On the Science-Policy Bridge: Do Spatial Heat Vulnerability Assessment Studies Influence Policy?

Human vulnerability to heat varies at a range of spatial scales, especially within cities where there can be noticeable intra-urban differences in heat risk factors. Mapping and visualizing intra-urban heat vulnerability offers opportunities for presenting information to support decision-making. For example the visualization of the spatial variation of heat vulnerability has the potential to enable local governments to identify hot spots of vulnerability and allocate resources and increase assistance to people in areas of greatest need. Recently there has been a proliferation of heat vulnerability mapping studies, all of which, to varying degrees, justify the process of vulnerability mapping in a policy context. However, to date, there has not been a systematic review of the extent to which the results of vulnerability mapping studies have been applied in decision-making. Accordingly we undertook a comprehensive review of 37 recently published papers that use geospatial techniques for assessing human vulnerability to heat. In addition, we conducted an anonymous survey of the lead authors of the 37 papers in order to establish the level of interaction between the researchers as science information producers and local authorities as information users. Both paper review and author survey results show that heat vulnerability mapping has been used in an attempt to communicate policy recommendations, raise awareness and induce institutional networking and learning, but has not as yet had a substantive influence on policymaking or preventive action.


Introduction
The direct and indirect health effects of a changing climate, along with an increasing frequency and intensity of extreme weather events such as storms, floods, cold spells and heat waves are considered to be this century's challenge for society [1,2]. Cities are of particular interest in the context of climate variability and change because this is where potentially a large number of people are exposed to threats from extreme climate events [3]. This is made more germane given current rates of urbanization with just over 50 percent of the world's population currently living in urban centers. It is estimated that nearly 9% of the world's population will be living in just 41 megacities by 2030 (UN, 2015). At the same time, the high concentration of people in urban areas, make cities a good place for innovation and social learning. For example initiatives such as the C40 Cities Climate Leadership Group (http://www.c40.org) have resulted in actions that will reduce greenhouse gas emissions and address climate risks and impacts locally and globally, through creating a network of megacities that share knowledge. In relation to this transformative approach to dealing with climate risk and as the paradigm of adaptive management becomes increasingly accepted, it is becoming clear that joint action at the community level, in tandem with top-down interventions from public health agencies are needed to promote resilience to climate change and other stressors [4].
Vulnerability assessments increase the understanding of complex processes and aim at providing decision support for stakeholders and government at a variety of levels. Ideally priorities in adaptation should target areas in greatest need. Spatial assessments of vulnerability are especially appealing to policymakers, as vulnerability maps provide a convenient way to communicate risk related information [5,6]. During the past two decades, the increase in data availability, computational processing and storage capability and the development of geospatial analysis techniques has advanced the possibilities of spatial assessments [7]. Pelling [8], Birkmann and Fuchs [9,10], provide a critical view of both deductive and inductive modeling assuming that vulnerability mapping can induce institutional learning at multiple levels and Hess et al. see it is part of a repository of tools that help to improve adaptive management and increase the resilience of local public health systems [11]. Vulnerability mapping has been developed for health risks related to hazards such as earthquakes [12], flood, infectious diseases [13] and others. For example, English et al. (2009) reviewed a number of environmental health indicators for multiple hazards associated with climate change (heat stress, flooding, fire, drought, allergens, infectious disease) that included vulnerability indicators among others [14].
Of the aforementioned range of hazards, heat or extreme high temperature events have begun to receive an increasing amount of attention because of their discernible impacts on health and infrastructure. The health impacts of recent heat events such as the 2003 European [15], 2009 Melbourne [16], 2010 Russian [17] and 2014 Japan heatwaves [18] have been considerable, either in terms of number of deaths or those hospitalized. Accordingly, and occasionally at the encouragement of local authorities, researchers have begun to explore the possibility of heat vulnerability mapping as an aid to the development of adaptive management strategies focused on heat as a health problem. Since Smoyer [19] and Wilhelmi [20] called for developing integrated, spatial vulnerability assessments for heat, at least 26 papers on this subject have been published. Theoretically, integrated spatial assessments of vulnerability to heat stress, with the intent to identify hot spots of vulnerability, should assist with indicating where health protection from extreme heat should receive major attention. Reid [21], Maier [22], Harlan [23] and Wolf et al. [24] have tested the performance of heat vulnerability indices (HVIs) in different cities and claim, because of the utility in relation to policy development, that they are more than theoretical constructs. Very recently, and therefore not included in this analysis, is Morabito's [25] mapping heat-related elderly risk index and surface temperature. Bao et al. [26] have provided a review of some HVIs, comparing in detail the different input variables.
The question however remains, whether policy makers use the outputs from vulnerability mapping studies. Given this, the goal of this paper is to comprehensively explore the extent to which the published literature on heat vulnerability mapping has been taken up by the policy development community. In relation to this overarching aim, specifically we address the following questions (1) how has the research concerning heat vulnerability indices (HVI) been conducted, in terms of methods and data used (inductive versus deductive)? (2) what are the limitations and potential problems of current approaches to developing HVIs; (3) what are the policy recommendations of the research? (4) are there discernible linkages between HVI research and policy application? (5) what is the degree of interaction and collaboration between the heat vulnerability research and policy making communities?
By exploring these questions through the critically reflective review and by surveying the authors, our hope is to identify whether current practice facilitates or impedes the inclusion of heat vulnerability (science) mapping in policy making. By analyzing the challenges and limitations in building and applying HVIs we make a contribution to the wider question of bridging the gap between (HVI-) research and policy making. Given the burgeoning number of vulnerability index development and mapping studies and the need for the information generated by such studies to be useable in a policy sense, the aim of this paper is to establish whether there is a gap between research and practice. Accordingly we were interested in posing and answering the question "is there evidence for the application of research findings from vulnerability index and mapping studies to policy development?"

