A typology of household-level adaptation to coastal flooding and its spatio-temporal patterns

The increased risk of flooding with rising sea levels (Nicholls and Cazenave 2010) and rapid population growth (Mc Granahan et al. 2007) indicates an urgent need for adaptation in coastal areas. It has been suggested that adaptation should be as low-regret (Wilby and Dessai 2010), flexible (Frankhauser and Soare 2013), and balanced in terms of costs and benefits (Hall et al. 2012) as possible. Nevertheless, it remains to be clarified how to achieve these targets in various regions of the world, especially in a proactive manner. Protection, as one possible type of coastal adaptation, is the most common approach of adapting to coastal hazards in many regions. During the 21st century it has shown to require substantial investments but also to be cost efficient for most developed coastlines (Hinkel et al. 2014; Nicholls 2009).

Accommodation can be an alternative to protection in many regions, particularly in those areas where protection is difficult to implement e.g. due to a complex coastal morphology, such as in deltas, or due to limited adaptive capacity in terms of technology and resources. When implementing accommodation measures, the use of land can be sustained, while existing structures and behaviours are modified. This can also involve the development of new structures, such as the construction of floating buildings or the modification of pre-existing structures to adapt to flood risk.

Of particular importance for accommodation strategies is the adaptation at the household level, as it can play a vital role in reducing the impacts of coastal flooding, mainly for the individuals themselves (Mendelsohn 2000; Tompkins and Eakin 2012). Integrating household-level adaptation into accommodation strategies also addresses the demand for assimilation of bottom-up stakeholder knowledge and top-down climate-impact projections (Mastrandrea et al. 2010). This would also be in accordance with the wider call for sharing the responsibility for adaptation (Adger et al. 2013).

Previous studies have explored household-level adaptation in coastal areas (e.g. Bichard and Kazmierczak 2011; Harvatt et al. 2011; Harwitasari and van Ast 2011; Molua 2012; Soane et al. 2010). It was shown that household-level adaptation to coastal hazards is widely applied by residents worldwide, despite being located in different regions and being surrounded by divergent institutional and social settings. The various measures range from every-day to high-cost actions.

Currently, however, there is no systematic grouping of adaptive behaviour of coastal households, nor is there knowledge on the measures different types of households are taking and how they differ between countries. This paper aims at bridging this gap in knowledge by developing a typology of households with regard to their adaptive behaviour. Such a typology can improve our understanding of which types of households are likely to implement which specific measures. This information can then be used for designing policy instruments, insurance schemes and risk communication tools that strategically target specific households.

The paper is organized as follows: Section 2 briefly reviews previous research on types of household-level adaptation. Section 3 describes the study area, the methodology applied and the sample. Section 4 presents the results. Section 5 discusses these results and section 6 concludes.

The analysis was carried out in two steps. First, we employed a cluster analysis in order to investigate whether we can identify categories of households likely to carry out a number of similar adaptation measures at the same time. Cluster analysis is a form of statistical analysis, which aims at classifying objects systematically. Cases or variables, which can be characterized by a set of solid values, are grouped into clusters according to their similarity. These clusters should be as homogeneous as possible internally and as heterogeneous as possible externally. We thus clustered the households that applied similar adaptation measures. In order to avoid chaining and outliers, we used the squared Euclidean distance as a proximity measure and the ward as a clustering algorithm (Kaufmann and Rousseeuw, 2005). To test the stability of results, different methods employing varying proximity measures and clustering algorithms were applied. As variables were binary, they were not standardized. Data records with missing values were excluded; thus, 75.1% of respondents were finally selected and grouped.

In a second step, we conducted uni- and bivariate statistical analysis and Χ² as well as Eta tests to analyse, whether the clusters attained (i) differ in their distribution between the sample sites, indicating that membership to a specific household type depends on the historical or institutional context; and (ii) could be explained through temporal variables on which data was also collected through the questionnaire, i.e. age of the respondent as well as length of residence in the house and at the coast. We hypothesise the household-level adaptation not to differ much spatially, as the exposure of coastal dwellers in the three study regions is similar. We further hypothesise the households to begin by implementing low-cost measures, e.g. reading information leaflets, and to only then conduct measures that involve higher expenditures, e.g. buying protective barriers for the doors of the house. This would imply temporal differences between type memberships.

The grouping of household-level adaptation behaviour into four categories gives some indication on the types of policy instruments to use in order to motivate household-level adaptation in coastal areas. In order to provide incentives for the implementation of structural measures, for example, such as storing pumps, elevating the house, constructing flood safe electricity supplies and heating systems as well as storing barriers, communication instruments have to specifically address the groups of the theoreticians and the minimalists, which make up more than half of the respondents of this study.

