INTRODUCTION

In the context of increasing frequency and severity of disasters, reducing risk and enhancing resilience are major societal challenges that have been prioritized at the highest levels through global policy compacts including the Sendai Framework, Paris Agreement, and Sustainable Development Goals (UN 2015, UNDRR 2015, UNFCCC 2015). With increasingly unprecedented frequency and severity of extreme weather events, driven by climate change and the inequitable distribution of social and ecological vulnerabilities both within and between communities, it is imperative to continue to expand the understanding of community resilience at local to global scales (World Bank 2021, IPCC 2023). Although resilience operates across multiple scales, understanding resilience at the community level is particularly important because many of the direct and indirect impacts of disasters are experienced at this level, and it is there that much effective action to build resilience can be taken (Keating 2020). Ecological systems cannot anticipate disturbances or disasters, yet communities can conceptualize such events and take action to manage them (Gunderson 2010). Understanding how to enhance resilience is far from trivial and the social-ecological systems (SES) literature, for example, suggests that resilience has to be understood as an emergent property of both human and environment interrelationships (see Faulkner et al. 2018 for a detailed discussion). Although there is now great attention to resilience by development and humanitarian organizations and policy across scales (Clement et al. 2024), a major information gap relates to empirical evidence on what builds community resilience over time (Florin and Linkov 2016, Linkov and Trump 2019).

Measuring community disaster resilience comes with its share of challenges (Schipper and Langston 2015, Asadzadeh et al. 2017, Cai et al. 2018, Jones et al. 2021). The last decade has seen a surge in efforts aimed at measuring resilience to a variety of hazards, efforts that have resulted in the development of many resilience measurement frameworks, tools, scorecards, indices, etc.; many of these, however, lack a theoretical framework and empirical validation, application, and up-take has been patchy (Cutter 2021, Tan 2021). The lack of definitional or methodological consensus (see, for example, Hahn and Nykvist 2017 in the context of SES) and absence of standardized, empirically validated approaches to resilience measurement undermine confidence by policy and decision makers who seek to translate outcomes into policy advice and implementation (Bakkensen et al. 2017, Cutter 2021, Jones et al. 2021). We suggest resilience thinking must be an integrative approach with regard to sustainability challenges and, as a consequence, this needs to be thought through across scales and dependencies, including corresponding complex mechanisms that can eventually cascade through different sub-systems (see Folke 2016 in the context of SES, Hochrainer-Stigler et al. 2020 in the context of systemic risks).

To bridge these information and practice gaps, the Zurich Flood Resilience Alliance (the Alliance), formed in 2013, developed the Flood Resilience Measurement for Communities (FRMC) approach (Keating et al. 2017). The first phase of the Alliance (2013–2018) saw the application, analysis, and validation of the first version of the FRMC framework in 118 communities between 2015 and 2017 (see Laurien et al. 2020 for a summary, Hochrainer-Stigler et al. 2021 for analytical insights). In the second phase of the Alliance (2018–2024), the framework was refined into the FRMC Next Gen tool and is currently in use in more than 292 communities (see Appendix 1 for more details). In the current and third phase, the FRMC has been further evolved into the CRMC (C for Climate), an approach that can measure community resilience to multiple hazards including flood, heatwave, and wildfire.

We present the first large-scale analysis of the data generated using the FRMC Next Gen version, collected in 292 communities from 20 developing countries between 2018 and 2022. Our focus is on quantitative findings of this global empirical analysis, building on the rich set of on-the-ground resilience indicators. Our analysis includes an assessment of the validity and reliability of the FRMC Next Gen framework. In addition, we query whether and how to organize the wealth of information of community contexts into a discrete set of clusters, for which we further develop the taxonomy developed in the first phase in Laurien et al. (2020). The theoretical underpinnings of the framework will be briefly discussed as well as advantages and limitations compared to other approaches identified. Special emphasis is put on different types of communities that can be empirically determined. This is especially important as analysis in phase 1 found that the dynamics of resilience are essentially different for different community types (Hochrainer-Stigler et al. 2021). With the larger FRMC Next Gen dataset (compared to phase 1) we are able to provide more nuanced, and empirically tested, perspective with regard to this question.

CHALLENGES AND METHODOLOGICAL APPROACH

A significant challenge of resilience measurement lies in taking a complex, multi-dimensional concept (Folke 2016) and operationalizing it in a measurable way (Alexander 2013). Measuring community resilience involves trying to anticipate, in the absence of a disaster event, which set of community characteristics, and ultimately indicators, will best predict resilient post-disaster outcomes. Holistic frameworks that presume dynamic interactive processes seek to determine which of the multitude of community dimensions, across many attributes and sub-systems, provide the most important resilience proxies (Cutter 2021). Achieving a balance between objective indicators and subjective assessments is essential to provide a comprehensive understanding of community resilience, however, integrating these dimensions into a unified resilience measurement framework is a complex endeavor (Keating et al. 2017).

