Social vulnerability in Chile: challenges for multi-scale analysis and disaster risk reduction

Socio-natural disasters can have profound consequences for countries exposed to natural hazards. Consequently, Disaster Risk Reduction (DRR) management and the development of techniques to measure social vulnerability, such as the Social Vulnerability Index (SoVI), are critical to comprehending and mitigating risk factors. However, the impact of considering different spatial scales to understand and analyze social vulnerability remains largely unknown. The objective of this research is to identify the factors that determine social vulnerability in Chile, the implications of using four different territorial scales, differentiating for urban/rural territory, and the implications in DRR. The research considers the SoVI method, using the national census and the socioeconomic household survey to construct 25 variables at the zone/locality levels, and the use of a GIS platform. On average, eight vulnerability components are defined per model, with an average explanatory variance of 71%. Our analysis shows that social vulnerability in Chile is highly conditioned by access to basic services, low educational level, quality of housing, and income levels. Furthermore, the use of SoVI has made it possible to determine that the use of different territorial scales is an opportunity and a tool for decision-makers that should be investigated for planning purposes and the design of DRR policies.


Introduction
The United Nations Office for Disaster Risk Reduction (UNDRR) defines a disaster as a "serious disruption of the functioning of a community or a society at any scale due to hazardous events interacting with conditions of exposure, vulnerability and capacity, leading to one or more of the following: human, material, economic and environmental losses and impacts." Here, three key concepts are presented: hazard, vulnerability, and exposure. According to UNDRR, a hazard is a "process, phenomenon or human activity that may cause loss of life, injury or other health impacts, property damage, social and economic disruption or environmental degradation," vulnerability is defined as "the conditions determined by physical, social, economic and environmental factors or processes which increase the susceptibility of an individual, a community, assets or systems to the impacts of hazards," and exposure is the "situation of people, infrastructure, housing, production capacities and other tangible human assets located in hazard-prone areas." The combination of these three concepts can most likely trigger a disaster (UNDRR, 2022).
Currently, the increase in socio-natural disasters at the global level has generated economic, social, and environmental impact in exposed territories (Tate et al. 2010). Between 1998 and 2017, there have been economic losses valued at $2.908 billion and 1.3 million human losses worldwide (Wallemacq et al. 2018). Understanding risk to natural hazards is therefore essential, especially in highly exposed countries to natural hazards.
Although the study of risk due to natural hazards was initially oriented toward the identification of the biophysical conditions of risk, in the past decade the study has been oriented toward an integrated approach. For example, Chamorro et al. (2020) proposes a conceptual framework for road network risk management, which considers a sustainable approach. In addition to simulating hazards, it considers not only physical aspects such as probabilistic models that assess the physical vulnerability of road assets, but also the social vulnerability of the population served by the road network. In the context of risk management, Koks et al. (2015) suggest social vulnerability as an essential element to achieve an effective, equitable, and acceptable strategies. Cutter et al. (2013) proposes the integration of social vulnerability using the SoVI as a weighting factor within the flood management carried out by the USACE.
Since the social dimension of risk, the vulnerability has had special attention in the context of socio-natural disasters (Birkmann 2007;Menoni et al. 2012), giving way to the development of multiple definitions given its interdisciplinary approach (Adger et al. 2004;Khan 2012;Lee 2014). However, various disciplines agree that vulnerability manifests itself as the susceptibility of a system or elemental product of its inherent state that is impacted by a disaster (Adger 2006;Cutter et al. 2003;Lee 2014;Mcentire 2005).
Among the dimensions of vulnerability, social vulnerability focuses on the study of the multiple underlying factors that exacerbate the effects of socio-natural disasters, such as preexisting social and economic conditions in society Lee 2014;Mcentire 2005). Thus, it is possible to elucidate several characteristics that manifest themselves differently in society and in the territory, which predispose it to suffer damage unequally from the impact of an extreme physical event, making it difficult to recover. Therefore, the study of social vulnerability will be linked to the concept of place, since it depends and is a reflection of the dynamics of the population in the territory and is a product of the interaction of the different factors that make up the reality (Cutter et al. 2003;Lavell and Maskrey 2014).
Considering this interaction, social vulnerability analysis provides a comprehensive understanding of risk, beyond hazard and exposure (Lee 2014), offering a multifaceted and multidimensional vision that highlights the importance of the social and territorial processes that are present in the different spatial-temporal scales (Cutter and Finch 2008;Khan 2012;Lee 2014;Menoni et al. 2012). The inclusion of social vulnerability, therefore, is required for the understanding of preexisting risk conditions, thus enabling the development of mitigation and adaptation policies geared in Disaster Risk Reduction (DRR) (Lee 2014;Cutter 1996;Cutter et al. 2003). Quantifying social vulnerability is complex, requiring methods and single and composite indicators to assess along it over time and space. As a result, the methods have been developed through inductive, deductive, and 1 3 high, directly affected numerous coastal and inland towns in the south-central regions of the country, generating extensive destruction of the territory, mainly along the coastal area of the country (Martínez et al. 2017).
This event in conjunction with other socio-natural disasters has led to the development of multiple risk and social vulnerability studies in the country. These studies, however, focused on specific territorial areas, temporal aspects, and specific hazards (Martínez 2014;Chamorro et al. 2020;Bronfman et al. 2021). As an example, Chamorro et al. (2020) present a first application of the SoVI methodology in Chile. Through a pilot case located in a reduced territory, social vulnerability is evaluated at the district level. It considers the municipality of Villarrica and Pucón, Region of La Araucanía, and the 2002 National Census is used as a database. As a result, Chile currently does not have a social vulnerability study to elucidate and understand the factors that contribute to risk along the country under different spatial and territorial perspectives. Just the study of Bronfman et al. (2021) has addressed the question of social vulnerability distribution in the country in order to identify exposure and potential susceptibility to natural hazards, however, this study was focused on the temporal changes of vulnerability more than on spatial changes and territorial scales.
In addition, the literature has dealt in a limited way with the role of territorial scales on the assessment of social vulnerability in various geographical contexts, even though the analyses of multi-scale contexts are essential to explain the distribution of social vulnerability and develop DRR policies applicable at different levels of action .
The objective of the research is to identify the components that determine social vulnerability in Chile based on the SoVI method and to verify the adaptability of the method for the evaluation of social vulnerability from a multi-scalar perspective. This could have implications for the development of management measures for disaster risk reduction, which are relevant to the levels of political decision-making. In turn, it could be an input for the development of risk studies, considering that the vulnerability assessment is designed for threats in general, since they could be adapted to the areas affected by a specific threat. The analyses were developed for four different territorial scales, including the differentiated analyses between rural and metropolitan areas. Data from the 2017 national census and the socioeconomic household survey, CASEN, are used. It is expected that a better understanding of social vulnerability under different territorial and spatial contexts will help in the design of DRR policies when accounting for the variability of the exposed population.