The international differences in adaptation policies described above may affect public awareness and perception of flood risk and, finally, adaptation behaviour. By comparing adaptation behaviour in Denmark, Argentina and Germany, significant differences in the number of respondents belonging to certain behavioural categories can be found. In our sample, Danish and German dwellers seem to be more informed and mentally prepared, whereas Argentinean respondents more often take specific structural measures. This result leads us to pose the question, which behaviour indicates a higher adaptive capacity.

We suggest the following alternative explanations for the differences encountered between the study areas: First, adaptation behaviour may depend on the social and institutional background. Second, the provision of risk information is likely to trigger a focus on relevant adaptation measures. Third, dealing with risks and local knowledge can vary between countries and therefore influence behaviour. Finally, adaptive behaviour can depend on the existence of and reliance on public adaptation. However, other implemented adaptation measures did not significantly differ between countries, indicating that an implementation of these measures does not depend on background, risk information, general dealing with risks or public adaptation.

Our results indicate household-level adaptation to be a process where stages of behaviour tend to follow a certain temporal sequence; during the first stage coastal dwellers tend to be minimalists. During the second stage, households shift to become theoreticians and in the third stage they become part of the comprehensive category. An exception pose those households that focus on the implementation of structural measures, whose decision to adapt does not appear to be related to age of occupants or duration of residence at the coast. Future research should focus on this category, as the structurals are, in addition to the comprehensives, undertaking structural measures, which can be more effective in preventing tangible damage than adaptation measures, which require low costs and efforts.

When interpreting the results of this study, it is important to note that the younger age groups are under-represented in our sample, as 68% of the respondents were over 50 years old. This under-representation most likely results from the fact that ownership of commonhold, and therefore an economic interest in adaptation, is more common at an older age. An indication for this argument is the small share of respondents with children living in the household (27%) and the high share of home owners (83%) in this survey. Furthermore, we noticed that some measures were only relevant for households, which currently have the means to undertake them. For example, moving expensive furniture to upper floors can be only realized if there is an upper floor. Cellars do not commonly exist at the Buenos Aires coast, thus the measure of barriers for basements were not included in the cluster analysis. Although we treated Danish and German respondents as one coherent group, they differed significantly in implementing seven measures. Furthermore, the sizes of the single samples differed; the Argentinean sample is smaller (29%) than the Danish (36%) and German (36%) samples.

Finally, common household adaptation measures should be valued according to their economic benefits. The reduction of costs depends on the type of adaptive measure; for example, temporary safety measures, such as door guards, reduce damage costs by 50% (Thurston et al. 2008). However, in this study, comparable adaptation measures (e.g. protective barriers) were found to be underrepresented in their implementation.

The proposed classification of household-level adaptation styles has potential both for practical applications and future research. In order to inform coastal dwellers about affordable and suitable adaptation, communication should target all categories of households. A typology of household-level adaptation can further assist in addressing specific target groups. For instance, risk communication through social media could make use of the classification and address younger age groups that often belong to the category of minimalists. Furthermore, by identifying the factors which have an effect on decisions concerning adaptation, we improve our understanding of how behavioural changes and investment decisions can be encouraged. The proposed typology can provide indications about unexplored drivers of household-level adaptation for future research (e.g. by further focusing on the characteristics of specific categories of households) and improve our understanding of existing patterns in household-level adaptation. The typological construction can also help to manage larger data sets containing information about household-level adaptation.

This paper proposes a typology of coastal households with regard to their adaptive behaviour. Based on questionnaire surveys carried out in coastal regions in Germany, Denmark and Argentina we have identified four categories of coastal households, which differ in their adaptive behaviour: the theoreticians, the minimalists, the structurals and the comprehensives. In most of the cases the respondents of this study belonged to the behavioural type which we termed the theoreticians, signalling interest in being informed, rather than taking structural measures. By analysing category affiliation geographically, further patterns emerged. Belonging to a specific behavioural type appears to depend on external characteristics, such as differences in the institutional background, in local knowledge or the general existence of public measures. Our findings show that household-level adaptation is, to a large extent, a process that develops over time. Coastal dwellers start with basic measures, then prepare themselves mentally and, with progressing age and duration of residence, start implementing more time- and effort-intensive measures. The small group of respondents which we termed the structurals constitutes a special case: group affiliation is not related to time. Future research could explore the reasons for concentrating on the implementation of structural measures, as these measures have a large potential in reducing damage.

This study emphasizes the importance of integrating household-level adaptation into coastal risk management, while at the same time considering the diversity of styles of household-level adaptation. We conclude that risk communication can only reach the entire group of households and increase versatile adaptation behaviour by addressing all behavioural types of households. The typology described is potentially applicable for describing behavioural types of coastal households in different areas and also in other environmental risk contexts.