1 Urban Vulnerability Assessment Through Indicators

Vulnerability has been widely conceptualized in several disciplines, resulting in a diversity of approaches (Adger 2006). One of the earliest definitions of vulnerability is the absence of a set of assets or resources (political, physical, natural, social, financial) that enables an individual or a group to come to terms with external stress or a damaging event (Chambers 1989). Following this definition, climate change adds to existing vulnerabilities that are structural in nature and often linked to poverty and inequality, bringing it into the equity context (Heinrichs et al. 2013). Adger and Kelly (1999) conceptualize vulnerability as the outcome of environmental, social, cultural, institutional and economic structures and processes that generate the uneven distribution of exposure, capacities and resources. Contextual vulnerability, in contrast, takes the multidimensional view of climate-society interaction into account and is influenced by contextual conditions such as dynamic social, economic, political, institutional and technological structures and processes (O’Brien et al. 2007). Here, the social system plays a prominent role. Furthermore, understanding what drives vulnerability is essential (Bergstrand et al. 2014). Vulnerability studies concerned with identifying assets and people that are most likely to be affected by natural hazards have a long history (e.g., Anderson and Woodrow 1991; Wisner et al. 2004; Cutter et al. 2003).

The specific ‘urban vulnerability’ to climate change and its conceptualization has seen only random investigation so far (Romero-Lankao and Quin 2011; Kuhlicke et al. 2012; Krellenberg et al. 2014). Research on vulnerability in cities has primarily focused on analysing hazard exposure (cf. Garschagen and Romero-Lankao 2013). We argue that the vulnerability concept would benefit from a more systemic perspective on the complexity of urban areas. Identifying a specific set of vulnerable objects, their complex drivers and their heterogeneity in spatial units of the city would allow to recognizing patterns of urban vulnerability (Pelling 2002).

Regarding the assessment of vulnerability, “there is a real need to develop a set of metrics to measure and compare the relative vulnerability of one place to another” (Cutter 2003, p. 7). Consequently, context-specific, indicator-based vulnerability assessments have been undertaken at city, community or neighbourhood level (e.g., Acosta et al. 2013; Müller et al. 2011; Scheuer et al. 2011; El-Zein and Tonmoy 2015), as well as at national level (e.g., Downing et al. 1995; Moss et al. 2001; Brooks and Adger 2003). Different scales call for different definitions of vulnerability and different methodologies for its assessment (Eriksen and Kelly 2007). Central to all indicator-based vulnerability studies is the indicator itself. The principal aim of using indicators is to reduce complexity by quantifying multidimensional issues through proxies (Hinkel 2011; Heink and Kowarik 2010). Composite indices with weighted averages of individual indicators are frequently used to allow for integration of several parameters that poses significant measurement challenges (see e.g. Birkmann and Wisner 2006).

Given the gap in urban vulnerability research and its related assessment, the primary objective of this paper is to develop an assessment methodology for urban vulnerability. This includes identifying the key requirements that place-based vulnerability indicators must meet. Accordingly, this article is guided by three research questions:

  1. 1.

    What indicators appropriately reflect the vulnerability of residents living in urban areas prone to climate related hazard?

  2. 2.

    What methodological approach allows for the assessment of overall urban vulnerability at different levels based on different data sources?

The methodological framework is applied in the Metropolitan Area of Santiago de Chile (MAS) which has a dry subtropical climate, with hot summers and winter rainfalls. Regional models predict higher median temperatures and less precipitation in the future but heavier concentrations of extreme events (McPhee et al. 2014). Ongoing changes in land-use and land cover have led to a significant loss of environmental services in certain areas of the city, such as storm water infiltration and heat mitigation (Müller and Höfer 2014). This, together with climate conditions reinforces seasonal hydro-meteorological hazards such as heat and flooding (Weiland et al. 2011). Flood hazard is defined as the likelihood of a potentially damaging flood event of a certain magnitude occurring within a certain period of time, whereas heat hazard is understood as an increase in surface temperature by one or more standard deviations above the mean surface temperature in an urban built-up area (Krellenberg et al. 2013). Flood and heat hazard exposure patterns vary by location as well as physical housing conditions and standards (Welz et al. 2014). Thus, vulnerability studies at the urban-regional level, e.g. Weiland et al. (2011), Müller (2012), Welz et al. (2014), Krellenberg et al. (2013) or Romero et al. (2012), have not yet applied an integrative framework for an overall vulnerability assessment. This exposes the gap in research on the susceptibility to climate related hazards and the attendant coping capacities, especially from a quantitative point of view.

The paper is organized as follows: Sect. 2 puts forward a conceptual perspective on vulnerability. Following the general discussion on quantitative vulnerability assessments, a context-specific indicator-based vulnerability assessment approach is presented. Section 3 introduces methods, place-based indicators and data sources around developing vulnerability indices, exemplifying them with the case study, the Metropolitan Area of Santiago de Chile. Section 4 considers the findings before concluding the article (Sect. 5) with a discussion on the feasibility of the methodology presented and its transferability to other cases.

2 Framing the Indicator-Based Vulnerability Assessment in the Context of Climate Change

To pursue the overall objective of this paper, we adopted a contextual perspective on vulnerability. It follows a bottom-up viewpoint and considers resident’s vulnerability to climate change, since our research shows that it is people who face climate related hazards under specific socio-environmental conditions. This refers to vulnerability as it emerged in the hazard and disaster community, as well as in development studies. Following Wisner et al. (2004), we understand vulnerability as the “characteristics of a person or group and their situation that influence their capacity to anticipate, cope with, resist and recover from the impact of a natural hazard” (ibid. 11). We define “urban vulnerability” as the interrelation between the spatial-structural (socio-environmental) conditions of an urban area exposed to a discrete and identifiable event in nature (or in society) with the underlying susceptibility and response profile of its inhabitants. Vulnerability is therefore a human condition generated by a hazard impact. It takes three components into account: exposure, susceptibility and coping capacity.

