A global comparison of community-based responses to natural hazards

Community-based disaster preparedness is an important component of disaster management. Knowledge of interventions that communities utilise in response to hazards is important to develop local-level capacity and increase community resilience. This paper systematically examines empirical information about local level responses to hazards based on peer reviewed, published case studies. We developed a data set based on 188 articles providing information from 318 communities from all regions of the world. We classified response examples to address four key questions: (i) What kinds of responses are used by communities all over the world? (ii) Do communities in different parts of the world use different kinds of responses? (iii) Are communities using hazard-specific responses? (iv) Are communities using a multi-hazard approach? We found that within an extensive literature on hazards, there is relatively little empirical information about community-based responses to hazards. Across the world, responses aiming at securing basic human needs are the most frequently reported kinds of responses. Although the notion of community-based disaster preparedness is gaining importance, very few examples of responses that draw on the social fabric of communities are reported. Specific regions of the world are lacking in their use of certain hazard responses classes. Although an all-hazards approach for disaster preparedness is increasingly recommended, there is a lack of multi-hazard response approaches on the local level.


Introduction
Natural disasters such as floods, storms and earthquakes put many people at risk, especially in coastal regions. Between 1994 and 2013, on average 218 million people were affected annually through losing their homes or livelihoods due to natural disasters, and 1.35 million number of people died over this period (Creed, 2015).
The intensity and frequency of hazardous events is increasing with climate change leading to growing numbers of annual disasters all over the world (IPCC 2014). The growing body of literature on disaster risk reduction, community resilience and adaptive capacity attests to the increased focus on finding 1 5 10 15 20 25 ways through which people can prepare for this, to reduce the impact of hazards and to increase the ability of local communities to cope with the consequences of hazardous events.
The systematic global approach to disaster risk reduction (DRR), which was initiated through the Hyogo framework (2005)(2006)(2007)(2008)(2009)(2010)(2011)(2012)(2013)(2014)(2015), has triggered much progress in disaster management in many parts of the world. However, the ongoing need for significant improvement of disaster mitigation measures, particularly in less developed countries, underpins the targets of the subsequently adopted Sendai framework for disaster risk reduction (2015)(2016)(2017)(2018)(2019)(2020)(2021)(2022)(2023)(2024)(2025)(2026)(2027)(2028)(2029)(2030). The 20 year review   The latter report also posits that more progress is needed in understanding how and why people, especially at the local level, are affected by disasters -so as to enable DRR strategies that are based on evidence rather than assumptions (Creed, 2015). The need for this is reinforced by a review by Gal et al. (2015), which reveals that most disaster risk research is academic and provides only limited evidence for policy improvement.
Although local communities have little influence on the hazard itself, communities have the potential to function and adapt successfully in the aftermath of disasters (Norris et al. 2008). Communitybased disaster preparedness is increasingly considered an important component of disaster management with the potential to increase community resilience through local-level capacity building (Allen, 2006). The actual assessment of community resilience remains a key challenge, however (Cutter et al. 2008), despite ongoing research to understand the nature of resilience to hazards more broadly.
One piece of the puzzle to improve community resilience to hazards is better knowledge of the kind of interventions that communities are already utilising in order to respond to hazards. Thus, this study aims to shed light on how people in communities are responding to hazards. This is done by classifying the kinds of responses used, into a typology providing an opportunity to understand the diverse character of community-based responses. A typology of responses also facilitates comparison of how people respond in different places or to different hazards.

Community-based responses to hazards
In the context of this paper, a 'response' is any intervention or activity that addresses the likelihood or consequences of hazards and thus aims to increase community resilience. Although contested, the term 'community' is defined in a place-based manner, to refer to a group of individuals and households living at the same location, i.e. a specific local-level geographic area, such as a village or group of villages, a municipality, a small city or a neighbourhood. Scale plays a major role in the consideration of resilience to hazards. While physical processes, such as climate change, that drive hazards and cause increased frequency and intensity of disasters, are global, the impacts are most acutely felt at the local level.
This paper systematically examines empirical information about local level responses to hazards based on peer reviewed, published case studies. Through this analysis and the typology approach, we are aiming to address four key questions: 1. What kinds of responses are used by communities all over the world? 2. Do communities in different parts of the world use different kinds of responses?
3. Are communities using hazard-specific responses? 4. Are communities using a multi-hazard approach?

