1. Introduction
Landslides affect exposed populations worldwide, causing damage, fatalities and injuries. In Portugal, landslides are quite frequent and have been responsible for damage and disruption on roads and buildings. Damage caused by landslides can be reduced with adequate policies and practices. For that, it is mandatory to assess landslide hazard and vulnerability to analyse the landslide risk in order to improve disaster preparedness and prevent losses [
1].
Most of the definitions of vulnerability to natural disasters agree to state that the vulnerability is: (1) multi-dimensional, as it has several facets (e.g., physical, social, economic, environmental, institutional); (2) dynamic, because it changes over time; (3) intrinsic of any community; (4) scale-dependent, given that it can be expressed at different scales from human or household to country resolution; and (5) site-specific, which implies that each study area might need its own approach [
2]. Nevertheless, these dimensions are rarely considered all together in a vulnerability assessment. Indeed, most of the scientific studies which assess the vulnerability to natural hazards examine only one facet of the vulnerability of the elements at risk—most of the time, either the social vulnerability or the physical vulnerability.
Social vulnerability characterises the inequalities which define the predisposition or susceptibility of social groups in the context of a disaster [
3,
4]. The concept of social vulnerability is complex [
5]; indeed, it is itself a multi-faceted entity and authors use this term with different meanings [
6], some of them are presented in the
Table 1.
Moreover, some approaches to assess social vulnerability are based on the assessment of intangible losses, whereas other approaches are based on the underlying socioeconomic factors that are responsible for vulnerability in a society [
5]. Social vulnerability is commonly measured by indexes which are based on a set of socioeconomic indicators (e.g., age, gender, disability) which can either be weighted (e.g., by expert judgement, analytic hierarchy, Principal Component Analysis (PCA), factor analysis or multiple regression models) or not (if all indicators are assumed to have an equal significance). The indicators are then combined, often by an additive combination when the factors are independent or by a multiplicative combination when the utility of one factor depends on another factor, or even by an association of both additive and multiplicative combinations [
12]. The obtained social vulnerability can then be combined with the hazard assessment to evaluate risk, e.g., [
13,
14].
Physical vulnerability is a functional relationship between process magnitude, the impacts on structural elements at risk and exposed values [
15]. Physical vulnerability of buildings is defined by the expected degree of loss resulting from the impact of a certain event. Its assessment requires the evaluation of different parameters and factors such as type of element at risk, resistance, and implemented protective measures (i.e., local structural protection) [
15]. In physical geography and in engineering geology, most of the studies consider the vulnerability to a hazardous event of a given magnitude as being the degree of loss of elements at risk expressed in a scale ranging from 0 (no loss) to 1 (total loss), e.g., [
16,
17,
18,
19,
20]. From this definition emerged a wide range of vulnerability assessment models, each study addressing vulnerability in a different way [
21]. These different models partly depend on the different scales of the vulnerability assessment [
22], and on the purpose for which the vulnerability assessment method was made. Some models are based on the collection and analysis of the registered damage in recent historic events, as it is usually done for relatively frequent events such as floods or earthquakes, which allows the construction of vulnerability curves by correlating the magnitude of the event and the degree of damage. When there is not enough available information regarding the caused damage, the physical vulnerability can be measured through expert opinions. For example, landslide experts can give their opinion on the potential damage that can be caused to a specific type of element at risk by a landslide that has a certain magnitude, and the vulnerability assessment would be based on their answers. This method is time-consuming because numerous experts must be interviewed to obtain results with a lower subjectivity.
In literature, few studies have the ambition to assess simultaneously several facets of the vulnerability to landslides. Léone was one of the pioneers as he proposed a method to assess the corporal (for people), structural (for material goods) and functional (for activities) facets of vulnerability to landslides and presented them in matrices [
23]. More recently, Shrestha proposed a method to assess the socioeconomic and physical vulnerabilities to landslides and floods in Putalibazaar, Nepal [
24]. In its study, total vulnerability was assessed by combining the physical and socioeconomic vulnerabilities which are functions of the landslide magnitude, the physical exposure indicators (number of households, total population, agricultural land area and road length), and the capacity of adaptation indicators (e.g., accessibility, health, communication). Perception of disaster by the population and by governmental and nongovernmental organisations was also assessed by questionnaires. Alexander proposed a method based on the vulnerability of buildings, structures, human lives and socioeconomic activities, which can be used at three levels: single asset; summed assets; and generalised assets [
25]. Puissant and co-authors created an index to assess the potential damage which can be caused by landslides [
26]. The index is called Potential Damage Index (DPI), and was elaborated to assess the potential physical injuries, structural and functional damage, and socioeconomic effects. This index has been used as a base in other studies: for example, Carlier and co-authors used it to assess the direct and indirect consequences of landslides in the Upper Guil catchment, an area of the PACA region in France [
27]. Recently, Murillo-Garcia and co-authors proposed a method to assess the vulnerability to landslides by combining the exposure level of the population and the infrastructures, the sensitivity of the population and its lack of resilience [
28]. The exposure is calculated from data regarding population and hazard, the sensitivity of the population is composed by social indicators (e.g., population density, youngest and oldest population, female population, indigenous population), and the lack of resilience is composed by indicators dealing with incomes, economically active population, health insurance and road characteristics. The combination of the different facets of the vulnerability considered in this study is interesting, though the physical vulnerability of the buildings is not assessed, and the possible application of the vulnerability assessment is not clearly defined.
The present study aims to combine the social vulnerability of residents and the physical vulnerability of buildings in order to analyse landslide risk in a municipality located in the Lisbon region, Portugal. It was developed on the continuation of previous studies in the Loures municipality, the first one regarding the landslide susceptibility assessment [
29] and the second devoted to the physical vulnerability of the buildings to landslides [
30]. The paper is organised as follows: First are described the data and methods used: (1) to assess the social vulnerability of the Loures municipality; (2) to assess the physical vulnerability of the buildings of Loures; (3) to combine physical and social vulnerability to landslides; and (4) to assess the landslide risk. Then, the presentation of the results is made as follows: (1) social vulnerability map; (2) five physical vulnerability maps (one per each landslide magnitude); (3) five combined vulnerability maps; (4) one table represents the landslide risk analysis of the resident population and a second table represents the landslide risk analysis for the buildings. Finally, a discussion about the results precedes some concluding remarks.
4. Discussion
In comparison to the social vulnerability values resulting from the study of Murillo-Garcia and co-authors [
28], the social vulnerability which was assessed at the BGRI scale in the present study is on average lower. This can be explained by the fact that Murillo-Garcia and co-authors [
28] considered the exposure of the population as a part of the vulnerability. Moreover, the original formulas were adapted, and the indicators are not always the same. Finally, the context is different, essentially by the absence of indigenous population in the Loures context.
Regarding the economic value of the buildings (
Table 7), the share of the economic values tends to follow the trends of the share of resident population shown in
Table 6. This is mainly because the distribution of the residential buildings (which were used in
Table 6) is fairly homogeneous among the total built environment (which was used in
Table 7).
The method used in the present study to assess the social vulnerability at the BGRI scale has the advantage to be applicable to any Portuguese municipality and it is not time consuming because the data used is provided by the INE and the Portuguese Tax Services and are available at the BGRI scale for the entire country. Moreover, the risk analysis presented here for 3-metre-deep landslides can be reproduced for any of the other landslide magnitudes for which the physical vulnerability was assessed. Finally, the risk analyses are summarised in tables which enable a quick interpretation of the findings.
However, the present study has several limitations; one of them is that the results were not validated, as it occurs in most of the geospatial modelling studies which focus on the construction, mapping and analysis of quantitative factors [
45]. Consequences data record after landslide occurrences in the Loures municipality is limited; the data are too few to build a reliable validation model of vulnerability and risk findings. Moreover, indicators relative to the health, the coping capacity and the risk perception are also important for a complete social vulnerability assessment [
45], and these indicators are not available in Portugal at the BGRI scale. Another limitation is that the model of social vulnerability provides results for the resident population, which must be close to the reality during the night-time; but during the daytime the population is not the same and the INE does not provide detailed data (e.g., age, employment) about the population which is present in the Loures municipality during the daytime. An estimate could be found by using a spatio-temporal model of the population distribution which uses population mobility statistics, as the one of Freire and co-authors [
46]. The population which is present in the Loures municipality during the daytime could be assessed in a future work. Regarding the susceptibility, the susceptibility map was prepared for landslides deeper than 1.5 m, while the risk analysis is limited to landslides with a depth of about 3 m. Therefore, there is a discrepancy between the susceptibility and the risk analysis, possibly leading to errors. However, the error is small because the landslide frequency strongly decreases with landslide size, which means that distribution of landslide susceptibility does not depend substantially on landslides deeper than 3 m. The last limitation regards the issues of the presentation of the results: first, the vulnerability classification is relative, which makes difficult the comparison of the results of this municipality with other municipality vulnerability which would be assessed with the same method; second, the combined vulnerability maps show the location of the vulnerable BGRI subsections without providing information about the number of residents which are in each subsection, and the stakeholders may have to overlap the population density on the combined vulnerability maps before making decisions to reduce landslide risk in the study area. Finally, the repartition of the residents per buildings was not known and was therefore estimated using a dasymetric distribution, which approximates the reality adding therefore a supplementary uncertainty. The stakeholders should take into consideration the propagation of uncertainty in the different steps of the methodology, which makes the results quite uncertain.
5. Conclusions
The major aim of the present study was to develop a method of vulnerability assessment including the social and the physical facets of the vulnerability, and which is applicable at the municipal scale in Portugal. The second aim was to use the vulnerability assessment to support the landslide risk analysis. Finally, the developed method should be useful to spatial planning, civil protection and insurance stakeholders.
Many methodologies developed for vulnerability assessment are difficult to implement because of the unavailability of data. In addition, the achievement of relevant data is frequently time-consuming as it implies extensive field work. The present study aimed to be easily and readily applicable in others Portuguese municipalities for which an inventory of the past landslides has been listed.
A social vulnerability assessment was made at the BGRI scale in the present study. Results coming from a previous study [
30] were used for the physical vulnerability assessment of the buildings. The combination of the two facets of the vulnerability was made by crossing the classes of the vulnerability assessment with the classes of the physical vulnerability at the BGRI scale in a matrix. The landslide risk was finally analysed considering the combined vulnerability provided by the combination of the physical vulnerability and the social vulnerability assessment, the landslide susceptibility, the exposure of the population and the value of the elements at risk. The 3-metre-deep landslide magnitude scenario was chosen because the risk is maximum in this case due to the relatively high frequency of such landslides and their substantial potential damage.
According to the susceptibility model (very high and high susceptibility classes), 75% of the future landslides should occur in 16.58% of the study area. This area gathers 0.9% of the resident population (the 0.47% of the Loures population who live in this area are highly vulnerable to 3-metre-deep landslides and the remaining 0.43% have a medium vulnerability) and 0.8% of the buildings, the economic value of which being EUR 146,170,000.
In terms of application, civil protection can be interested by the combined vulnerability developed in this study. Indeed, this vulnerability assessment provides the location of the more vulnerable population at a large scale crossed with the buildings which have a high physical vulnerability to different landslide magnitudes.