For the purposes of this research, exposure refers to the physical precondition of urban inhabitants to be affected by natural hazards (after Fuchs et al. 2011). We consider a number of aspects that render dwellings located in a hazard prone area within a city potentially exposed to flood or heat hazards. The key drivers include climate change, demographic and economic factors, and land-use changes that result is specific urban patterns. Susceptibility relates to the precondition of people with different exposure levels to suffer harm as a result of adverse conditions, such as inadequate infrastructure, physical disconnectedness or tenuous access to basic services. According to Birkmann et al. (2013), p. 8 “susceptibility (or fragility) describes the predisposition of elements at risk (social and ecological) to suffer harm”, referring to the reason why people and/or assets are exposed (cf. Cutter 1996; Balica et al. 2012). Coping capacity describes the immediate human response to a hazard occurrence (Birkmann et al. 2009). We refer solely to individual coping capacity evaluating available resources and actions or defence mechanisms, i.e., problem-solving techniques such as protective measures (cf. Wisner et al. 2004) that can be taken in response to a stressor within current structural constraints (Kelly and Adger 2000; Eriksen et al. 2005). In this way, we see coping capacity as a proximate, short-term and less strategic response to a hazardous event. It is highly context-specific and varies greatly between different places, groups, scales as well as over time, and therefore shows specifics in urban areas. In the same way, authors have also highlighted the role of institutions in determining changing levels of coping capacity (Pelling 2011; Berman et al. 2012). Thus, institutional capacity is not part of our vulnerability assessment thus forms part of the adaptive capacity of municipalities concerned.

With the aim of assessing urban vulnerability quantitatively, a definition of the system of analysis (what is vulnerable?), the valued attributes of concern (why is it important?), the external hazard (to what is the system vulnerable?), and a temporal reference (when?) is required (Füssel 2007; El-Zein and Tonmoy 2015). Pursuing this logic, several frameworks were developed to guide the vulnerability assessment systematically and provide a basis for comparative indicators and criteria development to measure the various dimensions of vulnerability (e.g., Schröter et al. 2005; Scheuer et al. 2011; Birkmann et al. 2013; El-Zein and Tonmoy 2015). In this line, our methodology provides a detailed assessment approach and defines a place-based set of indicators related to the specific local human environment system. Figure 1 shows the theory-based methodological framework of the proposed vulnerability assessment.

Fig. 1
figure 1

Indicator-based vulnerability assessment framework. Source the authors

Full size image

As an important pre-analytical step towards the subsequent urban vulnerability assessment, this framework first defines the urban structure of the area in terms of socio-environmental patterns. It considers that vulnerability to climate related hazards is unevenly distributed within cities and varies according to the prevailing socio-environmental factors such as accessibility to basic services, physical housing conditions, population density, vegetation cover and degree of impermeability, all of which determine the city in the context of climate change and its related hazards. These patterns are overlaid with hazard areas in order to identify those case studies that are in the focus of the urban vulnerability analysis at neighbourhood and household level. Unlike many other vulnerability analyses, the starting point of our approach is not the unequal access to resources and distribution of risks, nor does it take as read that poorer people are more vulnerable (Garschagen and Romero-Lankao 2013) or looking explicitly at vulnerability to poverty measures like e.g. Celidoni and Procidano (2015). Instead, we consider the socio-environmental conditions and the hazards as the point of departure for our analysis, and assess those areas in detail, integrating the three components of vulnerability (exposure, susceptibility and coping capacity). A place-based set of indicators is developed for each component and empirical methods for their assessment are defined. Data analysis includes the weighting and combining of selected indicators into composed indices (Krellenberg et al. 2014). Applying this methodology allows us to assess the prevailing degree of urban vulnerability to climate related hazard in the selected areas.

It is worth noting in this context that susceptibility is the least developed of the three vulnerability components, as confirmed by a review of the literature. Moreover, measurements for susceptibility and coping capacity are often less-than-adequate proxies and difficult to gauge. In some cases indispensable data is difficult to obtain. According to de Sherbinin (2014), the exposure aspect of vulnerability generally presents fewer problems. This could explain why most of the research undertaken so far has focused on the exposure side only (Garschagen and Romero-Lankao 2013; Romero-Lankao and Quin 2011; Birkmann et al. 2010).

3 Methods, Indicators and Data Sources

This section focuses on the methodologies and data sources behind the indicator-based vulnerability assessment. It describes data sources for the selection of place-based indicators and the methodology used for vulnerability index building and assessment.

3.1 Data Sources

Flood and heat hazard maps together with maps reflecting the socio-environmental conditions of the urban structure build the main basis for the detection of twelve case study areas within the MAS (see Fig. 2). The socio-environmental patterns are the result of a deductive approach based primarily on secondary data, e.g., the national census (INE 2002) and satellite imageries, using a combination of statistical, spatial autocorrelation and hierarchical classification techniques (cf. Krellenberg et al. 2014).

Fig. 2
figure 2

Case study areas located in six municipalities. Source the authors based on Höfer (2013) and Ayala et al. (1987)

Full size image

In these cases study areas, the majority of dwellings are low-rise detached houses with either one (58.2 %) or up to three floors (41.0 %) of varying physical condition that were found to be exposed to flood and heat in different ways. In some parts of the MAS, flooding occurs periodically during the winter after heavy rainfall and tends to inundate streets and ground floors of dwellings. Flood maps present hazard zones that are flooded at least once every 2 years resulting from natural stream and canal overflow, and the accumulation of storm water on the streets. The zones are anchored in the Metropolitan Regulatory Plan (PRMS). High flood hazard levels were selected to focus on the areas that suffer most. Heat maps are showing surface temperature derived from thermal satellite imagery with a geometric resolution of 30 m, which was processed to a heat hazard map indicating areas with one or more standard deviations above the average temperatures of the urban built-up area (Höfer 2013). Extreme heat is closely interrelated with land-use changes, as some areas have above-average air temperatures during warm dry summers (Müller and Höfer 2014).

With the aim of applying the vulnerability assessment in those areas showing diverse socio-environmental conditions and prone to either heat or flood or both hazards, an updated data base was needed as the latest available census for the MAS dates back to the year 2002. Intensive field mapping was undertaken to update the database in terms of material used for walls and roofs, type of housing construction, vegetation cover and yard trees, roof forms and protection walls. All of these characterizing housing conditions are of importance with regards to vulnerability to flood and heat hazard, particularly the exposure dimension. Field mapping was carried out between February and March 2014, and included the mapping of 5477 plots with reference to these housing conditions. A standardized “mapping code” was designed to allow for comparative analysis. The information obtained was pre-processed in Excel and SPSS before variables were merged to build sub-exposure indices and the final exposure indices (cf. Fig. 3 for more details on indicators and indices).

Fig. 3
figure 3

Variables and indicators of heat exposure. Source the authors

Full size image

A household survey was conducted in April 2014 in order to analyse the resident perspective on flood and heat hazard along the vulnerability dimensions of susceptibility and coping capacity and the identified indicators. A standardized questionnaire with open and closed questions was applied. As the sole prerequisite, interviewees had to be over 18 years of age, albeit not necessarily heads of household. Considering a confidence level of 95 %, a stratified sample of 2113 from the 29,052 target households in the twelve study areas was selected, 1274 of which had completed (61 %) the survey questionnaire. Open questions allowed residents to add comments and to list further aspects and variables relevant to the vulnerability assessment. Data gathered from the survey questionnaire was pre-processed in Excel and SPSS before variables were merged and susceptibility and coping capacity indices built for both heat and flood hazard.

3.2 Selection and Description of Place-Based Indicators

In order to assess urban vulnerability in the case study areas, a set of feasible indicators was selected for these areas (see Fig. 3), reflecting the specific local conditions. As the selection of the case study areas is based on socio-environmental conditions and hazards maps, the probability to be exposed under certain socio-environmental conditions in case of hazard occurrence is determining assumption of the vulnerability assessment. Consequently, the exposure side refers primarily to the housing conditions that determine the vulnerability of dwellings. Accordingly, indicators are describing physical housing conditions, vegetation, land cover and protection walls that allows for the assessment of potential exposure to hazard occurrence. The selection of indicators took into account that buildings show high evidence of vulnerability to climate change (Nikolowski et al. 2013). In this sense, physical housing conditions can hold vital information on living standards and indicate possible resistance to damage during a hazardous incident.

  • In the case of heat, thermal conditions are among the environmental factors that most affect people’s well-being and health, and are therefore in the focus of the exposure analysis of heat hazard. This considers different wall and roof materials that affect the indoor temperature of dwellings. Once solar radiation enters a building, it is transformed into heat; the more solar radiation enters, the higher the heat exposure. Our heat exposure indicators therefore take into account: “building material”, “building structure”, “roof form”, “roof material”, “vegetation cover” and “trees in foreyards”.

  • In the case of flood exposure, “housing type”, “protection walls” and “land cover” are added to “building structure” and “roof form”, acknowledging that exposure to flood and heat hazard is closely linked to the characteristics of the surface, land cover, shade, and their uses (Müller 2012). The indicators are strongly related to the prevailing level of imperviousness, which in turn impacts on the degree of exposure of the housing area to hazard. Surface sealing, e.g., reduces the absorption or retention of superficial rain-water runoffs, heightening flood incidents; in the case of heat, superficial heat storage increases.

A susceptibility “profile” of flood and heat hazard based on the household survey was created, to detect the reasons why people living in areas prone to hazards are particularly exposed. It takes the following indicators into account: “age structure of household” (all household members), “employment status” (of the interviewee), “socio-economic status” (of the household, GSE), “housing condition” (physical, COFIVI), “interruption in water supply”, “occurrence of hazardous incidences” (hazard return period) and “degree of affectedness” (by former events) (see Fig. 3).

  • The household age structure is a vital factor when it comes to the possibility of suffering harm. Various studies have revealed that children and the elderly seem more vulnerable to hazardous events, since their ability to thermo-regulate is limited (European Topic Center on Climate Change Impacts, Vulnerability and Adaptation 2012; Morrow 1999; Gladwin and Peacock 1997). Therefore, the indicator is composed of the age of all household members, classifies groups (<3, 4–10, 11–17, 18–64, 65–74, 75–85, >85 years old) and counts the shares of each age group. The indicator assumes that susceptibility is lowest if the share of all household members between 18 and 64 years old is highest.

  • The employment status gives information on the regular occupation of the respondent. The indicator assumes that higher proportions of people with low-skilled jobs such as housekeeping and childcare as well as retired persons are to be most susceptible in case of a hazardous event as they spend most of their time at home (Cutter et al. 2003). In order to get information on the employment status respondents were asked to indicate their primary type of employment of the last week.

  • Furthermore, the socio-economic status (GSE) of the household allows for conclusions to be drawn on household welfare and the socio-economic potential to prepare for the hazardous events. The GSE Index is widely used in Chile to evaluate the household socio-economic status and refers only to the educational level of the head of household and the possession of ten household goods. These parameter values are ranked and divided into five categories (ABC1, C2, C3, D and E)—with ABC1 referring to the highest socio-economic strata (Sabatini et al. 2010; Welz et al. 2014). As other studies have revealed, people of lower socio-economic status are generally seen to be more susceptible to heat-related mortality (Cutter et al. 2003).

  • The COFIVI Index (Welz et al. 2014) was used to measure physical housing conditions for the present study. The relevance of the latter to household susceptibilities lies in the construction materials used for roofs, walls and floors. Added to this in the case of heat hazard are specific hazard-related assets, such as the existence of either trees close to the building, balcony or terrace, air conditioning, solar protection or a swimming pool. It consists of five categories descriptive for the housing conditions: I (excellent), II (superior), III (ordinary), IV (deficient) and V (precarious). This vital information on physical housing conditions can be an indicator of living standards and hence of the predisposition of a household to suffer harm during a hazardous incident. Accordingly, in the household survey, people were asked to identify or determine within a predefined set of roof, wall and floor materials, those applicable for their own dwelling.

  • This further connects to the “interruption in water supply” indicator, which shows the number of blackouts during the previous year. It provides information on the accessibility of water as an essential component of basic technical infrastructure (Jean-Baptiste et al. 2011). Here, people were asked for the number of interruptions in water supply within the last 12 months.

  • Furthermore, the occurrence of hazardous events (hazard return period) and degree of affectedness were selected as indicators. The former refers to the likelihood to be affected by hazardous incidents such as flood and heat (estimated according to the time period a household has resided in a flood/heat-exposed area) (Birkmann et al. 2013), while the latter indicates the number of households affected by flood and heat during a specific time period in terms of health and physical damage. Both indicators are highly dynamic as they take former hazardous events into account. The components of the susceptibility index are composite indicators and require statistical pre-analysis.

Based on information gained from the household survey, the following coping capacity indicators were built: “preparedness/awareness” (of flood and heat hazard), “social networks” (and mutual support), “knowledge of protection measures”, “employment status”, “land tenure” (property), “use of public green spaces” and “perception of changes in land cover types” (see Fig. 3). Some are composite indicators, as in the case of susceptibility, and require statistical pre-analysis (for social network analysis calculations, see Welz 2014).

  • The level of preparedness/awareness is closely linked to the experience of hazardous incidents. This assumes that the more knowledge and information available, the higher the coping capacity (Cardona 2004; Müller 2012). In other words, experience of flood and/or heat is assumed to affect preparedness positively (Birkmann 2005). Here, different questions were asked and combined to one indicator. The questions relate to whether people have experienced flood and heat hazard since they are living in their dwelling, if they have experienced an increase, how well they felt prepared and if they would behave differently today in case of hazards occurrence.

  • The existence of diverse social networks facilitates the informal exchange of information, material, resources and mutual support. Social networks yield information on the degree of group cohesion at community level and thus on the possibility of preventive communal action and self-mobilization (Jean-Baptiste et al. 2011; Müller et al. 2011). An ego-centered network analysis was employed. Accordingly, the respondents were asked to identify their personal networks (naming up to four important contact persons) which they activate or not in case of emergency. They were asked if these contacts are relatives, friends, colleagues, neighbours, etc., where they live and how long they know each other. The diversity of each personal network was calculated. The network is assumed to be homogeneous if all named contacts are of the same category (e.g. all of them are his/her friend).

  • Knowledge of protection measures is a key indicator for coping capacity, since it evidences resistance to hazardous incidents. This includes knowledge of local hazard warning systems and/or emergency plans (Jean-Baptiste et al. 2011; Adger 1998), which can help facilitate evacuation activities and enhance the overall coping capacities of cities (European Topic Center on Climate Change Impacts, Vulnerability and Adaptation 2012). Therefore, by means of the household survey, information was gathered with regard to the alerts respondents receive from friends, neighbours, relatives, security services or police, or from public media. Furthermore, people were asked if they know any emergency system in place. Those respondents informed about the existence of early warning systems indicated also the specific type of warning system (e.g. community alarms, evacuation plans, public media etc.). Being informed improves the coping capacity, irrespective of the particular type of the system.

  • Employment status is also used in the susceptibility dimension, but interpretation differs. Within the coping capacity dimension, employment status means any labour relationship with a secure income, implying a certain capacity to take measures against a potential risk (Scheuer et al. 2011). A household with such an income is assumed to be more resilient and in a better economic position to cope with hazards (Adger 1998; Jean-Baptiste et al. 2011). Regular income is an indicator for the possibility of a household to save money for preparedness (Müller 2012). Since people are in many cases not able to give appropriate information on their overall household income we asked for the stability of income (e.g. self-employed, salaried employee, permanent work contract).

  • Land tenure refers to the ability to apply residential protection measures. Unlike house owners, tenants rarely have autonomy or a surplus to install mitigation measures and may therefore lack the capacity to cope with the consequences of a hazardous incident (Cutter et al. 2008; Jean-Baptiste et al. 2011).

  • The use and perception of urban public green spaces refers to heat hazard only. Parks, green lands and open areas play a significant role in the urban environment, since green spaces and their ecological benefits contribute to the reduction of heat stress and provide the community with amenities (Jean-Baptiste et al. 2011).

3.3 Indicator Reduction and Index Building

In order to allow for comparison of the three components of vulnerability, overall indices based on the selected indicators (see Sect. 3.2) were built. In order to make the different indicators equal to each other in value on the respective scale, an arbitrary numerical reduction was applied (Lazarsfeld 1993). Numerical weights for each variable were assigned along an axis of low to high exposure, susceptibility and coping capacity. The exact weighting of the variables is based on a review of the literature and expert knowledge, and reflects assumptions made (see Fig. 3). In terms of heat hazard, for example, a steel roof means high exposure, whereas a brick or concrete roof is more suited to predicting indoor temperatures and thus defined as “low exposure”. Consequently, each indicator is normalized between 0 and 1000 scores according to the respective accumulated frequencies (if ordinal scale) or equal points (if nominal scale), with 0 indicating the highest and 1000 the lowest degree with reference to exposure and susceptibility. In the case of coping capacity, weighting is reversed, with 0 indicating the lowest and 1000 the highest degree. Finally, an additive index as the arithmetic average of all indicators was employed in order to obtain a single score for each vulnerability component, likewise ranging from 0 to 1000. Applying this approach assumes that each indicator is equally important for the overall vulnerability assessment and acts largely independently on the target component. Due to the different empirical methods (field mapping and household survey) overall vulnerability is solely analysed at block level, based on average index values. In general, vulnerability is high when all three components approximate zero points and low when all of the values lean towards 1000 points.

4 Results of Vulnerability Assessment

The vulnerability assessment was performed in the case study areas only. To reach a more detailed level of spatial analysis within these areas, data was analysed at block level. The findings for each of the three vulnerability components, as well as for overall vulnerability are presented and discussed in the following.

4.1 Exposure

Regarding the potential exposure of residential buildings in the case of heat, the results reveal that the buildings were constructed with heat-resistant material and that the roof form does not contribute to heating the building. Regarding the roof form, 83 % of all mapped buildings have a double fall down roof, while 84 % of houses are built with brick or concrete (masonry). This is again reflected in the high mean values (over 800 points in 78 % of all residential buildings) (see Fig. 4) for each indicator in the case of heat hazard. In contrast, the building structure, roof material and vegetation cover indicators are characterized by low mean values ranging between 268 and 313 points. It can be assumed that they contribute to higher levels of exposure, since 75 % of all land plots feature very low levels of vegetation cover, while 87 % of houses are either row or semi-detached with roofs made of roofing felt or corrugated metal (steel). The statistical analysis findings show that the six indicators selected for heat all affect overall exposure. In general, the higher the values, the lower the level of exposure. When overall exposure is low (901–1000 points), all six indicators show more or less the same influence (almost the same number of points), emphasizing that both their general selection and their weighting is appropriate.

Fig. 4
figure 4

Exposure index for heat hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

For low levels of heat exposure (Fig. 4), the overall highest mean value achieved for vegetation cover is 793; all other indicators show mean values between 900 and 1000. It is the building material and roof form indicators that are anomalous, with high mean values in every exposure category, indicating that heat exposure can be prevented by consistently using good standard building material. The latter is therefore not the main driver of heat exposure. The same applies to the “roof form” indicator. Both indicators show low statistical relationships in the Pearson correlation coefficient (Fig. 4). All other indicator values increase with declining heat exposure. In other words, the higher the values, the lower the exposure. The Pearson correlation coefficient, in particular, allows us to see the statistical relevance of the “vegetation cover”, “trees in foreyard”, “building structure” and “roof material” indicators. Thus a higher percentage of vegetation (incl. trees) around a detached house with non-absorbent roof material potentially prevents heat exposure. When exposure is high, in contrast, vegetation cover and trees in foreyards show low or even zero values. This relationship is also confirmed by a strong correlation with the exposure index (after Pearson). High heat exposure is therefore clearly driven by the availability of vegetation, a theory-based assumption heavily reflected on the exposure index.

In terms of flood exposure, the highest mean values are found for the housing type indicator. With 929 points, it refers to the fact that 93 % of all land plots have stable housing conditions with high resistance potential in the case of flood hazard. All other indicators show low mean values, ranging from 159 points (roof form) to 441 points (land cover) (see Fig. 5). In particular, less than half (42 %) of all mapped land plots are equipped with trees in their foreyard, while 51 % of all land plots are completely paved (indicator “land cover”). This high level of imperviousness impacts on the degree of flood exposure and reduces the absorption or retention of superficial rain-water runoffs in case of flood incidents.

Fig. 5
figure 5

Exposure index for flood hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

Potential flood hazard indicators are reflected at different index levels in a similar manner to those for heat hazard exposure. The indicators on the flood exposure index—with the exception of “land cover”—range from 0 to 1000, confirming that all levels of exposure are represented. The “land cover” indicator is irrelevant when exposure is lowest. There is, however, evidence of a strong statistical correlation (after Pearson), that is, the lower the exposure, the lower the impervious cover of the residential area. This underpins the theory-rooted assumption that impermeability has a strong influence on flooding. The “housing type” indicator, on the other hand, displays no such linearity. It is characterized by mean values between 800 and 1000 for almost all exposure values. Accordingly, its correlation after Pearson is negligible. This means that residential houses are in a stable condition and not drivers of flood hazard exposure. We therefore advise removing this indicator from future exposure assessments. Similar evidence was found for the indicator “existence of protection walls”, i.e., a low correlation for various levels of exposure. It has fairly similar mean values (388 and 452) for two categories (high 101–200 and low 801–900). It is worth noting, however, that exposure would be much higher (in the category 101–200) if there was no protection wall. To sum up, potential flood exposure is primarily driven by the “trees in foreyard”, “land cover” and “building structure” indicators. In other words, where a detached house is surrounded by trees and a non-sealed surface, flood exposure is low.

4.2 Susceptibility

Susceptibility refers to a precondition to suffer harm in the case of flooding or heat. It can generally be stated that the susceptibility indicators show very good results in the case of potential flood and heat hazard. As in the case of exposure, high mean values indicate a low level of susceptibility. When overall susceptibility is low, all seven susceptibility indicators have a similar influence. The mean values of all heat indicators are concentrated in the middle range (400–600 points; see Fig. 6). The “interruption in water supply” indicator is an exception here, with very high mean values. This is due to the fact that 84 % of all surveyed households experienced no interruption in their water supply during the last 12 months. The lowest mean value (434 points), on the other hand, is assigned to socio-economic status (GSE) of the household. In total, 70 % of all surveyed households have a low or lower-middle socio-economic status. The indicators “age structure of household” and “employment status” are found in the middle range. Almost half of all households (48 %) are composed of less susceptible family members (adults between 18 and 64 years) and further 37 per cent show a very low share of children and/or elderly. Every second respondent (53 %) is employed and a considerable number—every third person—is housewife (32 %) spending most of day at home.

Fig. 6
figure 6

Susceptibility index for heat hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

In general, indicators on the heat susceptibility index range from 0 to 1000 with steadily increasing values when susceptibility diminishes, verifying that all levels of susceptibility are represented. This is confirmed by the Pearson correlation coefficient, which is particularly high and significant in the case of socio-economic status (GSE), employment status and housing condition (COFIVI) (see Fig. 6). It can therefore be assumed that these indicators are the core drivers of heat susceptibility. Moderate correlation was found in the “age structure of household”, “degree of affectedness” and “hazard return period” indicators. In the lowest susceptibility categories (101–200 points), the dynamic indicators “hazard return period” and “degree of affectedness” likewise show high mean values, indicating that high levels of susceptibility are alleviated by former experience of heat hazard. In total 51 % have experienced heat waves in the past and most of them date 1 year back. Every third respondent feels that the frequency of heat waves has increased in recent years. With regard to the degree of affectedness, almost two thirds (63 %) of the surveyed households were affected by heat hazard lasting >1 day. Heat stress was especially high in the residential building (also at night), at the work place and on the way to work in public transport. In consequence, people felt health problems and distresses such as skin burns, headaches, and respiratory disease. It can thus be concluded that in terms of heat hazard, the socio-economic status of the household affects its level of susceptibility. The lowest correlation is linked to the “interruption in water supply” indicator with high mean values (between 800 and 1000 points) in almost all susceptibility categories, thereby excluding it as a driver of heat susceptibility.

With regard to flood susceptibility, similar tendencies to those for heat were found. The main difference lies in the highest level of susceptibility (0–100 points), which is not evident in the case of flood hazard. It is again the socio-economic (GSE) and employment status that drives flood exposure. Higher levels of susceptibility are alleviated by the “hazard return period” and “interruption in water supply” indicators. The former in particular indicates that flood hazard occurs irregularly, i.e., in some winters only, and has shown no increase in recent years. Only one fourths (24 %) of all respondents have experienced flood hazard, half of which last winter. Furthermore, for the majority of the respondents, the hazardous event lasted half a day and mainly major roads of the neighbourhood were flooded. Consequently, residents were not able to reach their workplace as public transport was not operating. Overall this led to situations of psychological stress.

When susceptibility drops, all other indicators are equally relevant. These findings were also confirmed by the Pearson correlation coefficient. Again, the lowest correlation was found in water supply interruption, a moderate correlation refers to the hazard return period, degree of affectedness and household age structure, and finally, the highest and most significant correlation is linked to housing condition (COFIVI), employment status and socio-economic status (GSE; see Fig. 7).

Fig. 7
figure 7

Susceptibility index for flood hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

4.3 Coping Capacity

The immediate, short-term and less strategic response to flood and heat—individual coping capacity was analysed at household level. A first glance at the data reveals that unlike exposure and susceptibility, results for the coping capacity index and the related indicators show lower overall averages. The highest mean value was found in the “land tenure” (731 points) indicator. Values are highest in the case of fully paid owner-occupied flats or houses and those on instalment. This applies to 73 per cent of all surveyed households. The “preparedness/awareness” and “knowledge of protection measures” indicators also show low mean values (232 and 219 points, respectively), since both indicators fail to reach the maximum range of 1000 points (see Fig. 8). This could be because residents felt they were well-prepared for the last event.

Fig. 8
figure 8

Coping capacity index for heat hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

In the case of heat, people undertake very few measures to immediately cope with the hazard. This is similar to the “knowledge of protection measures” indicator, where approx. 70 % of surveyed households have no knowledge of existing emergency systems in their residential district and the relevant information is conveyed by only one of five possible sources (Radio/TV). The “perception of changes in land cover type” indicator had the lowest mean values (39 of 1000 points; see Fig. 8). This shows that only 8 % of all surveyed households associates increasing heat hazard with lack of green spaces, continuous construction and the location of their dwelling in, for example, the mountain basin. On the whole, knowledge or information on the direct causes of heat hazard and possible adaptation options is few and far between. Our statistical results allow for the conclusion that the coping capacity of the surveyed households in the case of heat hazard is low.

