An impact-chain-based exploration of multi-hazard vulnerability dynamics: the multi-hazard of floods and the COVID-19 pandemic in Romania

In light of the increased frequency of multi-hazards, the dynamics of vulnerability across time, space, and different hazards emerges as an intriguing but challenging research topic. Within multi-hazard contexts, both the impacts of hazards and mitigation strategies can augment vulnerabilities, adding layers to the complexity of multi-risk assessments. Delving into these interactions, this study aims to analyse new connections in rising vulnerability that result from impacts and adaptation options, as well as their implications, putting co-occurrent powerful river flood events and the COVID-19 pandemic in Romania under the magnifying glass, taking 2020 and 2021 as references. The proposed framework relies on an impact chain that was enhanced to include new elements (i.e. augmented vulnerabilities and derived impacts) and links (i.e. connections that describe the augmentation of vulnerability), which were also used to rank the vulnerabilities based on their augmentation. The impact chain draws on various data and information sources, including the scientific literature, the feedback of first responders, reports, legislative documents, official press releases, and news reports. This research work makes a significant contribution to the field of disaster risk reduction (DRR) by broadening the purpose of the impact chain, transforming it into a first-hand, semi-qualitative tool for analysing vulnerability dynamics.

1 Introduction

In the field of disaster risk reduction (DRR), the co-occurrence of natural hazards of various types and magnitudes amid the COVID-19 pandemic has increased the attention paid to potential synergies and asynergies between pandemics and other hazards (Terzi et al., 2022). Even before the pandemic, multi-hazard analysis had switched focus from analysing all the hazards that can affect an area in a given period of time, which is often called multilayer single hazard analysis (Gill and Malamud, 2014) or the “all-hazards-at-place approach” (Hewitt and Burton, 1971), to analysing the interactions among the hazards that overlap in time and space (De Angeli et al., 2022). This shift was supported by the Sendai Framework for Disaster Risk Reduction 2015–2030 (UNISDR, 2015) and the Paris Agreement (UN, 2015).

The interactions between the COVID-19 pandemic and co-occurrent natural hazards add layers of complexity to analysing vulnerability dynamics and constructing disaster risk management (DRM) models that factor in this dynamics. One of the complications arises from the necessity to adjust traditional natural hazard management approaches to new pandemic conditions, with implications for both the impacts and the adaptation options that can increase vulnerability. The scientific literature provides several examples (Andrews, 2020; Majumdar and DasGupta, 2020; UNDRR, 2020; Kassegn and Endris, 2021; Mangubhai et al., 2021; Mishra et al., 2021; Patwary and Rodriguez-Morales, 2022; Pramanik et al., 2022; Izumi and Shaw, 2022) that point out failures of hazard management, which stem from the fact that standard operational procedures were not adapted to pandemic conditions or from the fact that efforts to tilt the SARS-CoV-2 infection curve were not adapted to fit hazard management practices. In recent years, these misfits have gained traction in the field of DRM, being debated by numerous scientists (Frausto-Martínez et al., 2020; Quigley et al., 2020; Potutan and Arakida, 2021; Albulescu et al., 2022; Hariri-Ardebili et al., 2022). A counterexample is given by Mavroulis et al. (2022), who present pandemic-adapted practices of emergency response focusing on the cases of the earthquakes that hit different regions of Greece in 2020 and 2021.

This collection of negative and positive examples motivates the need for an in-depth understanding of the interplay of different hazards and of the spatial–temporal changes in exposure, vulnerability, and adaptation. It is only by gaining a profound understanding of these matters that we can develop new DRM models that account for pandemic conditions and acknowledge that all systems have limited and variable capacity (Terzi et al., 2022) and followed this up with improved multi-risk management (Potutan and Arakida, 2021; UNDRR 2020; Ashraf, 2021; Ishiwatari et al., 2020).

To date, scientific works on the interactions between natural hazards and the COVID-19 pandemic have primarily revolved around reporting observations, overlooking the effects on the dynamics of vulnerability. Many examples (e.g. Andrews, 2020; Majumdar and DasGupta, 2020; UNDRR, 2020; Kassegn and Endris, 2021; Mangubhai et al., 2021; Mishra et al., 2021; Patwary and Rodriguez-Morales, 2022; Pramanik et al., 2022; Izumi and Shaw, 2022) pertain to hydro-climatic hazardous events amid the pandemic, offering only factual documentation on their interactions. Narrowing down our focus to the flood hazard, the compounded impacts of flood events and the pandemic are largely unknown and have been described only tangentially or in brief (Simonovic et al., 2021; Patwary and Rodriguez-Morales, 2022; Pramanik et al., 2022; Turay, 2022), although the pandemic can augment typical health-related flood impacts (e.g. injuries, gastric problems stemming from water contamination, increased stress and/or anxiety) (Simonovic et al., 2021). Instead, more literature is available on the potential effects of flood events on the dynamics of COVID-19 cases (Frausto-Martínez et al., 2020; Mavroulis et al., 2021a, b; Albulescu, 2023). What is more, the augmentation or attenuation of vulnerability conditions by previous hazard impacts (be they floods, pandemics, or other hazards) was not considered in any case study and has only been documented in relation to long-term processes (de Ruiter and van Loon, 2022).

Given the increased frequency of co-occurrent or cascading hazards in recent years, the key position of vulnerability in multi-risk analysis has become more apparent, mainly due to the fact that the impact of multiple hazards and adaptive strategies have reshaped the spatial and temporal dynamics of vulnerability. This also raises significant challenges for risk management while reinforcing vulnerability's role in portraying disasters as human constructs (de Ruiter and van Loon, 2022). This study delves deeper into the changes in vulnerability under hazard-generated impacts, taking as a case study two co-occurrent, independent hazards (i.e. floods and the COVID-19 pandemic) that severely affected Romania. At the outset, it is necessary to clarify the role of impacts resulting from multiple hazards in shaping vulnerability, with illustrative arguments from the recent literature. These instances bring to light a notable research gap that requires investigation, as detailed in the following.

Hazards generate various impacts on exposed elements with certain vulnerability levels. A particular impact has the potential to alter the vulnerability conditions that underlie another impact, whether it is caused by the same hazard or a different one. Another way to frame this issue is that the impact of a hazard changes vulnerability conditions before the recovery process reaches its end, with significant implications for the manifestation of a different hazard (de Ruiter et al., 2020). This is also mentioned by Mohammadi et al. (2024) in relation to the functionality of a system: “Additionally, events of any size, no matter how severe, that occur after a destructive event may result in the system's functionality being reduced because the system will be more vulnerable than it was prior to the big event, due to the damages that have been imposed by the first big event.” The stated situation corresponds to the third type of dynamic vulnerability identified by de Ruiter and van Loon (2022), namely the changes in vulnerability during compounding disasters that are caused by a chain of events.

A particular situation is the one where the adaptation options or the structural measures implemented to reduce the risk associated with one hazard (de Ruiter et al., 2020) or the vulnerability to one hazard (Ward et al., 2020) have unwanted effects, increasing the risk associated with a second hazard and the vulnerability to this second hazard and leading to asynergies (de Ruiter et al., 2020). This means that multi-risk analyses become even more convoluted and that they have to account for interactions that act as both causes and effects, which is a tall order for both researchers and decision-makers (Reichstein et al., 2021) but also essential to consider in the recovery phase of the DRM cycle (Mohammadi et al., 2024).

This study aims to address the research gap regarding the dynamics of vulnerability in a multi-hazard context by analysing the increases in vulnerability that stem from hazard impacts and adaptation options, taking as a case study the co-occurrent extreme river flood events and the COVID-19 pandemic in Romania in 2020 and 2021. The proposed methodological framework relies on an enhanced version of the initial impact chain developed within the PARATUS project (PARATUS Deliverable 1.1, 2023) to document the 2-year unfolding of the two independent but co-occurrent hazards. This was upgraded to capture the shifts in vulnerability by enriching it with additional elements and connection types.

Impact chains are conceptual models designed to visualise, document, and analyse the interconnections between hazards, vulnerability, and exposure that ultimately give rise to a specific risk (IPCC, 2014b; Zebisch et al., 2017; UNDRR, 2022). In this study, we refined the model to focus on vulnerability dynamics in a multi-hazard context. Such efforts are vital for elaborating post-pandemic risk management plans that avoid inadvertently introducing additional sources of unforeseen vulnerability. Risk (or hazard) management can act on vulnerability conditions both ways (de Ruiter and van Loon, 2022): producing desirable results (i.e. by decreasing vulnerability) or unwelcome outcomes (i.e. by augmenting vulnerability). In certain cases, the risk management of a hazard was responsible for increasing the risk associated with another hazard (Ward et al., 2020; de Ruiter et al., 2021a, b), and there are fair chances that this will happen again if the dynamics of vulnerability in multi-risk situations is not properly understood.

This research work makes a significant contribution to the field of DRR by broadening the original purpose of the impact chain, transforming it into a first-hand, semi-qualitative tool for analysing vulnerability. The focus is on advancing its application to analyse the effects of hazard impacts and mitigation measures on vulnerability. The conceptual framework dwells on the argument of Otto and Raju (2023), who highlight that climate change should not be entirely blamed for climate-related disasters and that vulnerability conditions must be factored in when analysing impactful events. Placing greater emphasis on the vulnerability component highlights the necessity of understanding its dynamics across time and space (de Ruiter and van Loon, 2022) and even more so in multi-hazard situations. This can be achieved by expanding the scope of impact chains to give visibility to such shifts in vulnerability and, further, to diagnose past or present multi-hazard risk management and predict potential crises, shortcomings of management approaches, and the transformation of certain vulnerabilities into drivers of vulnerability.

2 Setting the scene

2.1 Flood risk in Romania

Floods are among the most common and impactful natural hazards that affect Romania, causing significant damage throughout the country. The EM-DAT (2023) database includes 102 natural hazardous events that occurred in Romania in 1900–2023, of which flood events represent almost 52 %. These floods resulted in more than 1700 deaths, more than 146 600 homeless people, over 1.64 million affected people, and total estimated damages of about USD 8.69 billion. This incomplete dataset, complemented by other European flood-related databases (e.g. HANZE v2.1. developed by Paprotny and Mengel, 2023, and Paprotny et al., 2023), highlights the prominence of floods among the natural hazards that occur in the country of reference.

Usually, river floods follow a seasonal pattern, with the largest events occurring in the late spring months and early summer due to the convergence of high rainfall amounts and snowmelt in mountainous areas. This water input increases the discharge of the main rivers (e.g. the Danube, Siret, Prut, Olt, Mureș, and Argeș rivers) as well as of low-ranking streams. The high level of flood hazard overlaps long-standing vulnerability conditions that have only partially been discussed in the scientific literature (Constantin-Horia et al., 2009; Constantinescu et al., 2015; Vinke-de Kruijf et al., 2015; Peptenatu et al., 2020): deforestation, the extension of the residential areas and transport networks in floodplains and other flood-prone areas favoured by inconsistent law enforcement, infrastructure-related failures (e.g. poorly performing, undersized urban sewage systems), and a reactive approach to flood management that neglects preparedness and does not properly understand what salient recovery involves (Mohammadi et al., 2024). In fact, the most recent National Synthesis of the Flood Risk Management Plan (2023) still focuses on generic (i.e. forest and bridge-related measures, about 50 %), structural methods (about 33 %) to reduce the flood risk at the national scale and also includes confusion about risk-related terminology (e.g. exposure, hazard, vulnerability).

The significant flood hazard and vulnerability levels result in a high flood risk that materialises once every few years into very impactful flood events. The flood risk is addressed by the Flood Risk Management Plans elaborated for the 11 Basinal Administrations that function at the county scale. On a national scale, flood risk management is coordinated by several organisations: the Ministry of Environment, Water and Forests; the National Administration of Romanian Water; and the National Institute of Hydrology and Water Management. These organisations are often criticised for their underperformance in managing flood risk by both scientists (Vinke-de Kruijf et al., 2015) and civil society, an attitude that is justified by the wreckage in the aftermath of large flood events that were forecasted and communicated by hydrological warnings.

Flood occurrences and the quantification of their associated impacts are not sufficiently documented in Romania, as evidenced by the absence of relevant official databases. Such information has to be obtained from alternative sources, like weather and hydrological forecasts and news reports. The flood events taken under analysis in this paper were identified using the hydrological warnings issued by the National Institute of Hydrology and Water Management during 2020–2021, which were corroborated with information from a national news platform. Multiple news reports were used for the validation of each extracted piece of information.

In 2020, there were five major flood events that necessitated the evacuation of people, all in June (16, 18, 19, 23, and 26 June). In the subsequent year, there were eight such events, of which two occurred in May (13 and 18 May), two in June (18 and 19 June), and four in July (15, 16, 19, and 20 June) (Albulescu, 2023). In addition, flood events that did not involve evacuation procedures but were still included in the impact chain occurred in January, August, and December 2021. The 2020–2021 flood events resulted in seven human casualties (Meteo Romania, 2020, 2021; HANZE v2.1., 2023). The spatial extent of the various impacts of the powerful river floods in 2020 and 2021 is presented in Fig. 1.

Figure 1 Spatial extent of the impacts of the extreme flood events that affected Romania in 2020 and/or 2021. Impacts: (a) human casualties; (b) displaced/(self-)evacuated people; (c) flooded/damaged households or houses; (d) damaged bridges; (e) isolated human communities; (f) railway transportation impairment; (g) damaged facilities/cutoff of electricity, gas, or water supply; (h) sewage system overflow; (i) fallen trees; (j) landslides (not as a flood impact but as secondary hazards co-occurring with or subsequent to floods, generating a range of severe impacts); (k) river water contaminated with rubbish; (l) dead/missing animals; (m) flooded croplands; (n) damaged cars; (o) disrupted tourism activities; (q) flooded business buildings; (p) flooded public buildings (including one hospital); and (r) disrupted ambulance service.

2.2 The COVID-19 pandemic in Romania

The first confirmed case of COVID-19 registered in Romania was recorded on 26 February 2020, and the first two deaths due to this disease occurred approximately a month later. Up to the beginning of June 2023, more than 3.4 million cases of COVID-19 and over 68 000 deaths were registered in the country of interest, of which 53 % and 86 %, respectively, can be traced back to the first 2 pandemic years (WHO Dashboard, 2023). This human toll unfolded in five pandemic waves (Fig. 2), of which the fourth one, starting in 2022, was the most aggressive.

Figure 2 The dynamics of new cases of COVID-19 in Romania with a focus on 2020 and 2021, plotted against the periods with and without restrictions and clusters of flood events (COVID-19 data source: WHO Dashboard, 2023).

Like in many other countries, the pandemic waves in Romania followed a seasonal pattern that was conditioned by temperature and humidity (Mecenas et al., 2020). Figure 2 indicates that the same seasonal pattern was followed by the COVID-19-related restrictions. As an immediate response to the emergence of COVID-19 cases, at the end of March 2020, the Romanian Government declared a national state of emergency (Decree no. 195/2020, 2023) and imposed a lockdown of increased severity compared to the ones implemented in other countries. This ended on 15 May 2020 and was followed by a 2-year national state of alert during which periods free of restrictions – which overlapped the summer months – alternated with periods of circulation restrictions for citizens that aimed to tilt the SARS-CoV-2 infection curve – which were specific to the cold season (Fig. 2).

Figure 2 shows that the flood events that occurred in June 2020 correspond to the beginning of a restriction-free period, which was followed by one with severe restrictions. The floods of January 2021 overlapped a period with restrictions for everyone, when wearing face masks was mandatory, circulation was prohibited between 23:00 and 05:00 local time, social gatherings were banned, and a large part of work was moved to virtual environments. Approaching May 2021, circulation restrictions were lifted only for vaccinated people, and it was not until 26 July that all COVID-19-related restrictions ceased. This means that the flood events that happened on 13 and 18 May 2021 overlapped a period with restrictions for unvaccinated people and that the ones in June–August corresponded to a restriction-free interval. The flood events of December 2021 occurred during a period of restrictions imposed on unvaccinated people.

3 Methodology

The proposed methodological framework aims to identify and analyse the augmentation in vulnerability conditions within a multi-hazard context. This framework dwells on impact chains as instruments for documentation, visualisation, organisation, and scientific inquiry, ultimately broadening their application to fit the objective of studying the dynamics of vulnerability – particularly the augmentation of vulnerability – and henceforth to turn them into diagnosis and prediction tools. With this addition, the documentary focus of the chain progresses to a more analytical stance, specifically geared towards identifying and tracking the transformation of specific vulnerabilities into drivers of vulnerability.

The next section presents two distinct workflows within the methodological framework (Fig. 3): building the impact chain initially developed by the authors within the PARATUS project (PARATUS Deliverable 1.1, 2023) – which was further strengthened by the first responders' input – and, secondly, its enhancement to account for vulnerability augmentation.

Figure 3 Methodological framework.

3.1 Building the impact chain

Impact chains represent conceptual models designed to facilitate the investigation of climate and disaster risk under a structured analysis framework for the risks associated with climate-related impacts (UNDRR, 2022). They have been used for elicitation, conceptualisation, analysis, and information-sharing purposes, as tools that explore and analyse the impacts of single hazards or multi-hazards specific to past or potential hazard events following different operational frameworks (e.g. expert workshop, desktop analysis, machine-generated) and taking into consideration different spatial and temporal scopes (Pittore et al., 2023). There are numerous examples of impact chains being integrated into vulnerability or risk assessments specific to climatic aspects (Becker et al., 2014; Schneiderbauer et al., 2020; Zebisch et al., 2017, 2021; Menk et al., 2022).

In this paper, impact chains were used as models of cause and effect (Menk et al., 2022) that were upgraded to capture the augmentation of vulnerability by hazard impacts and adaptation options, with limited participation of stakeholders (i.e. only integrating the feedback of first responders involved in flood emergency interventions). Unlike the scientific papers reviewed by Menk et al. (2022), this study does not integrate impact chains as tools for the assessment of vulnerability or risk pertaining to a climatic hazard but broadens their scope to focus on vulnerability dynamics within a multi-hazard context that involves a hydrological hazard (i.e. flood) and an epidemiological one (i.e. the COVID-19 pandemic). This approach aligns with the recommendation of Zebisch et al. (2021) that the “relatively linear and sectorial approach of impact chains could be widened to impact webs, which would include feedback relations and cross-connections”, which is also supported by Sparkes et al. (2023).

The structure of an impact chain includes elements that can be considered the fundamental units of a hazard-related context and the connections established between them. The elements can take the form of hazards, impacts, exposed elements, vulnerabilities, and adaptation options, defined according to the Sendai Framework Terminology on Disaster Risk Reduction (UNDRR, 2017). Given the central role of impacts, vulnerabilities, and adaptation options in the proposed vulnerability augmentation framework, we consider that their meaning should be highlighted here. In this paper, impacts particularly refer to the negative effects of a hazardous event or a disaster, while vulnerability represents the “conditions determined by physical, social, economic and environmental factors or processes which increase the susceptibility of an individual, a community, assets or systems to the impacts of hazards” (UNDRR, 2017). Adaptation options are measures meant to attenuate the negative impacts by addressing one or more vulnerabilities or impact mechanisms (IPCC, 2014a). These elements are organised in a chain-resembling structure that relies on different connection types: causes, affects, relates to, impacts, and mitigates. Detailed guidelines on how such connections were established within the PARATUS project are provided by Pittore et al. (2023) and PARATUS Deliverable 1.1 (2023).

Figure 3 illustrates a comprehensive breakdown of the construction of the impact chain through a combination of knowledge, data, and information extracted from a diverse range of sources: scientific papers, legislative documents, official press releases, reports, statistical datasets, and grey literature in the form of news reports (Fig. 3). The impact chain was implemented in Kumu, which is a powerful mind-mapping tool that allows for a variety of mapping settings (e.g. stakeholder, systems, social network, community asset, concept mapping), as well as import and export options (Kumu, 2023). Floods, the COVID-19 pandemic, and heavy rainfall were considered primary hazards within the impact chain, but only the first two are analysed in this study due to their significant impacts. Other secondary hazards (e.g. strong wind, landslides) co-occurred with the other hazards, but they were less impactful compared to the mentioned hazards in the analysed multi-hazard context.

The first phase of the building process (Fig. 3) relied on a literature review regarding the impacts of flood events and the pandemic, complemented by a supplement of examples specific to the flood events in 2020–2021 collected from studying the grey literature. As part of the first review, the most prominent scientific databases (i.e. Web of Science, Google Scholar, ResearchGate, and PubMed) were searched for relevant papers using the following keywords: “Covid-19 pandemic Romania”, “Covid-19 pandemic impact Romania”, “Floods Romania 2020”, and “Floods Romania 2021”. Next, during the exclusion phase, the titles and abstracts of the collected articles were analysed in order to select only the research works with a clearly defined and relevant aim, a thorough and methodologically validated analysis of the impacts of the hazards, and an adequate spatial and temporal focus. In the last phase of the literature review on impacts, content analysis was performed on the selected papers, and the relevant impacts were included in a database. The grey literature review was performed using a prominent online Romanian national news portal, Digi24 (2023). It was limited to the impact of floods and did not include the impacts of the pandemic. Supplementary documentation was necessary to extract pandemic impacts from legislative documents (Decree no. 195/2020, 2023), official press releases (CCR, 2022), reports (WHO, 2020b; HSRM 2021a, b; OECD and European Observatory on Health Systems and Policies, 2021; CDC, 2022; WHO Dashboard, 2023), and statistical datasets (Eurostat, 2021). Next, the Kumu design for appropriate flood impacts was enriched with photographs and maps depicting the spatial distribution of impacts in 2020, 2021, or both, at the county or local scale.

The last two phases of the construction process focused on the elicitation of vulnerabilities and adaptation options under an expert-knowledge-based approach. The identification of these elements was backed by scientific findings wherever possible in order to obtain a valid configuration of the chain (Fig. 3).

Regardless of their type, all elements and connections were integrated into Kumu with a short description, associated sources, and references. Cumulatively, the impact chain drew from 46 scientific papers (including 1 on the 595 first responders engaged in on-site flood management who gave feedback), 1 legislative document, 1 official press release, 1 Eurostat statistical dataset, 6 official reports, and 75 news reports.

An addition to the impact chain developed in the early stages of the PARATUS project (PARATUS Deliverable 1.1, 2023) was to integrate the feedback of 595 first responders involved in flood management in 2021, focusing on aspects concerning preparedness, coordination, and experience, upon extracting them from the study of Fekete et al. (2023). Their perception of the problems encountered during flood-related emergency interventions, potential improvements, cooperation among volunteers, provision of information about the deployment, and flood-affected infrastructure served as a basis for eliciting a new set of vulnerabilities and adaptation options.

3.2 Enhancing the impact chain

In the next phase, we extended the application of impact chains to explore the third type of vulnerability dynamics identified by de Ruiter and van Loon (2022), namely the changes in vulnerability conditions related to compounded hazards or, more accurately, the augmentation of vulnerability in a multi-hazard context. Elevating the impact chain from its mentioned original purposes to a tool for analysing vulnerability dynamics represents a pioneering research endeavour, standing out as an element of methodological novelty.

This broadening of the original application of the impact chain was done by (1) establishing new types of connections between the impacts/adaptation options and vulnerabilities, (2) introducing new types of elements (i.e. augmented vulnerabilities, derived impacts) based on these connections, and (3) ranking the vulnerabilities in the impact chain based on their augmentation. These steps were implemented to construct an enhanced impact chain, building on the previous version that documented the unfolding of the selected co-occurrent hazards in Romania in 2020–2021.

Figure 4 Conceptual framework of the new elements and links of the enhanced impact chain.

To set the stage for the implementation of the vulnerability augmentation framework, an in-depth analysis of the vulnerabilities in the impact chain was performed. The vulnerabilities were classified according to their related hazards, types, spatial scales, and links to specific adaptation options. This classification provided a better understanding of the contribution of vulnerabilities to the manifestation of flood and COVID-19 pandemic impacts. Further, the impact chain was enhanced by identifying new connection types between impacts/adaptation options and vulnerabilities (Fig. 4), drawing from the types of maladaptation to climate change and their implications for vulnerability proposed by Schipper (2020). The three types of maladaptation in question (i.e. rebounds vulnerability, shifts vulnerability, and creates negative externalities) were tailored to suit the multi-hazard context and complemented by a new connection type also relevant to the impact chain (i.e. deepens vulnerability). These new connections account for the augmentation of a vulnerability by a given impact or adaptation option in a way that could not have been prevented or precluded. The new connections are defined as follows:

• deepens (vulnerability) – the augmentation of a vulnerability by an impact, both relating to the same hazard;

• shifts (vulnerability) – the augmentation of a vulnerability to a certain hazard by an impact caused by a different hazard;

• rebounds (vulnerability) – the augmentation of a vulnerability by an adaptation option that aimed to attenuate an impact or vulnerability affecting a certain group/system but ended up increasing a vulnerability of that same exposed element;

• creates negative externalities – the augmentation of a vulnerability by an adaptation option that has adverse effects on anyone who is not targeted by it (Schipper, 2020).

To set up the new connections, each impact was studied from the perspective question, “which vulnerability can be augmented by this impact?” The adaptation options were also scanned according to the same adjusted question, with the goal of identifying possible unwanted effects of measures intended to lessen certain vulnerability conditions, as reported by de Ruiter et al. (2021a, b). The vulnerabilities that were connected with impacts and/or adaptation options through the above said new links that express vulnerability augmentation were transformed into elements called augmented vulnerabilities.

A noteworthy situation that emerged from the experience-based feedback of first responders is the one where certain vulnerabilities that influence the manifestation of impacts can also slow down or obstruct the implementation of adaptation options. Such instances were marked by a new type of connection called “slows down/obstructs”, established between vulnerabilities and adaptation options.

Within the new conceptual framework of the enhanced impact chain (Fig. 4), certain vulnerabilities, upon augmentation by an impact, can also act as impacts that further on amplify the very impact that increased the vulnerability in the first place. This process represents a positive feedback loop, where the initial impact augments a vulnerability that can be viewed afterwards a (derived) impact that will reinforce the first impact in the future. Such augmented vulnerabilities that also act as impacts were introduced in the enhanced impact chain as derived impacts and linked to the vulnerability element that they share their name with by “relates to” connections. These “relates to” links are not visible within the enhanced impact chain in Kumu in order to reduce the visual strain. Subsequently, the derived impacts were linked with the impact that deepened/shifted the corresponding vulnerability by a newly introduced type of connection referred to as “sharpens” (Fig. 4). These “sharpens” connections convey the message that the augmented vulnerability will intensify the impact that initially augmented the vulnerability, rendering this impact more prominent than before.

The ranking of the vulnerabilities based on their augmentation relied on the number of augmentation connections from impacts to vulnerability (i.e. deepens, shifts) and on the number of augmentation connections from adaptation options to vulnerability (i.e. rebounds, creates negative externalities). These were computed for each of the 26 vulnerabilities in the enhanced impact chain. Augmented vulnerabilities were transformed into Z scores based on the number of augmentation links they collect, showing the relationship of the values in terms of distance to the mean of the distribution. The basic principle we have assumed is that the further away a vulnerability is from the mean of the distribution, the more outstanding/augmented it is. Next, two rankings of the vulnerabilities were calculated based on the Z scores of the augmentation connections of impact–vulnerability and adaptation option–vulnerability. The final ranking was computed using a statistically weighted approach by attributing a weight of 70 % to the impact–vulnerability connections and a weight of 30 % to the adaptation option–vulnerability connections. The rationale behind the assigned weights lies in the fact that impacts augment vulnerabilities to a greater extent than adaptation options do, at a ratio of 54 % augmentation by impacts vs. 4 % augmentation by adaptation options, while 12 % is attributable to both impacts and adaptation options combined. The ascending order of the final ranking showed the extent to which the vulnerabilities were augmented overall, from the most to the least augmented.

4 Results

This section focuses on the augmentation of vulnerability stemming from flood and pandemic impacts and the adaptation options implemented to mitigate vulnerabilities and/or impacts. It starts with a classification of all vulnerabilities in the impact chain, followed by details on the augmented ones and on the vulnerabilities that also act as derived impacts. Finally, we rank the amplified vulnerabilities to identify those most significantly affected by the interplay of hazard impacts and adaptation options.

The intricate configuration of the impact chain does not allow for a proper visualisation within this paper, but it can be accessed online on the Kumu platform (https://kumu.io/cosminaalbulescu/ic-augment-reconstructed#impact-chain-on-floods-and-the-covid-19-pandemic-with-augmented-vulnerabilities, last access: 22 August ).

4.1 Classification of vulnerabilities

The enhanced impact chain includes 26 vulnerabilities upon integrating the perception of first responders. More than half (58 %) of these vulnerabilities were related to flood events, 23 % pertained to both hazards, and 19 % of them corresponded to the pandemic (Table 1). Most of the vulnerabilities contribute to prominent flood impacts such as the flooded/damaged houses or households, the flooded/damaged/blocked roads, the displaced/(self-)evacuated people, or multi-hazard impacts like increased stress or anxiety and the potential increase in new COVID-19 cases.

Table 1 Number and proportion of vulnerabilities by (a) hazard, (b) type, and (c) spatial scale in the enhanced impact chain.

The vulnerabilities were grouped according to their type, as described in Appendix A. More than a third (35 %) of them stemmed from failures of emergency management, while 19 % derived from failures of territorial planning or of medical management (Table 1). At the same time, the number of vulnerabilities associated with the coping capacity (15 %) or infrastructure (12 %) was rather low. It should be highlighted that this is a simple classification and that the range of vulnerabilities is more nuanced, including governance-related vulnerabilities (e.g. improper governance structure for effective flood management, flood management not adapted to the COVID-19 context, ineffective institutional communication) and development-related vulnerabilities (e.g. development of inhabited areas in flood-prone areas, development of infrastructure in flood-prone areas, poverty, depleted capacity due to seasonal patterns of hazards, low-quality construction materials, ineffective sewage system). In terms of spatial scale, most vulnerabilities were identified at the local level (69 %) and only 23 % were specific to the entire country (i.e. the national scale) (Table 1).

Only a third of the vulnerabilities were mitigated by adaptation options: three of them were related to the COVID-19 pandemic (i.e. low-performance medical system, insufficient medical personnel, insufficient intensive care unit (ICU) capacity), three others were related to both hazards (i.e. flood management not adapted to the COVID-19 context, ineffective institutional communication, uncooperative population), and the rest were related to floods (i.e. improper mapping and visualisation of affected areas, lack of equipment for first responders).

The 31 % rate of mitigated vulnerabilities shows that most adaptation options targeted impacts, which means that they produced short-term positive change, addressing the causes of the medical crisis and the multi-hazard vulnerabilities only to a limited extent and the flood vulnerabilities to an even lesser extent. The adaptation options that mitigated vulnerabilities related to the COVID-19 pandemic were the most numerous: four in the case of insufficient ICU capacity, three in the case of insufficient medical personnel, and two in the case of the low-performance medical system. The main adaptation options were related to the support provided by other states (in terms of medical equipment and staff), the transfer of COVID-19 patients to other countries, the establishment of new modular hospitals, and the hiring of additional medical personnel, all allowing the fight against the pandemic to continue.

All the other mitigated vulnerabilities were addressed by a single adaptation option, showing a unilateral approach. In the case of floods, several vulnerabilities were mitigated by an “umbrella” adaptation option that includes various actions specific to each context, namely the great capacity of first responders to develop creative solutions during crises and cope with new challenges. With a few exceptions (e.g. the RO-Alert SMS messaging system and hydrological warnings, which are part of early warning systems), most of these flood-related adaptations focused on alleviating the “symptoms” of the local crisis and did not address its root causes. For instance, during the flood event on 18 June 2020, riverbanks were heightened by firefighters with sandbags to prevent water from reaching the houses in proximity, in two settlements in Caraș-Severin County. Other examples of short-term, recovery-related adaptation options are the removal of fallen trees from streets and roads or of floodwater from households, buildings, or roads.

4.2 Statistical overview of augmented vulnerabilities and augmentation links

To identify the augmented vulnerabilities, 41 new connections (Appendix B) expressing different forms of vulnerability augmentation were established between the impacts that would potentially generate increases in vulnerability (i.e. deepens or shifts vulnerability links) or from the adaptation options with this effect (i.e. rebounds vulnerability or creates negative externalities), following an expert-based approach.

In the enhanced impact chain, 18 (69 %) out of 26 vulnerabilities were augmented, some of them more than once, by different impacts or adaptation options: 14 (54 %) were augmented by hazard impacts, 1 (4 %) was augmented by solely adaptation options, and 3 (12 %) were augmented by both impacts and adaptation options. The vulnerabilities that increased because of both elements were as follows: the uncooperative population, flood management not adapted to COVID-19 conditions, and shallow implementation of preventive measures.

The distribution of augmented vulnerabilities among the hazards is unbalanced: half of them are specific to floods, 28 % to the COVID-19 pandemic, and 22 % to both hazards. Also, the augmented vulnerabilities related to medical or emergency management account for 67 % of the total, and the other three categories (i.e. vulnerabilities related to coping capacity, infrastructure, or territorial planning) each account for less than 20 %. Most augmented vulnerabilities manifest at the local scale (67 %) and 22 % of them at the national level. Almost all vulnerabilities that were mitigated by adaptation options were also augmented either by hazard impacts (i.e. lack of equipment for first responders, improper mapping and visualisation of affected areas, low performance of the medical system, insufficient medical personnel, insufficient ICU capacity) or by both impacts and adaptation options (i.e. uncooperative population). The only mitigated vulnerability that was not also augmented was the ineffectiveness of institutional communication.

Almost half (49 %) of the new augmentation connections convey a deepening effect on vulnerability elements, and more than a third (37 %) augment vulnerability by shifting it from one hazard to the other (Fig. 5). The increases in vulnerability caused by adaptation options total about 15 %, with equal unwelcome effects (about 7 %) resulting from rebounding vulnerability and creating negative externalities. Details on the augmentation of certain vulnerabilities by impacts or adaptation options are provided in Appendix B.

Figure 5 Proportion of vulnerability augmentation connections in the enhanced impact chain.

4.3 Derived impacts

When augmented, certain vulnerabilities can function similarly to impacts, reinforcing the very impact that initially increased the vulnerability, forming a positive loop feedback composed of “deepens” or “shifts” links and “sharpens” links. Such augmented vulnerabilities with double status were duplicated in the enhanced impact chain and labelled as “derived impacts”, as detailed in Appendix B. Some vulnerabilities underwent multiple transformations into derived impacts because they acted as (derived) impacts in relation to more than one augmentation-generator impact (Fig. 6). This resulted in a larger number of cases where the augmentation of a vulnerability created derived impacts (15 cases) compared to the number of actual derived impacts (9) in the chain. The vulnerabilities that also act as derived impacts were low-performance medical system (reinforced as a derived impact three times), insufficient medical personnel (three times), insufficient COVID-19 testing capacity (two times), and uncooperative population (two times) (Fig. 6). On the other hand, the vulnerability called households at a short distance from river, insufficient/ineffective hard engineering infrastructure/measures, improper mapping and visualisation of affected areas, lack of equipment for first responders (including protective gear), and work overload on first responders were transformed into derived impacts only once.

Figure 6 The augmented vulnerabilities that function as derived impacts (DIs) more than once, linked through “sharpens” connections to the impacts that generated the augmentation.

Of the vulnerabilities that also act as derived impacts, 45 % pertain to floods, 33 % to the pandemic, and 22 % to both hazards. More than half (60 %) of the derived impacts are associated with “deepens” connections, suggesting that the augmentation of the vulnerabilities and their subsequent reinforcement as derived impacts are mostly related to the same hazard. All the identified derived impacts are detailed in Appendix B, with the focus in this section limited to the most significant ones (Fig. 6).

The augmentation of the low-performance medical system was caused by the effects of the pandemic on other diseases, the economic loss caused by both floods and the pandemic, and the economic challenges brought about by the pandemic (Appendix B, Fig. 6a). In the first instance, the COVID-19 pandemic delayed the provision of treatment for certain diseases (Cucu et al., 2021; Dionisie et al., 2022; Barbos et al., 2023) or accelerated the progression of diseases like kidney pathology (Trifanescu et al., 2022; Mureșan et al., 2022; Tudora et al., 2023). These circumstances exerted additional strain on the already-suboptimal medical system, contributing to the exacerbation of other health issues. In addition, the economic loss and the pandemic-related economic challenges have the potential to perpetuate the underfunding of the medical system, with negative effects on its performance. In return, the underperforming medical system is a cause of both economic loss (due to treatment delays and shortages of medical and human resources) and economic challenges stemming from its coping ineffectiveness.

The augmentation of insufficient medical personnel was linked to impacts like human casualties, the effects of the pandemic on other diseases, and increased stress or anxiety (Appendix B, Fig. 6b). The victims of COVID-19 included healthcare staff that became infected with the virus while attending to COVID-19 patients, which deepened the shortage of personnel and subsequently significantly altered their capacity to provide life-saving healthcare to the thousands of patients in need, therefore increasing the human death toll. Similarly, the surge in workload for medical personnel, resulting from aggravated diseases due to the COVID-19 pandemic, limited the availability of healthcare staff dedicated to tending to COVID-19 patients. Consequently, the insufficient number of medical personnel negatively affected the development of certain diseases, as timely and appropriate treatment was not administered. Lastly, the increased stress/anxiety temporarily affected the mental health and well-being of the medical staff, necessitating breaks in their duties. The temporary unavailability of their colleagues heightened the stress/anxiety levels among the remaining healthcare professionals, as well as among the general public, who was aware of the scarcity of medical human resources during critical times.

The insufficient COVID-19 testing capacity was augmented by the road transportation impairment resulting from floods and also by the disrupted ambulance service (Appendix B, Fig. 6c). During and immediately after floods, people were precluded from reaching COVID-19 testing centres and ambulances were prevented from reaching the people who requested to be tested at home. Both of these obstructions limited the testing capacity. In return, the limited testing capacity at the local scale forced people to undertake road journeys to the available testing centres located in other settlements, sometimes at great distances, resulting in traffic jams in numerous places and on numerous occasions.

In the analysed multi-hazard case study, the population became even more uncooperative because of the diminished trust in authorities and the increased stress/anxiety associated with both floods and the pandemic (Appendix B, Fig. 6d). The lessened credibility of authorities can be traced back to faulty pandemic management and the lockdown imposed in March–May 2020 (Džakula et al., 2022) and also to the economic problems resulting from both hazards. This increased reluctance to collaborate with first responders and authorities also undermined trust in authorities, establishing a positive feedback loop. Amid flood-related interventions, the escalation of stress or anxiety levels can make people fearful and less willing to collaborate with first responders, hindering rescue or evacuation operations. Conversely, this reluctant attitude of the population and associated difficulties can increase the stress/anxiety of the first responders on duty.

4.4 Ranking of augmented vulnerabilities

The last part of the analysis is dedicated to the ranking of augmented vulnerabilities under the proposed statistical approach, with the goal of pinpointing those vulnerabilities expected to experience the most substantial increase. By building augmentation links between impacts and vulnerabilities (i.e. deepens, shifts) and between adaptation options and vulnerabilities (i.e. rebounds, creates negative externalities), we can identify which vulnerabilities are expected to increase in the future and why. The increase in vulnerability refers to those levels of vulnerability that are expected to be higher in the future, provided that the next hazardous events will lead to similar impacts and that similar adaptation options will be implemented to mitigate those vulnerabilities.

This ranking, along with the corresponding computational values, is depicted in Table 2. The top three augmented vulnerabilities were uncooperative population, the low-performance medical system, and flood management not adapted to the COVID-19 context. The first and third are multi-hazard vulnerabilities that pertain to both floods and the pandemic, while the low-performance medical system is specific to the pandemic. In terms of type, the most augmented vulnerability relates to coping deficiencies, while the next two refer to either medical or emergency management failures. As for the scale of manifestation, the uncooperative population is a local-level vulnerability, while the other two manifest at a broader national scale.

Table 2 Ranking of vulnerabilities based on their augmentation. I – impact; V – vulnerability; Ao – adaptation option.

The mentioned vulnerabilities were followed by other management-related vulnerabilities, such as insufficient medical personnel, a lack of equipment for first responders, the shallow implementation of preventive measures, the insufficient COVID-19 testing capacity or ICU capacity, and the work overload on first responders, most of them relating to the COVID-19 pandemic (Table 2). The least augmented vulnerabilities are specific to the flood hazard (i.e. defective coordination of first responders from multiple counties, deforestation, households at a short distance from river, etc.).

When looking at the augmentation produced by impacts, the ranking resembles the final one (Table 2), which is to be expected due to the 70 % weight of the impact–vulnerability augmentation connections. The difference is that the flood management not adapted to the COVID-19 context, the insufficient medical personnel, and lack of equipment for first responders were augmented by impacts to equal extents, which also holds true for the next four vulnerabilities (sixth place) (Table 2). On the other hand, the vulnerabilities that were most augmented by adaptation options (to equal extents) were the shallow implementation of preventive measures and the absence of preparedness at the individual level. In the ranking of augmented vulnerabilities by adaptation options, these were followed by the uncooperative population and flood management not adapted to the COVID-19 context, which both occupy third place (Table 2). All other vulnerabilities were not augmented by adaptation options.

Although not augmented, the vulnerabilities at the bottom of Table 2 have the potential for escalation due to the fact that they were not addressed by any adaptation options. These vulnerabilities are as follows: assets at a short distance from river, depleted capacity due to seasonal patterns of hazards, development of inhabited areas or infrastructure in flood-prone areas, improper governance structure for effective flood management, ineffective sewage system, and low-quality construction materials. Except for the depleted capacity due to seasonal patterns of hazards (a multi-hazard vulnerability), all of these are specific to floods.

5 Discussion

The current study stands at the forefront of research, bringing into the spotlight the potential increase in vulnerability within the unprecedented co-occurrence of the COVID-19 pandemic and the multiple flood events that affected Romania in 2020–2021. The configuration of the enhanced impact chain shows a convoluted multi-hazard, wherein certain hazard impacts and adaptation options have an augmentation effect on underlying vulnerabilities. In return, some of the augmented vulnerabilities also act as derived impacts that reinforce the very impacts that increased vulnerability in the first place. In this sense, both hazards and what we do to mitigate them can be considered indirect generators of changes in vulnerability, with deep implications for how we approach multi-risk management.

In the presented case study, the enhanced impact chain shows that vulnerability is expected to increase based on the augmentation in different forms conveyed by the new links, as 69 % of vulnerabilities were augmented by either impacts or backfiring adaptation options or both. Another expected path to increasing vulnerability is related to the limited range of adaptation options that address vulnerabilities (only one-third of the vulnerabilities were addressed by mitigation measures). This means that (1) the unforeseen implications of impacts that act as vulnerability enhancers, (2) the wrongful action intended to mitigate vulnerability and/or impacts, and (3) inaction can set the premises for increased vulnerability levels that will render multi-risk management more difficult (Fig. 7).

