNs can be combined needed to explore and test how OSRA-BNs can contribute to different
with e.g. GIS for spatial risk assessment (Helle et al., 2016; Jolma et al., risk management frameworks, such as the guidelines provided by the
2014). Combining BN models with user-friendly interfaces, such as GIS, Society for Risk Analysis (SRA 2018), which are based on work by sci
helps to visualize outputs and also support the use of BNs as commu entific organizations. However, as noted by Aven and Ylönen (2019),
nication tools. Further, Dynamic Bayesian Networks (DBNs) can be used both industry-based standards and scientific-based guidance are needed
for representation of risk through time and space (Pollino and Hender to support the development and enhancement of risk science.
son, 2010). Both methods, however, are relatively time-consuming and
require extensive expertise for developing models. 6. Conclusion
5.3. Knowledge gaps and future work This study provides a global review of the OSRA-BNs models for
pollution preparedness and response and systematically analyses how
Most of the current work focuses on the environmental impacts of oil the models can help to implement the ISO 31000:2018 risk management
spill accidents in the Baltic Sea: only few OSRA-BN models exist for other framework in practice. The study also offers insight into the use of BNs in
sea areas. Limited attention has been paid to identifying and assessing oil spill risk assessment and management at a broader level, and suggests
environmental risks in sea ice conditions. New risks, such as the use of various avenues for future research.
liquefied natural gas (LNG) or biofuels and their impact on the envi The study indicates that OSRA-BNs can help to support many of the
ronment should be further explored. ISO 31000:2018 principles and that the models are especially useful for
Research on other than the environmental impacts of oil spills re risk analysis and evaluation, as well as communication. OSRA-BNs
mains exploratory. Models assessing the economic impacts focus on represent complex systems in a visual, easily understandable, manner,
comparing the cost-benefit or cost-effectiveness of different prevention and can utilize various knowledge sources (both quantitative and
and response measures (Helle et al., 2015; Montewka et al., 2013). qualitative information). Explicit treatment of uncertainty as well as
Similarly, further research is needed on the socio-cultural as well as communication about uncertainty is a vital part of the assessment and
health impacts of oil spills. In terms of risk analysis, i.e. the quantifi management of complex risks. BNs express uncertainty between system
cation of BNs, further attention should be paid on the limitations of variables in the form of probability distributions in a visual manner,
expert elicitation. Combining the views of multiple experts and/or other which aids risk communication. We also discuss the challenges related
relevant stakeholders may help in overcoming biases. Improving to OSRA-BNs. We suggest that transparent methods that allow also for
methods for the quantitative validation of results is also needed. For qualitative descriptions of risks and the system uncertainties are needed
analysing future risks, it might be useful to combine different types of to complement the current quantitative risk assessment models.
models to achieve a more complete picture of a system change, such as Finally, BNs for OSRA can support participatory management ap
the effects of an oil spill on the ecosystem, and the uncertainties related proaches that are based on continuous learning processes. We suggest
to the process (Laine et al., 2018; Uusitalo et al., 2016). Combining that OSRA-BNs can help to explore possible conflicting views over risks,
different models (deterministic, probabilistic, etc.) and types of risk facilitate collaborative problem framing, and promote social learning. In
assessments (quantitative, qualitative, participatory, etc.) could sum, we suggest that BNs aid robust decision making even in the case of
contribute to a more nuanced, transparent, and deeper understanding of systemic and multi-layered risks, such as oil spill risks. However, as
oil spill risks and the uncertainties related to the risks and risk participatory approaches do not necessarily lead to more inclusive
