Risk Analysis for Vessel Accident Prevention in Marine Areas: An Accident-Theoretic Perspective on Spatial Aspects of Risk

Abstract

Area-based marine management approaches aim to mitigate the risks and impacts of shipping on human safety at sea and on ecosystems in marine and coastal environments. Through various regulatory initiatives and policy practices, risk assessment has been established as an important element to support decision-making for area-based marine management. This chapter focuses on the use of risk assessment for supporting decisions to manage navigational risks through risk control measures such as the design of vessel traffic separation schemes, the selection and positioning of aids to navigation, and the definition of operational requirements from a vessel traffic management perspective. To facilitate further developments in this domain, this chapter provides a brief overview of risk analysis techniques currently promoted at the international level, and selected approaches proposed in the academic literature are outlined. A discussion is provided on these selected techniques, through the lens of accident causation theories, focusing on how aspects related to the marine space are conceptualized in these techniques. Finally, directions for future research and development are outlined.

1 Introduction

Risk assessment is increasingly embedded in the maritime regulatory context and in the shipping industry to support proactive decision-making to mitigate a variety of risks. For example, at the International Maritime Organization (IMO), the formal safety assessment (FSA) process was adopted in 2002 to support rule-making processes (IMO, 2002) and has been applied to assess the safety level of various ship types and suggest risk control measures based on risk-cost-benefit considerations (e.g., IMO, 2007, 2008). Goal-based ship design standards have also been introduced at the international level (Hoppe, 2005), for which risk-based approaches have been proposed to support the ship design process through structured and systematic analyses. Various methods and tools have been developed to support risk-based design (Papanikolaou, 2009), industry guidelines have been issued to support approval processes (KR, 2015), and risk-based ship design is an active area of academic research (Kujala et al., 2019). Risk assessment is also used in operational practices in the maritime industries. A prominent example of this is the industry practice of using risk assessment techniques and processes to satisfy requirements of the International Safety Management (ISM) Code to “establish safeguards against all identified risks” (IMO, 1993, para 1.2.2.2). Extensive industry guidance has been issued on how to use risk assessment to satisfy this requirement (INSB Class, 2010; IRS, 2018). Other contexts for which risk assessment techniques have been developed and guidance has been issued include maritime pollution preparedness and response (PPR) (Laine et al., 2021; Parviainen et al., 2021) and search and rescue (SAR) (Akbari et al., 2018).

Particularly relevant for area-based management of navigational shipping risk is the provision in the International Convention for the Safety of Life at Sea (SOLAS) that contracting governments should “undertake to arrange for the establishment of VTS [Vessel Traffic Services] and AtoN [Aids to Navigation] Services where the volume of traffic and the degree of risk justifies such services” (SOLAS, 1974). This has led the International Association of Marine Aids to Navigation and Lighthouse Authorities (IALA), in collaboration with national authorities, industry, and academic experts, to develop a toolbox for waterway risk management, to assess this “degree of risk”, and to consider measures to reduce these risks. Associated with this toolbox is guidance on how AtoN authorities can implement risk management within their activities and apply the techniques in the toolbox (IALA, 2022a). Training is also provided to support capacity-building through the IALA World-Wide Academy. The primary stated purpose of waterway risk assessment in this context is to assess navigational risks in maritime areas, to enable risk-based decisions to prevent accidents from occurring. Tools from the IALA Risk Management Toolbox have been used, for instance, to make recommendations to enhance waterway maintenance, install additional AtoNs, and improve operational reporting practices and information exchange (USCG, 2021; Nash Maritime, 2021).

The development of waterway risk identification and analysis techniques and related issues such as proposing risk acceptability criteria (Wang et al., 2022) and metrics to support cost-effectiveness analyses (Ventikos and Sotiropoulos, 2014), as well as proposing stakeholder processes to manage navigational risks in particular sea areas (Haapasaari et al., 2015), are active areas of academic work. Comprehensive literature reviews about maritime waterway risk models have been published by Li et al. (2012), Goerlandt and Montewka (2015), Lim et al. (2018), and Kulkarni et al. (2020).

In the wider academic risk and safety literature, there has been considerable attention to understand accident causation in socio-technical systems. This has led to the formulation of a number of so-called accident causation theories (Qureshi, 2007), which consist of a set of principles and mechanisms through which a socio-technical system transitions from a normal state of operation to a state beyond the normal operation conditions, in which a failure with unwanted safety implications has occurred. Some authors have discussed or analysed the relationship between waterway risk models and accident causation theories (e.g., Hänninen, 2014; Du et al., 2020).