Methods
We conducted a systematic review of case studies published in English language peer-reviewed literature. Each article describes one or more interventions that were implemented at the local level in places where people have been directly experiencing the impacts of hazards arising in many forms (e.g., floods, storms or earthquakes). We searched 9 different databases across multiple disciplines which required different search terms or keywords depending on the database used. Most search strings followed the format "(hazard OR storm OR hurricane OR cyclone OR earthquake OR flood …) AND communit*". Bibliographies of relevant papers were screened.
The search resulted in 1671 articles for full text analysis. The aim of the literature search was to find studies that were conducted after people had experienced the negative effects of a hazard event, and which provided information on how people responded to or coped with this event, whether immediately following the event or in the longer term afterwards. Therefore we only included articles that reported on (i) a hazard event or trend that has caused loss of life or health impacts or damage or loss to property, infrastructure, livelihoods and environmental resources; and (ii) a local community, as defined above. We excluded articles that (1) did not cite a specific event or geographic location, (2) described only potential but not actual actions, (3) provided assessments of risks, impact or vulnerability but no response actions, or (4) described results from modelling exercises. A total of 188 articles (11% of 1671) matched our criteria and were selected for content analysis (Table 1).
The date range of articles in our data set is 1982-2014, however the large majority of case studies fall within the period between 2001 and 2014. From the 188 articles, basic information was extracted about the community (name of the community, longitude, latitude, country), the hazard event/disaster (type of hazard, year, description or context, type of exposure, a description of the interventions/response actions and the aim of the intervention (to modify the event, to modify 3 5 10 15 20 25 30 people's vulnerability, to modify loss). The resulting dataset consists of descriptions of response actions. Each record in the database corresponds with a response action that has taken place in a specific community at a specific time.
We then coded these descriptions of response actions using a grounded theory approach to label the sentences or paragraphs discussing a particular response action, which resulted in 899 response codes. Similar responses were grouped first into 15 types and then into five classes according to the purpose of the actions or interventions that were taken. Because we used a qualitative approach, the number of types within each class varies.
For each type or class, we then counted how many communities have reported this kind of response action. Although the data is gleaned from published articles, we are not counting the number of articles but the number of communities for which response actions of a particular category have been reported.
In order to compare the frequencies of response classes between different types of natural hazards, we grouped the hazards into four types based on the EM-DAT classification of natural disasters over the period 1994-2014 (http://www.emdat.be/classification). EM-DAT classifies disasters into five groups: Geophysical, Meteorological, Hydrological, Climatological and Biological. Although not all hazard events in our data set are of disaster proportions, we adopted the same classification for the hazardous events, and here use the labels Earth, Weather, Water, and Climate for the four hazard types examined (excluding biological hazards here). Geographic analyses of reported responses is based on the United Nations' scheme of geographic regions.
A number of possible biases in the methods should be noted. First, because the sample of response examples has been taken from the peer reviewed English literature, it could be subject to a language bias. However, English is considered the international language of science (Tardy 2004;Kirchik et al. 2012) and although this notion may be problematic, it is not the subject of this paper. Second, it is possible that the response examples extracted from the articles represent the priorities of the authors rather than the realities of what took place in local communities. However, use of secondary data allows compilation of a global data set of local response actions, and the results obtained from that match findings of more specific research, which suggests that the data obtained do lead to a comprehensive and reliable global picture of community-based responses to hazards. Thus a data set based on peer reviewed English language articles is at the very least a useful starting point for development of a global perspective on community based responses to hazards.

