Impact Webs: a novel conceptual modelling approach for characterising and assessing complex risks

Identifying, characterising, and assessing the complex nature of risks are vital to realise the expected outcome of the Sendai Framework for Disaster Risk Reduction. Over the past two decades, the conceptualisation of risk has evolved from a hazard-centric perspective to one that integrates dynamic interactions between hazards, exposure, system vulnerabilities, and responses. This calls for a need to develop tools and methodologies that can account for such complexity in risk assessments. However, existing risk assessment approaches are hitting limits to tackle such complexity. To this end, we developed a novel complex-risk assessment methodology named Impact Webs, inspired by a conceptual risk modelling approach named Climate Impact Chains that integrates aspects of various other conceptual models used in risk assessments, such as causal loop diagrams and fuzzy cognitive mapping. Impact Webs are developed in a participatory manner with stakeholders and characterise and map interconnections between risks, their underlying hazards, risk drivers, root causes, and responses to risks, as well as direct and cascading impacts across multiple systems and at various scales. In this methodological paper, we show how we developed the Impact Web methodology, including how we derived which elements to include in the model, demonstrating the logic and visual output and listing the steps we followed for construction. As proof of concept, we present the results of a complex-risk assessment in Guayaquil, Ecuador, which investigated how COVID-19, concurrent hazards, and responses propagated risks and impacts across sectors and systems during the pandemic. Reflecting on the utility of Impact Webs, application in case studies demonstrates the methodology's usefulness for understanding complex cause–effect relationships and informing decision-making across different scales. The participatory process of developing Impact Webs with stakeholders uncovers critical elements in systems at risk, and helps to evaluate co-benefits and trade-offs of decisions by uncovering how the outcomes of disaster risk management practices affect people, organisations, and sectors differently. Offering a system-wide perspective for modelling, Impact Webs stand as a valuable methodological contribution for complex-risk assessment.

1 Introduction

Identifying, characterising, and assessing the complexity of risks are vital to realise the expected outcome of the Sendai Framework for Disaster Risk Reduction (UNDRR, 2022). As sectors and systems become increasingly interconnected, the space in which risks can cascade is expanding (Helbing, 2013; UNDRR, 2022). This has been starkly evident throughout the COVID-19 pandemic, where impacts have not just arisen in the health system, generated by the hazard, but also from the cascading effects of impacts and from societal responses through global lockdowns, with different regions suffering from vastly different consequences, depending on underlying societal vulnerabilities and the resilience of their systems (Hagenlocher et al., 2022). These characteristics are not limited to COVID-19, and have also been observed in other contexts, such as the compounding and cross-border effects of extreme climate events (Simpson et al., 2021; Zscheischler et al., 2018) or the global ripple effects of armed conflicts (Cui et al., 2023).

Over the past two decades, the conceptualisation of risk has evolved from a hazard-centric perspective to a more encompassing notion that integrates the dynamic interactions between hazards, exposure, vulnerability (IPCC, 2014) and, more recently, response risks, i.e. risks that can arise from responses to risks and impacts (Simpson et al., 2021; Ara Begum et al., 2022; IPCC, 2023; Hagenlocher et al., 2023). Different terminologies have been used to conceptualise these dynamic interactions, including cascading, compound, and systemic risks. In this paper, we use the term “complex risks” to encapsulate these different risk framings. Given that complexity is now understood as a defining feature of risks, single-hazard and single-risk approaches, while useful in certain contexts, are becoming increasingly insufficient for comprehensive disaster risk management (Simpson et al., 2021; UNDRR, 2022; Schlumberger, et al., 2024; Sett et al., 2024; de Ruiter and van Loon, 2022). This has been recognised by the Intergovernmental Panel on Climate Change (IPCC) in the Sixth Assessment Report, which notes that risks and responses, including their determinants, can all interact dynamically in shaping the complexity of climate risk (Ara Begum et al., 2022). Additionally, the Global Assessment Report 2022 (GAR 2022) from UNDRR stresses the importance of understanding and assessing the complex nature of risks as a key foundation for risk-informed decision-making (UNDRR, 2022). However, existing data-driven and quantitative modelling approaches are hitting limits to tackle such complexity. The combined effects of multiple hazards, threats, or shocks should not be assessed just through the addition of each of their impacts independently, but instead require systems approaches to understand risk and impacts (de Ruiter et al., 2020; Ara Begum et al., 2022; Hagenlocher et al., 2023; de Brito et al., 2024). There is therefore a need to develop methodologies that take a system-wide lens for analysis, which can account for how multiple hazards and vulnerabilities of systems and sectors interact to better understand complex risks.