However, no particular focus was given to how spatial aspects of risk are considered in waterway risk models or in the accident causation theories on which they build. In this chapter, this relationship will be explored through a selection of waterway risk analysis models. The conceptualization of the marine space is considered through four aspects: physical, environmental, infrastructural, and organizational, following knowledge obtained from ship accident investigations (Schröder-Hinrichs et al., 2011; Mullai and Paulsson, 2011; Puisa et al., 2018). The physical aspect concerns the layout of the waterway, addressing aspects such as the water depth, channel or fairway width, or the air draught for fairways with bridge spans. The environmental aspects address issues such as currents, waves, wind, and visibility conditions. The infrastructure aspect focuses on man-made structures or technological devices present in the waterway or sea area to facilitate navigation, for instance, aids to navigation (buoys, lights, etc.) and communication technologies to provide information to the different actors in ensuring navigation safety, for instance, Very High Frequency radio, automatic identification systems (AIS) transmission, etc. The organizational aspect focuses on the role and performance of individuals and teams working in the different organizations responsible for ensuring safe navigation, including the vessel’s master and ship personnel, pilots, and vessel traffic services (VTS) operators.

A better understanding of how different waterway risk models relate spatial aspects of risk to the occurrence of accidents can help academics better understand conceptual differences between approaches and reflect on open questions and uncertainties, through which further advancements in developing approaches to support area-based accident prevention can be made. The analysis and discussion can also help practitioners and decision-makers better appreciate the complexity of accident causation and consider some limitations of the existing approaches to waterway risk analysis.

To contextualize the work and to serve as an introduction to readers less familiar with the subject matter, Sect. 7.2 first gives a brief overview of the waterway risk models included in the IALA toolbox, contextualizing this in IALA’s suggested approach to risk management. Thereafter, Sect. 7.3 describes selected waterway risk models in some more detail. This includes a discussion on how these reflect the tenets of a particular accident causation theory and how spatial aspects of risk are reflected in the models. Section 7.4 provides a discussion on the findings, indicates directions for future academic research, and highlights some implications for practitioners and decision-makers.

2 International Recommended Practice for Waterway Accident Risk Management

2.1 IALA Risk Management Process

The overarching guidance document introducing the IALA Risk Assessment Toolbox is the IALA 1018 Guideline for Risk Management (IALA, 2022a). This builds on the IMO FSA process and the ISO 31000:2018 risk management standard (ISO, 2018), linking common practice for regulatory decision-making in the maritime shipping industry, with established risk management concepts and processes across industries. Figure 7.1 illustrates how these two processes are integrated. It shows how the IALA guideline adopts FSA’s analytical steps of hazard identification, risk assessment, risk control options, cost-benefit assessment, and decision-making recommendations, while using the ISO standard’s management processes of communication and consultation, monitoring and review, decision-making, and implementation of risk control options.

Fig. 7.1

Overview of the IALA risk management process, based on IALA (2022a)

The FSA part of the guideline, indicated in light grey in Fig. 7.1, focuses on producing information on waterway risks, identifying possible measures to implement to reduce their risks, and assessing whether these measures are cost-effective to reduce the waterway risks. In the G1018, three different strategies are proposed on how to practically implement these FSA-based processes. Strategy 1 focuses on small-scale assessments (e.g., marking a shipwreck) and includes only steps 1, 2, 3, and 5 of the FSA process. Strategy 2 addresses medium-scale assessments (e.g., new AtoN installations) and additionally contains step 4. Strategy 3 concerns large-scale assessments (e.g., planning a new offshore wind farm). This includes all steps of the FSA process, as well as an iteration aimed to assess and mitigate new hazards which could be introduced into the system, for example, stemming from the implementation of new routing measures. Thus, the selection of which strategy to follow depends on the scope and context of the problem for which a risk assessment is performed.

The other parts of the guideline are based on the ISO 31000:2018 standard and are indicated in white in Fig. 7.1. The parallel process of communication and consultation includes deliberation on the selected strategy (Strategy 1–3), the selection of the risk assessment tools, engaging in discussions with stakeholders as to what should be included in the scope of the analysis, and the collection of data, information, and expert judgments to support the analyses. Monitoring and review is a parallel process that involves periodically tracking and observing the risk management activities, evaluating the actual performance of the organization’s risk management processes, and assessing where and how continuous improvement actions are needed to improve how the risk management process is implemented in the organization. Decision-making should be risk-informed, so that the risk assessment results, the identified risk control options, and (if applicable in the selected strategy) the cost-benefit analyses are considered alongside with other concerns and considerations, such as legal constraints, stakeholder views, and the availability of resources. This decision-making should consider whether risks are at an acceptable level. If not, risk controls should be implemented to reduce the risk levels to be as low as reasonably practicable. Finally, when implementing the new risk control measures, the roles and responsibilities of different stakeholders should be agreed upon and appropriately embedded in organizational practices.

2.2 IALA Risk Assessment Toolbox

The IALA Risk Assessment Toolbox contains selected methods and techniques to support waterway risk assessments of AtoN authorities. This section briefly addresses these tools, outlining their main features and highlighting applicability for executing the different steps of the FSA-based risk assessment process described in Sect. 7.2.1.