Methodology
In order to address the above mentioned aim we identified and reviewed "heat vulnerability mapping papers" that had been published up until 15 April 2015; a total of 37 papers were identified. In addition we invited the lead authors of each paper to partake in a short online survey (see Appendix A) composed of 10 core questions designed to elicit information related to vulnerability assessment development and whether the resultant information was incorporated into the formation of any local to regional heat management strategies. In the following sections the methodology for identifying the 37 papers along with the related review criteria and the nature of the online survey questions are described. Paper review results and an analysis of author survey responses is then presented, followed by a discussion of the review and survey results in the context of main question posed by this study.

Paper Identification and Review Criteria
The criteria used in the literature search was informed by previous research undertaken by the authors [24,27,28]. Specifically we used the web resources PubMed.gov (US National Library of Medicine, National Institute of Health), Web of Science and Science Direct to identify heat vulnerability mapping papers. The following keywords were used to search for relevant peer-reviewed articles: Heat, vulnerability to heat, heat waves, and vulnerability mapping. We then screened the search results and kept the papers that applied geospatial techniques. The search was limited to papers published in the English language. Papers were then screened for information about the approach used, policy recommendation given and any indications concerning cooperation with local stakeholders or links between science and action. Information was extracted and summarized in three tables. Based on the extracted information, a qualitative classification of the nature of cooperation between the vulnerability index/mapping producers ("the scientists") and the decision makers was undertaken; the categories used were "Yes", "No" and "Vague". Further, the policy recommendations in these papers were also extracted (see Appendix B).

Author Survey
A survey was set up using an online tool for surveys (eSurv.org). The aims of the survey were to:  Tables 1, 2 and 3 (some authors have multiple papers). The questions covered data collection for the development of the heat vulnerability index, the interaction with local authorities, the impact of the work in terms of application of the index and awareness raised, as well as the whether the index was being used to support decision-making. Two questions explored if the authors plan to undertake further research on heat vulnerability or other vulnerability mapping. The consent to publish the survey results was also requested. Authors were invited to partake in the survey in May 2015 and two follow up reminders were subsequently sent to non-responding authors.