To this aim, we developed a novel complex-risk assessment methodology named “Impact Webs”. Impact Webs are inspired by a conceptual risk modelling approach named Climate Impact Chains (see Menk et al., 2022, for a review of applications) and draw inspiration from various other conceptual models used in risk assessments. Climate Impact Chains were originally developed for sectoral climate risk assessment (Schneiderbauer et al., 2013; Zebisch et al., 2023, 2021; Hagenlocher et al., 2018), in which elements of the model are assigned to the key risk components used in disaster and climate risk assessments of hazard, exposure, and vulnerability, and cascading effects are assigned as intermediate impacts. One critique of Climate Impact Chains is that they often depict a linear cause–effect relationship for a single sector or hazard and thus do not capture the complexity of system interactions well (Harris et al., 2022). With Impact Webs, we built on Climate Impact Chains, integrating aspects of system mapping approaches, such as causal loop diagrams (e.g. Coletta et al., 2024; Groundstroem and Juhola, 2021; Dianat et al., 2021; Rehman et al., 2019), fuzzy cognitive maps (e.g. Gómez Martín et al., 2020; Ahmed et al., 2018; Chandra and Gaganis, 2016), and Bayesian belief networks (e.g. Malekmohammadi et al., 2023; Scrieciu et al., 2021; Bashari et al., 2016; Giordano et al., 2013). With this, we aimed to integrate the key risk components in disaster and climate risk assessments with a systems-based perspective to identify, characterise, and map interconnections between risks, their underlying hazards, risk drivers, root causes, and responses to risks, as well as direct and cascading impacts across multiple systems and at various scales. Impact Webs aim to better account for the complexity of risk interaction, compared with Climate Impact Chains, by developing flexible and less linear conceptual models that can help to understand complex risks.

In this paper, we offer a new complex-risk assessment methodology in the form of Impact Webs, detailing how we developed it. To do this, we first conducted a scoping review of the literature on conceptual risk models that we drew inspiration from. Informed by the review, we identified constitutive elements for the model and developed a graphical structure. We then developed key steps for conducting a complex-risk assessment with Impact Webs, testing our methodology in five cases. These cases were Cox's Bazar humanitarian camp (Bangladesh), the Sundarbans region (India), a national-scale assessment (Indonesia), Maritime Region (Togo), and the city of Guayaquil (Ecuador). The complex-risk assessments investigated how COVID-19, concurrent hazards (e.g. hydrological, geophysical, climatological), and responses to them (e.g. restriction measures) interacted with underlying societal vulnerabilities to propagate risks and impacts across sectors and systems during the pandemic (Hagenlocher et al., 2022). COVID-19 was selected as the entry point for the risk assessments as the pandemic has been so diverse and cross-scale in its effects; therefore such an event was ideal to test a novel risk modelling approach for understanding complex risks. As proof of concept, we present the results and final output from one of the five test cases, showing an Impact Web and narrative storyline for the city of Guayaquil, Ecuador, during the COVID-19 pandemic. Guayaquil was selected to demonstrate our proof of concept due to the city's high vulnerability and exposure to the compounding effects of multiple hazards and the presence of many drivers of risks, creating numerous challenges for risk management, therefore making it a fitting case to showcase a new risk assessment methodology.

The remainder of the paper is structured as follows: In Sect. 2, we present the methodology for developing Impact Webs, which includes the scoping literature review of conceptual risk models, the constitutive elements we selected to populate the model, and the steps that were followed during the complex-risk assessments to construct an Impact Web. In the results in Sect. 3, we show our proof of concept, presenting the Guayaquil test case. In the discussion in Sect. 4, we reflect on the utility of Impact Webs, looking at strengths, limitations, and potential future research directions. We conclude in Sect. 5 with a synthesis of the paper, highlighting Impact Webs as a conceptual model that moves beyond single-risk or single-hazard assessment, which can be used as an approach for system-wide complex-risk assessment.

2 Methodology

In Sect. 2, we present our methodology to develop Impact Webs. We show our methodological pre-development, with a scoping review of other conceptual risk modelling approaches we drew inspiration from. We then elaborate on the elements that were selected in the model, introduce the five test cases, and present the steps we followed to construct an Impact Web.

2.1 Methodological pre-development: scoping review of conceptual risk models for inspiration

Given that we aimed to develop an approach that took a systems-based perspective for analysis to better understand complex risks, we conducted a scoping review of the literature on conceptual risk models that do this. The scoping review was non-systematic and not meant to be exhaustive. It was conducted to support methodological synthesis and inspire the concept development for our approach by looking at features of different methodologies that could be useful. A non-systematic scoping review was chosen as this type of review approach has advantages for developing new methodologies. Non-systematic scoping reviews allow for exploratory flexibility, drawing on grey literature, emerging studies, and the integration of methodological aspects that authors had used in past research. This supported creative synthesis by combining ideas from various disciplines (Munn et al., 2022). Texts were selected and reviewed based on the authors' own experience, expert judgement, and searching using the Scopus search engine. A general description of the approach's features is given, as well as the strengths and weaknesses in a complex-risk context. We also provide selected key references that inspired us (see Table 1).

Table 1 Overview of conceptual models used in risk assessments.

2.2 Lessons from the review

Different conceptual modelling methodologies have been applied across disciplines for assessing complex risks and have provided useful lessons for our approach. From the papers we reviewed, influence diagrams and Bayesian belief networks show usefulness to understand interactions of biophysical processes, such as extreme events, with additional dynamic inputs, such as interventions in response to risks or stakeholders' perceptions of risks and risk management decisions (e.g. Scrieciu et al., 2021). Causal loop diagrams and fuzzy cognitive maps provide a useful framework to examine interconnections and feedback effects between elements in one or multiple systems to support integrated and cross-sectoral decision-making (e.g. Hanf et al., 2025; Dianat et al., 2021) and Climate Impact Chains are effective for eliciting stakeholder knowledge, due to their flexible and relatively simplistic form, which is useful to develop shared system understanding and co-create policy recommendations, while their innovative focus on intermediate impacts makes them conducive to analysing cascading impacts (e.g. Sett et al., 2024). It is important to acknowledge that the approaches in Table 1 are not mutually exclusive and do cross over with one another. Methodological combinations of approaches are common and are adjusted to suit the decision context or setting of a risk assessment. For example, there is often integration between fuzzy cognitive maps, influence diagrams, and Bayesian belief networks.