The current toolbox contains six freely available methods. Table 7.1 briefly describes these, references the IALA Guidance document, and gives insights in the degree of required resources and skills and the type of output provided. To illustrate, the table shows that using the One Page Risk Assessment (OPRA) tool requires only low resources and skills, whereas in the IALA Waterway Risk Assessment Program (IWRAP) tool, the criteria are medium for both these aspects. This underscores why the IALA World-Wide Academy has developed training courses to support administrations in applying the IWRAP tool. Similarly, a navigation simulator (SIMULATOR) tool can provide quantitative outputs, whereas most other tools result in qualitative outputs, for example, the Ports and Waterways Safety Assessment (PAWSA), Simplified IALA Risk Assessment (SIRA) method, and IALA Risk Management Summary (IRMAS).

Table 7.1 IALA Risk Assessment Toolbox: overview and key features, based on IALA (2022a)

Table 7.2 illustrates the applicability of the IALA risk assessment tools for performing the steps of the FSA-based process shown in Fig. 7.1. As shown in the table, these tools are applicable especially for hazard identification (Step 1), risk assessment (Step 2), and estimating the effects of risk control options (Step 3) related to ports, waterways, and sea areas. These steps can be conducted, for example, by using the PAWSA or IWRAP tools. It is also seen that only SIRA, IRMAS, and OPRA tools are applicable to directly provide decision-making recommendations (Step 5), while none of the IALA tools can be used for cost-benefit assessment (Step 4). For that purpose, however, the original IMO FSA Guidelines can be utilized, and work is ongoing at IALA to provide further guidance on cost-benefit assessment.

Table 7.2 IALA Risk Assessment Toolbox: applicability for different FSA steps, based on IALA (2022a)

Overall, the current IALA Risk Assessment Toolbox provides a fairly comprehensive set of different risk assessment tools, which has been used extensively to produce risk-related information to support subsequently the decision-making process. Finally, the IALA Risk Management Guideline also provides references to other risk assessment tools, if additional information is required.

3 An Accident-Theoretic View on Spatial Aspects of Risk in Waterway Risk Analysis Techniques

3.1 A Brief Outline of Some Common Accident Theories

In the risk and safety literature, a significant question concerns how accidents happen. This is a difficult problem as a large body of empirical and multidisciplinary research has shown that a wide variety of physical, environmental, technological, psychological, and social factors and mechanisms contribute to the occurrence of accidents in socio-technical systems. Over the history of safety science, several attempts have been made to summarize these mechanisms in various levels of abstraction, leading to the existence of a number of so-called accident causation theories. Some of these are very briefly outlined below to support the subsequent discussion. It is important to note that several variations exist of each theory, with often nuanced but potentially significant differences in how exactly accidents are conceived to occur. For the present purposes, a high-level simplified description is considered sufficient to distinguish these theories and the waterway risk analysis methods based on these in the subsequent sections.

One of the oldest conceptions of accident causation is the accident pyramid, which is illustrated in Fig. 7.2a. This relational theory builds on the observation that different events with a gradation in severity levels occur in a system and uses observation to infer that these events are indicative of how susceptible a system is to experience accidents. A typical distinction between different events includes near-misses, minor accidents, and major accidents. The central mechanism in this accident causation theory is thus relational, that is, the notion that there is a more or less stable relationship between the number of near-misses, minor accidents, and major accidents occurring in a system. The model depicts a larger number of near-misses at the base, fewer minor accidents above, and the fewest severe accidents at the pyramid’s peak. This relational model suggests that for every severe accident, there are numerous near-misses and minor incidents. It is considered that understanding and addressing the near-misses at the base of the pyramid can effectively prevent more severe incidents and major accidents (Meyer and Reniers, 2016).

Fig. 7.2

Simplified graphical representation of some prominent accident theories, based on Meyer and Reniers (2016), Reason (1997), Leveson (2016), and Rae (2018)

A second theory is the linear accident causation model, of which one of the most well-known and widely used versions is illustrated in Fig. 7.2b. The central idea is that when a hazard (typically a physical phenomenon) can reach a target (e.g., a person, or a vulnerable environment), this leads to losses (e.g., injury or death, financial costs, environmental damage). Several layers of defence protect the target from being impacted by the hazard, so that this model (often referred to colloquially as the “Swiss Cheese” model) is more commonly referred to as the complex linear model, defence-in-depth model of accident causation, or the epidemiological model in the academic literature (Reason, 1997). Each layer in the model symbolizes a defensive barrier against accidents. When holes, which represent errors or failures in these layers (hence the “Swiss Cheese” analogy), align, a linear progression of events takes place, resulting in the occurrence of an accident. The linear character of this accident theory arises from the sequential alignment of these failures in the defensive barriers, considering how, when one layer fails, it leads to the exposure of the subsequent layer’s weakness, eventually creating a clear path for an accident. This linear view suggests that accidents result from a step-by-step series of failures in these defences. The theory considers accidents in a multi-faceted manner, emphasizing that it is the alignment of multiple failures rather than a single linear cause that allows accidents to manifest. In socio-technical contexts, typically considered failures include unsafe acts by a person (i.e., “human errors”), preconditions for unsafe acts (i.e., physical contextual conditions making it more or less likely for the “human errors” to occur), unsafe supervision (i.e., social conditions in the relationship between the person committing the unsafe act and managerial supervision), and organizational influences (i.e., factors stemming from the broader organizational context in which the work occurs, e.g., related to safety culture or safety training).