Results of the Literature Review
Based on the search criteria, 37 papers were identified. 21 papers were from the USA and Canada, twelve from Europe and four from Australia. The number of spatial units used for vulnerability mapping varied from 15 (local-government-area scale) to 92,000 (census-block scale). The GIS intrinsically implied possibility to "overlay" risk from hazard (high temperature), exposure to it and population vulnerability is further referred to as the "Crichton's Risk Triangle" approach [25,29,30]. A series of studies reduce the dimensionality of data of several input variables. The studies identified by the literature search are displayed in tables 1, 2 and 3 according to the approach used to construct the vulnerability index, namely: (1) "a-priori"/inductive index without testing with empirical health data (11 studies); (2) "a-priori"/inductive index and with testing with health data (6 studies, two are the testing part of studies from Table 1); and (3) a "bottom-up" approach to mapping health outcomes and subsequently exploring vulnerability factors (9 studies). These studies analyze health data to derive information about spatial vulnerability patterns from empirical data. We did not observe any HVI studies that adopt a deductive approach, which utilizes available theories or frameworks to derive an HVI-this may be because there is lack of a strong theory or framework for developing vulnerability indices [31], and vulnerability being a "place-based" concept which makes it rather context specific. Table 1 summarizes the details of 18 studies that use census data and methods of dimensionality reduction, such as Principal Components (PCA) or Factor Analysis, to produce an "a-priori" vulnerability index that is then mapped. Such an approach could be considered as a "top-down" approach. Tate refers to this type of research method as "inductive" [32], as it uses variables that directly or in a proxy sense represent heat risk factors, to construct vulnerability indices. Reid, for example, performs a Factor Analysis of ten variables (demographics, prevalence of air condition, vegetation cover from satellite images and diabetes rate) and combines the resulting components into a vulnerability index for urban areas in the US [33]. Some studies also include projections of climate and population and thus try to estimate future vulnerability [34,35]. In some cases, the heat/temperature hazard is mapped separately [28], in others it is integrated into the index [36]. Table 2 summarizes details of seven studies that apply the "top down"-approach as outlined above for Table 1 but further use spatial health data to test the performance of the index or elements of it. Wolf et al. [24] and Reid et al [21] in fact describe the testing/validation of vulnerability indices based on PCA or Factor Analysis presented in separate papers. Also risk-mapping studies [37] using health data are included here. Table 3 summarizes details of the remaining twelve studies with a "bottom-up" approach: They all use health data to weight components in a HVI or directly map adverse health outcomes (e.g., excess heat deaths or heat related illness) and then identify characteristics of the vulnerable populations using an array of statistical methods. The heterogeneous methods here range from General Linear and Mixed Models applied to different factors for heat mortality and heat distress calls [38], Principal Component Analysis with inclusion of mortality data into the analysis rather than for validation [39,40], identification of areas with excess mortality and subsequent identification of risk factors through principal component regression [41], mapping of heat related mortality rate ratios and evaluation of spatial association between variables that describe neighborhood-scale characteristics and excess deaths amongst the elderly [42], vulnerability and exposure modeling with standard linear regression relating temperature to morbidity and mortality indicators [43], a forward selection algorithm based on Bayesian information criterion [44] and a series of hierarchical Bayesian models to examine associations between temperature and morbidity [45]. Factor analysis of ten variables (demographics, prevalence of air condition use, vegetation cover from satellite images and diabetes prevalence); national coverage of urban areas; evaluation with health data in separate paper (see Table 2).
NO: Not specifically mentioned here, rather pure research.

[47]
Toronto, requested this information to support decision-making processes.

Canada
The SIMMER project and an evaluation report [48] are linked to this initiative.  Evaluation with health data in separate paper (see Table 2).

VAGUE: Cooperation with Greater London Authority (GLA)
and data providers is acknowledged. Further research on heat in London is ongoing but direct link to this work is not obvious.

[36]
Cologne area,      testing with all cause mortality.
NO: Work appears to be linked well with similar research in the US, but not with policy.
Chuang et al.

[27]
Phoenix, Arizona,   In 15 out of the 37 papers identified geo-referenced health outcome data is used. The type of health outcome data used in the studies listed in Tables 2 and 3 varies. There are, for example, different ways to identify "heat related" mortality: in some studies mortality includes "all causes", in other studies some specific causes of death (e.g., "external causes") are excluded. The same applies to morbidity measures such as (all or selected) hospital admissions and (all or selected) ambulance calls.
Indications relating to cooperation with local stakeholders or links between science and action were extracted and are summarized in Tables 1-3 and classified into "Yes", "No" and "Vague". This categorization reveals that most studies in group 1 (inductive/top down approach without health data) (6 out of 18 papers) give an indication about concrete cooperation with decision makers, whereas this can be confirmed for only one study (out of 7) in group 2 (inductive/top down approach with health data testing). Group 3 (bottom-up approach with use of health data) records 2 "Yes" (out of 12). For details see Table 4. The papers' information about policy recommendations is summarized in Tables A1-A3 in Appendix B.

Results of the Survey
Authors were invited to partake in the survey in May 2015 and two follow up reminders were subsequently sent to non-responding authors; 21 out of 34 authors completed the survey by closure of the survey in July 2015 (61% response rate). The replies were as follows: • Data collection: Most of the data were available for free (76%), a minor fee of less than US$100 or equivalent was charged to some (10%) and 14% paid a higher fee. Although not specified by the respondents, some of the high costs may be related to the acquisition of satellite images Data were available online for 29% of the respondents. Local authorities were supportive in data collection (48%) or helpful after considerable follow up by researchers (19%).
• Interaction with local authorities: The levels of interaction with local authorities varied. 86% reported that there was interaction with local authorities at different levels. 24% report much interaction (oral presentation/discussion at conferences, meetings or workshops, joint publications, email and phone), 38% "some" interaction and 24% "some but not much" interaction. 71% report that local officials commented on the vulnerability index, 29% did not know or did not get feedback. The overall tenor of the discussions and comments received by the survey respondents was considered fruitful and constructive. Only 14% of the respondents reported that there was no interaction. This probably applies to those studies with a pure focus on research where the exploration of the vulnerability assessment was the primary goal.
• Use of the analysis: Overall, respondents were positive about their index being applied. 71% think-to low, mid and high degree-that the respective vulnerability index is or will be used to support decisions on where to take action. Further, more than half of the respondents think that the results of the work are or will be applied in the local context: 14% thought the research results have already had significant influence in the local context and 43% saw or envisaged some local application. But there are also skeptics: 14% of respondents thought that the degree of application was very limited. 19% of respondents do think that results are not being applied in the local context and 10% of respondents do not know. 29% of respondents replied "I don't know" to the question "do you think the index is or will be used to support decisions on where to take action".
• Awareness raised: 90% of the respondents think that the respective work has increased awareness among the authorities and/or in the public to a low (38%), mid (33%) or high (19%) degree.
• Risk communication: 76% stated that the work has been used to communicate risks. In 40% the work was used by researchers to communicate risks to local agencies and/or experts or to the general public (30%). Only in 6% the local authorities were considered active in risk communication to the general public using the scientific work.
• Further research: 76% of the replying authors are planning to undertake further work on the topic of vulnerability to heat in the same or another urban area, 5% exclude this and 19% do not know. Consideration given to building a vulnerability index for other hazards is rather scarce, half of the respondents replied "no" or "do not know". Some also claim that similar criteria (social cohesion) define vulnerability to heat as well as other hazards. Others have done, know about or envisage vulnerability mapping for flooding and for several other hazards (severe storms, tsunamis, droughts, wild fires, disease vectors, earthquakes, land-slides environmental refugees, food shortages).