With Impact Webs, we drew on observed strengths in the literature we reviewed, aiming to create a model that is useful for (1) understanding interactions of impacts from extreme events and stakeholders' responses to them, as well as stakeholders' perceptions of risks and risk management; (2) examining interconnections and feedback effects in one or multiple systems to support integrated and cross-sectoral decision-making; and (3) eliciting stakeholder knowledge to develop shared system understanding and co-create policy recommendations. In order to achieve these aims with Impact Webs, we built on the hazard, exposure, and vulnerability framing of Climate Impact Chains, which is useful to understand how risks emerge from extreme events (e.g. Hagenlocher at al., 2018; Sett et al., 2024), expanding this to include such aspects as feedbacks and non-linear interconnections, which are well suited to fuzzy cognitive maps and causal loop diagrams (e.g. Hanf et al., 2025; Coletta et al., 2024; Ahmed et al., 2018). To do this, we included the dynamic interaction of multiple hazards, threats and shocks, multiple exposed elements, and the impacts to exposed elements. We additionally included the drivers and root causes of vulnerabilities to exposed elements in Impact Webs. Including drivers and root causes in the model helped us to understand not just what impacts occurred but also why they occurred (Blaikie et al., 1994; Wisner et al., 2004). Drawing on strengths of influence diagrams and Bayesian belief networks (e.g. Mühlhofer et al., 2023; Scrieciu et al., 2021), we included interventions in response to risks and impacts, as well as response risks arising from them. We did this as it was important for us to align with the most recent IPCC risk framing (Simpson et al., 2021; Ara Begum et al., 2022). All types of approach we reviewed use graphical methods to show cause–effect relationships, most commonly using arrows and symbols to signal a relationship and influence. We also adopted this, using graphical methods to show cause–effect relationships and feedbacks. The majority of studies we reviewed integrated some form of input from stakeholders. However, it was common for stakeholder participation to decrease with the increasing complexity of the method used, due to difficulties in communicating and facilitating the approach (Parviainen et al., 2019). This was an important lesson for us from the review. We aimed for a strong participatory approach that involved collaboration and integration of different expertise and knowledge. Given this, drawing on the strengths of Climate Impact Chains (Harris et al., 2022), we aimed to make the steps for developing the model simple so that stakeholders were not overwhelmed and could be highly engaged during the modelling process. This helped to identify key system elements that stakeholders felt were important, valued highly, and wanted to protect from risks and impacts, for example key economic sectors. A systems-focused perspective was commonly taken towards analysis in all approaches. We did the same for Impact Webs. The aim of taking a systems-focused perspective was to enhance system understanding and reduce uncertainty by modelling non-linear interactions and dynamics. We also wanted to model interactions across different scales (i.e. from global to local). Therefore, we expanded beyond a sectoral focus (e.g. drought risk for the agriculture sector), often used with Climate Impact Chains, aiming to capture cross-sectoral risks, impacts, and vulnerabilities and their influences between one another. Lastly, we observed in the majority of papers that the visual output of the model was also accompanied by narrative-based methods, used to explore and communicate findings (e.g. Hanf et al., 2025). We followed this by including a narrative storyline that described the findings of the assessment and described the Impact Web in a structured and relatable way.

2.3 Selection of constitutive elements in an Impact Web

Building on the lessons from the scoping review, here we present the elements that were selected for visualisation in an Impact Web (see Fig. 1 and Table 2). We elaborate on why these elements were selected, including the conceptual backing for choosing them and the system interactions we wanted to assess.

Figure 1 Elements and possible graphical structure of an Impact Web. While here we present our chosen graphical output of the conceptual model with computerised tools, an Impact Web could equally be made using a pen and paper, for example if being developed in a community workshop. The model maps the direct and cascading impacts and their interactions resulting from a biological and climate-driven hazard. These impacts trigger an intervention, which results in further negative and positive impacts (i.e. response risks), as well as a risk that did not manifest. Drivers of risk and root causes linked to why impacts emerge are also included. The elements in the model are predominantly focused on the local context; however, important regional and global interactions are included.

Table 2 Description of the elements used in Impact Webs, including the chosen visual representation in the model and examples.

2.3.1 Hazards, threats, and shocks

Conceptual risk models are developed to better understand impacts arising from a hazard, threat, or shock, such as hydrological extremes (e.g. flood and drought), biological hazards (e.g. COVID-19 or a cholera outbreak), or geopolitical aggression (e.g. a war or conflict). We wanted our model to improve understanding of compounding interaction, given the increasingly interconnected nature of multi-hazard impacts on sectors and systems (UNDRR, 2022). Therefore, we included multiple hazards, threats, and shocks to the system being modelled in Impact Webs.

2.3.2 Impacts

Impacts were the second element we included after our scoping review (Sett et al., 2024; Zebisch et al., 2023; Lawrence et al., 2020). This was done to identify direct negative impacts from hazards, threats, and shocks, as well as cascading impacts and any potential positive impacts that might have arisen (often as a result of interventions) in the system being modelled. Modelling cascading impacts, which arise through impact propagation (Mühlhofer et al., 2023; Carter et al., 2021), helped to understand the system's interconnectedness as linkages between sectors and sub-systems could emerge as connections were characterised. In the visualisation of the model (see Fig. 1), we do not make a visual distinction between direct and cascading impacts. There is, however, a conceptual distinction as every impact that is not directly connected to a hazard, threat, or shock can be understood as cascading. Additionally, through modelling impacts, the compounding effects of multiple hazards, threats, or shocks occurring simultaneously could be analysed (Simpson et al., 2023). We did make a visual distinction between negative and positive impacts, using crosses and ticks (see Table 2).