A third theory, which has gained significant support in academic and industry contexts in the last decade, is a systems-theoretic accident model based on control system theory. The Systems-Theoretic Accident Model and Processes (STAMP) is a comprehensive accident causation theory that shifts focus away from individual errors to a systemic view of accidents, which are considered to arise from complex interactions between multiple actors within a system. As illustrated in Fig. 7.2c, it takes a multi-layered hierarchy of controllers and controlled processes, along with command signals and feedback loops, as a basis of understanding the system’s functionality, which dynamically aims to maintain a safe state. The interplay between actors and their interactions in the system and the environment in which the system operates are considered key factors in accident causation. Causal factors stem from the mismatch between the system’s design and its operational demands, as well as from the dynamic unexpected adaptations in the system and its operative context. The resultant unexpected interactions, as well as the flawed control and feedback processes, are considered the primary mechanisms from which accidents emerge (Leveson, 2016).

The final accident causation theory considered here concerns risk indicators and is illustrated in Fig. 7.2d. These refer to measurable factors that act as warning signs within a system or environment, highlighting system components, functions, or characteristics which are considered to have a causal relationship to the possible occurrence of an accident. By monitoring and analysing these indicators, vulnerabilities or weak points within a system are proactively identified, enabling pre-emptive measures aimed at mitigating risks and preventing accidents. This accident causation theory does not necessitate combining these indicators as standing in specific relation to one another, and there is no necessary reference to an underlying linear representation of events or systemic interactions. In this sense, risk indicators do not explicitly depict the causal mechanisms through which an accident progresses (unlike the linear or systems-theoretic accident theories) but rely on a more implicit inference about the connection between the status of indicators and accident occurrence (Rae, 2018).

3.2 Techniques Based on Relational Accident Theories: Conceptualization of Space

In the IALA guideline G1018, there are currently no risk models or techniques included which build on the relational view on accidents as represented in the accident pyramid. Nevertheless, several risk models have been developed in the academic literature that build on a hierarchy of traffic conflicts to estimate the accident risk in a maritime traffic area, particularly for collision accidents.

A prime example of such a risk model is the navigational conflict technique proposed by Debnath (2009). The basic idea of this approach is that the severity of non-collision ship traffic encounters can be ranked and that this information can be used to derive the probability of a collision. The procedure to achieve this is schematically shown in Fig. 7.3. First, a vessel conflict operator is constructed using an ordered probit regression model of expert judgments on the risk level in vessel interactions. This risk level of ship-ship encounters uses well-known navigational proximity indicators DCPA (Distance to Closest Point of Approach) and TCPA (Time to Closest Point of Approach), making distinctions between day and night conditions and different vessel sizes. Vessel encounters are interpreted in five risk levels as indicated in Fig. 7.3, and a mathematical operator C(t)|S is defined. Second, this operator is applied in vessel traffic data for encounters involving a vessel conflict, and a measure C’_max is calculated. Finally, the collision probability P_X(A) is mathematically derived from the fitted distribution f(C’_max) to the empirical distribution p(C’_max). The threshold value τ_HR corresponds to the distinction between serious and non-series conflicts, that is, based on the risk score RS_m corresponding to the “high risk” level. Thus, the accident probability P_X(A) is derived from a monotonically decreasing function, which is conceptually similar to an accident pyramid in that this function represents vessel conflicts of increasing severity, with the area of the function above the defined threshold associated with the estimated accident risk.

Fig. 7.3

Overview of the navigational conflict technique to assess collision risk in waterways, based on Debnath (2009)

In this approach, the characteristics of the marine space are conceptualized as not having an influence on the accident risk level. The relative spatial relationship between two encountering vessels is accounted for in the vessel conflict operator as a perceived risk level by expert navigators. However, this relationship is independent from physical, environmental, infrastructural, or organizational characteristics of the marine space in which the vessel encounter occurs.

3.3 Techniques Based on Linear Accident Theories

In the IALA guideline G1018, a clear example of a model based on the linear accident theory is the IALA Waterway Risk Assessment Program (see IALA, 2022b). Furthermore, many risk models based on a linear view of accident causation have been proposed in the academic literature (see, e.g., Du et al. (2020) for a review). To serve as a basis for discussion, IWRAP is briefly explained below, focusing on collision accidents and making links to the closely related pertinent academic literature as appropriate.

This method aims to estimate the frequency of collision accidents in a waterway, as follows:

$$ \mathrm{f}={\mathrm{N}}_{\mathrm{G}}{\mathrm{P}}_{\mathrm{C}} $$

(7.1)

The calculation is based on a simplified sequence of events: the ship-ship encounter and the collision. Hence, the model represents a simple linear causality: first, two ships encounter each other, and second, if they fail to execute successful evasive actions, they collide. In IWRAP, the number of ship-ship encounters N_G is determined through a calculation based on probabilistic information of traffic flow characteristics (obtained through analysis of ship traffic data from the AIS). As an example, the formulation for crossing encounters is given below:

$$ {N}_G^{CR}=\sum \limits_i\sum \limits_j\frac{Q_i^{(1)}{Q}_j^{(2)}}{V_i^{(1)}{V}_j^{(2)}}{D}_{ij}{V}_{ij}\frac{1}{\sin \theta } $$

(7.2)

with V_ij the relative speed between the vessels and D_ij the apparent collision diameter, defined as follows:

$$ {D}_{ij}=\frac{L_i^{(1)}{V}_j^{(2)}+{L}_j^{(2)}{V}_i^{(1)}}{V_{ij}}\sin \theta +{B}_j^{(2)}\sqrt{1-{\left(\sin \theta \frac{V_i^{(1)}}{V_{ij}}\right)}^2}+{B}_i^{(1)}\sqrt{1-{\left(\sin \theta \frac{V_j^{(2)}}{V_{ij}}\right)}^2} $$

(7.3)

${Q}_i^{(1)}$ and ${Q}_j^{(2)}$ are the flow rates of vessels of subclasses i and j. L and B represent ship length and width, V the ship speed, and θ the angle between the waterways. The cross-waterway traffic distributions ${f}_i^{(1)}\left({z}_i\right)$ and ${f}_j^{(2)}\left({z}_j\right)$ integrate to unity for crossing encounters, but for overtaking and meeting encounters, the shape of these distributions affects the number of calculated encounters. The procedure to detect the number of encounters assumes that neither ship takes an evasive action prior to collision (Fig. 7.4).

Fig. 7.4

Overview of the logic underlying the IALA IWRAP tool, based on Pedersen (2010)

A widely used approach to determine the probability of an accident given an encounter, recommended in IALA’s IWRAP approach, is deriving this from accident statistics (Kujala et al., 2009), which leads to a recommended generic value for P_C of 1.2 × 10⁻⁴. Another approach to determine this probability, which has the advantage of allowing an analysis of the effects of risk control options as part of the causal chain, is using Bayesian network (BN) models (see, e.g., Friis-Hansen and Simonsen, 2002; Valdez Banda et al., 2016). A BN is a causal diagram that represents probabilistic relationships between variables and thus captures information how factors and events influence each other. It is a structured way to model dependencies, offering insights into the probability of an outcome based on the interconnected influence of various contributing factors. An important characteristic of BNs is that while complex causal dependencies can be represented, all links are linear, and no feedback loops between factors or events can be accounted for (Hänninen, 2014). Hence, referring to Eq. (7.1), when complex linear dependencies are modelled to estimate the probability of an accident given an encounter P_C using a BN, this approach relies on linear causal pathways influencing the occurrence of an accident by focusing on a sequence of events. This event sequence contains causally dependent events, as well as contextual causal factors influencing the probability of occurrence of the events in the sequence.

In this approach, a relative spatial relationship between two traffic flows of encountering vessels is accounted for in the way the number of ship-ship encounters N_G is determined. The characteristics of the marine space can furthermore influence the accident risk level if such spatial characteristics are represented in the causal network represented by the BN. For example, the model by Friis-Hansen and Simonsen (2002) includes spatially dependent environmental, infrastructural, and organizational factors such as visibility, weather conditions, traffic intensity, and the presence of a VTS. Other models to estimate the accident occurrence for grounding accidents (Mazaheri et al., 2016) and for accidents of vessels engaged in ice-going operations (Fu et al., 2023) similarly include event sequences and various spatially dependent physical, environmental, infrastructural, and/or organizational characteristics of the marine space in which the vessel encounter occurs, which are accounted for as causal factors influencing the occurrence of the events in the sequence leading to an accident.

3.4 Techniques Based on Systems-Theoretic Accident Models

In the IALA guideline G1018, there currently is no waterway risk assessment method based on systems-theoretic accident theories. Similarly, in the academic literature, there has until recently been little focus on developing navigational risk assessment methods based on these theories (Du et al., 2020; Kulkarni et al., 2020). To serve as a basis for discussion, a recently proposed risk analysis methodology for innovative remote pilotage operation in coastal areas, proposed by Basnet et al. (2023), is briefly outlined.

This method relies on the Systems-Theoretic Process Analysis (STPA) and BNs to identify and prioritize causal factors associated with hazards and accidents. The STPA technique is a hazard analysis method based on the STAMP systems-theoretic accident model (see Sect. 7.3.1). In addition to component failures, the technique focuses on understanding hazards occurring due to unsafe interactions of non-failing components. The STPA technique consists of the following steps: (i) defining the purpose of the analysis, that is, defining losses which are unacceptable to the stakeholders; identifying hazards at the system level, which can lead to the identified losses; and specifying the safety constraints to be satisfied to prevent the hazards; (ii) developing a model of the safety control structure (SCS), which is a hierarchical model showing the control actions and feedback loops between system components; (iii) identifying the unsafe control actions (UCAs) by inspecting the SCS using a set of guidewords; and (iv) identifying loss scenarios, that is, the causal factors that can lead to each UCA. In the method by Basnet et al. (2023), the information obtained through the STPA analysis is converted into a BN model, which then allows quantifying the influences of the causal factors to the losses.

