全球变化知识资源中心

To elicit cultural knowledge about climate change we employed the cognitive and psychometric methods of free listing, pile sorting, multidimensional scaling (MDS) and cluster analysis^{32, 33, 34, 35}. Together with ethnographic data, these methods allow us to visualize the content and structure of cultural knowledge about climate change. Specifically, we had individuals in each community sort terms related to climate change into piles, aggregated those piles, and then used MDS to visualize the relationships between the terms. We used ethnographic data, especially interviews with key informants (See Supplementary Information), to identify the meaning of word clusters and the cognitive dimensions that govern the overall distribution of data in the MDS plots (Supplementary Figs 1–3) and subsequently employed Johnson’s hierarchical cluster analysis (Supplementary Figs 4–6) to mathematically define word clusters in the MDS plots³⁶.

In consultation with study participants, we found that there were three clusters of terms that remained together in all three communities (Table 2), and eight terms that were placed in different clusters by different communities (Table 3; also see Supplementary Figs 7–9). The large extent to which communities shared cultural knowledge on climate change is supported by the significantly high correlation (measured by quadratic assignment procedure—see Methods) between community MDS plots (mean r = 0.707, p < 0.000). The three clusters of terms correspond to the three components of vulnerability as defined by the Fourth Assessment Report of the Intergovernmental Panel on Climate Change³⁷. That is, the green cluster includes words that describe risk or environmental impacts of climate change and include terms such as temperature, storms, floods and rising tides. The red cluster includes terms that relate to the communities’ sensitivity to climate change impacts, such as illness, ageing, fear and poverty. Finally, terms in the blue cluster are words that the community views as things that would affect their adaptive capacity to climate change, such as the federal government, jobs and relocating.

Crisfield’s MDS plot also differed from the other two by having more terms in the cluster that correspond to adaptive capacity. Whereas the study communities in St Michaels and Dorchester County both include five terms in the adaptive capacity cluster, Crisfield includes nine. A key adaptive capacity term for Crisfield study participants is family members. Whereas workshop participants in St Michaels and Dorchester County thought primarily about concern for the well-being of family members during a climate event, resulting in the term’s location in the sensitivity cluster, Crisfield residents had relied heavily on immediate and extended family members for assistance during and after Hurricane Sandy. Thus, in Crisfield, family members were not viewed as a community liability, but as a source of adaptive capacity.

The adaptive capacity cluster for Crisfield also includes God, knowledge and communication, which captures the importance of their deep faith, place-based knowledge, and social networks of communication during and after Hurricane Sandy. In contrast, workshop participants in Dorchester County grouped these words in the cluster corresponding to sensitivity, suggesting that community members perceive a relative lack of knowledge and access to government officials increases their sensitivity to climate change impacts. Furthermore, the location of God in the sensitivity cluster reflects their fear that their churches, all four of which are located near water bodies or tidal wetlands, will be lost to sea-level rise. Finally, in St Michaels, the terms God, knowledge and communication were grouped in a fourth, separate cluster. A possible ethnographic accounting for this result is that study participants in St Michaels perceived these terms to transcend the issue of climate change, with the result that they ended up being most similar only to each other using Johnson’s hierarchical cluster analysis.

A final important overall finding from these three MDS plots is that in all communities sensitivities to flooding are thought about as local, whereas adaptive capacities are extra-local. Some of the workshop participants expressed feeling uncomfortable navigating the techno-bureaucratic world of policymaking and regulation, and thus felt that they were isolated from the resources and expertise that could otherwise help them to better adapt to flooding from sea-level rise. The social and political isolation experienced by these communities is not something that is readily captured by quantitative vulnerability indices, yet is nevertheless an important contributor to local-level vulnerability.

Notably, race and age both contribute to the social and political isolation that has limited these communities’ access to sources of adaptive capacity at the extra-local level. Although race was not a term in the pile-sorting activity (see Methods), in individual interviews issues of injustice related to race did surface, revealing how historical and cultural legacies of discrimination have simultaneously discouraged African Americans’ participation in government decision-making processes and allowed their needs to be overlooked by government officials (see Additional Discussion of Results in Supplementary Information). Our ethnographic data further revealed that race can have differing impacts among seemingly similar communities; although race contributes to the vulnerability of all the study communities, it has impacted Crisfield to a greater degree. The advanced age of many in these communities also contributes to the difficulties they face in accessing resources for adaptation. Government and non-governmental agencies increasingly rely on online systems for dissemination of information and submission of applications for aid. Internet navigation is often more difficult for senior citizens, who have had less practice than those in younger generations. In addition, seniors may experience health problems that make it difficult for them to exert energy in reaching out to agencies that could otherwise enhance their adaptive capacity.

We have shown that systematically eliciting cultural knowledge about climate change and connecting it to a scientific framework of vulnerability can yield nuanced insights about local vulnerability. Although the qualitative methods we employed are relatively straightforward for identifying similarities and differences in the way communities group social–ecological factors related to climate change, interpretation of these results depended on consultation with community members. The results as presented here do not exactly reflect the understanding of vulnerability to sea-level rise of any one individual, but rather reveal each study community’s shared implicit and explicit understanding of vulnerability that influences behaviour and decision-making.

We find that the ways in which social–ecological factors affect local vulnerability can differ considerably even among communities classified as having an equally high vulnerability as measured by quantitative indices. Although the revealed similarities are useful for suggesting adaptation needs at a more regional level, the differences revealed by the MDS plots allow us to better understand the unique local experiences of vulnerability. Specifically, the different roles that social–ecological factors play in different communities re-emphasize the need for adaptation strategies to be tailored to the local circumstances. Understanding these nuanced differences in local vulnerability is a crucial precursor for policymakers to develop climate adaptation plans that will be flexible enough to meet diverse local needs. The methods employed in this study can be beneficially used towards that goal because they allow for expeditious analysis of the ways in which both quantifiable and non-quantifiable dimensions of vulnerability relate and are actualized in the local setting.

Finally, our finding that these African American communities feel isolated from sources of adaptive capacity located mostly outside their communities points to the need for policymakers to proactively reach out to these communities and provide them with the information, training and access to resources from which they could greatly benefit. In essence, this result suggests that enhancing democratic processes and actively engaging underserved communities in grassroots efforts for adaptation planning is key for reducing vulnerability among those who are most vulnerable. Such insights cannot be gained from vulnerability indices alone; a comprehensive understanding of vulnerability requires methodological diversity and an integrative approach that includes perspectives from physical, natural and social sciences.