2.3.3 Interventions

With Impact Webs, we wanted to characterise and assess how decisions in response to or anticipation of risks and impacts have impacts in systems. Drawing on aspects of fuzzy cognitive mapping, influence diagrams, and Bayesian belief networks, which are useful for modelling the effects of decision-making processes (Scrieciu et al., 2021), as well as the more recent framing of response risks (Simpson et al., 2021; IPCC, 2023; Hagenlocher et al., 2023), interventions were included in our conceptual model. When developing the model, both positive and negative impacts from interventions were included (e.g. the negative impacts that occurred because of COVID-19 lockdowns). The defined decision context and system boundaries denote the granularity of response risks and impacts included in the model, for example whether city-level or intergovernmental-level interventions are being mapped.

2.3.4 Risks that did not manifest

When modelling a system that has been affected by hazards, threats, or shocks, there can be potential adverse consequences that are avoided, often as a result of interventions (e.g. the risk of a healthcare system collapsing or a breadbasket failure). We included these in Impact Webs and named the element risks that did not manifest. These are conceptually different from positive impacts, as they are potential negative consequences that did not happen.

2.3.5 Drivers of risk

Understanding causality is a key rationale for disaster risk assessment (Oliver-Smith et al., 2017) and taking a systems approach facilitates looking into causal connections that can deepen the assessors' understanding of how and why impacts can emerge (Gómez Martín et al., 2020; Coletta et al., 2024). Therefore, an important element for our model was to look at what were drivers of risks and impacts in the system. Drivers of risks are processes or conditions that influence the level of risks and impacts by increasing levels of exposure and vulnerability or reducing the capacity of people to manage or adapt to risks. We were motivated to include drivers, as this asks the modeller to critically reflect on how and why societal functions, essential sectors, system elements, or stakeholders were adversely affected due to high susceptibility or low coping/adaptive capacity.

2.3.6 Root causes of risk and vulnerability

An additional step to further understand causality was to model root causes of risk and vulnerability. These are underlying factors that influence drivers of risk (Blaikie et al., 1994; Wisner et al., 2004; Zebisch et al., 2023). Including them supported exploring socio-economic and political structures and processes and choices that further explain why a particular community, sector, system, or place is at risk in the first place; this is important for designing risk management to be sustainable and lasting. Both drivers of risks and root causes are often distant spatially and temporally from the system under investigation (Wisner et al., 2004); they are, however, highly relevant when one wants to understand complex risks.

2.3.7 Connections between elements

Following the other conceptual modelling approaches we reviewed, we used graphical methods to show connections between our chosen elements and visualise risks. We selected arrows to indicate directional cause–effect relationships. Given the limitations of directed acyclic graphs used in many Bayesian belief networks and influence diagrams in showing feedback effects (Bashari et al., 2016), we took an approach more inspired by causal loop diagrams. This meant we could better demonstrate indirect effects and feedback loops (Groundstroem and Juhola, 2021), which is both more appropriate to a complex-risks context and helped us understand interconnectivity between elements. We used different colours and dashed arrows to show connections coming from drivers of risk and root causes of risk and vulnerability, compared with using black arrows for other elements in the model (see Table 2). This was done so that a quick and engaging visual distinction could be made for external stakeholders working with or viewing the model.

2.3.8 Scales

From our review, we did not find conceptual modelling approaches that were effective at demonstrating risk elements and their interactions across spatial scales. For example, a critique of Impact Chains and fuzzy cognitive mapping approaches is that they often have narrow definitions of system boundaries (Petutschnig et al., 2023; Ahmed et al., 2018). For Impact Webs, we included three spatial scales in our model (i.e. local, regional, and global), which were intended to model globally networked risks, as well as demonstrate risk drivers, root causes, and impacts that are often spatially distant but have effects in the local context (Helbing, 2013). As the test case study contexts where we made Impact Webs were geographically diverse (see Sect. 2.4), there was flexibility in how the “local” scale boundary was defined. For example, for the Cox's Bazar case, the local scale was defined as inside the humanitarian camp. Comparatively, the Guayaquil case focused on investigating the city municipality, whereas the Indonesia case was at the national scale.

2.4 Trial in test cases

Impact Webs were developed in five test cases to assess complex risks (Hagenlocher et al., 2022). This was done to trial our methodology with groups of stakeholders across diverse case study contexts. This had three purposes. First, it allowed for adjustment in the steps for construction (see Sect. 2.5) and improvement of the methodology through stakeholder feedback. Second, we could test Impact Webs across different locations, each with their own unique challenges and characteristics, building from the same entry point to see whether the approach was replicable and a useful risk assessment tool in different contexts. Third, we wanted to develop a methodology that was participatory; therefore, we needed to trial it with stakeholders to learn how they would engage with developing such a model. We trialled the methodology in the cases between June and September 2021, using COVID-19 as the entry “seed” element, building from there and adding additional elements to populate the model using desk study and stakeholder workshops. COVID-19 was selected as the first hazard to start building the model around as the pandemic had been a situation that challenged conventional risk and hazard settings, and was therefore a unique event in which to test a new complex-risk assessment methodology. The cases were chosen to cover a wide thematic range. In this paper we only present the final Impact Web for one of the five cases (Guayaquil, Ecuador), to demonstrate our proof of concept (see Sect. 3). The test cases were as follows.