The literature identifies three further challenges or limitations in relation to disaster resilience measurement. First, little effort has been made to integrate ecological components into community resilience measurement, and spatial scale and cross-scale dynamics have mostly not been considered in community resilience analysis (Chuang et al. 2018). Second, some of the most widely used measures of community resilience, such as the Social Vulnerability Index (SoVI) and Baseline Resilience Indicators for Communities (BRIC) model, rely on census or national-scale survey variables, whose availability differs from country to country, making global analysis difficult (Camacho et al. 2023, Cutter 2024). Finally, the factors identified solely based on multivariate analysis and their aggregation to estimate a single composite index may not be conceptually robust or consistent with the understanding of hazard-specific resilience and its drivers (Camacho et al. 2023).

The literature discusses many definitions of resilience (see for example Béné et al. 2014, Folke 2016, Faulkner et al. 2018) while the concept of resilience in general, and disaster resilience specifically, has evolved from a focus on ecological systems to a holistic perspective with multiple framings according to different disciplines including economics, risk science, ecological system theory, psychology, as well as engineering (Keating et al. 2014). As with resilience in general, the literature on disaster resilience, our focus, and its many definitions and conceptualizations have some key characteristics in common but, as indicated, have challenges with regard to operationalization and practical application.

As a step forward, Keating et al. (2017) outlined a conceptional framework of disaster resilience building on the system interactions between disaster risk, disaster risk management (DRM), and sustainable development (SD). It builds on a development-centric disaster resilience perspective. This approach as well as the corresponding definition of resilience is also used here: resilience as the “ability of a system, community, or society to pursue its social, ecological, and economic development and growth objectives, while managing its disaster risk over time, in a mutually reinforcing way” (Keating et al 2017:80). Based on this development-centric definition of disaster resilience, partners of the Zurich Flood Resilience Alliance translated the theoretical framework of disaster resilience into a practical framework for measuring disaster resilience. In the following, we proceed to discussing empirical aspects of the framework and the measurement tool, including key operationalization aspects (for further detailed discussions, we refer to Keating et al. 2017 as well as Laurien et al. 2020.)

The FRMC Next Gen Framework and Tool

The Next Generation FRMC approach, like the first iteration of the framework, is based on the so-called 5C framework: it includes 44 indicators called “sources of resilience” (or “sources” for short) that are distributed across and represent critical aspects of five complementary “capitals” (5C; see Appendix 2). It is built on the five capitals framing of the Sustainable Livelihoods Framework (DFID 1999). This framing emphasizes that community flood resilience is a multi-dimensional concept comprising elements across physical, social, human, financial, and natural aspects that interact over time to inform community well-being and disaster outcomes. The sources are selected for the roles they play in supporting community well-being, helping people on their development path and/or providing capacity to prepare for, withstand, respond to, and recover from floods.

Users collect and analyze data using the FRMC tool, a practical hybrid software application comprising an online web-based platform for setting up and analyzing the process, and a smartphone- or tablet-based app that can be used offline in the field for data collection. After data is collected on the app, it is uploaded to the web application. A grading team composed of the FRMC implementing team, community members, and sometimes other stakeholders such as local government representatives compare collected data to pre-determined grading rubrics to grade each of the 44 sources of resilience on an A–D scale (A being best practice, D being poor). For aggregation, A–D grades correspond to number scores as follows: D = 0, C = 33, B = 66, A = 100. The number scores of corresponding sources of resilience for each capital are then averaged to get an aggregate score for each capital; for example, if all sources of resilience in a capital group were graded “A,” the community would score 100 for that capital group. Graded results can be explored according to different “lenses” including the 5Cs (Fig. 1).

Importantly, the FRMC is a standardized approach to resilience measurement (e.g., not dependent on the location it is applied to) that can therefore be used across the globe (Fig. 2), a feature that is still often lacking in the resilience space. Consequently, it makes it possible to explore differences in resilience profiles across communities, track progress over time, and to learn and improve practices. The FRMC therefore provides a consistent benchmark against which to quantify flood resilience at the community level. Furthermore, it employs several data collection methods (household surveys, focus groups, key informant interviews, and secondary source data) and allows for the collection of data on community perceptions, knowledge, and capacities (Fig. 1, lower left). Use of data collection and software technologies are supported by online or in-person user-training and guidance resources, which help ensure systematic and consistent data collection and framework use. The online platform includes data analysis features that facilitate exploration of interconnections between results and preparation of reports that can be shared with community stakeholders (Fig. 1, right hand side; detailed information of the data acquisition process and approach used can be found in Appendix 3.)

Study Locations

The analysis presented here is based on FRMC Next Gen baselines conducted between 2018 and 2022. These assessments were done in 292 communities in 20 developing countries worldwide, covering a population of almost a million (approximately 966,600). From each country, at least four and up to 53 communities are represented. Locations of the communities and countries are presented in Figure 2.

FRMC users include NGOs, humanitarian organizations, and researchers. Users generally consider a variety of criteria when determining precise study locations and which communities to apply the FRMC in. These include history of previous flooding and communities being at high risk, communities’ need for external support, their location in a larger river basin (where applicable, this criterion was not considered with communities that suffer from coastal flooding, for example), and a community’s representativeness for their region and willingness to take part in the project. One predominant criterion is the presence and perception of high flood risk by the organization, local authorities, and the community itself. This criterion encompasses terms such as “risk,” “exposure,” and “vulnerability.” Organizations often seek to work in different parts of a watershed in order to advance integrated watershed management, addressing diverse needs, challenges, and opportunities across communities within the same watershed. The interest of local authorities and alignment with government initiatives are also factors in community selection in most locations. Some organizations aimed to collaborate with the government and fill gaps where their efforts fell short, fostering cooperation and complementarity. Several organizations emphasize co-benefits and addressing vulnerabilities, particularly as related to climate change, as an additional criterion for community selection (for further information about the total dataset used we refer to Appendix 4.)

Empirical analysis strategy

Because this is the first time that results from analysis of the FRMC Next Gen dataset are presented, we start with presenting some overall results on capitals and communities, including descriptive and exploratory analysis. We then discuss the internal consistency and reliability tests, possible dimensionality reduction, and sub-group analysis. Based on this we focus in on the main topic of this article, namely the cluster analysis and interpretation of taxonomy characteristics.

The methods employed are discussed in detail in Appendix 5, and here we provide a short summary only. To assess internal consistency and reliability we used the standard tests including Cronbach’s alpha, and for dimensionality reduction and sub-group analysis we focused on Principal Components Analysis. To identify possible clusters and taxonomies we used a variety of cluster analysis approaches, especially focusing on schemes that create a strong separation in similarities between clusters and strong similarities within clusters. Clusters are based on community resilience levels across the five capitals. Following the clustering process, clusters were defined by the characteristics of the communities within each cluster, using multinomial regression analysis and random forest models to test the relationship between clusters and with community characteristics. To achieve this, the data was split into 70% model training and 30% test datasets using stratified random sampling to test the model’s validity and performance. Confusion matrix and statistics (sensitivity, specificity, and balanced accuracy) were then used to test the accuracy of the cluster prediction by the models. Finally, we used the most significant model and socioeconomic indicators to explain cluster archetypes. In this way, we identified taxonomies and can interpret the taxonomy of community flood resilience as explained in the next section.

As Figure 1 (bottom) indicates, our focus in this paper is on the baseline analysis. There are, however, three additional pieces of data that make up the FRMC measurement approach. These are the post-event study, the intervention data record, and the endline study (a repeat of the baseline). The post-event study evaluates damages of, and community system performance in the event of, any flood disasters that occur in a community following the baseline study. The intervention data record documents the interventions done in the communities following the baseline. Finally, the endline study is a repeat of the baseline study, conducted 2–3 years after the baseline. These four parts are designed to provide a cohesive, empirical global analysis of community flood resilience over time. A further motivation for the present in-depth analysis of the baseline data is the need to establish a strong empirical understanding of the baseline in order to support future analysis of the undoubtedly complex dynamics (Hochrainer-Stigler et al. 2021) between baseline resilience, flood impacts, the effects of interventions, and finally endline resilience.

RESULTS

Community flood risk

We start with some overall insights into the communities in our sample. Most of the communities (73%) are rural, followed by urban (16%), and peri-urban (10%). River flooding is the most common flood type, occurring in 45% of the communities. Flash floods are the next most prevalent, occurring in 36% of the communities. Although surface flooding and coastal flooding are the most common flood type in 4 and 15% of the communities, respectively, for some countries these are a key flood issue, for example, surface flooding in Vietnam, Cambodia, Nicaragua, and Albania. Regarding frequency of flooding, just under half of the communities experience flooding more than once per year on average, with around 65%, on average, of houses affected. One third of communities experience a flood about once a year, and on average 60% of houses are flooded. Regarding the severity of previous floods, 76% of households across the communities reported that in the case of the worst flood they can remember, more than three quarters of the houses/buildings in the community were flooded.

Validity, reliability, and sub-group analysis

Here we provide a summary of the validity, reliability, and sub-group analysis, with detailed results available in Appendix 6. Face validity looks at whether the FRMC aligns with practitioners’ and communities’ understanding about the factors that contribute to and build community flood resilience. Practitioners widely agreed that the framework assesses community flood resilience as they perceive it, confirming the importance of all 44 resilience sources and finding no major gaps. For reliability, we evaluated the FRMC’s ability to consistently measure resilience across different contexts. The results show high reliability (C-alpha ≥ 0.7) for all capitals except natural capital, which slightly missed the threshold (0.69). Overall, the FRMC reliably measures resilience across all capitals and is valid for aggregation.

The PCA identified two to three sub-groups within each capital. Financial capital were grouped into two components: public and private financial capacity. Human capital had three components: first aid and WASH (water, sanitation, and hygiene) knowledge, flood exposure and evacuation awareness, and environmental management and governance. Natural capital split into physical status and services of resources and management efforts. Physical capital had three components: basic supplies during floods, utilities infrastructure, and early warning and emergency response infrastructure. Last, social capital showed three components: community structure, external flood response services, and DRM policies at national and community levels.

Community cluster analysis

As indicated in section 2, we analyzed the community resilience results using various cluster agglomeration schemes as well as similarity measures; here we report only our main findings. Using the hierarchical clustering method, we identified five distinct community clusters based on their flood resilience capital scores (see Fig. 3 for capital scores by cluster). The first important thing to note is that not only do the absolute levels of scores differ across clusters, but also the distributions of the different capitals are quite distinct. Moreover, the clusters highlight the differences in flood resilience between rural and urban areas. Correlations analysis between the five capitals can be found in Appendix 7. Here we focus on the details between our findings and the underlying resilience sources.

We detect cluster 1, which includes 99 communities and is considered the cluster of communities with the lowest resilience. Ninety-seven percent of communities in this cluster are rural communities primarily from Bangladesh, Kenya, Malawi, South Sudan, and Zimbabwe. Both private and public financial capacity components of these communities are very low. The relatively higher average human capital score is mainly due to a slightly better level of awareness of flood exposure, future risk, WASH, environmental management, and governance. However, other human capital aspects, namely knowledge of evacuation and safety, asset protection, first aid, and education commitment during floods, are very low. The natural capital dimension is also low because of degraded local ecosystems (both wild and managed), and little to no sustainable or regenerative management. Very low physical capital is due to poor utilities infrastructure, early warning systems, emergency response infrastructure, and low levels of basic supplies during an emergency. Social capital is measured to be low in these communities because of missing or inadequate community governance structures and representation, seen in low participation, low inclusiveness, lack of local leadership, and low mutual assistance. These communities also show low levels of external flood response and recovery services, and lack of national and community level DRM and integrated flood management policy and plans.

Cluster 2 contains 56 communities that exhibit slightly higher financial, human, and physical capital scores than natural and social capital. In this cluster, 70% are urban and peri-urban communities primarily from Senegal, the Philippines, Mexico, Jordan, and Bolivia. The public financial capacity of this cluster is moderate, but the private financial capacity is low. Higher awareness of flood exposure, future risk, WASH, environment management, governance, and asset protection improves human capital of this cluster as compared to cluster 1. However, knowledge of evacuation and safety, first aid, and education commitment during floods remain low. Sources of physical capital, mainly basic supplies during an emergency, utilities infrastructure, early warning system, and emergency response infrastructure are slightly higher than in cluster 1. Natural capital is low because of degraded natural environments and ecosystem services, despite some efforts in regard to their management. Social capital is also relatively low because of the status of community governance structures, and lack of DRM and integrated flood management policy and plans. The level of external flood response and recovery services is relatively better in cluster 2 communities, as compared to cluster 1.

Cluster 3 has 37 communities with capital score profiles that are somewhat the inverse of cluster 2. Seventy-six percent of these communities are rural communities primarily from Senegal, the Philippines, Mexico, Jordan, and Bolivia. Here, the scores for human, natural, and social capitals are relatively high compared to the rest of the sample, but financial and physical capitals are lower. Public financial capacity is moderate, while private financial capacity is very low. Physical capital is very low because of absent or inadequate utilities infrastructure, early warning systems, emergency response infrastructure, and low levels of basic supplies during an emergency. The average natural capital score in this cluster is higher because of better natural resource management and moderate levels of environmental health and ecosystem services provision. Social capital is higher because of better community governance structures, external flood response, and availability of recovery services. However, DRM and integrated flood management policy and plans remain below good standard. Sources of human capital are similar to cluster 2, with a high to moderate level of awareness of flood exposure, future risk, WASH, environmental management, and governance. However, there is limited knowledge of evacuation and safety, asset protection, first aid, and education commitment during the flood.

Cluster 4 has 92 communities and, along with cluster 5, shows higher scores compared to clusters 1, 2, and 3. Cluster 4 primarily includes communities from Vietnam, Nepal, India, Bolivia, Albania, and Montenegro. Average capital scores are generally higher than the previous three clusters, particularly human, natural, and social capital scores. These communities show high levels of awareness of evacuation and safety, flood exposure, future risk, asset protection, WASH, environmental management, and governance. Similarly, the level of first aid knowledge and education commitment during floods is moderate. Cluster 4 exhibits the highest average natural capital of all the clusters because of the better, or less degraded, state of natural resources and ecosystem services provision, together with stronger conservation and restoration. All aspects of community governance are moderate to high, with the exception of inter-community coordination, which remains low. Similarly, external flood response and recovery services are moderate to high. At both the national and community levels, DRM plans are at a moderate level, but integrated flood management policy is low. Public financial capacity is moderate, while private financial capacity is relatively lower. Physical capital sources are also at a moderate level in these communities.

Finally, cluster 5 is the smallest cluster, with only 8 communities from the Philippines and Vietnam, and exhibits high average capital scores as compared to the other clusters. Financial and physical capitals are the highest of all the clusters. This high financial capital is due to high public and moderate private financial capacity. Physical capital is strongest because of better emergency response infrastructure, utilities infrastructure, early warning systems, and basic supplies during emergencies. Human capital is high because of high awareness and knowledge levels. However, first aid knowledge and education commitment level during floods is relatively low. All aspects of community governance and level of external flood response and recovery services are moderate to high in this cluster, driving higher social capital. DRM policies and planning are at a moderate level, but integrated flood management policy is poor. The natural capital score is the lowest among all capitals for cluster 5 because of the low to moderate level of physical condition of natural resources and absent or limited efforts in their management.

Notably, clusters 1, 3, and 4 predominantly consist of rural communities, while clusters 2 and 5 are mainly composed of urban and peri-urban communities. However, around 24% and 20% of urban communities are also found in clusters 3 and 4, respectively, and around 30% of rural communities appear in cluster 2. This is primarily due to their similar capital score profiles, but it should be noted that there are exceptions within these clusters as communities can differ in other aspects beyond the capital scores. This qualification underlies the point that the clusters are not designed to be predictive or prescriptive about any individual community; they are an analytical tool to help organize and understand the volume of information about community flood resilience contained in this global dataset. Given this detailed information about the communities’ resilience sources for each cluster, next we explore the capital score-derived clusters in relation to the characteristics of the communities within them.

Taxonomy of community resilience

The description of the clusters above indicates that there are qualitative differences between the clusters. We now further analyze these by exploring the predominant socioeconomic characteristics in each cluster, to determine whether the resilience profiles have some common patterns with specific community characteristics. To do this, we used the identified clusters to run a multinomial regression model and a random forest model with socioeconomic characteristics as independent variables, separated into training and testing sets. We present the detailed results in Appendix 8.

We want to note that we also analyzed the geo-spatial distribution of the clusters. For this we tested both geographic distances and country locations as explanatory variables, for the clusters themselves and the distance matrix used to make them. Overall, we found no significant pattern, although (when taken as the sole explanatory variable) certain countries are more likely to have communities from specific clusters. Nevertheless, this significance is not present when socioeconomic variables like poverty or urbanization (community type: rural, peri-urban, or urban) are introduced. Further analysis using Voronoi cells showed no significance as well. In some cases, communities that are geographically close do indeed fall in the same cluster, however, we just as often see disparities between geographically close communities. Hence, socioeconomic community characteristics proved to be a much better predictor for clusters compared to geographical positioning. Moving forward, below we summarize the taxonomy of our identified community flood resilience clusters and their key characteristics, which are set out in Table 1.

Cluster 1: Rural communities with high risk and vulnerability, and low capacity

This type of community is mostly rural, with very low financial, natural, physical, and social capital, although with a slightly higher average human capital score due to increased awareness of flood risk and environmental management. Level of poverty is very high and women’s education is low. In the majority of communities, more than 50% of the households are living below the national poverty line; the percentage of women who have completed secondary education is below 25% in all the communities. Around 63% of the communities in this cluster report never having influence on decisions that are made at higher levels, while 32% have influence some of the time, and only 5% most of the time. These are also high flood-risk communities: around 30% of these communities typically experience flooding more than once per year, and 55% experience flooding about once per year. In these flooding events, more than 20% of houses are usually flooded in 82% of the communities.

Cluster 2: Urban communities with poor natural and social environments

This community type primarily represents urban communities that possess moderate physical, financial, and human capital, but lack adequate natural and social capital. Compared to cluster 1 communities, this community type is characterized by slightly lower poverty rates and higher women’s educational attainment. In most communities, 10–50% of households live below the national poverty line, and more than 25% of women have completed secondary education. Around 67% of the communities report that they sometimes have influence on decisions that are made at higher levels, and 7% report that they do most of the time. Still, 25% of the communities have no influence in higher-level decisions. Flood risk is high for this community type, with around 53% of the communities experiencing flooding more than once per year and 24% experiencing flooding about once per year. In these flooding events, more than 20% of houses are usually flooded in 38% of the communities.