Figure 7.5 shows the SCS for the case study of a remote pilotage operation in a coastal area, with further details concerning examples of control and feedback signals between the controllers and controlled processes given in Table 7.3.

Fig. 7.5

Safety control structure of remote pilotage operations in a coastal area, based on Basnet et al. (2023)

Table 7.3 Control and feedback signals in the safety control structure of remote pilotage operations in a coastal area, based on Basnet et al. (2023)

Applying the subsequent steps of the STPA technique to the SCS leads to an identification of the unsafe control actions and the causal factors. These are then causally connected to the system-level hazards, accidents and incidents, and losses using a BN model. The hierarchical structure of this STPA-based BN for the remote pilotage risk assessment in coastal areas is then developed, as illustrated in Fig. 7.6 and Table 7.4, and quantification of the nodes in the BN is performed based on incident and accident data and expert judgment. This allows an analysis of the most important causal factors in preventing the occurrence of system-level hazards, accidents, and losses.

Fig. 7.6

Schematic representation of the hierarchical structure of the STPA-based BN for remote pilotage operations in a coastal area, based on Basnet et al. (2023)

Table 7.4 Examples of losses, accidents and incidents, system-level hazards, unsafe control actions, and scenario causal factors for remote pilotage operations in a coastal area, based on Basnet et al. (2023)

This approach is based on a systems-theoretic view on accident causation, which focuses, as can be seen in Fig. 7.5, on the controllers, controlled processes, and the control and feedback signals between these. These are connected in a non-linear manner, with unsafe control actions, which lead to system-level hazards, being associated with the control and feedback processes as the basis for analysis. In turn, the causal factors are associated with these unsafe control actions, providing reasons why these can occur. In the provided example case study on remote pilotage in coastal areas, it is noteworthy that all causal factors are associated with deficiencies of the controllers or controlled processes (e.g., “fatigue”, “thruster unit failure”) or with the control or feedback signals between these (e.g., “lack of procedures or checklists”, “network failure”). Hence, this method focuses extensively on how infrastructural and organizational aspects of the marine space relate to the waterway risk. However, none of these causal factors are, in the given case study, related to spatially dependent physical or environmental characteristics of the marine space in which the vessels operate.

3.5 Techniques Based on Risk Indicators as Accident Theory

In the IALA guideline G1018, the Ports and Waterways Safety Assessment tool is a clear example of a risk assessment technique based on risk indicators. The PAWSA technique sets the basis for a systematic and highly collaborative process, engaging a diverse group of maritime experts, including vessel operators, regulators, pilots, port authorities, and other relevant stakeholders, to collectively identify, analyse, and evaluate risk factors within the specific waterway (IALA, 2022c).

PAWSA is undertaken in a structured, 2-day workshop where risks and potential mitigation measures are assessed based on inputs from local experts. During these workshops, waterway users and stakeholders discuss and estimate risks levels for 24 different risk factors, organized into 6 risk categories, which are represented in the waterway risk model shown in Fig. 7.7. The participants provide numerical values on a 1–9 scale to quantify their knowledge-based assessments of the status of each waterway risk indicator. Prior to the workshop, a facilitator team gathers, analyses, and prepares a summary of pertinent information to facilitate discussions. This can include, for instance, data on maritime traffic, cargoes, and maritime casualties; official nautical charts and publications; meteorological, hydrographic, and oceanographic records; and (if relevant) proposed or planned projects in or near the waterway in focus.

Fig. 7.7

Ports and Waterways Safety Assessment (PAWSA) waterway risk model, based on IALA (2022c)

The experts’ numerical ratings are weighted based on a rating of each expert team’s expertise, for which a weighting exercise is undertaken first. Subsequently, the baseline risk levels for each of the risk indicators are assessed, that is, the risks in the waterway are assessed without considering the existing mitigation measures. Thereafter, the effectiveness of the existing risk mitigation measures is assessed for each risk indicator. This results in the present level of risk, that is, accounting for the existing risk mitigation. The purpose of this is to evaluate the effectiveness of existing mitigation strategies in reducing the risk level for each factor in the model and to determine whether the risk mitigation strategies already in place adequately balance the resulting risk level. For those risk indicators which are found to be not adequately mitigated or balanced, additional mitigation measures are identified, and it is estimated how effective those new strategies would be to reduce the risk to acceptable levels.