• Cox's Bazar humanitarian camp (Bangladesh): Showcased COVID-19 and pre-existing social inequity in a challenging and fragile setting. The case highlighted characteristics of vulnerable people and communities living in highly dependent systems.

• Sundarbans region (India): Encompassed a strong multi-hazard perspective, demonstrating the concurrence of COVID-19 with tropical cyclone Amphan. The case exhibited the dynamic nature of complex risks by exploring the delay between causes and effects of impacts.

• National scale (Indonesia): Highlighted how COVID-19 and other hazards led to interconnected challenges on all fronts. The case had a special focus on the role of social protection.

• Maritime Region (Togo): Focused on the rural–urban and national–international interlinkages of systems and how they were affected by COVID-19 and concurrent hazards in a regional sub-Saharan context with high levels of poverty.

• Guayaquil (Ecuador): Gave specific insights into how COVID-19 and other hazards overwhelmed a densely populated, overcrowded, urban setting. The case presented characteristics of tipping points and showed how system dependencies from global to local scales created and reinforced vulnerabilities (see Sect. 3 for more).

2.5 Steps for constructing an Impact Web

Here we present the steps that we followed to construct an Impact Web (see Fig. 2) in the test cases, informed by lessons from the scoping review and from previous experiences among the research team of undertaking risk assessments.

Figure 2 Workflow of the steps that were followed for constructing an Impact Web. We trialled the approach in five test cases, which allowed for adjustment and improvement of the methodology as well as stakeholder feedback. The workflow followed a flexible stepwise methodology in five steps (scoping, identifying, and mapping a preliminary number of elements, workshops, and stakeholder participation, reviewing the model's logic and visualisation and drafting an accompanying narrative storyline). Workshop 1 allowed for new inclusions and the adjustment of already identified elements in the draft model. Once included, workshop 2 allowed for validating the logic and looking at entry points for risk management. This is shown in the figure through the circle of blue arrows, which indicates iteration in the model's development.

Step 1: scoping

Risk assessments are conducted in a specific setting to support decision-making processes. Following in the steps of risk assessments that have been successful in the past (e.g. Zebisch et al., 2023; Hagenlocher et al., 2018), the preliminary step for constructing an Impact Web was the scoping. Here, we defined objectives and the need for the multi-hazard risk assessments across each case, considering how the conceptual models could enhance understanding and inform decision-making that reduced risks. While systems theory denotes that system boundaries change, for example due to shifting climatic conditions (Steffen et al., 2015), practically, selecting the scale to model across the test cases helped to refine the decision context. This was done by looking at geographical or administrative boundaries to select the area of primary focus. We then identified critical societal functions, essential sectors, and key elements at risk in each of the cases, as well as key stakeholders that were engaged later in the process. Once this was defined, however, it was important that there was flexibility when populating the Impact Web with elements, given that we wanted to model cross-scale dynamics, including feedback effects, cascading effects, and globally networked risks that were identified outside the geographic boundaries of the test cases (Helbing, 2013; Sparkes and Werners, 2023).

Step 2: identifying and mapping a preliminary number of elements

While there are not restrictions in terms of the order for selecting the elements in an Impact Web, we found it was preferable to start from a limited number of key elements that we wanted to better understand and then progressively build up the causal connections. In our test cases, we wanted to understand multi-hazard interaction of COVID-19 and concurrent hazards, threats, and shocks; therefore, COVID-19 was the logical entry point. This perspective acknowledged that the system's complex relationships emerge more clearly when under stress, i.e. when direct and cascading impacts occur, the connections between them and hazards become more visible and therefore easier to observe. In this sense, the first number of elements functioned as “seeds” for identification of the system's interdependencies. We found building from key hazards, threats, and shocks as the “seed” elements facilitated following a more simplistic cause–effect chain at the start of construction, i.e., direct impacts arising from each of the hazards, threats, or shocks. From direct impacts, cascading impacts, then interventions and response risks, and finally drivers of risks and root causes followed. While Impact Webs eventually aim to map risk complexity, we found it difficult to start from the more complex interactions (i.e. feedback effects). Rather, starting with more simple connections is easier for the modeller and stakeholders to begin with, and the more complex interactions will emerge later as system understanding improves with desk study and more stakeholder interactions.

Step 3: workshops and stakeholder participation

Nearly all conceptual models that we reviewed integrated some form of stakeholder input, which was variable, depending on the decision context and complexity of the method chosen. Causal loop diagrams (e.g. Coletta et al., 2024) and Impact Chains (e.g. Sett et al., 2024), for example, generally elicit the integration of more stakeholder input than influence diagrams (e.g. Mühlhofer et al., 2023), which have a strong quantitative component. A key step in our approach was to draw on diverse knowledge from a range of expertise, which we did through application in test cases. In this way, the Impact Web would be co-created to develop a mutually agreed upon visual output of complex risks, as well as a shared heuristic of the system. Building on the preliminary number of mapped elements in Step 2, we held two workshops for each test case. The workshops were held with a range of different stakeholders, representing communities, policy, practice, civil society, academia, and governments. These stakeholders were identified in the scoping (Step 1). Workshop 1 focused on identifying new elements for the Impact Web, as well as reviewing the ones that had already been identified and mapped from the desk review (Step 2). After workshop 1, we included the new elements in the model, and held a second workshop to re-validate the logic and elements, as well as look at entry points for risk management. This stakeholder backstopping provided better understanding of otherwise unknown or missed model elements and their connections and helped to characterise the complex risk characteristics that could not be captured through desk study alone.