management. decision-making, further research is needed to study the “why”, “how”,
10
and “who” in participatory modeling. Aven, T., Ylönen, M., 2019. The strong power of standards in the safety and risk fields: a
threat to proper developments of these fields? Reliab. Eng. Syst. Saf. 189, 279–286.
https://doi.org/10.1016/j.ress.2019.04.035.
Author contribution Barton, D.N., Kuikka, S., Varis, O., Uusitalo, L., Henriksen, H.J., Borsuk, M., de la
Hera, A., Farmani, R., Johnson, S., Linnel, J.D.C., 2012. Bayesian networks in
environmental and resource management. Integrated Environ. Assess. Manag. 8,
Tuuli Parviainen: Conceptualization, Investigation, Writing - original 418–429. https://doi.org/10.1002/ieam.1327.
draft, Visualization.: Floris Goerlandt: Formal analysis, Writing - review Beegle-Krause, J., 2001. General NOAA oil modelling environment (gnome): a new spill
& editing, Visualization.; Inari Helle: Writing - review & editing, Visu trajectory model. In: International Oil Spill Conference Proceedings 2001, vol. 2,
pp. 865–871. https://doi.org/10.7901/2169-3358-2001 2-865.
alization.: Päivi Haapasaari: Conceptualization, Writing - review & Brugnach, M., Dewulf, A., Henriksen, H.J., van der Keur, P., 2011. More is not always
editing, Supervision.: Sakari Kuikka: Conceptualization, Supervision, better: coping with ambiguity in natural resources management. J. Environ. Manag.
Funding acquisition. 92 (1), 78–84. https://doi.org/10.1016/j.jenvman.2010.08.029.
Burgman, M., 2005. Risks and Decisions for Conservation and Environmental
Management. Ecology, Biodiversity and Conservation. Cambridge University Press,
United Kingdom.
Declaration of competing interest Chang, S.E., Stone, J., Demes, K., Piscitelli, M., 2014. Consequences of oil spills: a review
and framework for informing planning. Ecol. Soc. 19 (2), 26. https://doi.org/
The authors declare that they have no known competing financial 10.5751/ES-06406-190226.
COWI, 2007. Risk Analysis Oil and Chemical Spills in Danish Waters. HELCOM Response
interests or personal relationships that could have appeared to influence
Group 18. Apr. 2007. https://portal.helcom.fi/Archive/archive2/RESPONSE%20
the work reported in this paper. 8_2007_Presentation%201_Risk%20analysis%20Danish%20waters.pdf. (Accessed 6
November 2019).
Davies, A.J., Hope, M.J., 2015. Bayesian inference-based environmental decision support
Acknowledgements systems for oil spill response strategy selection. Mar. Pollut. Bull. 96 (1), 87–102.
https://doi.org/10.1016/j.marpolbul.2015.05.041.
This work resulted from the BONUS BALTIMARI project supported Düspohl, M., Frank, S., Döll, P., 2012. A review of Bayesian networks as a participatory
modelling approach in support of sustainable environmental management.
by BONUS (Art 185), funded jointly by the EU and the Academy of J. Sustain. Dev. 5 (1) https://doi.org/10.5539/jsd.v5n12p1.
Finland. The contributions by FG were supported in part by the project Döll, P., Romero-Lankao, P., 2017. How to embrace uncertainty in participatory climate
“Shipping Accident Oil Spill Consequences and Response Effectiveness change risk management—a roadmap. Earth’s future. Bognor Regis 5 (1), 18–36.
https://doi.org/10.1002/2016EF000411.
in Arctic Marine Environments (iSCREAM)”, funded by the Marine
European Food Safety Authority (Efsa), 2020. Environmental risk assessment. https://
Observation, Prediction and Response (MEOPAR) Network of Centres of www.efsa.europa.eu/en/topics/topic/environmental-risk-assessment. (Accessed 6
Excellence. IH was funded by the Helsinki Institute of Sustainability July 2020).
Fingas, M., 2011. Oil Spill Science and Technology, Prevention, Response and Cleanup,
Science (HELSUS), University of Helsinki. The work of SK was supported
first ed. Elsevier, Boston, MA.
by the Strategic Research Council (SRC) at the Academy of Finland, French-McCay, D.P., 2004. Oil spill impact modelling: development and validation.
under the project “Creative adaptation to wicked socioenvironmental Environ. Toxicol. Chem. 23 (10), 2441–2456. https://doi.org/10.1897/03-382.
disruptions” (WISE) (grant number 312627) and under the SRC project French McCay, D.P., Jennings Rowe, J., Whittier, N., Sankaranarayanan, S., Etkin, D.S.,
2004. Estimation of potential impacts and natural resource damages of oil. J. Hazard