Discussion
This literature review has shown that there have been numerous attempts to assess and map vulnerability to heat stress in urban areas. Among the 37 papers reviewed, almost half adopted a similar set of methods and variables that were established by a few foundation studies, such as Cutter [66]. This indicates that Cutter's inductive approach to developing a SoVI (Social Vulnerability Index) for environmental hazards in general, significantly influenced the majority of subsequent heat vulnerability work. The use of ready and often freely available (census) data and by now relatively accessible computation methods (PCA/factor analysis) has provided the opportunity to apply a "Cutter-type" approach to a variety of different urban areas. However, Tate [32] has indicated that the SoVI has limitations in measuring actual vulnerability; these weaknesses apply to some HVI as well. Indicator selection according to local availability, scale of analysis, measurement error, data transformation, normalization, and weighting as well as aggregation are all possible sources of index inaccuracy. In addition, the contextual effects from the complex interactions of a place's unique socio-ecological systems could also limit the predictability of generic vulnerability indicators [55]. Further, as Romero-Lankao [67] concludes in her meta-analysis, researchers often make rather subjective modeling decisions with little or no stated justification in urban heat vulnerability index studies. While she argues that subjectivity is not inherently a bad thing and inevitable to some extent, she underlines that the effects of subjective choices on the output index need to be assessed before applying and using it to prioritize action. Similarly, Tate [32] claims that uncertainty and sensitivity analysis is often missing in such research. This is largely borne out by the review of studies presented in this paper.
While uncertainty originating from model development procedures is acknowledged in some papers (e.g., [68,69]), the implications that uncertainties associated with vulnerability assessment hold for policy development and decision making are not addressed explicitly. In fact, the way in which many of the papers present the science in relation to policy fits the linear communication model [70] such that information is provided with the expectation that there will be uptake by a relatively amorphous policy community. Some researchers have referred to this as the "loading dock" approach [71] to the provision of scientific information for policy development and have expressed the inappropriateness of this method, especially in the case of communicating scientific content that may possess elements of uncertainty [72]. In a similar vein, Knaggard [73] highlights the issues associated with evidence-based decision-making in the context of a paradigm of rationality. Despite the best intentions of some of the studies reviewed here, the fact that uncertainty in relation to policy and decision making is not addressed unambiguously raises the possible spectre of unusable science or non-actionable knowledge [74,75]. Although this situation is regrettable it is not irretrievable as the opportunity exists to adopt some imaginative approaches to assessing or simulating the effects of uncertainty on policy development in future studies that build on those reviewed here. For example van Pelt et al. [72] discuss the utility of intermediaries or boundary objects in acknowledging and communicating uncertainty and Head [76] outlines strategies for addressing the challenges arising from so called "wicked problems", a category of problem in which the impacts of heat on society comfortably sits. Dany et al. [77] and Berkhout et al. [78] both emphasize the critical role that stakeholder opinions play in how research and policy can develop in a mutually beneficial way and Reed et al. [79] suggest a number of principles that should be followed for effective practice of knowledge exchange.
Given that many of the heat vulnerability indices reviewed here have been developed in isolation of local knowledge, not only about micro-climates but also known socially determined hotspots of vulnerability, the suggestion by Kniveton et al. [80] that integrating local and scientific knowledge can be beneficial for addressing uncertainty, is particularly pertinent. Millner et al. [81] also describe how expert elicitation can assist with gauging the impact of uncertainty on decision making.
While there is agreement on the awareness raised (only 10% are not sure about this), several respondents claim that the application of the developed index at the local level is limited. Only 14% think that the research results have had a significant influence in the local context. When it comes to the use of the index to support decisions, merely 5% of the respondents clearly stated "yes, to a high degree"; 29% replied "I don't know" and 43% (24%) replied yes, to a mid (low) degree. This shows that the mapping and index development does seem to have a positive effect in awareness raising, but less so in triggering action or supporting decisions. Alternatively, the researchers are simply not informed to what extent their scientific results are used for decision support.
The inherent subjectivity and uncertainty of the index development methods, although known but not explicitly expressed by the index developers, could be the reason for limited uptake. Accordingly Tate underlines that the addition of uncertainty analysis to the index construction process is therefore an important step toward improving the quality of the next generation of social vulnerability indexes. Epistemic uncertainty is associated with all vulnerability models. The lack of its assessment and portrayal does not deny its existence [32]. Clearly a challenge for the vulnerability index development community, who want their science to be applied in policy making, is how to quantify uncertainties and effectively communicate these. This remains an area to be explored.
Along these lines Romero-Lankao [67] points to another weakness of vulnerabiltiy mapping science when she writes: "equally fundamental dimensions and determinants of vulnerability are ignored just because of lack of data and (index development) omits any attempt to gain ethnographic knowledge of behavioural norms, social networks and risk perceptions that are equally relevant to understanding urban vulnerability." Klinenberg [82] has shown this in his "social autopsy" for Chicago as well. Social cohesion as a protective factor is something that appears even more difficult or impossible to "measure" as a determining component of vulnerability. New perspectives or approaches, such as applying qualitative methods that capture for example, measures of social cohesion, may improve the performance and acceptability of vulnerability indices. Social cohesion could even be of relevance when victims of heatwaves do not belong to the classic risk groups (single, frail elderly) as Duneier [83] showed for Chicago.
Hinkel [31] has confronted the scientific limitations of vulnerability indicators and the issues they profess to address. He concludes that vulnerability indicators (and indices) are only good for identifying vulnerable people, communities or regions. Other common expectations regarding indicators such as identifying mitigation targets (and adaptation actions), raising awareness, allocating adaptation funds, monitoring adaptation policy and conducting scientific research he suggested cannot be fulfilled. This means that the studies identified here seem to best serve purposes (raising awareness, conducting research, identifying adaptation action) that they were not originally intended to serve. Hinkel [31] points to the problems intrinsic to risk communication: accordingly vulnerability indicators (and indices) are not the right means to raise awareness of climate change.