Step 4: review of model and visualisation

After collecting stakeholder inputs across the five test cases, an important step was to review the model among the research authors. This included in-depth structuring of the information gathered in the workshops and cross-referencing it from available literature sources gathered in the desk study. Where possible, we also refined the number of elements, for example by clustering two elements that represented the same issues. This was done to reduce the model's complexity and ensure that the final visual could be an effective communication tool. We also reviewed causal connections and the logic behind them, reflecting to understand what this meant in a system's context, thus enhancing our own understanding of complex risks. We then reworked the graphical design to create visual and causal connections that could be simpler to follow.

Step 5: drafting narrative storyline

As a final step to accompany the Impact Web model, a narrative risk storyline was drafted for each test case that described the model and its connections in a narrative format. This helped to communicate, in a descriptive and engaging manner, the complex model output that resulted from following the previous steps, making it more engaging and useful to direct risk management decisions for both experts and non-experts (Hanf et al., 2025; van den Hurk et al., 2023). The storylines were drafted by the research authors, explaining the key aspects in the Impact Web and findings from the complex-risk assessment. This was done after the authors had completed the desk review and stakeholder workshops and reviewed the model.

3 Results: proof of concept

Here we present our result, showing our proof of concept detailing the final output of an Impact Web and narrative storyline from the Guayaquil, Ecuador, test case. Next we elaborate on why Guayaquil was chosen for our proof of concept in this paper.

3.1 Complex risks linked to COVID-19, concurring hazards, and responses in Guayaquil, Ecuador

Here we show our proof of concept, presenting the results and final outcome of one of the test cases, from Guayaquil, Ecuador. We only show the results of one case in this paper as our aim has been to demonstrate how we developed the methodology. Selecting Guayaquil to showcase Impact Webs highlights the outcomes of steps 4 and 5 in Fig. 2.

Step 1: scoping

We developed an Impact Web to study risks and impacts emerging from the COVID-19 pandemic and concurrent hazards, threats, and shocks in the city of Guayaquil, Ecuador. Guayaquil was selected due to its high population density, high levels of poverty and inequality, large informal work sector, overcrowded housing, and high exposure to climate-related and geophysical hazards (Hallegatte et al., 2013). These factors make the city's inhabitants vulnerable to the compounding effects of multiple hazards and present challenges for risk management that are exacerbated by limited financial resources at both municipal and national levels. These factors additionally have numerous and compounding drivers of risks and root causes, making this an important case in which to undertake a complex-risk assessment. We used COVID-19 as the “seed” element for developing the Impact Web as this hazard has been so diverse in its effects across communities, sectors, and economies, which additionally provided important lessons for the application of a novel conceptual risk modelling approach using a systems-focused lens. It was decided that taking a case study at the city scale supported in defining system boundaries and decision context, for which COVID-19 has been cross-scale and highly dynamic (Hagenlocher et al., 2022).

Steps 2–4: Impact Web of Guayaquil, Ecuador

Figure 3 presents the final conceptual model of the complex-risk assessment in Guayaquil. The Impact Web visualises (i) multiple interacting hazards, threats, and shocks across various scales; (ii) the identification of different risks/impacts for communities, sectors, and societal functions, as well as their interconnections and cascading effects; and (iii) their underlying risk drivers, as well as (iv) the root causes behind underlying risk drivers, some of which can be spatial and temporally distant from newly emerging risks/impacts. Further, the Impact Web model also maps (v) risks and impacts linked to responses (e.g. policy interventions aimed to reduce risks), as well as (vi) risks that did not manifest due to the interventions.

Figure 3 Impact Web for the test case of Guayaquil, Ecuador. The conceptual model visualises complex risks and impacts linked to the COVID-19 pandemic, concurring hazards, and the responses to it, as well as interconnections between system elements and drivers and root causes of risks.

Step 5: narrative storyline for Guayaquil, Ecuador

The first confirmed case of COVID-19 in Guayaquil was identified on 29 February 2020. Driven by the city's high population density, challenges with overcrowded housing, and unpreparedness in the health system, there was a rapid rise in cases and hospitalisations. Governmental policies of austerity in the 5 years prior to the pandemic meant that hospitals and healthcare facilities were understaffed and under-equipped. The lack of personal protective equipment resulted in a high number of cases and deaths among healthcare workers, which put further pressure on a health system that was already burdened by increases in vector-borne diseases due to seasonal flooding exacerbated by climate change. From the compounding effects of multiple hazards and cascading impacts that emerged, the health system reached a tipping point and collapsed, tragically resulting in a large number of bodies being left in the streets, hospitals, and care homes. This significantly increased psychological stress for the city's residents. In March of 2020, Guayaquil had an excess mortality rate five times that of the same month in the previous year and the highest COVID-19 mortality rate of any Latin American city.