"SmartSea" (grant number 292 985). Mater. 107 (1), 11–25. https://doi.org/10.1016/j.jhazmat.2003.11.013.
Garza-Gil, M.D., Surís-Regueiro, J.C., Varela-Lafuente, M.M., 2006. Assessment of
economic damages from the Prestige oil spill. Mar. Pol. 30 (5), 544–551. https://doi.
Appendix A. Supplementary data
org/10.1016/j.marpol.2005.07.003.
Gelman, A., Carlin, J.B., Hal, S.S., Dunson, D.B., Vehtari, A., Rubin, D.B., 2013. Bayesian
Supplementary data to this article can be found online at https://doi. data analysis. In: Texts in Statistical Science, third ed. Chapman and Hall/CRC, Boca
Raton.
org/10.1016/j.jenvman.2020.111520.
Goerlandt, F., Montewka, J., 2014b. A probabilistic model for accidental cargo oil
outflow from product tankers in a ship–ship collision. Mar. Pollut. Bull. 79 (1),
References 130–144. https://doi.org/10.1016/j.marpolbul.2013.12.026.
Goerlandt, F., Montewka, J., 2015a. Maritime transportation risk analysis: review and
analysis in light of some foundational issues. Reliab. Eng. Syst. Saf. 138, 115–134.
Afenyo, M., Veitch, B., Khan, F., 2015. A state-of-the-art review of fate and transport of
https://doi.org/10.1016/j.ress.2015.01.025.
oil spills in open and ice-covered waters. Ocean. Eng. 119, 233–248.
Goerlandt, F., Reniers, G., 2017. An approach for reconciling different perspectives and
Afenyo, W., Jiang, C., Ng, A.K.Y., 2019. Climate Change and Arctic Shipping: A Method
stakeholder views on risk ranking. J. Clean. Prod. 149, 1219–1232. https://doi.org/
for Assessing the Impacts of Oil Spills in the Arctic. Transportation Research Part D.
10.1016/j.jclepro.2017.02.129.
https://doi.org/10.1016/j.trd.2019.05.009.
Goerlandt, F., Khakzad, N., Reniers, G., 2017a. Validity and validation of safety-related
Aguilera, P.A., Fernández, A.R., Fernández, R., Rumí, R., Salmerón, A., 2011. Review:
quantitative risk analysis: a review. Safety Science, Risk Analysis Validation and
bayesian networks in environmental modelling. Environ. Model. Software 26 (12),
Trust in Risk management 99 (November), 127–139. https://doi.org/10.1016/j.
1376–1388. https://doi.org/10.1016/j.envsoft.2011.06.004.
ssci.2016.08.023.
Aps, R., Fetissov, M., Herkül, K., Kotta, J., Leiger, R., Mander, U., Suursaar, U., 2009a.
Goerlandt, F., Lu, L., Valdez Banda, O.A., Rytkönen, J., 2017b. An approach for an
Bayesian inference for predicting potential oil spill related ecological risk. WIT
integrated assessment of maritime accidental oil spill risk and response for the
Trans. Built Environ. 108, 149–159. https://doi.org/10.2495/SAFE090151.
Northern Baltic Sea. In: 17th International Congress of the International Association
Aps, R., Herkül, K., Kotta, J., Kotta, I., Kopti, M., Leiger, R., Mander, U., Suursaar, U.,
of the Mediterranean, Lisbon, Portugal, 9-11.10.2017.
2009b. Bayesian inference for oil spill related Net Environmental benefit Analysis.
Goerlandt, F., Reniers, G., 2018. Prediction in a risk analysis context: implications for
WIT Trans. Ecol. Environ. 126 (12), 235–246. https://doi.org/10.2495/CP090211.
selecting a risk perspective in practical applications. Saf. Sci. 101, 344–351. https://
Aps, R., Sawano, N., Hamada, S., Fetissov, M., 2010. Bayesian inference in oil spill
doi.org/10.1016/j.ssci.2017.09.007.
response management. WIT Trans. Inf. Commun. Technol. 43, 35–46. https://doi.
Haapasaari, P., Mäntyniemi, S., Kuikka, S., 2012. Baltic herring fisheries management:
org/10.2495/RISK100041.
stakeholder views to frame the problem. Ecol. Soc. 17 (3), 36. https://doi.org/
ASCE, 1996. State-of-the-Art review of modelling transport and fate of oil spills.
10.5751/ES-04907-170336.
J. Hydraul. Eng. 112, 594–609.
Haapasaari, P., Dahlbo, K., 2014. Minimizing Risks of Maritime Oil Transport by Holistic
Asselt, M., Rotmans, J., 2002. Uncertainty in integrated assessment modelling. Climatic
Safety Strategies (MIMIC). Final Report. Kotka Maritime Research Center, Kotka.
Change 54 (1), 75–105. https://doi.org/10.1023/A:1015783803445.
https://doi.org/10.13140/2.1.3773.7604.
Assmuth, T., Hilden, M., 2008. The significance of information frameworks in integrated
Haapasaari, P., Helle, I., Lehikoinen, A., Lappalainen, J., Kuikka, S., 2015. Proactive
risk assessment and management. Environ. Sci. Pol. 11, 71–86.
approach for maritime safety policy making for the Gulf of Finland: seeking best
Aven, T., 2012. The risk concept—historical and recent development trends. Reliab. Eng.