This approach is clearly based on risk indicators, which are used as a basis for an expert-based deliberation on the effectiveness of risk mitigation measures and on the need for additional ones. From Fig. 7.7 it is seen that several indicators concern spatial aspects of the waterway, especially in the categories “Navigational conditions” and “Waterway conditions”. Environmentally dependent indicators include wind conditions, water movement, and visibility restrictions. Physical characteristics include the dimensions, configuration, and bottom type of the waterway and the presence of obstructions in the waterway. Infrastructural and organizational aspects relevant to the marine space are also included in the indicators, where, for example, the configuration of the waterway can include technological support for navigation, and the vessel and traffic conditions allow experts to focus on the performance of vessel operators and other waterway users. In practical applications of the PAWSA approach, such as in (USCG, 2021, 2023), experts reflect on these indicators and make observations and recommendations tailored to the local conditions, for instance, commenting on defective aids to navigation or sharp bends in port fairways (related to the “Configuration” indicator) or on the effects of background lighting on visibility due to the presence of commercial infrastructure near the waterway.

4 Discussion

The analysis of the waterway risk analysis methods promoted at the international level shows that these are based on a diverse set of accident causation theories. Currently, only methods based on complex linear accident and indicator-based theories are included in the guidance on waterway risk analysis by IALA. In the academic literature, methods have additionally been proposed which are based on relational (accident pyramid) and control systems perspectives. These different theoretical lenses through which accident risks are analysed consider the physical, environmental, infrastructural, and organizational aspects of the marine space in different ways.

In the presented method based on the relational accident theory (Sect. 7.3.2), no aspects of the marine space were considered in assessing the risk of ship collision. In the method based on the complex linear accident theory (Sect. 7.3.3), selected aspects of the environmental, infrastructural, and organizational context are considered in assessing the occurrence of a collision accident. In this method, these spatial aspects are conceived to have a direct causal relationship to the occurrence of an accident. In the control systems-based risk assessment method (Sect. 7.3.4), there is a very strong focus on infrastructural and organizational aspects of the marine space, whereas the presented case study does not include causal factors related to physical or environmental aspects of the marine space. In this theory, the causal factors are related to unsafe control actions, which are associated with feedback loops between the different actors in the safety control structure. Even though these causal factors are linked through linear causal relationships in the Bayesian network represented in Fig. 7.6, the underlying mechanisms of accident causation represented through the analysis focus on the non-linear interactions between actors. Finally, in the presented indicator-based methodology (Sect. 7.3.5), all considered aspects of the marine space are represented, that is, physical, environmental, infrastructural, and organizational aspects can be addressed by the experts in the workshops. In practical applications of this method, the local knowledge of domain experts will direct the focus on particular issues relevant to the case.

The fact that different waterway risk assessment methods are based on different mechanisms to understand accident causation and the observation that different methods include different causal factors to assess the occurrence of an accident raises questions about how reliable and valid these approaches are. In other words, if different methods rely on different factors to make statements about the risk level in a waterway or sea area, which method gives the “correct” result? Or more generally, which method can be relied on to provide useful information to reduce the risk level? In this regard, it is perhaps surprising that, even though relatively many methods have been proposed in the academic literature, there has only been limited focus on the reliability and validity of waterway risk assessment methods (Du et al., 2020). Research where the results of different applications of quantitative methods have been compared for selected case studies has furthermore indicated that different methods can lead to significantly different results (see, e.g., Goerlandt and Kujala, 2014; Rawson and Brito, 2022). This observation does not imply that waterway risk assessments have no use, but it should warn decision-makers not to rely exclusively on one quantification-oriented method and to seek information on risks in wider qualitative processes. This is consistent with views in the wider literature on quantitative risk analysis that such analyses are best thought of as systematic arguments (Apostolakis, 2004) and as a basis for discussion and shared understanding by different stakeholders, not as an accurate representation of an underlying true risk (Aven and Heide, 2009; Rosqvist, 2010).

More generally, the complexity of accident causation and the different theoretical lenses through which accidents in socio-technical systems can be understood have proven a rich ground for academic debate. In particular, the relational accident pyramid and the linear accident theory have received significant criticism. These are challenged, for instance, for their adherence to the “common cause hypothesis”, according to which different types of events (near misses, minor accidents, and major accidents) have the same underlying causes, whereas evidence suggests that this hypothesis does not hold in complex systems. Another critique on the relational theory is that evidence suggests that the ratio between the events in different categories is not stable, so that making a mathematical abstraction to derive an estimate of major accidents based on observed near-misses is questionable (Dekker, 2019). The linear theory has been criticized for its simplistic view that complex systems can be divided in separate components and that the occurrence of events can be reduced to a simple failure of such individual components, which does not account for component interaction failures and the conflicting goals of actors in the system (Leveson, 2016). Similarly, while indicators can be a fruitful way to gain insight in the status of systems and serve well as a basis for discussions between stakeholders, the value of such indicator-based methodologies relies on the indicators being representative of the factors which are, in fact, indicative of the occurrence of accidents. Therefore, relying on well-grounded empirical and/or theoretical approaches to decide what indicators to include in the method is important (Rae, 2018).