Economic disruptions from the intervention to close international borders were particularly severe in Guayaquil due to the city's high dependency on the port. The closing of borders triggered economic shutdown, with widespread adverse effects on employment and livelihoods. Due to the lack of job retention schemes, many citizens, a lot of whom were already living in poverty before the pandemic, were left without income-generating opportunities. These impacts were exacerbated for the large informal employment sector in Guayaquil. Due to the limited availability of space per person, driven by the high population density and overcrowded housing, lockdown interventions and social distancing were difficult to follow for a large segment of the population. As seen in many places, there were also sharp increases in domestic and gender-based violence during lockdown. As Guayaquil is a food-producing city, one risk that did not manifest as a result of lockdowns was disruption in the food supply chain and food shortages that were prevalent in some other cities in the region.

State coordination challenges and reliance on international guidance, which was unclear and contradictory in the early stages of the pandemic, meant there was a lack of an integrated, cross-sectoral, and multi-scale response between Guayaquil's and Ecuador's public institutions. The national government maintained a centralised COVID-19 testing system, which hindered the effectiveness of city institutions in setting up early detection and monitoring systems, such as contact-tracing and testing facilities. The unclear guidance from the World Health Organization resulted in the output of unclear information at the national level, which was one of the factors that contributed to the spreading of misinformation throughout digital networks. One positive impact that arose from state coordination challenges was the strengthening of public and private sector cooperation.

In response to the economic disruptions, the government of Ecuador brought in more austerity measures. Furthermore, corruption allegations were brought against some city and state-level actors for capitalising on the emergency healthcare situation. These factors saw increasing societal distrust in the government, which was already underlying. This came to fruition in Guayaquil when a societal tipping point was reached in May of 2020, resulting in widespread protest and civil disobedience.

Key risk drivers identified included initial unpreparedness in the health system, high population density, overcrowded housing, economic dependency on the port, and state coordination challenges linked to reliance on international guidance, among others. These risk drivers influenced cascading response risks, including widespread negative economic effects of lockdown and closures of international borders, as well as an increase in societal distrust, with subsequent protest and civil disobedience, which were in part due to further austerity interventions in response.

A number of considerations for risk management emerged from developing the Impact Web for Guayaquil. These include focusing attention, resources, and efforts towards multi-sectoral and multi-scale coordination across public and private institutions, as well as ensuring strong reach and availability of social protection mechanisms and investment in risk monitoring and data systems. The case also highlights that clear guidance and risk communication are key to building societal trust during times of crisis.

4 Discussion

With Impact Webs, we integrated Climate Impact Chains with aspects of system mapping approaches. In doing this, we aimed to close gaps in current conceptual models of risks, by characterising dynamic interactions between hazards, exposure, vulnerability, response risk drivers, and root causes (IPCC, 2023), improving our understanding of complex risks by following a flexible stepwise methodology. In this discussion we reflect on the strengths and limitations of making Impact Webs. While in this paper we only present one test case, our discussion reflects on lessons we learnt through the process of developing the methodology and from across all of the cases. We also provide future research directions.

4.1 Strengths

The application of the Impact Web methodology in case studies showed that the approach is useful to conceptualise, identify, and visualise networks of interconnected elements across different systems and sectors. The conceptual model's suitability for mapping the interactions of multiple concurrent hazards with multiple pre-existing drivers of risks and root cases helps to uncover underlying societal vulnerabilities and is useful to derive storylines of how interconnected risks and impacts emerge from a hazard or shock events. In the context of Guayaquil, the Impact Web and its accompanying narrative storyline characterise how COVID-19 revealed vulnerability in the health system, resulting in lockdowns that subsequently affected many other systems and exacerbated already existing economic, domestic, and governance challenges in the city and country. Taking COVID-19 as the “seed” element for our Impact Web resulted in constructing a more simplistic cause–effect chain at the beginning of the modelling exercise, which could be useful for replicability. Given the model's effectiveness for mapping an event as complex as COVID-19, this suggests that one could equally develop an Impact Web to understand complex climate change risks. Moreover, modelling five test cases with a flexible approach towards the “local scale” (e.g. a humanitarian camp in Cox's Bazar, a city scale in Guayaquil, a regional focus in Togo, the Indian Sundarbans, and a national scale in Indonesia) suggests that one could create an Impact Web to meet the needs of a variety of decision contexts. For example, one could create a model to assess complex risks for a river basin, a town, or even a specific community.

Applying a systems-focused lens towards analysis and mapping elements in the conceptual model, the developer of an Impact Web and the stakeholders engaged gain a more comprehensive overview of complex risks in the system they are mapping. While the final visual and the narrative storyline is the output, it is the process of developing an Impact Web that stimulates critical reflection in the modeller and involved stakeholders, which is one of the key outcomes. In each of the test cases, many of the stakeholders involved in the workshops entered with expertise in one specific sector or to share their own lived experience. However, many participants gave verbal feedback that, after collaboratively working on the model together in the workshops, they had learnt from each other and now better understood impacts and drivers outside of their areas of expertise, and thus had a better understanding of complex and cross-sectoral risks. Moreover, involving stakeholders throughout the modelling process can help identify key agents who can act as a catalyst for change (Renn et al., 2022; Özesmi and Özesmi, 2004). These can be, for example, stakeholders who perceive more causal relationships or options to change the system. Working with stakeholders to co-create the model can widen the lens for identifying critical elements, such as feedback effects and trade-offs, which can then be further analysed. Additionally, taking a participatory or bottom-up approach for the risk assessment brings in perspectives that can influence top-down decision-making.