Cluster 3: Rural communities with high capacity but low income and poor physical infrastructure

Similar to cluster 1, this community type primarily consists of rural communities with relatively better human and natural capital, but low physical, financial, and social capital. In the majority of these communities, more than 60% of households are living below the national poverty line. However, women’s educational attainment and influence on decisions are slightly higher than in cluster 1. Between 10 and 50% of women have completed secondary education in most communities. In regard to decisions made at higher levels, 62% of the communities have influence sometimes, and 11% have influence most of the time; around 27% have no influence. Flood risk is relatively moderate in these communities, with around 41% of communities experiencing flooding more than once per year, and 24% experiencing flooding about once per year. This flooding results in more than 20% of houses usually being flooded in 68% of the communities.

Cluster 4: Less vulnerable rural communities

This community type primarily consists of rural communities that exhibit better capital scores compared to cluster 1 and 3 rural community types. Poverty is less than for cluster 3 but still prevalent, with 20–60% of households living below the national poverty line in most communities. Between 10 and 50% of women in the majority of the communities have completed secondary education. Seventy-six percent of the communities have influence over decisions made at higher levels sometimes, 8% influence decisions most of the time, and only 16% have no influence. Flood risk is moderate in these communities, with 45% of communities experiencing flooding more than once per year and 23% experiencing it about once per year. During floods, more than 20% of houses are usually flooded in 60% of the communities.

Cluster 5: Less vulnerable urban communities

This cluster consists of urban communities with lower poverty rates and higher women’s educational attainment. People living below the national poverty line are below 30% in all communities. Similarly, in the majority of communities more than 50% of women have completed secondary education. Around 25% of the communities have influence over decisions made at higher levels sometimes and 13% have influence most of the time. However, 62% of the communities report never having influence on these decisions. Flood risk is quite high in these communities: around 88% of the communities experience flooding more than once per year and the remaining 12% experience it once yearly. During those flood events, only 5–10% of the houses are flooded in around 75% of the communities, and in only 25% of the communities, more than 20% of the houses are usually flooded.

DISCUSSION

The literature has shown that disaster resilience is strongly case and context specific. While accepting these findings, our analysis also empirically shows that various clusters of community resilience can be distinguished using statistical analysis. We identify five types of community clusters with different resilience profiles. This taxonomy supports a nuanced understanding of different community types classified according to settlement type/density, poverty, education, socio-political influence, and flood risk exposure. This result supports the conceptualization of resilience as a “multifunctional” concept, with inherent complexity and dynamics (Wilson 2008).

The clusters defined here are similar to the four community types identified by Laurien et al. (2020) based on the FRMC phase I data. Compared to Laurien et al. (2020), in this study we were able to further disaggregate communities that have similar aggregated levels of resilience but differ in their resilience scores across capitals (their resilience profiles). Specifically, we found that clusters 2 and 3 exhibit similar resilience scores but cluster 2 is characterized by higher physical capital and lower natural and social capitals, while cluster 3 shows lower physical capital and higher natural and social capitals. Our taxonomy indicates the existence of some overall “resilience structures” across diverse sets of dimensions and supports the argument that resilience thinking has to be understood as an integrative approach for dealing with sustainability challenges (Folke 2016). This includes integration between system dynamics and scale (see the work on panarchy by Gunderson and Holling 2002) as well as methodological integration across scientific disciplines, two targets for integration with regard to resilience that are in need of attention.

In this regard, the identified taxonomy of community resilience and its foundation in indicators of community characteristics can provide insights into the multiple functionalities of communities and their development trajectories. Because different clusters have different resilience profiles, or distributions of capitals, we suggest that resilience-enhancing efforts ought to be diverse when enhancing across community types (Hochrainer-Stigler et al. 2021). This insight supports targeting of interventions that meet increasingly urgent needs for bolstering capitals depending on the community type. Caution needs to be exerted as goals and targets, for example with regard to adaptive capacity, have normative connotations, associated ontologies and value systems, which do not always overlap with perspectives held by stakeholders, risk bearers, or policy makers.

Our analysis shows that although resilience measurement is a time-consuming and resource intensive task, socioeconomic characteristics can be more easily gathered, therefore making it possible to provide indications of which cluster a community is likely to belong to. As a consequence, our analysis has the potential to cultivate a foundation for shared understanding of flood resilience, thus providing an analytical platform for relationship building with and between community members, local and national governments, development practice, and international policy. Bringing this insight together with FRMC user interviews, we argue that the FRMC approach thus can foster systems thinking, which is fundamental to resilience and which supports informed decision making (albeit with important caveats). Such a systems-thinking approach could also assist in developing a deeper understanding of what it means to engage in flood resilience-enhancing processes and programming by exploring gaps and strengths across the range of sources of resilience and recognizing the wide range of sectors involved in resilience. Such systems-based analysis can further assist in the identification of both co-benefits and maladaptive consequences of various activities on the individual as well as at system level.

CONCLUSION

We presented validation and a resilience taxonomy analysis using the FRMC Next Gen baseline data set, which includes 292 communities from 20 countries across the world and provided insights into the measurement and characterization of community flood resilience. Using a PCA and Cronbach Alpha analysis for validation, we concluded that the FRMC Next Gen approach of measuring 44 sources of community flood resilience and aggregating those to financial, human, natural, physical, and social capitals leads to valid and reliable results. Each capital can be disaggregated into two to three components representing different aspects of the capitals. We presented an empirically based taxonomy related to flood risk and community socioeconomic characteristics, where, using multinomial and random forest methods, we identified five distinct community clusters based on their capital score profiles. This taxonomy provides a way to group communities by similarities and differences between absolute and distributional resilience levels and socioeconomic community characteristics. For programs lacking a foundational quantitative resilience baseline analysis our results may be a useful additional metric for informing and monitoring resilience programming.

Finally, we emphasize that our approach focuses on broad-based characteristics and does not circumvent the need for determining appropriate resilience-enhancing decisions at the community level with deep community and stakeholder engagement, which will always be case specific. In addition, communities are always embedded within larger systems, and there are different advantages as well as limitations at each scale for enhancing resilience. Yet, as we suggest, integrating generic insights associated with different community types with case-specific contextualization may help to support community-level engagement, programming, and intervention implementation.

LITERATURE CITED

Alexander, D. E. 2013. Resilience and disaster risk reduction: an etymological journey. Natural Hazards and Earth System Sciences 13:2707-2716. https://doi.org/10.5194/nhess-13-2707-2013

Asadzadeh, A., T. Kötter, P. Salehi, and J. Birkmann. 2017. Operationalizing a concept: the systematic review of composite indicator building for measuring community disaster resilience. International Journal of Disaster Risk Reduction 25:147-162. https://doi.org/10.1016/j.ijdrr.2017.09.015

Bakkensen, L. A., C. Fox‐Lent, L. K. Read, and I. Linkov. 2017. Validating resilience and vulnerability indices in the context of natural disasters. Risk Analysis 37:982-1004. https://doi.org/10.1111/risa.12677

Béné, C., A. Newsham, M. Davies, M. Ulrichs, and R. Godfrey‐Wood. 2014. Resilience, poverty and development. Journal of International Development 26(5):598-623. https://doi.org/10.1002/jid.2992

Cai, H., N. S. Lam, Y. Qiang, L. Zou, R. M. Correll, and V. Mihunov. 2018. A synthesis of disaster resilience measurement methods and indices. International Journal of Disaster Risk Reduction 31:844-855. https://doi.org/10.1016/j.ijdrr.2018.07.015

Camacho, C., P. Bower, R. T. Webb, and L. Munford. 2023. Measurement of community resilience using the Baseline Resilience Indicator for Communities (BRIC) framework: a systematic review. International Journal of Disaster Risk Reduction 95:103870. https://doi.org/10.1016/j.ijdrr.2023.103870

Chuang, W. C., A. Garmestani, T. N. Eason, T. L. Spanbauer, H. B. Fried-Petersen, C. P. Roberts, S. M. Sundstrom, J. L. Burnett, D. G. Angeler, B. C. Chaffin, L. Gunderson, D. Twidwell, and C. R. Allen. 2018. Enhancing quantitative approaches for assessing community resilience. Journal of Environmental Management 213:353-362. https://doi.org/10.1016/j.jenvman.2018.01.083

Clement, S., J. Jozaei, M. Mitchell, C. R. Allen, and A. S. Garmestani. 2024. How resilience is framed matters for governance of coastal social‐ecological systems. Environmental Policy and Governance 34(1):65-76. https://doi.org/10.1002/eet.2056

Cutter, S. L. 2021. Urban risks and resilience. Pages 197-211 in W. Shi, M. F. Goodchild, M. Batty, M.-P. Kwan, and A. Zhang, editors. Urban informatics. Springer Singapore, Singapore. https://doi.org/10.1007/978-981-15-8983-6_13

Cutter, S. L. 2024. The origin and diffusion of the social vulnerability index (SoVI). International Journal of Disaster Risk Reduction 109:104576. https://doi.org/10.1016/j.ijdrr.2024.104576

Department for International Development (DFID). 1999. Sustainable livelihoods guidance sheets. DFID, London, UK.

Faulkner, L., K. Brown, and T. Quinn. 2018. Analyzing community resilience as an emergent property of dynamic social-ecological systems. Ecology and Society 23(1):24. https://doi.org/10.5751/ES-09784-230124

Florin, M. V., and I. Linkov, editors. 2016. IRGC resource guide on resilience. EPFL International Risk Governance Center (IRGC), Lausanne, Switzerland.