Results & Discussion
Our data set includes information for 318 communities from all regions of the world, reflecting the fact that some of the 188 articles refer to more than one community each. Asia has the highest research coverage, both in terms of number of articles (table 2) and in terms of the number of communities (162, i.e. over half of the total) for which data is reported. Africa has the lowest research coverage, with only 5% of articles, and reporting on a total of 12 communities. The most studied countries in their respective regions are USA (86% of Northern America) followed by UK (45% of Europe) and Australia (39% of Oceania). Since, as noted above, only a small proportion of reviewed case studies (11% of 1671) provide empirical information on local responses to hazards, very little data is available across all regions (which corresponds to similar findings by Gal et al. 2015) but in particular for the African region.

What kinds of responses are used by communities? A typology of response actions
The results of the analysis of response examples into types and then into major classes are shown in activities, these kinds of responses could be grouped together. However, we feel that the salient property that sets evacuations apart from other disaster responses is the physical removal of people from the hazard area. We thus grouped evacuation with migration under the heading Mobility.
Interventions that focus on moving people out of harm's way or resettling victims in a new location after a disaster, can negatively affect the social relations and identities among community members.
In some cases, these potential negative effects are addressed by moving the entire community. These kinds of relocations, although technically referring to the movement of people, were grouped not in Mobility but within the class relational wellbeing.

Relational wellbeing
Responses focused on relational wellbeing draw on the fabric of the community for preparedness, for protection during a hazardous event or for the recovery afterwards. These responses typically involve collective action and interactions among community members, e.g., utilising kinship networks or formal community networks. In some cases, the fabric of the community is strengthened as a result of the disaster, e.g. in New Zealand where people "developed a real sense of community and doing things together" by supporting others who had lost family members in an earthquake disaster (Gawith 2013, see also Zahari and Ariffin 2013).

Awareness of hazards and risks
This class of responses includes two types. The type information and data includes actions to gather and communicate information about the threat conditions. Examples of this type range from the monitoring of an approaching cyclone via television, radio and internet (

Guidance and governance
This class comprises three types of responses pertaining to decision making. The type governance includes actions that handle, direct, govern, or control aspects of human hazard interaction ranging from land use planning regulations to reduce risk of wild fires in Australia (Buxton et al. 2011) to earthquake building codes in Pakistan (Ainuddin et al. 2014) to a local government policy in USA requiring that local residents raise their houses or risk that the entire community loses access to insurance (Colten et al. 2008). In Taiwan a community is officially designated as a "driftwood art area" in order to boost local artists as part of recovery from flooding (Wang et al. 2013). Planning

Infrastructure
Responses that focus on infrastructure for physical hazard defense are of two types: hard protection based on engineering efforts, or utilizing ecological properties for green protection. Examples of hard protection range from tying down houses and using safety features, such as hurricane shutters, clinching and rafter anchorage in Saint Kitts and Nevis (Hobson 2003) to flood defences such as sandbags, dykes or breakwaters (Esteban et al. 2013a)  We found that in our dataset, some 40% of communities employed interventions from the class Individual & Material Wellbeing (Fig. 2). The second largest class is Guidance with 20% of communities using responses from that class. The classes Awareness (12%), Relational Wellbeing (14%) and Infrastructure (14%) are reported less frequently. Some 56% of communities (177 out of a total of 318) utilise responses from more than one class.