As the conceptual model accounts not only for negative impacts but also for how policy responses and societal reactions to policies can lead to additional positive outcomes, as well as unintended consequences, i.e. risks arising from responses (Simpson et al., 2021), Impact Webs are useful to reflect on positive and negative outcomes of previous disaster risk management practices. The inclusion of interventions and response risks and impacts additionally allows for the identification and management of trade-offs or maladaptation that can occur through decision-making processes. While the outputs of an Impact Web do not quantify the severity or probability of such trade-offs, the approach is informative by revealing sometimes unclear or more nuanced relationships between decisions and negative outcomes in the system being analysed. The visual and accompanying narrative storyline can thus inform policy and risk management through learning from past impacts and how these have or have not disrupted critical societal functions (Hanf et al., 2025). They are additionally effective for pre-intervention evaluation and for communication purposes (Termeer et al., 2017; Wiebe et al., 2018).

4.2 Limitations

Given the complexity of interconnected systems and the ambiguity of system boundaries, it is not possible to characterise all interconnections using Impact Webs. These models are a simplification of reality and only the most prominent outcomes are derivable. These prominent outcomes are shaped by the developers' own inherent biases, although the participatory approach aims to reduce this by providing a mutually agreed upon heuristic of complex risks in a system. In consideration of this, it is important to acknowledge that participatory modelling is an exercise in which power dynamics come into play. Therefore, this should be considered when identifying key agents as catalysts of change. Communicating that the model is a simplification of real-world interactions, as well as who it was developed by and with, to decision-makers is important, to ensure these factors are considered in policy making.

Even though we recommend standardised constitutive elements and steps for construction, given the sheer variety of effects originating from one or multiple hazard events, no one Impact Web would be replicable, even if it were developed for the same hazards at the same scale and focus, if modelled by different stakeholders. Where to define the boundaries of the systems being mapped is vague, along with which elements are selected for the model, depending on stakeholders' views on key protection targets and societal functions. A system is usually defined according to its elements within defined system boundaries (i.e. endogenous system elements) and outside of its boundaries (i.e. exogenous system elements) (Sillmann et al., 2022), which are selected based on the scale and objectives of analysis. However, given that we developed a model with COVID-19, which affected all corners of society and did not occur within defined boundaries, as the seed element, it was difficult to know where to stop. This challenge could equally arise when developing an Impact Web in a multi-hazard multi-risk climate change context, where the cascading impacts of events are also felt across sectors and scales (van den Hurk et al., 2023). This “messiness” of complex and ongoing cascading effects that the Impact Web sheds light on is a challenge for policy, which often requires sectoral and spatially defined targets, and can equally render the direct visual output of an Impact Web difficult to engage with.

An additional challenge concerns how the outputs of the conceptual model can be integrated with quantitative data for further analysis. While the logic for our model drew inspiration from reviewing data-driven models, including fuzzy cognitive maps, influence diagrams, and Bayesian belief networks, our approach instead combines stakeholder inputs, desk review, and the outcomes of historic events to arrive at a characterisation of how the system under investigation has been affected. As data limitations are often a challenge when modelling socio-ecological systems, analytics on interactions in a multi-hazard context would be difficult.

4.3 Future research directions

A number of questions emerge from the application of our methodology that would benefit from further research. Following the steps for construction enhanced our own understanding of complex risks in the systems under investigation and the outputs are useful to communicate complexity. However, a number of modelling considerations remain to be explored that are important for disaster risk management, such as temporal dimensions, critical vulnerability moments (de Ruiter and van Loon, 2022), and system tipping points (Lenton et al., 2023). Bridging conceptual models with quantitative modelling approaches, as well incorporating lessons from methods that tackle different aspects of complex risks in more depth, such as vulnerability dynamics (e.g. Albulescu and Armaș, 2024), would be useful in this regard. Additionally, while the model is effective for assessing risks and trade-offs of interventions, a more structured decision-focused approach and methodology to see how Impact Webs can provide comprehensive entry points for disaster risk management and climate change adaptation would be useful. For example, pathway methodologies have been applied to evaluate risk management decisions in complex systems (Schlumberger et al., 2024; Haasnoot et al., 2013; Werners et al., 2021). Thus, integrating conceptual risk modelling with a pathways approach is one avenue that warrants further exploration. Understanding and mapping risk complexity is only useful if cascading effects and systemic risks can be minimised, for example through decoupling unnecessary connections across sectors. Moving from complex-risk assessment to complex-risk management needs further attention in order to strengthen the resilience of systems.

5 Conclusions

This paper offers a new conceptual modelling approach called Impact Webs, which involves identifying, characterising, and mapping complex risks. The inadequacy of single-hazard and single-risk approaches in the face of global challenges like COVID-19 and climate change emphasises the need for comprehensive risk assessments that account for interconnectivity. Impact Webs are one such methodology in an emerging field of research to do this. Their application in test cases identified critical links between multiple hazards, responses to them, drivers of risk, and root causes, as well as pre-existing societal vulnerabilities. The conceptual model provides a more nuanced understanding of how risks propagate through systems, offering valuable insights into potential feedback effects, trade-offs, and key agents that can act as catalysts of change and influence risks in a system. While the approach contributes to improving complex-risk assessment, a number of future research directions presented in this article would further advance the methodology. These include bridging the conceptual model with data-driven approaches and transitioning from complex-risk assessment to complex-risk management that strengthens systemic resilience. In the evolving and interconnected landscape of communities and societies, disaster risk reduction and climate change adaptation must account for complexity. The Impact Webs approach stands as one valuable contribution to realise this, offering a system-wide perspective for complex-risk assessment.