Figure 7 Conceptual paths of rising vulnerability: (a) augmentation of vulnerability resulting from hazard impacts, (b) augmentation of vulnerability resulting from misfiring adaptation options, and (c) perpetuation of vulnerability due to inaction. t0 denotes the present moment, and t1 denotes the future moment.

5.1 Conceptual paths of rising vulnerability

The first conceptual path refers to the impacts of the flood events and the pandemic (Fig. 7a). These mainly reinforce deeply rooted vulnerabilities, like the reluctance of the population to collaborate with first responders and/or authorities (Fekete et al., 2023) or the low performance of the Romanian medical system, which has been widely reported in the literature (OECD and European Observatory on Health Systems and Policies, 2021; Lupu and Tiganasu, 2022; Popescu et al., 2022). The top three most impact-augmented vulnerabilities also include deficiency in aligning flood management with pandemic conditions (Table 2), which can be associated with local increases in new COVID-19 cases (Albulescu, 2023) and is expected to cause further issues in similar future multi-hazard scenarios unless amended. Other top impact-augmented vulnerabilities are related to medical or emergency management failures (i.e. insufficient medical personnel, lack of equipment for first responders). All of the above said vulnerabilities were addressed by various adaptation options, but most of them produced short-term effects and were not part of larger vulnerability-reduction-oriented schemes. Several examples include clever on-the-spot solutions implemented by first responders to engage with the uncooperative population, the hiring of additional medical staff and volunteers during the pandemic, and the support received by Romania from other countries in terms of medical resources.

The second line along which the augmentation vulnerability propagates is established when adaptation options misfire and end up increasing vulnerabilities (Fig. 7b). The most augmented vulnerabilities in this regard concern the preparedness phase of DRM: the shallow implementation of preventive measures against the COVID-19 pandemic and the absence of preparedness at the individual level when confronted with floods (Table 2). In Romania, the low level of preparedness was associated with an external locus of control (Armaș, 2008; Armas et al., 2015; Albulescu et al., 2021), and it was also reported by first responders who performed interventions during the floods of 2020–2021 (Fekete et al., 2023). The analysis unravels the possibility that these vulnerabilities related to coping capacity can evolve into vulnerability drivers. Against this background, a major gap emerges between the efforts undertaken by first responders in the response phase and the lack of interest on the part of citizens, who take no or little action to prepare to withstand floods or to prevent the spread of the SARS-CoV-2 virus during the preparedness phase.

Another thing to consider is that the top three impact-augmented vulnerabilities are the same as the ones that rank vulnerabilities based on the combined augmentation effects of impacts and adaptation options. However, the vulnerabilities that were augmented by both impacts and adaptation options (to different extents) are the uncooperative population, the lack of adaptation of flood management to pandemic conditions, and the shallow implementation of preventive measures against the pandemic (Table 2). In future multi-risk management plans, special emphasis should be placed on addressing these vulnerabilities, particularly given that the first two are related to both hazards.

The third conceptual path of increasing vulnerability is through inaction (Fig. 7c), standing out since the number of vulnerabilities (26) is 2 times larger than the ones of adaptation options (13) and only about a third of the vulnerabilities were targeted by adaptation options. When looking at the entire enhanced impact chain, a striking imbalance is highlighted: most flood-related mitigation efforts focused on impacts rather than vulnerabilities, while pandemic-related adaptation options primarily addressed vulnerabilities rather than pandemic impacts. The only flood-related vulnerability addressed by adaptation options is the improper mapping and visualisation of affected areas. This means that human communities might be equally or more vulnerable to floods in the future. What is more, even the adaptation options that mitigated the flood impacts mostly provide short-term solutions (e.g. the heightening of riverbanks with sand banks to prevent or limit the flooding of houses or households) or have negative unforeseen effects (e.g. the RO-Alert SMS messaging system or the hydrological warnings that can reduce the motivation of the people who are not located in an area affected by a particular flood event to prepare for future floods or to undertake COVID-19 prevention measures, as described in Appendix B).

On the contrary, many of the key pandemic vulnerabilities were tackled by adaptation options (e.g. low-performance medical system, insufficient medical personnel, insufficient ICU capacity), and the same can be stated for multi-hazard vulnerabilities (e.g. flood management not adapted to the COVID-19 context, ineffective institutional communication, lack of equipment for first responders, uncooperative population) (Appendix A). Nevertheless, the brighter perspective described here is overshadowed by the fact that the very same vulnerabilities (except the ineffective institutional communication) were augmented by hazard impacts and/or adaptation options.

This approach leaves deeply engrained vulnerabilities to floods unaltered (e.g. the location of households and/or assets at a short distance from river, the improper governance structure for effective flood management, the shallow implementation of the absence of individual flood preparedness) but ready to resurface during future hazardous events. In other words, the implemented adaptation options belong to the response and/or recovery phase of the DRM, and no initiatives have been undertaken in the preparedness phase. What is worse, as argued above, certain adaptation options augment the two prominent vulnerabilities specific to the preparedness phase (e.g. the shallow implementation of COVID-19 preventive measures and the absence of individual flood preparedness measures).

The reactive approach is typical of developing societies or of early, one-dimensional flood management approaches (Scott et al., 2013), complemented by an external locus of control of the population (Armaș, 2008; Armas et al., 2015). Sound risk mitigation requires integrating preparedness for future hazards into the recovery process (Johnson and Jensen, 2023), all with a high degree of flexibility (White and Haughton, 2017), but such efforts were absent in the presented case study. Therefore, the unbalanced DRM-phase distribution of the adaptation options has prominent implications for the dynamics of vulnerability in the sense that it allows vulnerability to perpetuate and further contribute to future hazard impacts.

Another aspect to ponder is that the depleted capacity due to seasonal patterns of hazards, although not augmented, was not addressed by any adaptation options. Both floods and pandemic waves follow seasonal patterns, allowing human communities to prepare for their impacts (to some extent) by following a predictive but tight timeline. Considering the unaddressed vulnerabilities, together with the short-sighted nature of the adaptation options, human communities affected by the COVID-19 pandemic did not fully recover until the occurrence of floods, until the next pandemic wave, or perhaps not even from one flood event to the next. In this context, it can be expected that the overall vulnerability level will increase, since the recovery process is not only slow (de Ruiter and van Loon, 2022) but also fragmented.

The short time intervals between pandemic waves, which unfold during the cold months of the year (Fig. 2), and the clusters of flood events at the end of spring and beginning of summer require expedited mitigation efforts and updated multi-risk management plans that account for the particularities of the co-occurrent hazards. This is particularly important because the most prominent adaptation option is the great capacity of first responders to develop creative solutions during crises and to cope with new challenges. This is the only adaptation option that mitigates multiple top-level augmented vulnerabilities that pertain to both hazards: the uncooperative population, flood management not adapted to the COVID-19 context, lack of equipment for first responders including protective gear, and ineffective institutional communication. The umbrella adaptation option covers a large spectrum of mitigation actions thought about and implemented by first responders on the spot to cover for the lack of resources or specific protocols. This means that there are no adaptation options that account for the challenges imposed by the two independent but co-occurrent hazards, highlighting a lack of vision in the current risk management plans applied in Romania.

5.2 Contribution and novelty

Although vulnerability dynamics has gained traction over the last decades, interest in vulnerability dynamics within multi-hazard contexts has particularly surfaced since 2020 and discussions have remained at a theoretical level (de Ruiter and Van Loon, 2022), with no case study to date. Moreover, few studies have investigated the interactions between flood hazards and the COVID-19 pandemic (Simonovic et al., 2021; Patwary and Rodriguez-Morales, 2022; Pramanik et al., 2022; Turay, 2022). This paper addresses a double research gap, aiming to advance our understanding of both vulnerability variations against a multi-hazard background and compound impacts of the two hazards of interest.

The methodological framework proposed to reach this goal carries multiple elements of novelty, as it enhances the impact chain to account for the fluctuations in vulnerability by establishing new element and connection types and taking an in-depth look at the double status of certain augmented vulnerabilities (i.e. those that also act as derived impacts). The enhanced impact chain is a readily available operational tool suitable for replication across various multi-hazard contexts, time frames, spatial scales, and geographic settings. This upgraded version of the chain can extend the list of methods for vulnerability dynamics modelling put together by de Ruiter and Van Loon (2022), also emerging as a solution to the issue raised by Tilloy et al. (2019): “We believe there is a need to not only study case studies inclusive of multi-hazard interrelationships but to generalise to more inclusive frameworks that are applicable to a broad range of hazards and locations.” The dual functionality highlights the capability of the methodological framework to account for both changes in vulnerability and the interconnectivity of multi-hazard impacts.

It should be noted that the present analysis on the augmentation of vulnerability against a multi-hazard background is an initial research work. Prospective avenues for research include the development of a model of systemic vulnerability in a multi-hazard context, which will be tested on multiple impact chains, including the enhanced one discussed in this study, to further validate its effectiveness and applicability.

5.3 Limitations and constraints

Pursuing scientific rigour and transparency, the limitations of the study have to be acknowledged too. The case study aimed for a comprehensive analysis of the multi-hazard of interest, drawing on various data and information sources. However, this is only as comprehensive as it can be given the fact that there are no official sources that detail the impact of flood events. Also, the exact quantification of the impacts is constrained by the lack of official data. Along the same lines, the absence of information on the COVID-19 preventive measures implemented during flood evacuation procedures and inside emergency shelters raises uncertainties that are integrated into the impact chain. Another shortcoming concerns the limited time range, which does not cover the entire pandemic period but only its first 2 years. It should be mentioned that 2022 was a dry year in Romania (Iuga, 2022; Toreti et al., 2022a, b), implying that flood occurrences were scarce. In addition, Albulescu (2023) reports that there were no flood events that required the evacuation of the population in the first 8 months of 2022 (including the flood season in Romania). A fourth limitation regards the tangled configuration of the impact chains, which does not allow for a figure-based visualisation in the paper. Nevertheless, the visualisation available via the Kumu link (https://kumu.io/cosminaalbulescu/ic-augment-reconstructed#impact-chain-on-floods-and-the-covid-19-pandemic-with-augmented-vulnerabilities, last access: 22 August 2024) has the advantage of interactive manipulation of connections and elements, as well as access to the descriptions, source types, references, maps, and images embedded in the impact chains. A comprehensive understanding of the paper is facilitated by engaging with the online platform.

The inclusion of stakeholders in the construction of the multi-hazard impact chain is limited to the feedback provided by first responders who performed on-site emergency interventions during the floods of 2021 (Fekete et al., 2023). Future research directions should focus on a broader involvement of different stakeholders in order to maximise the benefits of co-produced knowledge and refine the details specific to the multi-hazard context from a transdisciplinary perspective. A notable methodological limitation refers to the lack of testing against other case studies and external validation, which we plan to address in the future by applying the methodological framework to other impact chains focusing on different multi-hazard situations. Finally, the paper provides a limited view on the dynamics of vulnerability, relying on only two temporal pictures captured by the initial impact chain and the enhanced version of it. In the future, the development of impact chains within the same multi-hazard context but for multiple years and the tracking of the augmentation of vulnerability across multiple temporal snapshots will yield more nuanced results that can also be validated with narratives from the grey literature. Some of these methodological limitations are inherent to impact-chain-based analyses, as highlighted in the literature review performed by Menk et al. (2022).

6 Conclusions

Since the start of the decade, the co-occurrence of natural hazards amid the COVID-19 pandemic has presented unparalleled challenges that have demanded a new way of approaching multi-risk management and adaptability to both public health crises and the impacts of various natural hazards. This increase in multi-hazard frequency has taught us valuable lessons that we still have to study in order to reduce our vulnerability in the face of future similar multi-hazard events.

We posit that particular attention should be dedicated to understanding the dynamics of vulnerability within a multi-hazard context and that we still have to develop tools for analysis focusing on the fluctuations in vulnerability across hazards, time, and space. In pursuit of this goal, we enhanced the impact chain regarding the multi-hazard of the floods and COVID-19 pandemic that affected Romania in 2020–2021, transforming it from a documentation tool to one that can capture the dynamics of vulnerability. The main enhancements are the introduction of new types of connections between the impacts/adaptation options and vulnerabilities, the introduction of new types of elements (i.e. augmented vulnerabilities, derived impacts), and the ranking of vulnerabilities based on their augmentation. The key findings of the paper can be summarised as follows:

• In a multi-hazard context, vulnerability can be augmented by both impacts and adaptation options in ways that can be captured by an impact chain, but it can also perpetuate over time due to inaction to address it.

• Certain augmented vulnerabilities can also function as impacts (here called “derived impacts”) that sharpen the impact that initiated the augmentation of that vulnerability in the first place.

• In the case study of the floods and the COVID-19 pandemic in Romania (2020–2021), vulnerability is augmented mostly by hazard impacts and, to a lesser extent, by adaptation options. This is explained mainly by the surface-level approach to multi-hazard management, which lacks sufficient integration of adaptation options capable of generating positive or negative effects on vulnerabilities.

• The most augmented vulnerabilities (by both impacts and adaptation options) in the proposed impact chain are uncooperative population, the low-performance medical system, and flood management not adapted to the COVID-19 context.

• The most augmented vulnerabilities by adaptation options alone (i.e. shallow implementation of preventive measures and absence of preparedness at the individual level) show that the implemented mitigation strategies can undermine preparedness for both floods and pandemics.

These results reinforce the idea that old ways will not solve new or reinforced problems and that a proper understanding of all components of multi-risk – and especially of those that can be mitigated (i.e. impacts and vulnerabilities) – is key to improving multi-risk management. The impact chain brings to light the shallow approach of multi-hazard management in Romania, which fails to cover all three DRM phases (i.e. preparedness, response, recovery), to account for the co-occurrence of multiple hazards, or to rise to the challenges faced in the last few years. Such situations drive the need for an improved “multi-hazard approach and inclusive risk-informed decision-making” mentioned in the Sendai Framework for Disaster Risk Reduction 2015–2030 (UNISDR, 2015). Although such goals were set before the COVID-19 pandemic, their achievement is still an ongoing process, the progress of which hinges on our understanding of the dynamics of multi-hazard vulnerability.

Appendix A

Table A1 Vulnerabilities grouped by hazard, type, and scale.

Appendix B

Table B1 Details on the new connection types and derived impacts included in the enhanced impact chain. The asterisk (*) marks cases where the impact in the first column or the vulnerability in the second column relates to both floods and the COVID-19 pandemic. In such cases, the type of augmentation can expressed by both “deepens” and “shifts” connections, and the choice is based on the explanation given in the fourth column.