This was also confirmed by the accumulated values of each indicator according to the classified coping capacity index (see Fig. 9). The highest coping capacity levels (801–900 and 901–1000 categories) are never reached. With reference to indicator distribution in the different coping capacity categories, it emerges that the “preparedness/awareness” and “perception of changes in land cover type” indicators are not distributed linearly, leading to an insignificant correlation in terms of the Pearson correlation coefficient. Whereas urban public green spaces are generally seen as a hazard-coping measure in the case of heat, they become irrelevant where the overall coping capacity is highest. Here, however, the Pearson correlation is likewise high, indicating that the use of public green spaces is a coping capacity driver. The findings show that the highest correlations (after Pearson) were found in the “land tenure” and “social network” indicators. The latter, referring to the social diversity (heterogeneity) of networks, plays a major role on the coping capacity index. When social networks are homogeneous, coping capacity is low.

Fig. 9
figure 9

Coping capacity index for flood hazard and related indicators, mean values and Pearson correlation coefficient. Note The x-axis shows the average of overall exposure across a classified scale; the y-axis shows mean values (absolute values) of each indicator for the respective exposure group. Source the authors

Full size image

Flood hazard coping capacities show similar tendencies to those found for heat hazard. Again, it is the “land tenure” and “diversity of social networks” indicators that affect coping capacity most. The majority of all social networks are totally homogenous (56 %). Therefore, it is supposed that information on coping strategies comes from the same group affiliation and is less diverse. The principal difference lies in the higher mean values of the “perception of changes in land cover type” indicator (203 points) and the linearity of the “preparedness/awareness” indicator. The former shows a significant (even low) Pearson correlation coefficient, indicating that the perception of changes in land cover type contributes to higher coping capacity levels (see Fig. 9). This is rooted in the fact that 47 % of the surveyed households associate changes in land cover type (street design, construction activities, lack of canalization) with increasing flood hazard. Thus, in contrast to heat, knowledge does exist on the direct cause of flooding. In case of preparedness/awareness, 39 % of the respondents felt that they were not well prepared during the last hazardous incidence. Therefore half of them would implement prevention measures (e.g. supply with food products and isolate the building) to mitigate and minimize the effect of future events. The experience with a flood event clearly influenced the perception of preparedness, because 64 % feel better prepared for future hazardous incidents although they have only rarely implemented adaptation measures. This is also reflected in the indicator “knowledge of protection measures”, where only approx. 24 % of surveyed households have knowledge of existing warning systems in their residential district and the relevant information is conveyed by community alarm systems (60 %). Knowledge regarding the implementation of physical and individual hazard prevention measures is rarely available. Overall, when coping capacity is highest, only four of the six indicators are relevant: diverse social networks, knowledge of protection measures, employment status and land tenure. Hence immediate response options are less relevant for higher coping capacity levels than, for example, social and economic conditions.

4.4 Overall Vulnerability at Block Level

Our aim was to analyse the effect of the three vulnerability components, i.e., exposure, susceptibility and coping capacity, on the overall vulnerability of the areas under review. This helps (a) to evaluate the appropriateness of the indices for overall vulnerability analysis and (b) to gain insights into the spatial distribution of vulnerability. Index value dispersion is analysed by means of box plots and the correlation of each index with overall vulnerability scores, allowing us to determine its explanatory power.

Against the backdrop of the findings in the previous sections, Fig. 10 points to the following conclusions. With reference to vulnerability to heat, overall exposure ranges from 175 to 947 points with a median value of 526 points. It is symmetrically distributed in a flat manner, indicating that all heat exposure profiles (from low to high) are represented. This applies only in part to susceptibility, as values vary considerably from 304 to 871 points with a median value of 552 points. Here, too, we found symmetrical distribution with a concentration of scores in the middle range, indicating moderate susceptibility to heat. In contrast, scores for the coping capacity index are less scattered than other indices, ranging from 97 to 562 points with a median value of only 337 points. This indicates a narrow symmetric distribution and a low coping capacity profile. On the whole, the results can be interpreted as moderate vulnerability to heat hazard at block level, driven by low coping capacity levels.

Fig. 10
figure 10

Distribution of vulnerability indices for flood and heat hazard. Source the authors

Full size image

In terms of vulnerability to flooding, Fig. 10 shows a different picture. The exposure index is less scattered, with scores ranging from 259 to 779 points and a median value of 429 points. Distribution is asymmetrical with a tendency towards high exposure. The susceptibility index, on the other hand, shows a narrow symmetrical distribution with values varying in the upper range from 392 to 980 and a median value of 622 points. Consequently, the overall susceptibility to flooding is relatively low. The distribution of coping capacity in the event of flooding is fairly similar to that of heat hazard, with scores varying in the lower range from 126 to 598 points and a low median value of 361 points. This narrow symmetric distribution again produces a low coping capacity profile for flood hazard in all residential blocks. It can be stated that high levels of vulnerability to flood hazard are driven by a low capacity to cope but also by low to medium exposure values. At the same time, low susceptibility alleviates high levels of vulnerability to flooding.

Taking the coefficient of determination (r 2) of the three indices and of overall vulnerability to flood and heat hazard into account, vulnerability is first and foremost related to exposure and susceptibility. The exposure index, for example, explains 65 % of vulnerability to heat and 53 % of vulnerability to flooding. With reference to the susceptibility index, 46 % explains the vulnerability to heat and 56 % to flooding. These results suggest that the higher the vulnerability to heat, the more influences the exposure components, e.g., the physical housing condition and its surroundings. In case of vulnerability to flood, the results point to an important influence of the susceptibility in terms of the socio-structural dimensions of the household condition. Both indices are therefore highly significant (p value = .000) and have substantial explanatory power. In contrast, only 23 % of vulnerability to flooding and 26 % of vulnerability to heat are explained by the coping capacity index, although the values are highly significant (p value = .000). Consequently, apart from very low coping capacity levels, the index does not provide sufficient explanatory power to vulnerability. This could be due to indicator selection or weighting and will be discussed in the concluding section below. To sum up, vulnerability to flooding and heat is highly dependent on exposure and susceptibility profiles.

5 Conclusions

The indicator-based assessment approach for urban vulnerability presented in this article followed a deductive methodology. It was also driven by the availability of existing data obtained from field mapping and a household survey. The indicators are context-specific and therefore allow for the assessment of vulnerability in a predefined urban area. Indicator weighting was based on assumptions gleaned from a review of the literature and expert knowledge. This approach has the potential to assess vulnerability to flooding and heat, as was tested in twelve selected case study areas within the MAS.

Our findings lead to conclusions that are both content-related and methodological. In terms of content, the data we analysed shows that overall vulnerability to flooding and heat hazard in the case study areas concerned is moderate. In general, exposure is distributed across all levels, susceptibility shows moderate to low levels, and coping capacity very low levels. Results for both hazards differ slightly, while overall vulnerability to flooding and heat is closely associated with the respective exposure and susceptibility profiles. Socio-environmental conditions of the selected case study areas have an impact on resident’s vulnerability, in particular in terms of exposure. It is assumed that lower socio-environmental conditions of an urban area shape higher degree of resident’s vulnerability. Thus, urban vulnerability is likely to increase. This way, the assessment approach contributes to the development of a specific ‘urban vulnerability’ perspective with a more systemic perspective on the complexity of urban areas.

In terms of methodology, the proposed set of place-based indicators tested in the case study areas is a first approximation of urban vulnerability in terms of exposure, susceptibility and coping capacity. The results reveal that exposure and susceptibility are well framed whereas coping capacity remains a challenge. To some extent this coincides with research undertaken in other settings that has often shown susceptibility and coping capacity measurements to be less-than-adequate proxies (cf. de Sherbinin 2014), since they are difficult to measure or difficult to obtain. It is highly recommended, notably in relation to the coping capacity profile, that the inclusion of other indicators be verified in the interests of generating a more comprehensive picture of vulnerability to flood and heat. We therefore propose to strengthen particularly three aspects of assessing the coping capacity: economy, mobility and preparedness/awareness. Regarding economic aspects, an indicator addressing the total monthly average income of a household could analyse the ability to be self-sufficient during a hazardous incident. This also includes the amount of savings and debts as an indicator for the ability of recovery. Expending the hazards insurance to any kind of health insurance could give light to the coping capacity in terms of health impacts. Including mobility indicators could help to evaluate the individual mobility patterns in case of an hazardous event (e.g., provision of vehicle, additional apartments, accessibility/proximity to health care services and supermarkets etc.). In terms of preparedness/awareness, psycological and physical aspects could complement the existing analysis. Indicators such as calmness or selfconfidence in extreme situations could help to better sketch the abilities to cope with adverse consequences of a disaster. Moreover, the level of physical preparedness would allow to gain inside into actions taken to be prepared (e.g. family recovery plan, keeping flashlight and shoes beside bed, keeping supplies, knowing how to use radio transmitter, presence of cell phones and internet in the household and any kind of training related to cope with disasters) all of which are strongly related to levels of awareness.

As other vulnerability assessments have illustrated, results depend heavily on the weighting of the different indicators. El-Zein and Tonmoy (2015) also see the process of data aggregation through indicator weighting as an ongoing methodological challenge. We therefore pursued a two-sided evaluation approach. First of all, we sorted our results statistically to gauge assumptions for the selected indicators that led to the respective weighting. Secondly, we discussed weight determination with decision-makers (cf. Scheuer et al. 2011). This transparency led to an increase in credibility and allowed for validation of the selected indicators, the indices and, ultimately, the results. In this way, we explored several factors that contribute to different levels of vulnerability and now have a solid basis at hand for the proposal and discussion of context-specific adaptation options with stakeholders at municipal level (see also Schröter et al. 2005). From initial workshops held with each municipality we can already say that the presented approach is a key instrument for the depiction of different levels and concrete locations of vulnerability and of the work on appropriate adaptation measures to respond to these vulnerabilities. It provides a comprehensive database with detailed information on why people live in areas defined as prone to flood and heat hazard, the degree to which they are exposed and susceptible, and how able they are to cope with those hazards and their impacts. This allows municipalities to base their interventions on a proven database. It is of the utmost importance to continue the discussion with local stakeholders in order to further test what has been statistically assessed within the confines of this study and ultimately achieve spatial application under real-world conditions. This would give stakeholders a better grasp of why interventions in terms of climate change adaptation are essential in different areas of the MAS.

In general terms, the approach underpins that working with different data sources such as general statistics, field mapping or household surveys has its merits if methodologically combined in a comprehensive manner. It provides a methodology to quantify vulnerability that sees the assessment of all three components of vulnerability in the form of three single and finally merged indices of exposure, susceptibility and coping, allowing for conclusions on overall vulnerability. The statistical results reveal that the overall framework is feasible and transferable to other cities.

Some general criticism of working with vulnerability indicators has been voiced. Hinkel (2011), for example, faults in many cases the lack of deductive arguments for the appropriate aggregation of variables into indicators. He concludes that indicator-based approaches are only appropriate at local scales and when systems are narrowly defined. Local indicators cannot be generalized, since vulnerability is context-specific (O’Brien et al. 2007). Hence it is the assessment methodology that allows for transferability, as also discussed in this article. Cutter (2003) sees the development of vulnerability indicators as frequently hindered by the lack of conceptual development of, for example, the most suitable metrics and scales. Furthermore, translating vulnerability into quantitative metrics tends to reduce its impact and conceal its complexity (Alwang et al. 2001).