practices. Mar. Pol. 60, 107–118. https://doi.org/10.1016/j.marpol.2015.06.003.
Syst. Saf. 99, 33–44. https://doi.org/10.1016/j.ress.2011.11.006.
Hamilton, S.H., Fu, B., Guillaume, J.H.A., Badham, J., Elsawah, S., Gober, P., Hunt, R.J.,
Aven, T., Heide, B., 2009. Reliability and validity of risk analysis. Reliab. Eng. Syst. Saf.
et al., 2019. A framework for characterising and evaluating the effectiveness of
94 (11), 1862–1868. https://doi.org/10.1016/j.ress.2009.06.003.
environmental modelling. Environ. Model. Software 118 (August), 83–98. https://
Aven, T., Reniers, G., 2013. How to define and interpret a probability in a risk and safety
doi.org/10.1016/j.envsoft.2019.04.008.
setting. Saf. Sci. 51 (1), 223–231. https://doi.org/10.1016/j.ssci.2012.06.005.
Hassler, B., 2011. Accidental versus operational oil spills from shipping in the Baltic sea:
Aven, T., Renn, O., 2009. On risk defined as an event where the outcome is uncertain.
risk governance and management strategies. Ambio 40 (2), 170–178. https://doi.
J. Risk Res. 12, 1–11.
org/10.1007/s13280-0100128-y.
Aven, T., Renn, O., 2010. Risk Management and Governance - Concepts, Guidelines and
Applications, first ed., vol. 60. Springer http://www.springer.com/us/book/9783
642139253.
11
HELCOM, 2001. Restricted Use of Chemical Agents and Other Non-mechanical Means in fi/Lists/Publications/OpenRisk-Guideline-for-pollution-response-at-sea.pdf.
Oil Combating Operations in the Baltic Sea, HELCOM Recommendation 22/2. (Accessed 30 October 2019).
Helsinki Commission. Lagring, R., Degraer, S., de Montpellier, G., Jacques, T., Van Roy, W., Schallier, R., 2012.
Helle, I., Lecklin, T., Jolma, A., Kuikka, S., 2011. Modelling the effectiveness of oil Twenty years of Belgian North Sea aerial surveillance: a quantitative analysis of
combating from an ecological perspective - a Bayesian network for the Gulf of results confirms effectiveness of international oil pollution legislation. Mar. Pollut.
Finland. The Baltic Sea Journal of Hazardous Materials 185 (1), 182–192. https:// Bull. 64 (3), 644–652. https://doi.org/10.1016/j.marpolbul.2011.11.029.
doi.org/10.1016/j.jhazmat.2010.09.017. Lecklin, T., Ryömä, R., Kuikka, S., 2011. A bayesian network for analyzing biological
Helle, I., 2015. Assessing Oil Spill Risks in the Northern Baltic Sea with Bayesian acute and long-term impacts of an oil spill in the Gulf of Finland. Mar. Pollut. Bull.
Network Applications. Doctoral thesis. University of Helsinki. https://helda.helsinki. 62 (12), 2822–2835. https://doi.org/10.1016/j.marpolbul.2011.08.045.
fi/bitstream/handle/10138/155153/assessin.pdf?sequence=1. Lehikoinen, A., Haapasaari, P., Storgård, J., Hänninen, M., Mäntyniemi, S., Kuikka, S.,
Helle, I., Ahtiainen, H., Luoma, E., Hänninen, M., Kuikka, S., 2015. A probabilistic 2013a. Multi-disciplinary Lego-bricks: building an integrative metamodel for policy
approach for a cost-benefit analysis of oil spill management under uncertainty: a analysis by using Bayesian Networks, 13. International Council for the Exploitation
bayesian network model for the Gulf of Finland. J. Environ. Manag. 158 (August), of the Sea Council Meeting 2013/Q.
122–132. https://doi.org/10.1016/j.jenvman.2015.04.042. Lehikoinen, A., Luoma, E., Mäntyniemi, S., Kuikka, S., 2013b. Optimizing the recovery
Helle, I., Jolma, A., Venesjärvi, R., 2016. Species and habitats in danger: estimating the efficiency of Finnish oil combating vessels in the Gulf of Finland using bayesian
relative risk posed by oil spills in the northern Baltic sea. Ecosphere 7 (5). http://o networks. Environ. Sci. Technol. 47 (4), 1792–1799. https://doi.org/10.1021/
nlinelibrary.wiley.com/doi/10.1002/ecs2.1344/full. es303634f.
Hänninen, S., Rytkönen, J., 2006. Transportation of Liquid Bulk Chemicals by Tankers in Lehikoinen, A., 2014. Bayesian Network Applications for Environmental Risk
the Baltic Sea, vol. 595. VTT Publications, pp. 3–121. Assessment. Doctoral thesis. University of Helsinki. https://helda.helsinki.fi/bitstre
Hänninen, M., 2014. Bayesian networks for maritime traffic accident prevention: benefits am/handle/10138/135820/bayesian.pdf?sequence=1.
and challenges. Accid. Anal. Prev. 73 (December), 305–312. https://doi.org/ Li, P., Chen, B., Zhang, B., Jing, L., Zheng, J., 2014. Monte Carlo simulation-based
10.1016/j.aap.2014.09.017. dynamic mixed integer nonlinear programming for supporting oil recovery and
Hänninen, M., 2015. Bayesian Network Modeling of Potential Patterns in Maritime devices allocation during offshore oil spill responses. Ocean Coast Manag. 89
Safety Performance. Doctoral thesis. Aalto University. https://aaltodoc.aalto. (March), 58–70. https://doi.org/10.1016/j.ocecoaman.2013.12.006.
fi:443/handle/123456789/15104. Li, P., Cai, Q., Lin, W., Chen, B., Zhang, B., 2016. Offshore oil spill response practices and
IRGC, 2017. Introduction to the IRGC Risk Governance Framework, Revised Version. emerging challenges. Mar. Pollut. Bull. 110 (1), 6–27. https://doi.org/10.1016/j.
EPFL International Risk Governance Center, Lausanne. marpolbul.2016.06.020.
IMO, 2010. Manual on oil spill risk evaluation and assessment of response preparedness. Liu, Z., Callies, U., 2019. Implications of using chemical dispersants to combat oil spills
http://www.imo.org/en/Publications/Documents/Newsletters%20and%20Mailers/ in the German Bight – depiction by means of a Bayesian network. Environ. Pollut.
Mailers/I579E.pdf. (Accessed 28 February 2020). 248, 609–620. https://doi.org/10.1016/j.envpol.2019.02.
ISO, 2009. ISO/FDIS 31000:2009: Risk Management - Principles and Guidelines. Lu, L., Goerlandt, F., Valdez Banda, O.A., Kujala, P., Höglund, A., Arneborg, L., 2019.
International Organization for Standardization. A Bayesian Network risk model for assessing oil spill recovery effectiveness in the
ISO, 2018. ISO 31000:2018. Risk Management – Guidelines. International Organization ice-covered Northern Baltic Sea. Mar. Pollut. Bull. 139, 440–458. https://doi.org/
for Standardization. 10.1016/j.marpolbul.2018.12.018.
Jardine, C., Hrudey, S., Shortreed, J., Craig, L., Krewski, D., Furgal, C., McColl, S., 2003. Mazaheri, A., Montewka, J., Kujala, P., 2016. Towards an evidence-based probabilistic
Risk management frameworks for human health and environmental risks. J. Toxicol. risk model for ship-grounding accidents. Saf. Sci. 86 (July), 195–210. https://doi.
Environ. Health, Part B 6 (6), 569–718. org/10.1016/j.ssci.2016.03.002.
Jensen, F.V., 1996. An Introduction to Bayesian Networks. UCL Press, London. Mazaheri, A., 2017. A Framework for Evidence-Based Risk Modeling of Ship Grounding.
Jensen, F.V., Nielsen, T.D., 2007. Bayesian Networks and Decision Graphs. Springer, New Doctoral thesis. Aalto University. https://aaltodoc.aalto.fi:443/handle/123456
York. 789/26625.
Jolma, A., Lehikoinen, A., Helle, I., Venesjärvi, R., 2014. A software system for assessing Merrick, J.R.W., Van Dorp, R., 2006. Speaking the truth in maritime risk assessment. Risk
the spatially distributed ecological risk posed by oil shipping. Environ. Model. Anal. 26 (1), 223–237. https://doi.org/10.1111/j.1539-6924.2006.00708.x.
Software 61 (November), 1–11. https://doi.org/10.1016/j.envsoft.2014.06.023. Miraglia, R.A., 2002. The cultural and behavioral impact of the Exxon Valdez oil spill on
Juntunen, T., Rosqvist, T., Rytkönen, J., Kuikka, S., 2005. How to model the oil the native peoples of Prince William Sound, Alaska. Spill Sci. Technol. Bull. 7 (1–2),
combatting technologies and their impacts on ecosystem: a bayesian networks 75–87.
application in the Baltic sea. In: International Council for the Exploitation of the Sea Mitchell, R.B., 1994. Intentional Oil Pollution at Sea: Environmental Policy and Treaty
Council Meeting 2005/S:02. Compliance. MIT Press, Cambridge, MA.
Khan, F., Yang, M., Veitch, B., Ehlers, S., Chai, S., 2014. Transportation risk analysis Montewka, J., Goerlandt, F., Zheng, X., 2015. Probabilistic meta-models evaluating
framework for Arctic waters. In: 33rd International Conference on Ocean, Offshore accidental oil spill size from tankers. In: Information, Communication and
and Arctic Engineering, vol. 10. Polar and Arctic Science and technology. https:// Environment: Marine Navigation and Safety of Sea Transportation. https://doi.org/
doi.org/10.1115/OMAE2014-23421. 10.1201/b18514, 231–41.
Kelangath, S., Das, P.K., Quigley, J., Hirdaris, S.E., 2012. Risk analysis of damaged ships Montewka, J., Weckström, M., Kujala, P., 2013. A probabilistic model estimating oil spill
– a data-driven bayesian approach. Ships Offshore Struct. 7 (3), 333–347. https:// clean-up costs – a case study for the Gulf of Finland. Mar. Pollut. Bull. 76 (1), 61–71.
doi.org/10.1080/17445302.2011.592358. https://doi.org/10.1016/j.marpolbul.2013.09.031.
Kingston, P.F., 2002. Long-term environmental impact of oil spills. Spill Sci. Technol. Nelson, J.R., Grubesic, T.H., Sim, L., Rose, K., Graham, J., 2015. Approach for assessing
Bull. 7 (1), 53–61. https://doi.org/10.1016/S1353-2561(02)00051-8. coastal vulnerability to oil spills for prevention and readiness using GIS and the
Klemola, E., Kuronen, J., Kalli, J., Arola, T., Hänninen, M., Lehikoinen, A., Kuikka, S., blowout and spill occurrence model. Ocean Coast Manag. 112 (August), 1–11.
Kujala, P., Tapaninen, U., 2009. A cross-disciplinary approach to minimising the Nelson, J.R., Grubesic, T.H., 2018. Oil spill modeling: risk, spatial vulnerability, and
risks of maritime transport in the Gulf of Finland. World Rev. Intermodal Transp. impact assessment. Prog. Phys. Geogr.: Earth Environ. 42 (1), 112–127. https://doi.
Res. 2 (4), 343–363. https://doi.org/10.1504/WRITR.2009.026212. org/10.1177/0309133317744737.
Knudsen, O.F., Hassler, B., 2011. IMO legislation and its implementation: accident risk, Nevalainen, M., Helle, I., Vanhatalo, J., 2017. Preparing for the unprecedented —
vessel deficiencies and national administrative. Mar. Pol. 35, 201–207. https://doi. towards quantitative oil risk assessment in the Arctic marine areas. Mar. Pollut. Bull.
org/10.1016/j.marpol.2010.09.006. 114, 90–101. https://doi.org/10.1016/j.marpolbul.2016.08.064.
Kontovas, C.A., Psaraftis, H.N., Ventikos, N.P., 2010. An empirical analysis of IOPCF oil Nevalainen, M., 2019. Preparing for the Unprecedented : Moving towards Quantitative
spill cost data. Mar. Pollut. Bull. 60 (9), 1455–1466. https://doi.org/10.1016/j. Understanding of Oil Spill Impacts on Arctic Marine Biota. Doctoral thesis. Helsinki
marpolbul.2010.05.010. University. https://helda.helsinki.fi/handle/10138/301453.
Kuikka, S., Varis, O., 1997. Uncertainties of climatic change impacts in Finnish Nevalainen, M., Vanhatalo, J., Helle, I., 2019. Index-based approach for estimating
watersheds: a Bayesian network analysis of expert knowledge. Boreal Environ. Res. vulnerability of Arctic biota to oil spills. Ecosphere 10 (6), e02766. https://doi.org/
2, 109–128. 10.1002/ecs2.2766.
Kuikka, S., Haapasaari, P., Helle, I., Kulmala, S., Mäntyniemi, S., 2011. Experiences in OECD, 2013. OECD Environmental Risk Assessment Toolkit: Tools for Environmental
applying Bayesian integrative models in interdisciplinary modelling: the Risk Assessment and Management - OECD. http://www.oecd.org/env/ehs/risk-ass
computational and human challenges. In: 19th International Congress on Modelling essment/environmental-risk-assessment-toolkit.htm. (Accessed 6 July 2020).
and Simulation (MODSIM2011, pp. 2135–2141. Parviainen, T., Lehikoinen, A., Kuikka, S., Haapasaari, P., 2019. Risk frames and multiple
Kuikka, S., Helle, I., Lehikoinen, A., Rahikainen, M., Vanhatalo, J., Mäntyniemi, S., ways of knowing: coping with ambiguity in oil spill risk governance in the
Hänninen, M., 2013. Experiences in developing risk analysis tools for oil spills in the Norwegian Barents Sea. Environ. Sci. Pol. 98, 95–111. https://doi.org/10.1016/j.
Gulf of Finland – how to create learning systems to oil spill risk analysis?. In: envsci.2019.04.009.
Proceedings of the Asian Simulation and Modelling Conference (ASIMMOD2013), Pascoe, S., 2018. Assessing relative potential economic impacts of an oil spill on
19-21.1.2013, Bangkok, Thailand, pp. 19–25. commercial fisheries in the Great Australian Bight using a Bayesian Belief Network
Kwakkel, J.H., Walker, W.E., Haasnoot, M., 2016. Coping with the wickedness of public framework. Deep Sea Res. Part II Top. Stud. Oceanogr. 157–158, 203–210. https://
policy problems: approaches for decision making under deep uncertainty. J. Water doi.org/10.1016/j.dsr2.2018.08.011.
Resour. Plann. Manag. 142 (3), 01816001 https://doi.org/10.1061/(ASCE) Pearl, J., 1988. Probabilistic Reasoning in Intelligent Systems: Networks of Plausible
WR.1943-5452.0000626. Inference. Morgan Kaufmann Publishers, San Francisco.
Laine, V., Goerlandt, F., Baldauf, M., Mehdi, R.A., Beşikçi, E.B., Koldenhof, Y., Al Pearl, J., MacKenzie, D., 2018. The Book of Why: the New Science of Cause and Effect,
Quhali, M.A., Backer, H., 2018. OpenRisk guideline for regional risk management to first ed. Hachette Book Group, New York.