Geographical context
Chile is administratively composed of 16 regions, subdivided into 346 municipalities, with the city of Santiago located in the Metropolitan Region as the capital of the country. Due to its geography marked by its latitudinal extension, the National Production Development Corporation (CORFO) envisages a division of five macro-zones: Norte Grande, Norte Chico, Centro, Sur and Austral, each of which contains certain homogeneities of geographical and economic characteristics of the territory (National Development Corporation, 2013), as shown in Fig. 1.
In demographic terms, the country's last national census states that 17,574,003 people live in Chile, 48.9% corresponding to the male population and 51.1% to the female population, with an average annual growth rate of 1.06%, while the number of homes amounts to 6,486,533 units (INE, 2018).
Because of its geographical location on an active continental margin and more than 4000 miles of Pacific Ocean coastline, the country is exposed to numerous geological hazards. However, other types of natural hazards of hydrometeorological and biological origin are also present (Arenas et al. 2010). These hazards combined with the forms of occupation of the territories concur to present risks that have materialized in large disasters, generating very high social and economic impact.
Between 1980 and 2011, Chile suffered losses, resulting from disasters of socio-natural origin, close to 1.2% of its annual GDP (SVS, 2012). The worst situation occurred with the earthquake and subsequent tsunami of 27 February 2010, causing more than 500 deaths and economic losses estimated at $30 billion US, approximately 15% of that year's national GDP (UNISDR 2015).
High levels of loss have made it clear that Chile has no territorial order that considers existing levels of risk. On the contrary, the multiple natural hazards have not been properly considered by the laws of urban planning and construction, thus increasing the weakness and the absence of instruments of territorial order that consider these events. Romero (2014) suggests that the inadequate integration of Disaster Risk Reduction (DRR) policies into land use and urban planning in Chile poses a threat to the country and its people.
Also, no national Land Management Law exists in Chile, nor is there integrated watershed management, ecological or environmental planning, that regulates the land use, the location of the population, and the development of infrastructure in hazardprone areas. Therefore, in the country, there are no effective measures to neutralize or mitigate the risk levels to which it is exposed (Campos 2010;Romero 2014). Although Chile has historically approached its disaster management in a reactive manner, in recent years it has made significant progress in risk prevention and preparedness. One of the main achievements has been the creation of the "National Policy for Disaster Risk Management," an indicative instrument at the national level in this matter, which provides a guiding framework for the various sectoral and territorial initiatives that are necessary to effectively reduce the risk of disasters.
In addition, the government agency in charge of disaster management has been transformed, with the closure of the National Emergency Office (ONEMI) and the creation of the new National Disaster Prevention and Response Service (SENAPRED). This new service aims to plan, organize, and prevent natural disasters through territorially decentralized work, in collaboration with local communities ranging from municipalities to provinces and regions, following the guidelines of the aforementioned National Policy.

