A methodological approach for mapping and analysing cascading effects of flooding events

ABSTRACT

Current local or regional flood risk assessments, as required by the EU flood risk directive, rarely account for cascading effects due to interdependencies between critical infrastructures. However, it is essential to consider these effects, as they may severely impact areas outside the immediate flood risk area. The main purpose is to present and problematize a method (AB-CEM) for mapping and analysing cascading effects due to floods, aiming at also being relevant for other spatially widespread hazards. The method development and the pilot study, in Sweden, reveal that there is a prominent practical need for methods for mapping and analysing critical infrastructure interdependencies and cascading effects. Another key finding is that the method process and its results can serve as an important basis for decision-making about proactive and reactive efforts related to geographically extensive hazards. We further conclude that there is a recurring problem regarding sensitive and secret data. More specifically, the conflicting interests of information availability and information security concerning critical infrastructures, which needs to be resolved at the national level and communicated through clear guidelines. The method is a much-needed step towards accounting for cascading effects of floods in practice.

KEYWORDS:

• Cascading effects
• flood risk assessment
• critical infrastructure
• interdependencies
• GIS
• Sweden
• flood directive

1. Introduction

Flooding is one of the leading causes of disaster damages worldwide (Wallemacq & House, 2018), and the damages are likely to increase due to climate change (IPCC, 2014). To manage floods in the EU, the European Council issued a directive on the assessments and management of flood risks in 2007 (the Flood Risk Directive or FRD in short), which obligates each EU member to perform a national assessment and develop a plan to manage flood risk. The assessments are structured around flood hazard and flood risk maps, where flood extent, water depth, affected inhabitants, and affected economic activities are shown (European Union: European Commission, 2007). Floods can lead to both direct consequences for society, caused by water in the inundated area, and indirect consequences, which are induced by the direct consequences. The induced indirect consequences are often referred to as cascading effects (Nones & Pescaroli, 2016). Cascading effects are common when Critical Infrastructure (CI) is damaged, which can lead to indirect consequences that extend beyond the inundated areas and affect parts of society that are far from the primary hazard. One factor that contributes to this phenomenon is that CI often spans over large geographical areas. Another factor is that failures in one critical infrastructure can propagate to other infrastructures due to interdependencies (Johansson et al., 2015; Luiijf et al., 2010; Rinaldi et al., 2001). These adverse effects have been observed for example in the 2007 floods in the UK, where people were affected far beyond the flooded areas, rendering 42,000 people without power and 350,000 without freshwater (Pitt, 2008). Another study, of New York after Hurricane Sandy, revealed that a flood event directly affected 11% of the transportation system, and indirectly affected an additional 19% of the system by comparing GIS-layers of the direct consequences (e.g. flooded buildings or subways) with the reported failures (Haraguchi & Kim, 2016). These findings indicate that indirect consequences can be substantial and, in this case, the major part of the total consequences in a flooding event. However, national flood risk assessments tend to only consider the direct consequences and omit the cascading effects (Kystdirektoratet, 2015; MSB, 2013; SEPA, 2015). Thereby, they exclude the significant indirect consequences a flood can give rise to, which in turn can lead to misguided decisions.

It is a challenging task to take cascading effects into account in risk and vulnerability assessments at a regional and local level. Firstly, while there are methods for spatial and network analyses to analyse cascading effects (Pant et al., 2016; Poljanšek et al., 2012), they often require a certain level of expertise and are not readily applicable for crisis management practitioners given available resources at the local or regional level. Moreover, such methods tend to focus on national-level infrastructure. Secondly, there is a general lack of suitable data for analysing cascading effects, which makes it challenging to apply advanced methods and techniques (Nones & Pescaroli, 2016). Thirdly, the data and expertise required are scattered across a multitude of both public and private organizations, making it a challenge to conduct all-encompassing analyses from a data collection, as well as a collaborative and security point of view.

Other challenges are associated with the FRD specifically. For example, the directive does not provide any guidance on which metrics should be used to measure flood risk, for example, whether individual or societal risk measures should be considered (Mostert & Junier, 2009). Additionally, the FRD only specifies an assessment of consequences for the rather overarching categories of human health, the environment, cultural heritage and economic activity (European Union: European Commission, 2007). However, some of these types of consequences, in particular those that are intangible in nature, such as psychological trauma or loss of trust in authorities, are more challenging to assess or compare (Merz et al., 2010). Because most rivers cross administrative borders (e.g. municipal, county or national borders), studies on implementations of the flood risk directive call for standardization of flood risk assessments and presentation (van Alphen et al., 2009), and identify the need to reduce technical differences between countries, for example, how GIS data is stored (Müller, 2013). Müller (2013) also argues that data availability is critical for successful flood risk management.

There is usually some awareness of interdependencies within critical infrastructure organizations, but it is often a rather tacit form of knowledge. Therefore, expert knowledge needs to be more accessible, structured, and shareable to enable a better analysis of interdependencies between systems between infrastructures and the impact of these in various threat scenarios. One approach is to extract interdependency data from existing databases of spatial data or past events. However, these methods are quite cumbersome and have been designed towards highly specific sets of databases and types of data (Abdalla et al., 2007; Cheng, 2017; Johansson et al., 2015; McNally et al., 2007; Shih et al., 2009).