Folke, C. 2016. Resilience (republished). Ecology and Society 21(4):44. https://doi.org/10.5751/ES-09088-210444

Gunderson, L. 2010. Ecological and human community resilience in response to natural disasters. Ecology and Society 15(2):18. https://doi.org/10.5751/ES-03381-150218

Gunderson, L. H., and C. S. Holling. 2002. Panarchy: understanding transformations in human and natural systems. Island, Washington, D.C., USA.

Hahn, T., and B. Nykvist. 2017. Are adaptations self-organized, autonomous, and harmonious? Assessing the social-ecological resilience literature. Ecology and Society 22(1):12. https://doi.org/10.5751/ES-09026-220112

Hochrainer-Stigler, S., C. Colon, G. Boza, S. Poledna, E. Rovenskaya, and U. Dieckmann. 2020. Enhancing resilience of systems to individual and systemic risk: steps toward an integrative framework. International Journal of Disaster Risk Reduction 51:101868. https://doi.org/10.1016/j.ijdrr.2020.101868

Hochrainer-Stigler, S., S. Velev, F. Laurien, K. Campbell, J. Czajkowski, A. Keating, and R. Mechler. 2021. Differences in the dynamics of community disaster resilience across the globe. Scientific Reports 11:17625. https://doi.org/10.1038/s41598-021-96763-0

Intergovernmental Panel on Climate Change (IPCC). 2023. Climate change 2023: synthesis report. Summary for policymakers. Contribution of Working Groups I, II and III to the Sixth Assessment Report of the Intergovernmental Panel on Climate Change. Core Writing Team, H. Lee and J. Romero, editors. IPCC, Geneva, Switzerland. https://www.ipcc.ch/report/ar6/syr/

Jones, L., A., M. A. Constas, N. Matthews, and S. Verkaart. 2021. Advancing resilience measurement. Nature Sustainability 4:288-289. https://doi.org/10.1038/s41893-020-00642-x

Keating, A. 2020. Measuring and building community disaster resilience: essential for achieving Sendai. Pages 169-190 in M. Yokomatsu and S. Hochrainer-Stigler, editors. Disaster risk reduction and resilience. Springer Singapore, Singapore. https://doi.org/10.1007/978-981-15-4320-3_9

Keating, A., K. Campbell, M. Szoenyi, C. McQuistan, D. Nash, and M. Burer. 2017. Development and testing of a community flood resilience measurement tool. Natural Hazards and Earth System Sciences 17:77-101. https://doi.org/10.5194/nhess-17-77-2017

Keating, A., R. Mechler, J. Mochizuki, H. Kunreuther, J. Bayer, S. Hanger, I. McCallum, L. See, K. Williges, S. Hochrainer-Stigler, and C. Egan. 2014. Operationalizing resilience against natural disaster risk: opportunities, barriers, and a way forward. Zurich Flood Resilience Alliance, Zurich, Switzerland.

Laurien, F., S. Hochrainer-Stigler, A. Keating, K. Campbell, R. Mechler, and J. Czajkowski. 2020. A typology of community flood resilience. Regional Environmental Change 20:24. https://doi.org/10.1007/s10113-020-01593-x

Linkov, I., and B. D. Trump. 2019. The science and practice of resilience. Springer International, Cham, Switzerland. https://doi.org/10.1007/978-3-030-04565-4

Schipper, E. L. F., and L. Langston. 2015. A comparative overview of resilience measurement frameworks: analyzing indicators and approaches. Overseas Development Institute, London, UK.

Tan, S. B. 2021. Measuring community resilience: a critical analysis of a policy-oriented indicator tool. Environmental and Sustainability Indicators 12:100142. https://doi.org/10.1016/j.indic.2021.100142

United Nations (UN). 2015. The Sustainable Development Goals. UN, New York, New York, USA. https://sdgs.un.org/

United Nations Framework Convention on Climate Change (UNFCCC). 2015. Adoption of the Paris Agreement. Report No. FCCC/CP/2015/L.9/Rev.1. UNFCCC, New York, New York, USA. https://unfccc.int/documents/9064

United Nations Office for Disaster Risk Reduction (UNDRR). 2015. Sendai Framework for Disaster Risk Reduction 2015-2030. UNDRR, Geneva, Switzerland. https://www.undrr.org/publication/sendai-framework-disaster-risk-reduction-2015-2030

Wilson, G. A. 2008. From ‘weak’ to ‘strong’ multifunctionality: conceptualising farm-level multifunctional transitional pathways. Journal of Rural Studies 24:67-383. https://doi.org/10.1016/j.jrurstud.2007.12.010

World Bank. 2021. Financial risk and opportunities to build resilience in Europe. World Bank, Washington, D.C., USA. https://hdl.handle.net/10986/35685