Definition of scales and adaptation of SoVI for Chile
To analyze the scope and variability of SoVI for the different study scales, the following procedures were performed: First, the census zone and locality, indicative of urban and rural territory, respectively, were defined as the unit of analysis, being referred to as "ZONALOC." Second, the following scales of the study were defined: "National," "Macro-zonal," "Regional," and "Metropolitan." The final scale permits differentiating the characteristics of the territories, for the different scales, according to their urban or rural character. In this way, each of the models will be analyzed with the complete territories and, in turn, without the inclusion of the metropolitan territories, allowing us to observe the vulnerability of rural spaces (Table 1).
The "Macro-zonal" scale responds to the fact that each natural event presents an affected area determined by the physical characteristics of the territories, which do not necessarily respond to the political-administrative borders of a country. In this way, the macro-zonal scale focuses on the analysis of each of the five large macro-zones of the country (Norte Grande, Norte Chico, Centro, Sur, and Austral), which are part of a geographic and political division of the country, which allows an understanding of the national territory based on its physical and natural differences. (Contreras and Chamorro, 2019). On the "Regional" scale, the country's sixteen administrative regions are considered as independent models of vulnerability; four of the most urbanized regions were considered for this research (see Table 1).
Finally, the last scale defined was the "Metropolitan" which evaluates large metropolitan areas independently of the rest of the territory. This idea arises as a response to the search and necessity for a differentiated valuation of the rural and urban territory, understanding the origin and particularity of vulnerability in these places (Fekete et al. 2022). In this way, the independent evaluation of both areas allows a more faithful observation of the reality of each territory, considering a priori that rural areas tend to concentrate the highest levels of vulnerability as a result of the least number of resources and infrastructure they possess. Four of the cities with the largest population in Chile were defined as metropolitan areas, concentrating approximately 47% of the population according to the 2017 national census (INE, 2018), each belonging to the regions under analysis. To produce more accurate results, analysis models were replicated at the national, macro-zonal, and regional levels, with the exclusion of metropolitan cities from the models. It is worth noting that the macro-zonal analysis was only replicated for Norte Grande and Centro regions as they are the only macro-zones with metropolitan areas and therefore have a change in analysis.
By comparing and analyzing models that consider the inclusion of metropolitan areas with those that do not, we can gain a better understanding of vulnerability levels and identify greater differentiations. This is because when evaluating both areas together, rural areas could be overestimated and present higher vulnerability values, while in the opposite case, urban areas could be underestimated with lower levels of vulnerability, by containing the largest number of services, accessibility, and social conditions. In this way, separating both types of territory would make it possible to identify the existing differences in each of the territories separately, this being of special interest in the face of threats that may be of greater affectation in each of the territories, such as drought in territories rural, for example.

Data preparation
To implement the SoVI index, the data sources available to the country were defined and analyzed, the main criterion being that the databases were freely accessible.
The first corresponds to the 2017 census prepared by the National Statistical Institute and containing microdata for population, housing, and households. This basis holds data for the whole of the country in different units of disaggregation; from them, the "ZON-ALOC" analysis unit was selected, achieving the collection of 21 variables (see Table 2).
The second database corresponds to the National Socioeconomic Characterization Survey 2015 colloquially called as CASEN, prepared by the Ministry of Social Development, which is important for its collection of social and community data. It should be noted that at the time of the development of this research, the CASEN 2015 survey was the latest version available. This base holds data in different units of analysis; however, these do not reach the level of disaggregation of the census database. Therefore, the smallest unit of analysis available was the "COMUNAL" level, which was selected for the extraction of variables, achieving the collection of four additional variables (see Table 2). However, for these data, analyses were carried out by spatial analysis at the "ZONALOC" level.
The 2017 census cartography was used as a third database and as a spatial input for the visualization and zoning of the data, being mainly integrative support for the databases and data obtained from the models. As for the preparation of the data, the variables went through a process of transformation to the percentage or average values for each unit of analysis, to allow comparison between them despite the demographic differences. Subsequently, they were standardized to Z-values with mean 0 and standard deviation 1, allowing comparison of variables expressed in different units of measurement. Finally, the orientation of the variables concerning the vulnerability was analyzed, identifying 23 variables with a positive orientation and 2 with a reverse or negative orientation "Average monthly per capita income" and "Percent of the working population." These last two variables were reversed, leaving the total variables with a positive orientation toward vulnerability to simplify the understanding of the components, as a new alternative to the original SoVI.
Regarding the selection of variables, it was prioritized that these cover different areas of vulnerability, recognizing eight dimensions from the bibliographic review. Once this was defined, the data collection was carried out from the various data sources. It should be noted that the variables were selected for threats in general, without considering any specific, based on what was indicated in previous studies, and adapted to the Chilean national context.

Socioeconomic status
This dimension reflects the ability to absorb and cope quickly with losses to a disaster and to have greater resilience to impact. It is also reflected in "environmental justice," which seeks to assist socially marginalized populations who are most vulnerable and who are often relegated to settlement areas with increased exposure to hazards and without access to quality housing.

Educational
The educational dimension reveals the ability of the population to understand information about emergency plans and alerts, knowledge of existing hazards, and measures to avoid dangerous situations (Cutter et al. 2003;Frigerio and De Amicis 2016).

Employment and occupation
Vulnerability is reflected through the population dealing with the primary and secondary economic sectors, mainly in self-employed workers who may be severely impacted disasters due to the loss of means of production and the need for monetary capital to resume work. This in turn leads to a migration of workers to other sectors and accepting lowerskilled jobs, thus reducing economic income. In terms of employability, a high percentage of the unemployed population's problems can be exacerbated after a disaster, contributing to a slow recovery from the disaster. Generally, and as shown throughout Chile, this population has a lower capacity to deal with the disaster (Cutter et al. 2003;Frigerio & De Amicis 2016;Menoni et al. 2012).

Demographics (age/gender)
Age reflects the population's capacity for evacuation, considering that the dependent population has less autonomous mobility than the workforce. This problem also manifests in special care requirements. Gender, meanwhile, manifests itself in greater vulnerability in the female population, due to family responsibility, differences in employability levels, lower salary and decision-making power, among others (Cutter et al. 2003;Khan 2012;Lee 2014;Menoni et al. 2012;Tate et al. 2010).

Social capital
The interaction of the population through social networks is usually associated with rapid access to warnings and information about preparatory and emergency actions (Cutter et al. 2003;Khan 2012;Lee 2014;Menoni et al. 2012).

Ethnicity/migrants
The concentration of ethnic and migrant population represents an increase in vulnerability due to language and cultural barriers, low knowledge of existing hazards, reduced access to post-disaster recovery funds, and high level of exposure to hazards when located in highrisk areas (Cutter et al. 2003;Frigerio and De Amicis 2016;Tate et al. 2010).

Housing quality
The quality of residential construction affects the potential loss and recovery. High-value houses require a great economic resource to replace them, and lower-quality homes are destroyed very easily and are less resistant to hazards (Cutter et al. 2003;Tate et al. 2010).

Access to basic services
Lack of access to pre-disaster services can be exacerbated after the event, decreasing the population's resilience capabilities and generating greater financial spending on the population and the State, to gain access to resources (Cutter et al. 2003).

Data analysis and spatialization
The "Principal Component Analysis" (PCA) was used for the identification of the components of vulnerability. This method reduces the number of variables and is especially useful when, within the dataset, the variables have a certain level of correlation, and in addition, allows evidence of any change in vulnerability over time (Cutter et al. 2003). As for the criteria for this analysis, a rotation "Varimax" was considered, while those variables that were grouped, i.e., common factors shared by more than one variable (eigenvalue > 1), were defined as components. For interpreting the components, variables with a factor load greater than or equal to 0.3 or less than or equal to −0.3 were considered.
Obtaining the components with a linear additive model was done to achieve the SOVI score for each unit of analysis. As Cutter et al. (2003) mentions, SOVI is a model that makes an equal contribution of each component to the vulnerability index for each unit, without going through a weight allocation for the identified components. Subsequently, the representation of the phenomenon continues through zoning; thus, the SoVI values were classified into five levels of vulnerability according to the criteria proposed by Frigerio and De Amicis (2016) and Guillard-Gonçalves et al. (2014) in their research in Italy and Portugal, respectively. These levels were based on standard deviation ranges (see Table 3), to allow a comparison of the values obtained in the models.
Finally, the Moran index was obtained for each model carried out, to understand the changes in spatial patterns of vulnerability according to changes in the territorial scale. In this way, the changes can be oriented toward the dispersion or concentration of vulnerability values when the index approaches −1 or 1, respectively.

Implications of multi-scale analysis in SoVI evaluation
It was identified between seven and nine vulnerability components, per model from PCA analyses, with an average explain variance of 71% for models that considered metropolitan and rural areas (Table 4), 70% for those models without metropolitan areas (Table 5), and 61% for the models that evaluated just the main metropolitan areas of the country (Table 6).
In Appendix A, the complete matrices obtained for further analysis are presented, while Tables 4, 5, and 6 present an excerpt of the main components of each model, which represents more than 35% of the explanatory power of each model.
These results allowed to determine that eight components stand out for their presence in all the models developed (see Table 7), concentrating the highest levels of experience explained in each of model.
From the different scales studied, it was observed that the models at the National level (with and without the inclusion of metropolitan areas) delivered eight vulnerability components with a total variance of 68% and 67%, respectively. It was determined that the principal components shaping the vulnerability of the country, in both models, were the characteristics associated with "Quality of housing" and "Access to basic services," such as access to drinking water network and sewage network. These two variables concentrated the greatest explanatory variance, followed by the components of "Population with low educational level" and "Socioeconomic." Regarding the macro-zone scale, it was observed that models with metropolitan areas did not have the same principal components for the five macro-zones under study. However, the components "Population with low educational level," "Concentration of dependent population," "Quality of housing and basic services," "Socioeconomic," and "Concentration of older adults" are highlighted.
For its part, the models for Macro-zone Norte Chico and Macro-zone Centro without the metropolitan areas obtained a better adjustment of the results with those obtained from the complete macro-zonal model. In this way, the components with the highest explanatory power achieved were "Population with low educational level," "Concentration of older adults," and "Quality of housing and basic services." At the regional scale, the models obtained, with and without the inclusion of the metropolitan areas (see Tables 4 and 5), delivered components of similar interpretation to models of National scale. However, among the regional models, it is possible to find certain differences in the components as well as in the levels of variance obtained.
The regional models with metropolitan areas (see Table 4), make it possible to point out that the vulnerability is made up of such components, which do not present in the same way in different regions under study. These components are "Concentration of older adults," "Population with low educational level," "Multidimensional poverty," "Concentration of dependent population," "Severance," and "Quality of housing and access to basic services." While regional models that do not include metropolitan areas (see Table 5), consistently concentrate the same three principal components of vulnerability: "Quality of housing and access to basic services," "Concentration of adult population," and "Population with low educational level." Finally, the vulnerability assessment in the main metropolitan areas of the country (see Table 6), also highlighted among its principal components the "Quality of housing," the "Multidimensional Poverty," and the "Concentration of dependent population." However, these are not uniform for all cities and have variations between them, where components such as "Population with low educational level" and "Socio-Economic" appear differently across the metropolitan areas.
In short, it is observed that, for the scales evaluated, the component with greater explanatory power is repeated. However, as shown in Appendix A, the remaining components vary between the models analyzed, mainly when considering the differences between models with and without metropolitan areas, which provide additional information on the characteristics of the vulnerability typical of each territory. The vulnerability is due to the presence of a population under 15 years of age who has not completed primary education (basic education) Dependent population concentration Has a positive relationship between the concentration of the older adult population and the younger population, who are more prone and vulnerable to natural hazards Socioeconomic Concentrates age relation variables that relate directly to economic and educational variables Multidimensional poverty Consists of variables like those of the socioeconomic component, but here the multidimensional poverty variable associates with organizational and community characteristics of the population Major adult population concentration Considers an aging population as an impact factor on levels of vulnerability because of its high economic and social dependence, which increases in the face of a hazard Severance Characterizes the vulnerability of the population according to the inability to respond economically to a hazard

Zoning of social vulnerability
From the zoning of the values obtained from the SoVI, it was possible to analyze the spatial distribution of the vulnerability in Chile and its behavior in the different models of spatial scales. The zoning of the models in the Coquimbo region showed that, considering the metropolitan areas, the vulnerability varies profoundly at the three scales assessed (see Fig. 2). Note that, at the national level, the high vulnerability values are scattered between the southern and central sectors of the region, while the low values are also scattered. On the macro-zone scale, the high vulnerability values were concentrated toward the center of the territory, while the other vulnerability values were found dispersed in the region. The regional scale presented a concentration of vulnerability values, getting a high value in the Moran index of 0.53, where the metropolitan area (zoom) concentrated the lowest levels of vulnerability.
As for the results of the models without metropolitan areas of the Coquimbo region (see Fig. 2), it is observed that the levels of vulnerability are visually more dispersed for the three scales, a consequence of eliminating the influence that metropolitan areas exert on the rest of the territory, which is consistent with a lower value of the Moran index of 0.23. This situation has as its origin the high concentration of resources and services that usually concentrate in these areas, among other variables, which does not Fig. 2 Zoning of the level of vulnerability estimated by SoVI for the Coquimbo region. Source: Author's elaboration coincide with the reality of rural areas. Therefore, these models capture those components typical of rural vulnerability, without the influence of metropolitan areas.
Finally, the evaluation of the Coquimbo-La Serena conurbation yielded different results than those obtained in the models that contemplate metropolitan and rural areas, allowing visualization of the different vulnerabilities that occur within the city with sectors marked by the high and low vulnerability.
The zoning of social vulnerability in the Bio-Bio region (see Fig. 3) exhibits similar behavior to that described in the Coquimbo region. Models with metropolitan area integration see spatial variations in vulnerability levels. It is possible to see a higher concentration of high and very high vulnerability in rural areas, reaching a high value in the Moran index of 0.73, the product of the influence of the Greater Concepción, which comparatively concentrates fewer vulnerability characteristics than the rest of the region and thus presenting the lowest levels of vulnerability.
The above is confirmed with the results obtained from models without metropolitan areas, which show a decrease in the overall levels of vulnerability in rural areas. However, it is also possible to see spatial differences between the three scales. At the National level, the region is estimated to have low levels of vulnerability compared to the rest of the country, while considering a scale of macro-zone, it is evident that the region is more vulnerable than other regions that make up the Centro geographic macro-zone.
As for the measurement of the Greater Concepción area, analysis showed similar behavior to that obtained in the urban area of the Coquimbo region, reflecting the different realities that Similar to the previously discussed regions, the models of the Valparaiso region (see Fig. 4) and the Santiago Metropolitan region indicate similar vulnerability profiles (see Fig. 5). Therefore, it is possible to appreciate the spatial changes for each scale and the influence of the metropolitan area on the rest of the territory. In turn, the low levels of vulnerability that metropolitan areas concentrate by being considered with the rest of the territory change when analyzed independently. These differences are repeated throughout the country.
Regarding what was obtained in the Moran index, it was determined that the models with metropolitan areas have a higher concentration of vulnerability values compared to the models that do not, with an average Moran index of 0.39 versus a value of 0.24, respectively. This is explained because of the influence that metropolitan areas exert on rural areas and vice versa, where the former concentrate the lowest values of vulnerability when evaluated together.

Understanding the social vulnerability in Chile
According to the results of this research, social vulnerability in the country is explained by the incidence of four relevant components or indicators: "Concentration of dependent Fig. 4 Zoning of the level of vulnerability estimated by SoVI for the Valparaíso region. Source: Author's elaboration population (older adults)," "Quality of housing," "Access to basic services," and "Population with low educational level." The dependent population linked to the group of older adults can be related to the process of "demographic transition" that has been experienced in much of the world and which has gradually transformed the age distribution of the world's population in favor of the most advanced (United Nations 2007). Chile has not been oblivious to this process as its population has been progressively aging over the past few decades. While this symbolizes the success of the human development process in lowering mortality levels, from the point of view of social vulnerability, it is a factor influencing the magnitude of socionatural disasters especially when combined with other components of vulnerability (Cutter et al. 2003;Rofi et al. 2006;Villalobos 2017). In this context, this age-group is linked not only to the lack of physical capacities for evacuation scenarios but also to a certain degree to socioeconomic characteristics such as greater economic dependence, which can influence the level of access to services and resources. They also require a higher level of social support and support networks, being, therefore, the concentration of older adults, a relevant group when managing measures to reduce vulnerability in the country.
Socioeconomic conditions stand out, in turn, as one of the main dimensions of vulnerability in this research, being consistent with numerous other studies (Birkmann 2007;Cutter et al. 2003;Rofi et al. 2006). This dimension is relevant because, in the context of disasters, the socioeconomic level explains into how well communities absorb losses and build greater resilience (e.g., the ability to contract insurance and social safety nets) (Cutter et al. . Severance levels also stand out within the country's vulnerability components, and these have a direct impact on both the global and local economies, since it is in the family nucleus where lack of access to resources and access to basic services such as clean water and sewer networks are highly sensitive issues (Birkmann 2007).
Comparing these results at the subnational level with those in Brazil and Portugal, the research team found great similarity in the causes that influence social vulnerability. In Brazil, for example, poverty stands out as the main component of social vulnerability, linked to the aging population, its associated economic and social dependence, low educational levels, the high level of overcrowded households, and illegal and informal work (Loyola et al. 2016). In Portugal, poverty is deeply linked to low levels of education and illiteracy, coupled with high levels of education dropout due to economic conditions (Guillard-Gonçalves et al. 2014).
The socioeconomic component, on the other hand, is directly linked to the heterogeneity of the urban development of the country, which has left rural and agricultural areas relegated to less development, lower levels of education and poor conditions of construction of housing (Guillard-Gonçalves et al. 2014). These components have a major impact on vulnerability due to the inequality gaps they generate and that are linked, especially since smaller and rural communities are unable to rebuild after damage that can leave a natural hazard, as well as to economically assume the loss of communication networks, water, transport and infrastructure, which in turn are linked to access to medical services (Cutter et al. 2003).
This context explains why the component on housing quality and access to basic services is presented as a prominent indicator in Chile's vulnerability, given the level of housing built with precarious materials and poor construction directly associated with the socioeconomic conditions of the population. In the context of disasters, these conditions relate to a lower capacity to deal with natural hazards, exposing both people and material goods to major damage and thus to ongoing vulnerability (Palermo et al. 2013).
In this regard, the latest tsunami vulnerability studies in Chile, for the cases associated with the 2010 (M-8.8) and 2014 earthquakes (M-8.2) have established that the quality of housing is a critical factor influencing the magnitude of the damage (Martínez et al. 2012;Aránguiz et al. 2014;Martínez and Aráguiz, 2016;Martínez et al. 2017). In Dichato, a coastal town, the damage from the 2010 tsunami resulted in the failure of brick masonry infill walls and lightly reinforced concrete columns; these were a major damage factor for the entire town's population (Martínez et al. 2017). The 2010 tsunami in Chile also affected many coastal locations that depended on the extraction of resources from the sea or seasonal tourism. The townspeople had to change their economic activities, and many experienced complex reconstruction processes due to the social cost (Rojas et al. 2014;Martinez 2014). The primary sector of the economy, therefore, is another component that directly increases the levels of vulnerability, since, after disasters, any extractive industry suffers a long recovery period that ends in an increase in unemployment (Loyola et al. 2016).
As for the component of access to basic services, there are large differences between urban and rural areas. While in the models that make up both areas this component is one of the main predictors of vulnerability, when evaluating metropolitan areas independently only this component could be identified in the Coquimbo-La Serena conurbation. This relates to the fact that 200,000 people in rural areas of Chile still receive insufficient water supply, following the minimum supply standards established by the World Health Organization (Donoso et al. 2015), in contrast to the global reality where according to WHO-UNICEF data (2017), there has been a double increase in access to safe drinking water in rural areas, while in urban areas access has remained unchanged (96% of urban populations have access to safe water). These differences originate from the institutional and regulatory frameworks for the supply of drinking water in Chile that are completely different for urban and rural areas. While the service in urban areas is provided mainly by private water companies, the provision of drinking water in rural areas is implemented by the State (Ministry of Public Works) (Molinos-Senante et al. 2019).
It is worth delving into the fact that people in rural areas often struggle to maintain government and commercial operations in normal times, and they do not have the capacity or resources to deal with disasters at the time or after it. In addition, poor conditions of access to basic services in turn respond to the suboptimal conditions that concentrate in these areas, such as low population density and large service areas (Wedgworth et al., 2014in Molinos-Senante et al. 2019. In this way, the results obtained allow us to understand the predictors of the vulnerability of the country in general, but above all they provide an opportunity to bring an approach to local problems that may be triggering the vulnerability at small scales. Examples of this may be found in coastal areas, where a synergy is currently developing on the urbanization of areas with little suitable for development area and high soil values, generating social exclusion, marginalization, and elitism of the territory (Hidalgo and Arenas 2009;Mansilla and Fuenzalida 2010) conditions that have built socio-territorial inequalities promoted by the dispersed growth model (urban sprawl) (Hidalgo and Zunino 2011;Martínez et al. 2020a). This situation may be explained by and relates to much of the vulnerability components that were obtained at the national level, but which manage to be represented and identified spatially, only when analyzed at smaller scales.

Advantages and limitations of SoVI's multi-scale application
Among the advantages of SoVI is the ease of replicating the method, since it does not require complex processes and allows researchers to spatialize the phenomenon . However, despite this positive aspect of the system, one of the great difficulties for its reproduction, in underdeveloped or developing countries like Chile, is to have open access databases, which process data on different spatial scales. This contrasts with the situation of the good experiences obtained in developed countries such as the USA, China, and Portugal that have a large number of up-to-date data sources.
As for the application of SoVI at different scales of analysis, uncertainty has been recognized on small scales (Loyola et al. 2016). However, in Chile, SoVI was applied on a small scale, at the local level in the bay of Cartagena (central Chile), where the socioeconomic context of the population was adequately reflected using census data (Martínez et al. 2020). What was observed in this study made it possible to verify that SoVI, being a comparative method, is highly sensitive to changes in scale.
This research presented great flexibility in the analysis of vulnerability for the different levels evaluated. In this sense, spatial changes of vulnerability are a consequence of the relativity of the method which, being measured in standard deviations, is affected by changes in the study areas as well as by changes in the scales of analysis. Therefore, the vulnerability index of a zone may vary according to the study area being evaluated, for example, a zone may have a lower vulnerability within a region, but an increase in these levels when compared to the country level. Hence, the importance of these models being associated with studies geared to specific levels of action in terms of public policy or for decision-makers.

3
The use of the macro-zonal and regional scale has the advantage of identifying the main components of vulnerability among components with greater explanatory power; however, it also allows, through the components with less variance, to reveal those local characteristics that are hidden or homogenized at higher scales, giving the user or decision-makers knowledge closer to the local reality of the territory. In particular, the macro-zonal scale has the advantage of providing a better adjustment of social analysis to the scale or radius of effects presented by long-range natural hazards such as tsunamis and earthquakes. In this way, this scale helps to evaluate socially those territories that are under permanent hazard, recognizing the components that mark the vulnerability of these zones and their spatial behavior.
In addition, the metropolitan scale presents the advantage of spatially identifying and representing the different levels of vulnerability that cities concentrate, which are usually left as homogeneous areas of low vulnerability when evaluated at scales that consider the entire territory, i.e., urban and rural combined. The same issue holds for models that did not integrate urban areas into the assessment, highlighting the vulnerability characteristics that rural areas concentrate. Therefore, these evaluations as mentioned by Cutter et al. (2016) help policymakers understand the differentiated origin of both zones and their different challenges, giving decision-makers a tool that makes it possible to direct public policy actions directly to the vulnerability index of each area of the territory. In this context, the results obtained from the differentiated evaluation of the territories constitute a favorable first approximation, and it is necessary to carry out further methodological studies that capture the subtleties that mark the differences between and within these territories especially in rural areas.
Regarding the difficulties or limitations encountered using SoVI, these relate to validation especially at local scales, wherein the detail is increased. Another limitation is its difficulty in reflecting the dynamic nature of the vulnerability, i.e., the method does not allow reflecting on space-time changes that mark vulnerability but present as a static value at a given moment (Zhou et al. 2014). However, research has made progress on the issue of space-time in countries such as China and the USA, and more recently in Chile, by conducting three vulnerability analyzes for the years 1992, 2002(Cutter and Finch 2008Zhou et al. 2014;Bronfman et al. 2021). Finally, since the method depends on the data and variables collected, limitations were presented in its implementation because Chile has had unsuccessful experiences in the generation of census data, with the 2012 Census being discarded by the executive, and the most recent of 2017 is very limited in terms of availability and number of variables, which coincide with previous research on vulnerability Zhou et al. 2014).
A relevant aspect to consider for future research is the feasibility of integrating sociological, ideological, of perception or psychological variables, which generate interesting application prospects, especially in the ability to recognize types of reaction to a (Ruan and Hogben 2007). Social vulnerability in the same way responds to characteristics that cannot be extracted just from state sources such as the census, for example. In this sense states sources do not incorporate data is not collected on perception of risk, density of critical infrastructure, or social capital measured in connections and networks within the community (Pelling 2003;Zhou et al. 2014).