Conclusions
Heat vulnerability index development and associated risk mapping has developed significantly during the past decade. This is a result of the recognition that levels of vulnerability to heat vary within the population and that certain combinations of a range of risk factors may conspire at the individual to community level to increase the level of vulnerability to heat during extreme temperature events. As for other natural hazards, the visual presentation of vulnerability using coloured maps is experienced and seen as an effective communication device to raise awareness and as a possible input into policy and decision making. The increasing availability of high resolution spatial data at the intra-urban scale and advances in GIS technology have created the possibility for the natural hazards and social determinants of risk communities to create information that will assist with protecting the health of the vulnerable. The hypothesis that information at higher spatial resolution is better absorbed for protective action warrants further testing. However this normative view, based around potentialities and aspirations must be tempered by a positivist view of the current situation. The analysis presented in this paper has attempted to do this.
Although a considerable effort has been invested in heat vulnerability mapping research, the analysis presented here, based on a review of available literature and a survey of vulnerability index mapping paper authors, demonstrates the persistence of an unambiguous gap between vulnerability mapping science and policy. At the heart of this issue is a lack of understanding of how to mainstream this type of science into decision making, notwithstanding of course the issues that something as complex as vulnerability can be captured in a single index and that methodological issues related to vulnerability index construction need to be resolved before proceeding to policy development. Other substantive issues highlighted by this analysis include how effective or accurate (in terms of predicting impacts or outcomes) does a vulnerability index need to be in order to be "reliable"? The answer to this question not only lies with improving the science of vulnerability index based predictions of health responses but rests with the ethics of what is a socially acceptable level of vulnerability assessment inaccuracy. Allied with this is the issue of uncertainty. In relation to this, our analysis of vulnerability mapping papers and author survey results reveal that not only is there an inherent uncertainty associated with the indices on which vulnerability mapping is based, but a critical analysis of how uncertainty may cascade through the decision making and policy development process is lacking in the heat vulnerability mapping literature. This may relate to the nature of interaction that the vulnerability index development community has with policy developers in local authorities, in that it is one-way in the direction of end users, as opposed to a truly reciprocal interaction, which may assist with improving the understanding and communication of uncertainty.
While the results of our analysis point to a number of commonalities within the vulnerability index and mapping literature that may militate against mainstreaming of research outcomes into policy development, we are hopeful that the gap between science and policy will lessen. This will be achieved by both science and policy makers working within a co-production of knowledge paradigm and developing a greater mutual understanding of the barriers, constraints and limitation faced by both communities in terms of what they are striving for, such as the achievement of social justice and the reduction of inequalities in the context of the societal impacts of extreme heat events. In addition to this paper, to gather researchers active in this field to develop common guidelines could be a way forward.
Lastly, it should be acknowledged that this being an emerging field of research with more studies appearing during finalization of the manuscript, this review might not have captured all the published literature on recent heat vulnerability mapping. Further, 39% of the contacted authors did not reply to the questionnaire. Accordingly the findings from the analysis presented here should be treated as only indicative of current heat vulnerability mapping activity, application and use of the research results. To what extent these findings for heat vulnerability mapping may be applicable, or not, to other hazards, is another interesting question to be explored.