Currently, there appears to be an academic trend towards understanding accidents based on systems-theoretic views on accidents (Dekker, 2019; Kulkarni et al., 2020). This can be explained because these theories better account for mechanisms which have been observed in accidents, such as the presence of conflicting goals, systems failures due to erroneous interactions between system components, and the multi-level hierarchies of safety controls to dynamically maintain a system in a state of safety. As described in Sects. 7.2 and 7.3, the international guidelines for waterway risk assessment, however, do not currently include methods based on systems theory.

Therefore, it may be a fruitful area of future work to develop and test systems-theoretic waterway risk assessment methods which meet the requirements of different stakeholders, for instance, in terms of the required skill level and the required resources for executing analyses. Other questions which would benefit from future research include how decision-makers use the results of waterway risk assessments to guide their decision-making, for example, whether they use results of different analyses and under which conditions. It is also an open question how decision-makers and stakeholders understand accident causation and how this understanding affects their preference for a specific assessment method or the credibility they assign to its results. As mentioned above, increased research on the reliability and validity of specific waterway risk assessment methods would also be beneficial to better understand the limits and value of the methods in particular case studies. The effectiveness of using different methods as a basis for decision-making would also benefit from future study, especially when they rely on different accident causation theories and when they stress different aspects of the marine space (physical, environmental, infrastructural, and organizational) as causal factors. Also considering the status of the international guideline on waterway risk assessment, particularly Table 7.2, future work on developing guidelines for cost-benefit analysis and risk acceptance would also be useful to support decision-making processes. Whereas some methods allow consideration of risk control options, developing guidelines for selecting risk control options for a particular context and to assess their effectiveness would also be beneficial. Finally, referring to other chapters in this book that address the issue of traditional knowledge and indigenous knowledge systems (see Chap. 6, this volume), a fruitful area of future scholarship relates to how such knowledge can be constructively aligned with the results of risk assessments based on Western scientific views on accident causation.

5 Conclusion

This chapter has provided an overview of risk assessment methods for preventing accidents in waterway areas. The international recommended practice for waterway accident risk management is described, including the overall process and a high-level description of the tools included in the IALA Risk Assessment Toolbox. Thereafter, a brief description of some common accident theories is presented, and selected examples of waterway risk assessment methods are outlined to illustrate how these accident theories are reflected in practical risk assessment models and techniques.

The conceptualization of space in terms of physical, environmental, infrastructural, and organizational characteristics relevant to accidents in marine areas is investigated for a selected set of examples for the relational, complex linear, systems-theoretic, and indicator-based accident causation theories. Models based on different theories do not put equal focus on these spatial characteristics, and the related causal factors have a different relation to the accident causation mechanisms in the different theories. In the presented vessel traffic conflict technique, which is based on the relational theory (accident pyramid), no features of the marine space are accounted for in estimating the occurrence of accidents. In the IWRAP model, which is based on the complex linear theory of accident causation, selected environmental, infrastructural, and organizational factors are included in the model. These spatial characteristics are conceived as having a direct causal relationship to the occurrence of a sequence of events, which ultimately results in a navigational accident. In the presented case study on risk assessment for remote pilotage operations, which is based on a control-systems theoretic view on accidents, there is a very strong and elaborate focus on infrastructural and organizational aspects of the marine space. The focus of this analysis is on the unsafe controls in the non-linear interactions between actors in multiple hierarchical levels. In this view, the occurrence of these unsafe control actions is perceived to be influenced by infrastructural and organizational aspects of ship operations in a given marine space. However, in the presented case study on remote pilotage operations, no causal factors related to physical or environmental aspects of the marine space are considered. Finally, in the indicator-based method, no specific causal mechanism is defined through which accidents are considered to manifest. The indicators in the PAWSA method allow experts of various stakeholder groups relevant to the given waterway to identify challenging conditions in the area in a collaborative workshop setting. In this method, all characteristics of the marine space (physical, environmental, infrastructural, and organizational) can be addressed through discussions on the 24 risk indicators included in the PAWSA waterway risk model.

The presented findings have several implications. First, users of the results of waterway risk assessments should understand and consider the limitations of risk assessments. These should primarily be understood as systematic arguments about hazards and risks in waterways, and additional information beyond the risk quantifications should be sought to support decision-making, including the evidence base on which the risk assessments build. Second, considering that waterway risk assessments rely on different accident causation theories and include different causal factors related to the marine space, questions about the reliability and validity of waterway risk assessment methods would benefit from increased academic work. Similarly, increased focus on the selection, use, and credibility of waterway risk assessments in practical contexts would be a fruitful research direction. Finally, future work can be directed towards developing new approaches based on systems-theoretic views on accident causation; developing and documenting guidance on cost-benefit analysis, risk acceptance, and selection of risk control options for waterway risk management; and proposing approaches and processes to constructively align the results of waterway risk assessment based on Western scientific views on accident causation with traditional knowledge and indigenous knowledge systems.