improve European pollution preparedness and response at sea. http://www.helcom.
12
Peterson, C.H., Rice, S.D., Short, J.W., Esler, D., Bodkin, J.L., Ballachey, B.B., Irons, D.B., Sperotto, A., Molina, J.L., Torresan, S., Critto, A., Marcomini, A., 2017. Reviewing
2003. Long-term ecosystem response to the Exxon Valdez oil spill. Science 302 bayesian networks potentials for climate change impacts assessment and
(5653), 2082–2086. https://doi.org/10.1126/science.1084282. management: a multi-risk perspective. J. Environ. Manag. 202 (November),
Phan, T.D., Smart, J.C.R., Capon, S.J., Hadwen, W.L., Sahin, O., 2016. Applications of 320–331. https://doi.org/10.1016/j.jenvman.2017.07.044.
Bayesian belief networks in water resource management: a systematic review. Stern, P.C., Fineberg, H.V., 1996. Understanding Risk: Informing Decisions in a
Environ. Model. Software 85, 98–111. Democratic Society. National Research Council, Division of Behavioral and Social
Pollino, C.A., Hart, B.T., 2008. Developing bayesian network models within a risk Sciences and Education, Committee on Risk Characterization, p. 264.
assessment framework. In: 4th International Congress on Environmental Modelling USEPA, 1998. US Environmental Protection Agency. Guidelines for Ecological Risk
and Software. Barcelona, Catalonia, Spain. Assessment. USEPA Risk Assessment Forum. EPA/630/R-95/002F, Washington
Pollino, C.A., Henderson, C., 2010. Bayesian Networks: a Guide for Their Application in (DC).
Natural Resource Management and Policy, vol. 14. Landscape Logic. Technical Uusitalo, L., 2007. Advantages and challenges of bayesian networks in environmental
Report. https://www.utas.edu.au/__data/assets/pdf_file/0009/588474/TR_14_ modelling. Ecol. Model. 203 (3–4), 312–318.
BNs_a_resource_guide.pdf. Uusitalo, L., Lehikoinen, A., Helle, I., Myrberg, K., 2015. An overview of methods to
Rae, A., Alexander, R., 2017. Forecasts or fortune-telling: when are expert judgements of evaluate uncertainty of deterministic models in decision support. Environ. Model.
safety risk valid? Safety Science, Risk Analysis Validation and Trust in Risk Software 63 (January), 24–31. https://doi.org/10.1016/j.envsoft.2014.09.017.
management 99, 156–165. https://doi.org/10.1016/j.ssci.2017.02.018. Uusitalo, L., Korpinen, S., Andersen, J.H., Niiranen, S., Valanko, S., Heiskanen, A.-S.,
Rahikainen, M., Hoviniemi, K.-M., Mäntyniemi, S., Vanhatalo, J., Helle, I., Dickey-Collas, M., 2016. Exploring methods for predicting multiple pressures on
Lehtiniemi, M., Pönni, J., Kuikka, S., 2017. Impacts of eutrophication and oil spills ecosystem recovery: a case study on marine eutrophication and fisheries. Continental
on the Gulf of Finland herring stock. Can. J. Fish. Aquat. Sci. 74 (8), 1218–1232. Shelf Research, Assessing marine ecosystems health, in an integrative way 121
https://doi.org/10.1139/cjfas-2016-0108. (June), 48–60. https://doi.org/10.1016/j.csr.2015.11.002.
Raiffa, H., Schlaifer, R., 2000. Applied Statistical Decision Theory. University of UNCTAD, 2019. Review of Maritime Transport 2019. United Nations, Geneva.
California: Wiley, p. 356. https://unctad.org/en/pages/PublicationWebflyer.aspx?publicationid=2563.
Reed, M., Aamo, O.M., Daling, P.S., 1995. Quantitative analysis of alternate oil-spill (Accessed 6 November 2019).
response strategies using oscar. Spill Sci. Technol. Bull. 2 (1), 67–74. https://doi. UNISDR, M., 2009. UNISDR Terminology for Disaster Risk Reduction. United Nations
org/10.1016/1353-2561(95)00020-5. International Strategy for Disaster Reduction, Geneva, Switzerland. https://www.un
Reed, M., Johansen, O., Brandvik, P.J., Daling, P., Lewis, A., Fiocco, R., Mackay, D., isdr.org/files/7817_UNISDRTerminologyEnglish.pdf. (Accessed 2 March 2020).
Prentki, R., 1999. Oil spill modelling towards the close of the 20th century: overview Valdez Banda, O.A., Goerlandt, F., Kuzmin, V., Kujala, P., Montewka, J., 2016. Risk
of the state of the art. Spill Sci. Technol. Bull. 5, 3–16. management model of winter navigation operations. Mar. Pollut. Bull. 108.
Refsgaard, J.C., van der Sluijs, J.P., Højberg, A.L., Vanrolleghem, P.A., 2007. Uncertainty Valdez Banda, O.A., 2017. Maritime Risk and Safety Management with Focus on Winter
in the environmental modelling process – a framework and guidance. Environ. Navigation. Doctoral thesis. Aalto University. https://aaltodoc.aalto.fi/han