The SoVI in territorial planning for disaster risk reduction
The benefits of the SoVI have been mentioned in various articles; however, it is important to highlight its importance as a fundamental tool for disaster risk management, since it can guide the creation of policies and structural measures for a country, as well as specific measures aimed at local areas. This ability to recognize the causes or predictors of vulnerability and its ability to be zoned enable it to be a reliable input for the generation of risk matrices.
On the other hand, research provides a flexible tool for decision-makers, helping them manage the territory from the various scales of planning. In this way, in the Chilean context, it allows the central government to draw general lines for policies considering the national scale, and for governors and regional representatives to consider the macro-zonal or regional scale and, finally, to local authorities, such as mayors, the implementation of local assessments on a communal scale.
From the point of view of territorial planning in Chile, there is broad consensus on the difficulty of linking the risk areas and territorial planning instruments that govern urban land uses, which are usually not very effective in considering criteria of urban and social resilience criteria (Martínez et al. 2017(Martínez et al. , 2019. In this sense, having a multi-scale vulnerability analysis has the advantage of being able to dynamically relate to other risk elements, such as the threat, since it allows the scale of impact of a threat to be associated with one of the scales of risk analysis, vulnerability, providing an input for the different decision-makers. This way, incorporating multi-scale vulnerability analysis for the development of risk models allows decisionmaking and risk management to be linked to planning instruments from the local scale to the central level.
Further, given the high recurrence of disasters caused by natural hazards in the country, reconstruction processes are also called into question in their role of incorporating resilience into urban design and occupation strategies, critical aspects that influence the configuration of new areas of risk (Rahman and Kausel 2013;Martinez, 2014;Martínez et al. 2017 and. Therefore, the incorporation of results such as those exposed and the replication of methods such as SoVI provide the opportunity to include social criteria, which contribute to the decrease in levels of vulnerability and increase the resilience capacities of the population.
In the same way, incorporating vulnerability analyzes into multihazard risk studies is presented as a need that must be addressed in future research, considering that the interaction of multiple hazards can have a greater potential impact on the population, infrastructure, the environment, and the economy of a region.
On the other hand, the occupation of Chile has been accompanied by loss of critical ecosystems and habitats (wetlands), especially in coastal areas, which are also related to environmental vulnerabilities with strong connection to the previous causes, especially socio-environmental inequities (Salinas and Pérez 2011;Rojas et al. 2017;Martínez et al. 2020). The ecological impact from urban growth in Chile and the rest of Latin America is severe (Darrah et al. 2019;Martinez et al., 2020). In line with the 2030 Agenda and the Global Sustainable Development Goals (SDGs), cities are intended to focus on territorial planning and urban management to promote social inclusion, increasing resilience and sustainability over time. This aspect, while not included from a methodological point of view in the schemes to measure resilience, is found in the conceptual frameworks dealing with the relationship between vulnerability, resilience, and adaptive capacity, all relevant aspects to consider in future research ).

Conclusions
Social vulnerability in Chile, considering different spatial scales, is conditioned in greater proportion by socioeconomic aspects such as the concentration of the dependent population, especially in the group of elderly population, the quality of housing, access to basic services, and the population without primary education. The levels of vulnerability, therefore, are not randomly found in an area, but respond to the characteristics of the area and vary according to the scale of analysis.
The application of SoVI in Chile captured those differences from the identification of the components that make up the social vulnerability, distinguishing those that mark the general trend of the country from those that mark the trend at the local level. Moreover, the method presented a first favorable approach to the independent treatment of urban and rural areas, obtaining a more comprehensive representation of the vulnerability in each of these areas, when it was not influenced by the rest of the territory. In this way, the measurement visualizes different vulnerabilities that are concentrated within cities, and not only the homogeneous values of low vulnerability that are obtained when compared to rural areas, while, for rural areas, changes in the composition of the vulnerability could also be observed when individually assessed, mainly shown by low access to basic services.
According to this research, actions on vulnerability reduction should be geared toward measures focused on improving the baseline conditions of the Chilean population identified in the vulnerability composition. For this, structural and indicative measures must be implemented with a criterion of decentralization.
As for economic terms, consideration should be given to the implementation of measures that consider the vulnerability of the Chilean economy, which is based on extractive markets. These sectors are dependent on developed countries, but also on normal operating conditions, which are deeply altered by socio-natural disasters.
SoVI was highly sensitive in spatial analysis for the three scales used, but above all it was sensitive to the variations obtained by independently evaluating rural and urban territories. While this made it difficult in the first instance to identify the scale that is in line with the country's reality, it made it possible to determine that the test of different scales presents an opportunity that should continue to be investigated and provides a tool for decision-makers commensurate with their level of management. In this way, smaller territorial-level management allows a higher level of agreement with reality, while larger-scale management loses this scope but enables managing from a panoramic view and drawing general guidelines for the reduction in vulnerability.
In summary, SoVI is an essential tool for decision-makers from the RRD perspective. First, it allows identifying the components underlying the risk, which are largely drivers of the vulnerability, thus enabling its control and management (Lee 2014). Second, its incorporation at the different levels of management equips authorities to implement regulatory measures through the construction of risk matrices and their inclusion in the zoning of the territory.

Appendix A.
Matrix of principal component analysis of realized models.