Acknowledgments
We thank the participants to the survey and five anonymous reviewers for constructive comments.

Author Contributions
Tanja Wolf generated the idea for the paper and wrote the main parts of the text. She developed the survey in close collaboration with Wen-Ching Chuang. Wen-Ching Chuang performed the set up of the survey in e-Surv and provided the summary of the survey results. Wen-Ching Chuang identified some of the papers for the literature review in a separate paper and provided the full text of most papers as well as constructive comments that shaped the article. Glenn McGregor provided input on an earlier version of the paper and added substantive sections to the paper during the revision phase.

Conflicts of Interest
The authors declare no conflict of interest.  Table A1. Policy recommendation from "a-priori"/inductive heat vulnerability mapping.

Reference Policy Recommendation
Vescovi et al.
2005 [34] "The most important aspect of our results is the geographical designation of specific zones where people are expected to be at risk in a warmer climate.
Specific measures concerning mainly the elderly should be put in place for these regions so that relief can be provided immediately in the event of a heat wave". (p. 77) Lindley et al.
2006 [46] "…the method also provides a mechanism through which areas suitable for further neighbourhood scale assessment and potential adaptation strategies can be determined. An analysis of the nature of hazards and vulnerabilities within cities and other urban areas is clearly a useful basis for tailoring planning and design strategies to the specific needs of the affected community". (p. 565) Reid et al.
2009 [33] "With further validation at the local scale and evaluation with health outcome data, our methodology and results can help target resources for intervention". (p. 1735) Rinner et al.

[47]
The recommendations include creating multiple representations of vulnerability indicators, indices and hot spots in order to avoid issues resulting from geographic aggregation and scale effects, variable selection, and the input parameters of cluster analysis and multi-criteria methods.
Kershaw and Millward 2012 [49] Our results highlight the value to public health organizations of in situ meteorological data when evaluating potential vulnerability during extreme heat events. (p. 7340) Chow et al.
2011 [50] "Anticipate increased heat-related emergency dispatches calls during heat wave events and tailor effective measures for them (e.g., more Spanish-speaking responders or specialized elderly medical aid centers). Policies to improve social cohesion and integration within neighborhoods via widespread dissemination of heat-stress mitigation information in different languages". (p. 15) Wolf et al.
2013a [28] "…the index presented here needs to be tested as a reliable a priori predictor of health outcomes such as mortality or ambulance call out. This will be the focus of future work". (p. 67) Depietri et al.
2014 [36] "Our analysis showed that, while the higher vulnerability of the population of Cologne to heat waves is concentrated in the city center, policies that aim to tackling it should also take into account the connections and interactions between the city center, the surrounding districts and its hinterland, reducing the susceptibility of lower status social groups and enhancing ecosystem management". (pp. 115-116) El-Zein and Tonmoy 2015 [51] "outranking procedures, previously only applied to decision-making problems, can be used for vulnerability assessment and may provide a better approach for teasing out policy-relevant information from uncertain vulnerability data. " (p. 216) Buscail et al.
2012 [29] "We recommend, however, using the health risk index together with hazard and vulnerability indices to implement tailored programs because exposure to heat and vulnerability do not require the same prevention strategies". (p. 8) Aubrecht and Özceylan 2013 [52] "Applying a very granular approach at a high level of spatial detail enables the detection of hotspot areas within cities. (…) It can therefore provide valuable decision support in directing risk mitigation measures which in a heat stress context particularly implies increasing the local communities' adaptive capacity". (p. 74) Table A1. Cont. Table A2. Policy recommendations "a-priori"/inductive heat vulnerability mapping using health data to test the index.