Model. Software 22 (11), 1543–1556. https://doi.org/10.1016/j. dle/123456789/26239.
envsoft.2007.02.004. Van Wee, B., Banister, D., 2016. How to write a literature review paper? Transport Rev.
Regan, H.M., Akçakaya, H.R., Ferson, S., Root, K.V., Carroll, S., Ginzburg, L.R., 2003. 36 (2), 278–288. https://doi.org/10.1080/01441647.2015.1065456.
Treatments of uncertainty and variability in ecological risk assessment of single- Varis, O., Kuikka, S., 1997. Joint use of multiple environmental assessment models by a
species populations. Hum. Ecol. Risk Assess. 9 (4), 889–906. https://doi.org/ bayesian meta-model: the baltic salmon case. Ecol. Model. 102, 341–351.
10.1080/713610015. Venesjärvi, R., 2012. Assessment of the Impacts of an Oil Spill on the Populations of
Regan, H.M., Colyvan, M., Burgman, M.A., 2002. A taxonomy and treatment of Common Guillemot (Uria Aalge) and Long-Tailed Duck (Clangula hyemalis) – an
uncertainty for ecology and conservation Biology. Ecol. Appl. 12 (2), 618–628. Expert Knowledge Based Bayesian Network for the Gulf of Finland. MSc. Thesis.
https://doi.org/10.2307/3060967. University of Jyväskylä, Department of Biological and Environmental Science, p. 82.
Ringbom, Henrik, 2018. Regulation of ship-source pollution in the Baltic sea. Mar. Pol. Venesjärvi, R., 2015. Knowledge-Based Marine Conservation in Oil Spill Risk
98, 246–254. https://doi.org/10.1016/j.marpol.2018.09.004. Management. Doctoral thesis. University of Helsinki. https://helda.helsinki.fi/bitstre
Sebastião, P., Guedes Soares, C., 2007. Uncertainty in predictions of oil spill trajectories am/handle/10138/161292/knowledg.pdf?sequence=3.
in open sea. Ocean. Eng. 34 (3), 576–584. https://doi.org/10.1016/j. Ventikos, N.P., Vergetis, E., Psaraftis, H.N., Triantafyllou, G., 2004. A high-level
oceaneng.2006.01.014. synthesis of oil spill response equipment and countermeasures. J. Hazard Mater. 107,
Sepp Neves, A.A., Pinardi, N., Martins, F., Janeiro, J., Samaras, A., Zodiatis, G., De 51–58. https://doi.org/10.1016/j.jhazmat.2003.11.009.
Dominicis, M., 2015. Towards a common oil spill risk assessment framework – Walker, W.E., Harremoës, P., Rotmans, J., Van der Sluijs, J.P., Van Asselt, M.B.A.,
adapting ISO 31000 and addressing uncertainties. J. Environ. Manag. 159 (August), Janssen, P., Krayer von Krauss, M.P., 2003. Defining uncertainty: a conceptual basis
158–168. https://doi.org/10.1016/j.jenvman.2015.04.044. for uncertainty management in model-based decision support. Integrated Assess. 4,
Shrader-Frechette, K.S., 1991. Risk and Rationality: Philosophical Foundations for 2003. http://resolver.tudelft.nl/uuid:fdc0105c-e601-402a-8f16-ca97e9963592.
Populist Reforms. University of California Press, p. 272. Wirtz, K.W., Baumberger, N., Adam, S., Liu, X., 2007. Oil spill impact minimization
Society for Risk Analysis (SRA), 2018. Society for risk analysis glossary. https://www. under uncertainty: evaluating contingency simulations of the Prestige accident. Ecol.
sra.org/risk-analysis-overview/glossary/. (Accessed 8 September 2020). Econ. 61 (2), 417–428. https://doi.org/10.1016/j.ecolecon.2006.03.013.
Spaulding, M.L., 1988. A state-of-the-art review of oil spill trajectory and fate modelling. Wirtz, K.W., Liu, X., 2006. Integrating economy, ecology and uncertainty in an oil-spill
Oil Chem. Pollut. 4 (1), 39–55. https://doi.org/10.1016/S0269-8579(88)80009-1. DSS: the Prestige accident in Spain, 2002. Estuarine, coastal and shelf science.
Spaulding, M.L., 2017. State of the art review and future Directions in oil spill modelling. Merging Engineering and Science in Marine Environmental Model Applications 70
Mar. Pollut. Bull. 115 (1), 7–19. https://doi.org/10.1016/j.marpolbul.2017.01.001. (4), 525–532. https://doi.org/10.1016/j.ecss.2006.06.016.
13

Journal of Environmental Management

Uploaded by

Copyright:

Available Formats

Journal of Environmental Management

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Journal of Environmental Management

Uploaded by

Copyright:

Available Formats

Journal of Environmental Management 278 (2021) 111520

Contents lists available at ScienceDirect

Journal of Environmental Management

Implementing Bayesian networks for ISO 31000:2018-based maritime oil

Fig. 1. Framework for classifying oil spill response

Table 1 Table 1 (continued )

You might also like