Thus, there is an overall lack of practical methods to gather data on and analyse cascading effects for flooding events. This paper aims to contribute towards this by presenting a method for mapping and analysing potential cascading effects due to floods in a local and regional context. Moreover, it aims to be relevant to the practical work with the European FRD. The development process behind the method is based on influences from design theory (Carlsson et al., 2011) and performed in close cooperation with relevant stakeholders. Since an essential part of the work with FRD is GIS-based, the method also facilitates the use of GIS as an analytical and visual tool. In short, we term this method AB-CEM (Area-Based Cascading Effect Method).

2. Background

This section presents scientific and grey background literature related to the EU flood risk directive, interdependencies and cascading effects, and GIS-based consequence analysis.

2.1. EU flood risk directive

The EU directive on the assessment and management of flood risks requires EU member states to submit a report on flood risks every six years. The six-year cycle is divided into three steps: (1) perform a preliminary flood risk assessment, (2a) create flood hazard maps and (2b) flood risk maps, and (3) develop flood risk management plans (European Union: European Commission, 2007). The preliminary assessment looks at river basin characteristics, such as water and elevation levels, physical measurements, and past floods, to assess the potential impacts of future floods. For areas with significant flood risks, the member states are required to produce hazard maps of flood extent, water levels, and preferably water velocity. The maps must include floods with a low probability (or extreme event scenarios), a medium probability (defined as less or equal to a 100-year return period) and a high probability where it is deemed appropriate (European Union: European Commission, 2007). In Sweden, hazard maps are produced using a theoretical maximal flood, a 100-year flood or a 50-year flood for the respective probability levels and the resulting inundation levels are output as a raster with deterministic values (MSB, 2013). The hazard maps are then used to make a flood risk map of potentially affected inhabitants, economic activity, potential pollution sources, and other relevant consequences (European Union: European Commission, 2007). Lastly, flood risk management plans coordinated on river basin level are developed where cost–benefit, flood extent, floodplains, soil, water management, spatial planning, land use, nature conservation, and navigation and port infrastructure are taken into account. There are, however, no regulations or guidelines on which types of consequences should be prioritized or how they could be compared in, for example, a cost–benefit analysis.

The implementation of the FRD in Sweden is led by the Swedish Civil Contingencies Agency (MSB). This agency is also responsible for coordinating and supporting DRM efforts in Sweden from an all-hazards perspective. In practice, much of the work related to the FRD is performed at the local and regional levels, as the FRD is structured around catchment areas. The FRD risk assessments do generally take some consideration of CI into account. For example, in Sweden, impacted infrastructures such as hospitals, important roads or police stations within flood hazard areas are identified (MSB, 2013). In Denmark, simple descriptions of infrastructure in the flood hazard area are given (Kystdirektoratet, 2015), that is, mainly identification of infrastructures at risk. Scotland takes it one step further by also assessing direct economic damages on critical infrastructures like roads, railways and other utilities (e.g. electric substations, radio towers, and mines) (SEPA, 2015). A general drawback with these examples is that they do not consider the cascading effects that can occur when CIs are flooded.

2.2. Interdependencies and cascading effects

A dependency can be described as a relationship between two infrastructures, where the state of one correlates with the state of another (Rinaldi et al., 2001). An interdependency is then technically a relationship when two infrastructures are dependent on each other, either directly or through feedback-loops (Rinaldi et al., 2001). However, it is common to use interdependency for both kinds of relationships, as will be done in this paper. Whenever interdependencies exist, and infrastructure is directly affected by an event, it can lead to indirect effects in other infrastructures. The propagation of effects between CI is often referred to as a cascading effect or a cascading failure (Arvidsson et al., 2015; Hilly et al., 2018; Johansson et al., 2015; Nones & Pescaroli, 2016; Rinaldi et al., 2001). Interdependencies can thus be seen as the mechanism that allows for cascading effects.

There are several methods to analyse interdependencies and cascading effects between CIs. The methods can be divided into three broad categories based on their approaches: expert-based methods, empirical methods or simulation-based methods.

The approach of expert-based methods is to consult CI experts and inquire them to estimate the strength of interdependencies between CI, for example, through surveys (Laugé et al., 2015; Moon et al., 2015; Toubin et al., 2012), interviews (Chang et al., 2014) or workshops (Chang et al., 2014; de Bruijn et al., 2016; Deltares, 2018; Moon et al., 2015). Although it is common to support some form of analysis, the focus instead lies on describing and mapping interdependencies and estimating cascading effects for plausible future events.

Empirical methods often focus on gathering information about critical infrastructure interdependencies based on past events. A frequently used approach is to gather data about events through, for example, newspapers, the internet or official incident or investigation reports, structure the data in some form of database, and then analyse and draw a conclusion about the studied phenomena. The data is often analysed statistically to investigate, for example, the prevalence of cascading effects between CI, the strength of cascading effects or the kinds of interdependencies involved in the event (Johansson et al., 2015; Luiijf et al., 2010; McDaniels et al., 2007).

The simulation-based methods use models of CIs to study the effect of interdependencies and how cascading effects arise. Simulation methods typically rely on some form of expert-based or empirical data as input and use it to construct one-to-one scale, infrastructure models. This approach is generally more demanding compared to the expert-based and empirical methods and often requires more profound expertise within, for example, engineering, mathematics or computer science. Several kinds of models are relevant for studying interdependent CIs, for example, Agent-based models (Basu et al., 1998; Dudenhoeffer et al., 2006; Ehlen & Scholand, 2005; Kaegi et al., 2009), System-dynamic models (Brown et al., 2004; Min et al., 2007; Sterman, 2000), Network-based models (Espada Jr et al., 2015; Hines et al., 2010; Johansson et al., 2011; Johansson & Hassel, 2010; Lee et al., 2007; Viavattene et al., 2015; Zio & Sansavini, 2011), National Economic Models (Barker & Santos, 2010; Haimes et al., 2005; Kelly et al., 2016; Setola, 2007; Svegrup et al., 2019; Xu et al., 2012) and Flow-based models (Johansson et al., 2017; Svegrup et al., 2017).

2.3. GIS-based consequence analysis

Applying spatial analysis in a GIS environment is a well-established approach for predicting flood hazards and estimating flood consequences, both in scientific literature and in FRD assessments. Often, predicting a flood hazard relies on technical methods such as sophisticated hydraulic modelling which can be integrated with GIS, for example, Arc Hydro (Maidment & Morehouse, 2002), or performed in conjunction with specialized software, for example, MIKE 11, HEC-RAS or LISFLOOD-FP (Jain et al., 2018).

A common approach to estimating consequences with GIS is to perform an overlay analysis and thus be able to estimate the number of affected persons and objects within a hazard area, for example, residential population, buildings, roads, railroads, hospitals, water treatment plants, cultural heritage sites, national parks (Arvidsson et al., 2021). This approach is applied for various hazards with a clear spatial extent such as floods (Haraguchi & Kim, 2016; Johnston et al., 2014; MSB 2018), earthquakes (Poljanšek et al., 2012; Tamaro et al., 2018), storms (Zhang et al., 2018) or explosions (Armenakis & Nirupama, 2013; Kulawiak & Lubniewski, 2014). Estimating the damage inflicted on an object can be based on, for example, assumptions, expert judgements, statistical averages or fragility curves. Estimating the number of objects within an area is a simple GIS operation and requires relatively limited data input, for example, land use or demographic databases may be sufficient. A common limitation is that these approaches usually do not include analyses of cascading effects.

The GIS toolset also shows potential for supporting more advanced approaches to consequence analysis such as modelling and simulating interdependencies and indirect consequences to CIs (Arvidsson et al., 2021). A few studies have combined spatial analysis with CI modelling and simulation approaches (Espada Jr et al., 2015; Pant et al., 2016; Poljanšek et al., 2012; Zhang et al., 2018). These examples of modelling or simulation of indirect consequences in flooding events have their advantages, as they can account for cascading effects in a more detailed fashion. However, a drawback is that they often require expert knowledge, large datasets, and a substantial amount of computational resources. For example, to model the UK railway system and its interdependencies to other infrastructure in detail, a quite extensive network with more than ten thousand nodes and edges was used (Pant et al., 2016).

3. Framework for development

This section presents the general framework for the development process of the proposed method.

Because a method, essentially, is an artificial construct, the development of methods can be seen as a design problem (Simon, 1996). Carlsson et al. (2011) suggest that there are four activities associated with design research: (1) to identify and clearly articulate problem situations and desired outcomes, (2) to review existing theories, knowledge and data that can be used to support the design proposition, (3) to propose or refine design theory and knowledge, where the design is formulated extensively and grounded in existent theories, and (4) to test the design theory and knowledge. In this paper, the focus is on proposing a method rather than proposing general design theories. Therefore, ‘design theory and knowledge’ in the third and fourth activity above were replaced with ‘method’. The following section describes how our method development was guided by and relates to these design research activities; an overview is presented in Figure 1.

Figure 1. Schematic overview of the method development phases (1–5). Ellipses represent research activities, and rectangles represent inputs to or outputs from the research activities. The colours represent different types of research activities; blue for formulating design problem and desired outcomes, green for reviewing existing methods and practical work, orange for proposing or refining the method, and purple for incremental testing of the method. For each workshop, the represented organizations are listed along with the number of stakeholders in parenthesis.

4. Development process

This section presents the research activities that were conducted in the process of developing the AB-CEM, from the initial problem formulation to the final design.

The problem was identified and formulated in collaboration with the Swedish Civil Contingencies Agency (MSB), the agency that is responsible for coordinating and supporting FRD and DRM efforts in Sweden. As outlined in the introduction, the problem is that ongoing work relating to the FRD in Sweden rarely take any consideration for cascading effects in the flood risk assessments. Therefore, the desired outcome of the method was to provide a structured and practically applicable approach to analyse cascading effects that is relevant for the work with the FRD, which also constitutes the first design criterion. A second criterion was that the method also should relate to the Swedish work with the Directive on European Critical Infrastructures. A third criterion was that the method should take advantage of geographical information systems (GIS). The fourth criterion was that the method should be flexible enough to be applicable to various hazards and in different contexts so that it can support cascading effect analyses and more detailed consequence analysis in risk- and vulnerability-oriented work in general. The method was designed based on these criteria.

The development process was divided into five phases, the first four was accompanied by a 3–6-hour long workshop and the fifth was a pilot study in southern Sweden. The general structure of the workshops was guided, plenary discussions on the focal topic(s) with 7–11 participating stakeholders from various administrative levels. A list of stakeholders attending the workshops is presented in Figure 1. The national level was represented by the Swedish Civil Contingencies Agency (MSB), Statistics Sweden (responsible for collecting official statistics) and Statkraft (largest renewable energy provider in Europe). On the regional level, the stakeholders consisted of safety officers and regional planners from county administrative boards and the police. Lastly, the local level was represented by safety officers from municipalities, rescue services and local utility companies for water, sanitation and power.

The first development phase consisted of a review of relevant scientific and grey literature, which is presented in the background section of this paper. To complement the scientific knowledge and to gain insight into the current flood risk assessment practices and challenges in Sweden, input was attained through a workshop on a regional flood consequence analysis arranged by practitioners from the Lagan river catchment area (a river located in southern Sweden). A secondary objective of the workshop was to establish connections with regional and local practitioners to identify relevant partners. Based on the reviewed literature and the insights from the workshop, an initial design of the method was proposed.

In the second phase, the workshop focused on evaluating the initial design and selecting a test case with input from both MSB and regional practitioners. The Värnamo municipality was deemed suitable as a test case since the national flood risk assessment identifies it as one of the 18 areas in Sweden with the highest flood risk (MSB, 2011). The test case was used to discover and address the logical or practical flaws of the method through internal testing. The internal test was performed using input from the second workshop, the local risk and vulnerability analysis, the regional flood risk assessment, and other documents supporting these analyses. This process led to the first refinement, and a second version, of the method.

The third development phase aimed to test the feasibility and the applicability of the second version of the method. At the workshop, the method together with a short example application from the test case was presented to practitioners from Värnamo municipality and counties within the Lagan river catchment area. The discussions focused on the relevance of AB-CEM, on local managers’ ability to access or gather the required information and on perceived challenges to apply the method. The input from this workshop was used to produce a third version, which also was summarized in a report.

The fourth development phase was performed by presenting the method in a fourth workshop with stakeholders, and revisions were completed based on minor inputs from this workshop and written feedback on the report. At this point, AB-CEM was considered ready for a small-scale, applied pilot study.

The fifth development phase consisted of a pilot study where AB-CEM was applied in a municipality in southern Sweden, Karlshamn, and evaluated by persons external to the development group (Andersson & Carlström, 2020). In the study, AB-CEM was used to map cascading effects of a flood and a drought scenario by interviewing local authorities and CI operators. The draught scenario was added to explore the applicability of the method for different hazards. The included stakeholders are Karlshamn Municipality (local government), Karlshamn Harbour (transportation and fuel supply), Södra Cell Mörrum (pulp and mass industry contributing to local district heating production and regional electricity production), Uniper – Sydkraft Thermal Power (local district heating and regional electricity production) and Karlshamn Energy (electricity distribution, district heating distribution, broadband internet, water distribution and sanitation). Based on the pilot study, minor changes and tuning of the method were implemented. The final version of the method is presented in the following section.

5. The resulting method: AB-CEM

The fundamental aim of the AB-CEM is to improve existing flood risk assessments, by supporting the mapping of potential cascading effects (i.e. supporting improved analyses in step 2b in the FRD). Thus, a prerequisite is that a scenario and an impacted area have been identified in a hazard assessment, for example, a flood, and that hazard maps are available. However, the method also aims at supporting cascading effects analyses for other types of hazards and threats, which means that information gathering for this purpose is also supported by the method. The method focuses on making local and regional expert knowledge regarding the effects of flooding more explicit through the use of interviews or workshops with both societal and infrastructure actors affected by a potential scenario.

AB-CEM consists of eight steps that should be performed for the scenario, as shown in Figure 2. The steps aim to provide a structured workflow for local and regional actors who wish to include cascading effects in their flood risk assessments. Additionally, supporting documents were developed to assist the work further. The following sections describe the steps and documents in further detail. The examples and the conclusions drawn from the studied test case are also included to aid the description of the method. The test case was used for the sake of method development, hence no results from the application of the developed method are presented.

Figure 2. Overview of AB-CEM.

5.1. Step 1 – input from existing material and experts

The first step is to collect available material on the scenario and to engage with local experts such as safety managers and CI operators in the analysed area. The material can consist of, for example, previously performed consequence analyses, flood maps, risk maps or related risk and vulnerability analyses. The primary goal of this step is to reuse existing material and knowledge. It is also of importance to complement the written material with input from experts, for example, through interviews, workshops or informal talks, to produce as reliable and comprehensive input as possible for the analysis. Here it is also of importance to clarify if the flood potentially triggers other hazards, such as landslides, ground collapses or ground heaves (Gill & Malamud, 2014). If so, the method should be also be applied for these hazards to achieve as a holistic account of the consequences as possible.

5.2. Step 2 – identify directly affected CI

In the second step, the directly affected CIs are identified based on the flood scenario; however, in principle, any spatially widespread event can be used. The impacted CIs can be identified in different ways, although GIS is highlighted as an essential toolbox to identify directly impacted CIs, for example, through regional flood risk maps like the ones produced within the FRD. The information gathered in Step 1 should be helpful, but often it is necessary to consult local or regional experts such as CI operators, rescue services, municipal departments or county administrative boards.

The directly affected CIs constitute the foundation for identifying and mapping cascading effects. Although it is reasonable to prioritize certain CIs for further analysis, the recommendation is to document all directly affected CIs, even if the consequences seem insignificant at face value. Even less severe consequences in an individual infrastructure have the potential to cause significant cascading effects, or, at least, to contribute to an increased understanding of cascading effects.

5.3. Step 3 – register CI

In the third step, the aim is to provide context for the identified consequences and interdependencies. Therefore, it is necessary to include a description of the initiating event and of how the CI is affected by the particular event, that is, is it directly affected by the initiating event or indirectly affected through interdependencies with other CIs? Additionally, registering which sector the CI belongs to, what activity within the CI is affected and a unique identifier of the affected CI provides further context and can be used for organizing the results. Lastly, ownership and contact information should be noted for further inquiries or updates.

5.4. Step 4 – consequence assessment

With the context in place, it is time to assess the consequences of a disruption in the registered CIs. The assessment is preferably done together with experts of the respective systems, that is, the CI operators managing the systems at the regional or local level. When assessing the consequences that arise, one should consider the effects not only for the CI but also what the impact might be on human health, economic activity, the environment or cultural heritage, as is regulated in Sweden (MSB 2018). It should be noted that these categories have quite broad definitions and therefore should include intangible effects such as injuries, psychological trauma or loss of trust in authorities, in addition to more tangible effects, such as quantifiable economic consequences. As such, these kinds of consequences should be documented as far as is reasonably practicable using both qualitative and quantitative descriptions. Other countries might regulate flood risk assessments differently, which could affect the necessary consequence categories.

Since CIs often have wide service areas (e.g. hospitals or emergency services) or are a part of geographically widespread networks (e.g. power transmission lines, fibre networks, railways, or highways), a disruption in one part can give rise to consequences far beyond the flooded area. Therefore, the approximate geographical extent of the impacts should be mapped and geocoded, if possible. The mapping can be used to support visualization of the consequences and strengthen the analysis, for example by using it to identify additional indirect consequences for CIs and society outside of the immediate flood area. A practical way to approximate the extent is to bring maps to the interview or workshop with experts on the CI so that the interviewees can draw the estimated affected area on a map, which later can be digitalized in GIS.

An important parameter to assess is the time it will take until CIs can return to normal operations given the analysed scenario. Although it is difficult to estimate the duration of a failure with precision, it is an essential factor to consider when identifying cascading effects. For example, a short power outage might not affect a hospital with backup generators initially, but a more prolonged power failure could prove problematic.

It is also helpful to note if the consequences are specific to the currently analysed scenario, for example, a flood, or if they could arise during other kinds of scenarios, for example, an earthquake or antagonistic attack. This kind of information is vital for the reusability of the gathered material for other scenarios. For example, the consequences of a closed preschool might be similar even if the causes differ, as both a flood or a long-term power blackout can force a preschool to close. In the example of a closed preschool, it is reasonable to assume that the consequences are relatively independent of the event that caused it, and therefore, the consequence analysis can be reused for other scenarios.

During interviews and workshops, it is not uncommon that there will be suggestions on measures that can help mitigate the consequences or shorten the duration of the consequences. They should also be written down, as they can prove to be valuable as input to the flood risk management process. Both technical measures, for example, walling in a particularly vulnerable electric substation, and organizational measures, for example, having a plan for relocating preschool children, should be considered.

5.5. Step 5 – interdependencies with other CIs

Based on the consequence analysis, this step aims at assessing the effects the failure of the analysed CI might have on other CIs, and if the CI is dependent on another CI. The effects on other CIs are central for this analysis; therefore, it is essential to make interviewees or workshop attendees reflect on their organization’s interdependencies. A straightforward question such as ‘What are your interdependencies with other critical infrastructures?’ is rarely fruitful. Instead, it is advisable to inquire based on counterfactual questions, for example, ‘Given a flood of 30 cm, what services or infrastructures would you rely on to restore your operations to normal?’ or ‘How would your organisation be affected if you lose a service “x” from CI “y”?’. The aim is to encourage counterfactual reasoning, that is, to evaluate alternative, not factual, possibilities and their consequences (Hendrickson, 2008).

5.6. Step 6 – compile a single CI into form

When Steps 2–5 have been performed for a particular CI, the information should be compiled into a form that has been developed. The form provides a structured overview of the consequences for the studied CI and enables a systematic gathering of the data. The main headlines in the form are presented in Table 1. In practice, the form can also serve as a guiding document during interviews or workshops in Steps 2-5. The dashed line from Step 6 to Step 2 in Figure 2 indicates that Steps 2–6 should be repeated for every directly affected CI. The dashed line from Step 6 to Step 3 in Figure 2 indicates that Steps 3–6 should be repeated for every indirectly affected CI. In some instances, an indirectly affected CI will lead to cascading effects of the second, third, fourth, or higher order. To map the entire dependency-chains and indirect effects is possible, but available time or other resources might be factors limiting the depth that can be achieved. In such instances, it is recommended to analyse the first and second-order effects. Eliminating or mitigating cascading effects in the first or second order, will likely also reduce many higher-order effects, as they per definition rely on other cascading effects to occur. In Step 6, a geographical description of the consequences should be included, if this has not been done previously.

Table 1. Headlines of the form, description of the content, and fictive examples.

	Headline	Content	Example
Step 3	CI sector	The sector the CI belongs to. In Sweden CIs are categorized in 11 sectors.	Energy
	CI activity	What activity in the sector is affected?	Electric distribution
	Unique identifier	E.g. a name, number or other kind of designation	TST01, NST109, NST112, NST124
	Owner organization or name	The organization responsible for the affected CI	‘B-on’
	Contact information	Contact person at the responsible organization	‘John Smith’
	Description of the event	A summary of the scenario	A 100-year flood in Lagan river, which affects several CIs and inhabitants in Värnamo
	Description of the effects on CI	A summary of the effects on the CI	Power supply is affected as a transformer station would get flooded together with several substations
	Directly or indirectly affected	If the CI is directly affected by the scenario or indirectly affected through interdependencies	Directly affected
Step 4	Duration	The best estimate of the duration of a failure, given the scenario or impacts from other CIs	Estimated to 36–72 h, given the scenario. The company estimates a general restoration time for a substation of 12–36 h and a transformer station to 24–72 h.
	Societal consequences	Impact on human health, economic activity, the environment or cultural heritage from the studied CI	Human health: 200 households without power Economic activity: Restoration work, change of components etc. estimated to 500–750 thousand SEK for the energy company. 30 local businesses without power. Environment: no estimated impact Cultural heritage: no estimated impact
	Geographical extent	At minimum a description of the extent, preferably a GIS layer	Most neighbourhoods in central and eastern parts of Värnamo will be affected (see Figure 3)
	Suggested measures	Suggested measures from the operator	Protective walls around the main transformer and installation of pumps. Substations deemed too expensive to protect.
Step 5	Interdependencies with other CIs	A description of the interdependencies that were identified in the interviews or workshops. Which dependencies do the CI have and how will the disruption affect other CIs?	Dependent on accessible roads and working communication, e.g. emergency channels. One, preschool, ‘The Pear’, and two health clinics, ‘South’ and ‘East’, are dependent on power from affected stations.

5.7. Step 7 – compile all CIs into a database

In this step, the information from the individual forms is gathered in a database. The database is a necessary step towards an overview of the cascading effects. An Excel workbook or similar should be sufficient for the needs of a local or regional flood risk assessment. Excel also benefits from being a simple and widely known software and can easily be exported to more sophisticated software for further analysis. The Excel sheet can also be seen as an embryo for a more sophisticated database, where forms are registered directly into the database, for example, through a web-based interface. The column headlines and numbering correspond to the content of the individual forms, see Table 1, and each row should represent one form. This allows for several analyses to be aggregated in the same document and provides a structured overview of both the direct and the indirect consequences of affected CIs.

5.8. Step 8 – overarching analysis and visualization

The last step aims to perform an overarching analysis and visualization of the compiled consequences of the scenario. One straightforward approach is to summarize the expected consequences of the scenario, using the database from the previous step. However, spatial analyses and geographical visualizations tend to provide more understandable and comprehensive support for decision-making, for example, on proactive measures regarding threats, and risks, for example, natural disasters, crime, blackouts and floods (Blom et al., 2013).

One can, for example, use spatial analysis to identify areas that get affected by a high concentration of consequences stemming from multiple affected critical infrastructures. Additionally, the systematic identification of indirectly affected CIs leads to the inclusion of effects that manifests outside the flooded area. Overall, this enables analyses of proportions between the flooded area, the direct effects and the indirect effects due to interdependencies.

We recommend visualizing the societal consequences both in a map format and in a Cascading Effect (CE) diagram (Arvidsson et al., 2015). Essential information includes for example the hazard area, location of flooded infrastructure, location of indirectly affected infrastructure, and the geographical extent of the impacts and buildings. To be able to show a temporal progression of the cascading effects, estimated durations of impacts are needed. The map could include all identified consequences for an overview, see Figure 3, or be split up into several layers or maps for more detail, for example, highlighting particularly impactful cascading effects or displaying temporal progression. In a CE-diagram, the same consequences are instead organized by CI and the cascade order, see Figure 4. The diagram provides a clear overview of the CIs involved (left column), the interdependencies (arrows), the cascade order (subsequent columns), and the consequences that arise (boxes).

Figure 3. Visualizing societal consequences on a map. The flood scenario is based on simulations from the national flood hazard assessment. The consequences are fictive.

Figure 4. Visualizing societal consequences in a CE-diagram. The consequences are fictive.

6. Discussion

Several design choices were made during the development process of the AB-CEM. The method aims to be relevant for work related to the FRD and the Directive on European Critical Infrastructures. Therefore, a great deal of consideration has been taken to keep it useful and practically feasible for the targeted practitioners at local and regional levels within the Swedish DRM system. The testing of AB-CEM and feedback received from the different workshops, using the test case as a basis for discussion, and experiences from the pilot study were fundamental in the iterative refinements of the method, in particular for evaluating its applicability and accessibility. The trade-offs made and the insights gained through the development of the proposed method are discussed here.

The purpose of the AB-CEM is to improve current practices concerning consequence analysis by supporting a systematic mapping of critical infrastructure interdependencies and the cascading effects that arise due to spatially widespread events, such as flooding. To not consider cascading effects in a consequence analysis leads to an underestimation of risks, which in turn could lead to wrongful prioritization or underinvestment in risk mitigation. Given the intended users, a method based on collecting data through interviews or workshops was deemed the most suitable. The expert-based approach was also confirmed to be the most favoured approach among the stakeholders at the feedback workshops. This expert-based approach can be complemented by using empirical methods, which aim at investigating past events and drawing generalized conclusions regarding interdependencies (Johansson et al., 2015; Luiijf et al., 2010). Moreover, simulation-based methods to aid the estimation of cascading effects could also be further explored (Ouyang, 2014). However, such approaches were not deemed feasible at the moment considering the target group and the context of the method, mainly due to the in-depth knowledge required to apply such methods properly. They also rely on extensive input data, which often is not readily available. The use of expert knowledge as a basis for analysis comes with inherent uncertainty on the accuracy of the information. As is, the AB-CEM does not explicitly account for this. However, to achieve take one step towards including expert uncertainty, the analyst can request experts to give interval estimates (and thereby quantifying the uncertainty) or consult several independent experts when possible (Morgan, 2014). Through the widespread use of the AB-CEM to gather more extensive datasets, more rigorous approaches could be used to provide more in-depth analyses of cascading effects and give valuable scientific input.

Describing the consequences in a spatial format is argued to be beneficial both from a scientific and a practical perspective. In scientific literature, the lack of relevant data seems to restrict the use of more advanced spatial analysis methods in applied flood risk management. The data scarcity could be remedied, to some extent, with a widespread application of the developed method. The workshops revealed a generally positive attitude towards extending the usage of GIS for analysis and visualization of cascading effects to aid both critical infrastructure work and risk and vulnerability analyses in a local and regional context. Through the conducted workshops, it was also concluded that necessary GIS-competence to support the application of the AB-CEM generally existed at the municipal and regional levels in Sweden, although the competence level may vary from municipality to municipality. Lastly, the importance and usefulness of maps for conveying information to support decision-making were also identified through the workshops, along with some concerns about the risk of misinterpreting maps.

Therefore, to provide an additional visualization technique to complement the maps, the AB-CEM also suggests arranging the identified interdependencies and cascading effects in a CE-diagram (Arvidsson et al., 2015). A benefit of CE-diagrams is that they visualize the order in which effects propagate, something that is difficult to display in a static map format. A clear presentation of the order of events is helpful for decision-making, as the consequences, and the subsequent cascading effects can be more easily understood. To tackle cascading effects efficiently, mitigating or eliminating low-order cascading effects should be prioritized, as the measures in theory also will prevent downstream cascading effects. This could, for example, be done by introducing buffers, such as auxiliary power supply in critical facilities, or redundancy, such as additional and spatially separated fibre cables for rerouting electronic information.

An important question for the applicability of AB-CEM was if the data needed for the suggested method was available in some form. The stakeholders in the workshops confirmed that, on a general level, the required data either existed or was deemed plausible to compile. The pilot study, in its applied case, corroborates these general views, as the study managed to systematically identify several cascading effects using existent information from the interviewed experts (Andersson & Carlström, 2020). However, several stakeholders raised concerns regarding the confidentiality of some of the required data, in particular, related to CIs. For example, the municipal safety officer involved in the pilot study considered AB-CEM to potentially reveal too much sensitive data (Andersson & Carlström, 2020). Hence, confidentiality considerations could hinder the collection and aggregation of the data required to get a holistic picture of the cascading effects that arise during flood events. A few workshop stakeholders also expressed frustration over the, in many instances, contradicting goals: a thorough flood risk assessment on the one hand and keeping sensitive details about CI secure on the other hand. According to the stakeholders, the problem is exaggerated by unclear guidelines on classifying information in Sweden. Other CI researchers have also raised the confidentiality issue (Bekkers & Thaens, 2005; Fekete et al., 2015; Shih et al., 2009), and it thus deserves further attention and efforts to develop appropriate strategies to tackle this challenge.

The value of mapping and understanding cascading effects in flood events is supported both in the scientific literature (Hilly et al., 2018; Nones & Pescaroli, 2016; Poljanšek et al., 2012; Rinaldi et al., 2001) and by the stakeholders involved in this study. The stakeholders also highlighted the importance of being able to analyse cascading effects and to include indirect effects in all types of risk assessments, not only flood risk assessments. An additional general request was that the method should be applicable for analysing other types of hazards. The desired flexibility of the method extends its applicability and increases the chances of being perceived as useful and applied by practitioners. This flexibility was also confirmed through the pilot study, where the method was successfully applied for a draught scenario (Andersson & Carlström, 2020). The method only requires that an initial hazard analysis has been conducted that successfully identifies CIs directly affected by a hazard. Hence the AB-CEM is likely also applicable for a range of other types of hazards signified with a geographical extent, for example, earthquakes, storms, or forest fires, although further testing is required.

The initial identification of impacted CIs in the suggested method relies heavily on the initial hazard analysis. Therefore, it is important to keep in mind that limitations, flaws or uncertainties in the hazard analysis will have implications on the outcome of AB-CEM. In the context of the FRD, the use of deterministic hydrological models in Sweden assumes that an area shares the same probability of being flooded given the input parameters. In reality, there are several kinds of uncertainties related to inundation modelling, including the choice of model structure, parameters, and inputs (Teng et al., 2017). As a consequence, when applying AB-CEM some buildings and CIs near flooded areas might be considered safe based on the hazard analysis even though there is some likelihood it will be flooded (thus underestimating the risk). There can be similar limitations with for example hurricane (Coles & Simiu, 2003) or wildfire models (Benali et al., 2016). Therefore, a step towards improving flood hazard maps in Sweden is to incorporate, and clearly outline, the uncertainties involved.

The internal testing successfully enabled the identification and structuring of a limited number of cascading effects by using the suggested method on already existing written material, such as local and regional flood analyses, and risk and vulnerability analyses. A general finding during the testing was that the analysed documents contained minimal information regarding cascading effects, even the description of the direct consequences of the flooding generally lacked clarity, depth, and consistency. Unclear descriptions can, for example, affect the willingness to take political action. Using ‘loss-of-life’ in a risk assessment is perhaps more likely to lead to political action compared to ‘number of affected people’ which is commonly used in FRD risk assessments (MSB 2018). Moreover, any geographical extent of the consequences beyond the initial flooded area was missing in the material. Hence, it is apparent that there is a need for a practical method to combat these deficiencies.

The AB-CEM is thus deemed to be both useful and feasible for mapping interdependencies and cascading effects at the local and regional levels for flood events and other types of hazards with a geographical extent. The successful application of AB-CEM in the pilot study is a positive indication of its feasibility. However, it is important to note that the developed method remains to be more rigorously tested. Additional research of interest is to apply the method in another context than Sweden and to develop it further by more comprehensive incorporation of temporal and uncertainty aspects. The analysis and visualization of the data gathered through the use of the method is another important area with significant potential for further research and refinements. The proposed method, however, is deemed as a promising step towards improving current consequence analyses by systematically including cascading effects in flood risk assessments.

7. Conclusion

The main purpose of this paper is to present and problematize a method (AB-CEM) for mapping and analysing cascading effects among critical infrastructures due to interdependencies during specifically flood events but also for a range of other types of hazards. The AB-CEM provides a systematic approach and guidance towards mapping cascading effects. By requiring geocoding of the collected data it could also be used for more advanced spatial analyses. The development process and the pilot study reveal that there is a great practical need for methods for mapping and analysing critical infrastructure interdependencies and the cascading effects that arise due to spatially widespread hazards, such as flooding. Another important finding from the tests of AB-CEM is that it can serve as a basis for decision-making about both proactive and reactive efforts in connection to geographically extensive hazards. We can also conclude that there is a problematic issue regarding sensitive and secret data, more specifically the conflicting interests of information availability and information security concerning critical infrastructures. This issue must be resolved at the national level, for example by providing more detailed guidelines. The paper also contributes by highlighting the necessity and relevance of this research area through document analyses and workshops with stakeholders. AB-CEM was mainly developed with practitioners in mind and the participating stakeholders in this study concluded the method to be both practically feasible, applicable and useful in a Swedish context. The method is a much-needed step towards integrating an analysis of cascading effects in practical flood risk assessments, such as the ones demanded by the EU flood risk directive. Further testing of the method, the development of suitable GIS-oriented analyses and visualization techniques, and challenges relating to confidentiality issues with collecting and aggregating this type of data are deemed highly relevant for further research.

Acknowledgements

The Swedish Civil Contingencies Agency provided the funding enabling the presented research through the Centre for Critical Infrastructure Protection Research (CenCIP); their support is greatly acknowledged. The authors further extend gratitude to all the involved stakeholders for providing crucial insights and support for the development and evaluation of the method. Lastly, we wish to thank Emma Andersson and Daniel Carlström for performing the pilot study using AB-CEM as a part of their master thesis in risk and safety management.

Disclosure statement

No potential conflict of interest was reported by the author(s).

Additional information

Funding

This work was supported by Myndigheten för Samhällsskydd och Beredskap.