Do communities in different parts of the world use different kinds of responses?
We found that, with the exception of Europe, all regions of the world have the class Individual and Material Wellbeing as the most frequently reported (table 3) (2017) found that significantly lower standards for flood protection and damage mitigation policies apply in the USA even for some main coastal metropolises such as New York City and New Orleans.
Indeed, flood risk management in the US is centered on flood insurance (Gouldby et al. 2017) in spite of the continued failure of the National Flood Insurance Program to achieve its objectives (Knowles & Kunreuther 2014). If we assume that the frequency of Guidance & Governance responses says something about the emphasis on policy and planning in a region, then we can conclude that Europe is more pro-active than similarly-developed Northern America, where there may thus be a need for more community-level responses of this class.
Looking next at responses of the class Awareness, we find that this is the least frequently used class of responses in Oceania and in Asia, with 8% of the community case studies from Oceania and 18% of Asian communities using responses of this kind. These are much lower frequencies compared to all other regions. Awareness responses aim to increase response capacity through improving understanding of hazards and better disseminating information about the threat conditions. Although it is possible that these kinds of responses are under-reported in studies conducted in these regions, our findings may also reflect a need for more efforts to increase hazard awareness in this region. In 9 5 10 15 20 25 30 the case of Australia, this conclusion is in fact supported by the literature (e.g. Box et al. 2013;Sewell et al. 2016).
Finally, the use of Relational Wellbeing responses draws on collective action and the social fabric of the community for preparedness or protection against hazards. Overall, responses of this class are much less frequent (table 3), even though, for example, "community-based disaster preparedness" is gaining importance (Allen 2006) and emphasises the need to address social and political aspects of vulnerability (Allen 2006;Blaikie et al. 1994). Oceania is the region with the largest fraction (44% of all communities in our dataset) reporting community-level responses of the class Relational Wellbeing.
Case studies discuss communities from Melanesia, Polynesia, Australia and New Zealand. Although it is not clear why relational wellbeing responses are more prominent in Oceania than in the other regions, this result does suggest that case studies from Oceania may provide a useful source of information to planners and decision makers who seek to enhance relational wellbeing in an effort to build social capital at the local level and to increase the resilience of their communities (Allen 2006).

Are there differences in responses to the different hazard types?
In order to answer this question, we examined whether the kinds of responses used differed between Similarly, Infrastructure responses are most frequently reported in connection with water hazards, but hardly ever in connection with climate hazards (2%). Responses from the class Guidance and Governance are more often reported for earth (45%) or water (43%) hazards than for weather (27%) or climate (25%). Awareness type responses are reported least frequently for weather hazards (15%).

Are communities using a multi-hazard approach?
Of the 318 communities in our data set, 53 (17%) are reported to have experienced hazard events from more than one hazard type (table 5). Although 17% is a surprisingly small fraction of the data, it does not imply that all other communities do not experience multiple hazards, but rather that the articles The World Health Organisation Strategic framework for emergency preparedness recommends an allhazards approach for disaster preparedness that includes hazard-specific measures where necessary (WHO 2017). Nonetheless, the creation of hazard-specific response plans has been the global norm (PAHO 2011) and our results suggest that community-based responses to increase hazard preparedness are mostly hazard specific.

Conclusion
To our knowledge this study is the first comprehensive global overview of community-based responses to natural hazards. At the core was a typology of hazard responses, which emerged from intentionally global, there is no other feasible method of collecting this kind of data, at least not to our knowledge. If the scope of the study was smaller, say regional or national, then it would be feasible to utilise additional or different data sources, such as grey literature or community surveys. But these sources are not appropriate for a global overview. On the other hand, a global study with intensive resources and participation by all nations of the world could gather more extensive data, but this has not happened. Thus we engaged in an extensive comprehensive review and extracted response actions from the articles which we are using as proxies for what is going on in the world.
As shown in section 3, the analysis herein was suitable to address the four key questions posed at the start of the paper. Our results resonate with the findings of studies that are more specific in terms of geographic area or type of hazard. This helps to validate our approach, and facilitates comparison of the findings of local case studies, within a global context.
Our data does not suggest what is the best response decision in a given situation. Rather, the typology presented here may be useful also for communities and community-focused agencies to structure their decision making and planning of response actions. A global overview of community response activities is also useful for national governments to aid decision making around hazard policies and how to best support and build local level response strategies, and for international aid agencies to see what kind of response actions need to be strengthen in different parts of the world.
Our results have several implications for research and policy: