1 Introduction

The interaction of natural phenomena with humans does not affect all societies in the same way. The amount of social, economic, material and environmental losses is proportional to the vulnerability of lands and communities (Vargas 2002). Since the middle of the twentieth century, when the concept of vulnerability was introduced in the risk analysis terminology, the definition of vulnerability has evolved. These changes have been based on (i) the characteristics of the community where it has been applied and (ii) the moment when it has been defined. Therefore, the new meanings of vulnerability correspond to the different ways of conceptualizing and quantifying it. In this way, each new definition has added new content (Weichselgartner 2001; Fuchs et al. 2007).

This means that when a new study related to this subject begins, the initial step is to define the term vulnerability. In the present study, the definition is based on the precedent characterization performed by UNDRO (1980), Vargas (2002) and Kumpulainen (2006). In this way, vulnerability has been considered as the susceptibility of an element or group of elements under risk that is the result of a natural or anthropogenic hazard of a determined magnitude. It is important to note that this vulnerability is measured on a scale ranging between 0 and 1 and is estimated by two components. The first is the degree of exposure faced by social, physical, economic and environmental factors. The second is the capacity for resistance against the phenomena. This resistance includes the capacity of protection and immediate reaction and the capacity of recuperation or resilience (Tascón-González 2017). This last term is defined as “the capacity of a system, community or society exposed to a threat to resist, absorb, adapt and recover from its effects in a timely and effective manner, which includes the preservation and restoration of its basic structures and functions” (UNISDR 2005). Therefore, this definition considers resilience as a factor of the vulnerability (Berkes 2007), in contrast to those authors who consider the resilience as an antonym to vulnerability (Adger 2000). Then, it is important to know how to identify, manage and reduce disasters, but most of them are unavoidable and the only way out to diminish the effects of these phenomena is to increase the resilience of the society.

The vulnerability of societies has increased in recent decades (Lundgren and Jonsson 2012). The main reason is the population growth and the extension of the infrastructure network over traditional flood areas that previously have not been human-occupied. Consequently, this incursion has meant an increase in exposure of the societies. The communities and their belongings have become weaker than before and may be hit more often by extreme events. At the same time, the population is changing their risk perception, and they feel that the number of extreme events is rising. This new awareness stems from the development of new technologies and the free circulation of information. Both allow the detection and knowledge of any natural event that occurs at any point of the globe and give the impression that the frequency of extreme events is increasing (Velev and Zlateva 2012; Kryvasheyeu et al. 2016; Sim et al. 2018; Wang and Ye 2018). This means that not always there is an increment in the number of extreme events but in the number of people and beings exposed to them and in the spreading of information.

Vulnerability is a multidimensional concept. For this reason, estimating the vulnerability of a hazard may be analysed from different points of view, mainly economic, social, environmental, physical, administrative or political. In some societies, the main view could even be religious (Gaillard and Texier 2010; Gianisa and Le De 2018). To address all these interpretations, the best method is to evaluate each vulnerability type by a group of factors. The primary characteristic of these factors is that they are observable and identifiable before an event occurs. This situation leads the society to study the factor through the prevention, prediction and reduction of the vulnerability to counteract its negative influence. Possible factors in social vulnerability include age, foreign population and disabled residents (Kappes et al. 2012; Dunning and Durden 2013; Terti et al. 2015; Aksha et al. 2019).

At the same time, factors are informed by different indicators (Balica et al. 2009). An indicator is defined as a variable that captures the factor’s condition or part of it in a determined place and time. In general, indicators are statistical data that synthesize the primary information that the factors describe. Due to the difficulty of achieving this goal with just one indicator, in most cases, two or more indicators are used to define one factor. Therefore, the best indicator is the variable that summarizes or simplifies basic information, makes visible a phenomenon of interest and quantifies, and measures and communicates relevant information (Gallopín 1997). Following the abovementioned social factors, as an example, a possible indicator of foreign population would be the percentage of foreigners or the number of tourists per each 10,000 inhabitants.

So far, the literature shows that social vulnerability is the most common approach treated in risk analysis (Scheur et al. 2011; Koks et al. 2014; De Loyola et al. 2016; Noradika and Lee 2017; Tascón-González et al. 2020). Nevertheless, interest in physical vulnerability is growing. It has a long tradition in seismic risk analysis, and this approach is spreading over other hazards such as floods (Creach et al. 2016; Mazzorana et al. 2014; Bigi et al. 2021). Physical vulnerability focuses on the response of buildings and infrastructures or basic elements when a hazard occurs (Papathoma-Köhle et al. 2022). Consequently, the main analysed factors are building age, building material, building height, and the existence of infrastructure networks related to telecommunication, energy and transportation (Bisbal et al. 2006; Holand et al. 2011; FEMA 2011; Miranda and Ferreira 2019; Malgwi et al. 2020; Usman et al. 2021; Malakar and Rai 2022). These last ones are the set of services over which the productive structure of a society is based. They facilitate services, consumption activities and social relationships, which allow regional development and quality of life.

The implementation of this type of analysis demands large amounts of data. Some of them belong to citizens, so field enquiries are needed. This can be done when the number of buildings is low, but it can only be afforded by companies (such insurance companies) or Governmental Administration when the study area is a town. Other type of data is stored in governmental datasets. Since 2009, there has been an important movement among Governmental Administrations to give open access to their data (Abdelrahman 2021). This has led to the creation of the International Open Data Charter (ODC 2015) in which 170 countries agreed to several principles as a frame of norms for how to publish governmental and public data. These principles could be summarised as follows:

  1. (1)

    Open by default: it remarks that Governments should justify when data are not opened in juxtaposition to the default closed data that they have offered until now.

  2. (2)

    Timely and comprehensive: open data only is valuable when they are published on time and in a comprehensive way.

  3. (3)

    Accessible and usable: data should be free of charge and easy to find as well as it should follow a well-known file format.

  4. (4)

    Comparable and interoperable: data should be stored following some standards and with metadata that allow comparing among different sources.

All of them should encourage transparency to improve government performance. At the same time, it is expected that the use of open data in enterprises promotes their innovation and economic development. This new scenario poses challenges to researchers to analyse the fitness of the large open government data in their studies, such as the physical vulnerability of a natural hazard.

Therefore, in this paper, we propose a method to map the physical vulnerability against floods and its implementation in a middle-sized town, Ponferrada (Spain). This city is located 10 km downstream of a dam. This proximity to the infrastructure means that the citizens of Ponferrada could suffer significant damage as a result of a potential dam break.

In contrast to other studies related to physical vulnerability, our analysis will not focus on the damage suffered by the infrastructures when the flood has already occurred, but it will estimate the real vulnerability and identify the weakest infrastructures against a flood before the event happens (Ezell 2007). The objective is to facilitate land planning and to prepare a resilient society.

Until now, most of the vulnerability results have been presented at the municipal scale (Weichselgartner 2002; Ruiz 2011; Aroca et al. 2017) or analysing buildings or infrastructures, but not both at the same time (Erena and Worku 2019; Singh et al. 2019; Cheng et al. 2022). Thus, this document will analyse the public data sources of infrastructure and buildings to be used at an urban parcel scale and determine the vulnerability of a large area such as the municipality of Ponferrada.

2 Study area

The study area, the municipality of Ponferrada, is located in northwest Spain (Fig. 1). The total vicinity has approximately 67,000 inhabitants, and it is the head of the region called El Bierzo (in the province of León, northwest of Spain). To understand the proposed analytical method, it is important to note that Ponferrada is divided into 6 districts and 45 census sections (Fig. 2) because not all of the available data used in this study were at the urban parcel scale.

Fig. 1
figure 1

Location of the study area with the simulated water sheet and the districts of the town (Image source: PNOA 2020)

Full size image
Fig. 2
figure 2

Census sections in Ponferrada

Full size image

The Sil River is the main water course in the area and meets two tributaries in the middle of the city. The overall drainage network has flooded the town several times. This is partly the reason for building three dams upstream of the town, together with agricultural water demands. The first dam is in the Boeza tributary with 2 hm3 of storage capacity, and it is out of the scope of the present study. The other two dams, Bárcena (341.5 hm3) and Fuente del Azufre (4 hm3), are in the Sil River. The present contribution focuses on evaluating the physical vulnerability of the town in the potential scenario that both dams break when they are at their maximum storage capacity. This scenario could have different origins. The first one would be for hydrological reasons. Bárcena is a loose material dam, and this type of dam presents a higher probability of breaking due to erosion of the protective coating when the crest is overtopped (Comité Nacional Español de Grandes Presas 2016). Another possible origin of the dam break would be structural failure. For this reason, the hydrological administration performs regular controls. Despite this oversight, some recent dam failures, such as those at Derna (Libya) and Braskereidfoss (Norway) in 2023, have raised the possibility that ageing is affecting dams beyond their economic and physical design life (Ho et al. 2017). One additional origin would be seismic activity (Ministerio de Medio Ambiente 2001). In this area, several faults cross along the Bárcena dam, associated with Cenozoic thrusts (Martín-González and Heredia 2011a, 2011b), but they are categorized as low hazard according to the instrumental and historical records. These faults are associated with other structural and geomorphological features that played a role in the geomorphological development of the drainage pattern and the landscape configuration during Quaternary times (Mínguez 2015). Finally, there is the possibility of an anthropogenic origin, such as an act of terrorism.

The choice of this study area was reinforced by its geomorphological characteristics. As Fig. 1 shows, the town is located in the district 02, at the outlet of a confined reach of the Sil River, where the channel changes from a steep and narrow valley traversing more resistant igneous rocks, towards a wide area with friable sedimentary lithologies with a low slope. Without the dam, these areas are considered highly susceptible to floods. The geomorphological change from a narrow to a wide landscape lead to an expansion of the water sheet when floods occur. At the same time, the change in slope facilitates the runoff slowdown, which impedes water evacuation and increases flood damage. Therefore, considering the potential dam break and the location of the study area, the town presents a high flood hazard.

However, not all of the town is at risk. Definition of the flood sheet extension was based on previous studies focused on the hydraulic modelling of the collapse of both dams (CHMiño-Sil 2012). Their results show an area inside the boundaries of the flood sheet where 45,338 inhabitants of the total 67,367 municipal inhabitants live (Tascón-González 2017). This affected area presents a wide range of modelled water height values. They move from the 85 m at the outlet of the confined river to the 25 m in the river downtown and the 15 m in the river at the south of the city.

3 Method and data sources

The methodological approach of using indicators to define the different factors that are involved in physical vulnerability allows comparison of the different building scenarios inside of a town or among towns. As in all types of vulnerability, the analysed factors may be classified into exposure and resistance. The factors associated with exposure reflect how an ecosystem is submitted to the effects of a natural event and its duration, in this case a flood. Related to physical vulnerability, this ecosystem refers to infrastructure characteristics. On the other hand, the resistance factors show the capability of the society to protect itself, the capability of its immediate reaction and the capability to recover the former functioning of the society. These last factors may incorporate indicators related to land planning, such as the existence of floodplains allocated to alleviate peak discharge or the construction of artificial levees, river dikes or dams. At the end, both groups of factors relate to each other by Eq. 1 based on the method of Balica (2012):

$$PVI = \left( {EL/R} \right)/10$$
(1)

where PVI is the physical vulnerability index, EL is the exposure level and R is the resistance. The final division by 10 is performed to achieve a PVI value inside the interval 0–1.

In the scientific literature it is observed that the most studied indicators are those referring to construction materials (Behanzin et al 2015; Krellenberg and Welz 2017; Fatemi et al. 2020), condition of the building (Thouret el al. 2014; Carlier et al. 2018; Xing et al. 2023), age building (Fernandez et al. 2016; Sadeghi-Pouya et al. 2017; Usman et al. 2021) and number of floors (Blanco-Vogt and Schanze 2014; Papathoma-Köhle et al. 2019; Leal et al. 2021). It should be noted that the determination of the construction materials indicator is carried out in study areas with a low number of buildings where it is relatively easy to carry out field work. This is performed through surveys where the damage suffered by the building is analyzed, depending on the material, once the flood has occurred. Unluckily, there is not a database with such information in the study area and the number of buildings is too large to get this type of data through a survey. Therefore, this paper does not include this indicator, although to get this type of data is strongly recommended for smaller areas.

The present method proposes an analysis divided into the following four steps:

  1. (1)

    Calculation of the different indicators. For the exposure level, the studied indicators are building age, percentage of the building affected by the flood based on the water height and percentage of harmed roads. For the resistance level, the researched indicators are the number of construction defences, the number of floodplains allocated to reduce downstream floods and the existence of land planning or emergency plans.

  2. (2)

    Normalization of each indicator. Each indicator has been converted to a scale of 0–1 to be able to combine and compare all of them without considering their different nature. In the case of exposure indicators, the value 1 means the highest vulnerability or the weakest infrastructure or building. The resistance indicators follow the inverse correlation. Two mathematical methods are used to normalize the data:

  3. •

    The correlation of percentages, by dividing the real value by 100. This method is applied in those indicators where data is in percentage.

  4. •

    The increasing linear function with a positive slope, which is applied to the rest of the indicators since their values are proportional to the degree of exposure or resistance. The normalized values will be calculated following the Eq. 2:

    $${\text{y = }}\left( {\left( {{\text{x }} - {\text{ Min}}} \right)} \right)/\left( {\left( {{\text{Max }} - {\text{ Min}}} \right)} \right)$$
    (2)

where y is the normalized value, x is the indicator value and min and max are the minimum and maximum potential limits of the indicator scale: a value lower or equal than the minimum will correspond to 0 and a value equal or higher than the maximum will correspond to 1.

  1. (3)

    Estimation of physical vulnerability exposure and resistance based on indicator weights. These weight values were the result of an Analytic Hierarchic Process (AHP) analysis carried out by Tascón-González (2017), in which she sent a survey to an expert panel comprising researchers, engineers and civil protection agents. Finally, she worked with 15 surveys that showed high consistency in the answers.

  2. (4)

    Application of Eq. 1 to calculate the physical vulnerability index.

In the following subsections, the first two steps are described for each chosen indicator. The two last steps are described together at the end of the section (Fig. 3).

Fig. 3
figure 3

Flow chart of the proposed method

Full size image

3.1 Exposure indicator 1: building age

Buildings have evolved throughout history because of lifestyle changes and technological development. Modern buildings have adapted to the complex combination of social, cultural and productive dynamics. Therefore, there are significant changes between the old and the more recent buildings, with the oldest buildings being more vulnerable.

To estimate the real value of this indicator, the required input data came from cadastre or land registry administration. In Spain, this administration maintains an inventory of physical details pertaining to cadastral parcels, including their surfaces, locations, uses, shapes, cartographic representations, as well as information on the type and quality of constructions, among other factors (Cadastral Electronic Site 2023). These records adhere to the guidelines of the European Directive INSPIRE/Directive 2007/2/EC, Infrastructure for Spatial Information in Europe. The data is organized into three distinct datasets (Directorate General for Cadastre 2016). The first dataset focuses on cadastral parcels, which represent individual areas of land with uniform property rights and unique ownership. It contains the geometric information that outlines the boundaries of cadastral parcels in both rural and urban areas. In this study, the urban polygons are also referred to as urban parcels indistinctively. The spatial resolution of the urban map is 1:1000 or higher. The second dataset stores the addresses and provides the location of properties based on identifiers such as street names and numbers, cities, and more. Therefore, this dataset primarily consists of attribute data rather than geographical information. Finally, the third dataset is dedicated to buildings and comprises geographical data that represent the geometry of the footprint of the buildings, along with a set of attributes stored in tabular format. All this information is currently available as open-source data.

Consequently, the vector map of the urban parcels and the alphanumerical table related to the building age were free downloaded to estimate the building age indicator. There was no problem to join both data in a Geographical Information System (GIS) because there was a unique correspondence between the buildings and the cadastre parcels, so the real value of building age was easily mapped (Fig. 4).

Fig. 4
figure 4

Map of building construction year

Full size image

The second step, the normalization, included data related to building conditions for residential use: ruinous, bad, deficient and good. The proposed method related each term with a different normalization value (Table 1).

Table 1 Normalized values for building conditions for residential use
Full size table

The first problem arising was that these data were distributed by districts, not by cadastre parcels. Hence, there were data on the percentage of buildings in a district that were in ruinous, bad, deficient or good condition for each year of construction (Table 2). The normalization of these data was achieved after calculating a weight mean of the Table 1 normalized values for each district and year (Table 2). Based on the year of construction of each urban parcel, these normalized values were assigned to each cadastral unit This means that all parcels that were built during the same year in a district presented the same normalized value (Fig. 5).

Table 2 Example of the Spanish national statistical institute database related to building conditions
Full size table
Fig. 5
figure 5

Normalized value map of building conditions

Full size image

The second problem was that there are only data about building conditions in residential parcels. The other areas aimed at commercial or industrial use did not show this type of data. The authors assumed that the situation of residential buildings was common to the rest of the buildings, so the same Table 2 was used to classify commercial and industrial constructions.

3.2 Exposure indicator 2: building percentage covered by water

The objective of this indicator is to estimate the percentage of the building affected by the water depth. This means that at the same spatial coordinates, the lower the building height, the higher the physical vulnerability. To estimate the real value of this indicator, we divided the work into three parts. The first was to define the height of the buildings, differentiating those floors that were above and under the ground. The second was to estimate the water depth reached at each building. Finally, the percentage of the total building covered by the flood was calculated.

Again, part of the input data was acquired from the cadastre administration: the same polygon map with all the urban parcels that was used for building age indicator and another alphanumerical table. In this case, a digital elevation model (5 m × 5 m), a polygon layer with the modelled water surface covered by the flood and a layer with 53 points with the modelled water depth value were also used.

The data needed to calculate the number of floors belonging to each building—above and under the ground—were available in the alpha numerical table. Each floor above the ground appeared in the alphanumerical table as positive values (1 for the first floor, 2 for the second, etc.). This means that several registers of the table corresponded to a unique building, one register per floor. Therefore, a new table was created with two fields: one corresponding to the urban parcel number (named as Refcat) and the other corresponding to the maximum value of the floors of each urban parcel. Something similar happened with the floors under the ground, but they appeared as negative values. The process was repeated, and a new field with the minimum value per parcel polygon was added to the recently created table. According to Ponferrada local urban norms, each floor has a 2.5 m vertical distance between the ceiling and the ground. A half metre was added considering the construction materials needed between the two consecutive floors. As a result, the height of the building was calculated by multiplying the number of floors above and under the ground by 3 m (Table 3, Fig. 6).

Table 3 Example of some records in the new table with heights above and under the ground per building. The Refcat field corresponds to the urban parcel number
Full size table
Fig. 6
figure 6

Map of building heights (m)

Full size image

The next step was the estimation of the water flood depth in each building. In this case, as in many studies, there was a map with the flood extent, but there was not a continuum map with the water depth data (de Moel et al. 2009). Only 53 points with estimated water depth and spread over the town were collected from the previous flood modelling study based on a potential dam break (CHMiño-Sil 2012). This number was integrated into the present work with those points corresponding to the flood extent perimeter modelled by the Water Administration. These last points conformed to the water limit, so it was considered that they had a minimum water depth (0.01 m). The values of water height at all the points were added to the terrain altitude extracted from the actual DEM, and afterwards, these water heights above sea level were interpolated by applying the natural neighbour kriging algorithm. The results showed a raster map with water altitudes above sea level. The last step consisted of subtracting the altitude values from the DEM. At this point, the distributed value of water height above the ground in the entire town was achieved (Fig. 7).

Fig. 7
figure 7

Map of interpolated water depth (m)

Full size image

The goal of the following stage was to estimate the percentage of each building that was covered by the flood water. As an initial statement, it was considered that if a building with floors above ground was affected by the flood, the water also affected all of its underground floors. Consequently, the first step was to assign a water depth at each building. Based on statistical overlay analysis, the mean water height was assigned to each parcel of the cadastral vector map. Hence, this map had two types of data in its attribute table: the height of the building above the ground and the water depth. Next, a new field was created with the result of the building height minus the mean water height. If the result was negative or equal to zero, all of the building was affected, including those floors beneath the ground surface. If the result was positive, this meant that only part of the building was affected, although the underground floors were always considered affected. Based on these data, it was possible to estimate the percentage of the building covered by the flood water (Eq. 3) and to represent it through a map.

$$\% \;building = 1 - \frac{{H_{f} }}{{H_{b} }}*100$$
(3)

where Hf is the number of metres of the building free of flood water and Hb is the overall building height (including the underground floors).

To normalize this variable and to achieve a map with 0–1 values, it was only necessary to divide the pixel value by 100 (Fig. 8).

Fig. 8
figure 8

Normalized value map of the building percentage covered by water

Full size image

3.3 Exposure indicator 3: percentage of harmed roads

Physical vulnerability considers not only the building factor but also the infrastructure factor as the representation of the kind of services that allow regional development. This means that in flood areas, road safety implies flood risk management (Benedetto and Chiavari 2010). Consequently, roads built in flood areas are more vulnerable to hazard impacts. For this reason, the indicator chosen to represent the transportation network vulnerability was the potential percentage of roads harmed by the flood.

This indicator distinguishes the number of road kilometres that are covered by the flood water surface compared to the number of road kilometres built in the study area. To estimate its real value, the following data sources have been consulted: the Spanish Works Ministry (Ministerio de Fomento 2017) and the image databases of World Imagery (2017) and World Street Map (2017). As can be inferred, the base unit of this analysis was the municipality, not the urban parcels. For the final study, the data set of the percentage of harmed roads at the municipal level has been assigned at the urban parcel level, since the value of this indicator does not depend on the unit of analysis.

The easiest way to normalize this indicator represented by percentage is dividing the real value by 100, where values of 1 and 0 indicate the maximum and the minimum vulnerability, respectively (Tascón-González et al. 2020).

Finally, other infrastructure indicators were researched, such as telephone or electrical network, but none of them had free data available. For this reason, they were rejected from being incorporated in the present study. However, if available, it is strongly recommended to include them.

3.4 Resistance indicators

None resistance indicators were present in Ponferrada. There were no resistance measures that can minimize the effects of the flood wave if the dam breaks, especially if the dam size and proximity to the town are considered. Only previous land planning might assume flood resistance for buildings and infrastructures, and it did not exist in Ponferrada at the moment of the present study. For this reason, a minimum value of 0.1 was assigned as the normalized value, since division with value 0 would not be possible.

3.5 Physical vulnerability index

To estimate the PVI, the first step was to calculate the exposure level. For this purpose, different weights were assigned to each indicator (Table 4) following the AHP analysis (Tascón-González 2017).

Table 4 Value assignment to indicators according to the AHP method (source: Tascón-González 2017)
Full size table

The AHP method first considers a set of evaluation indicators and generates a weight for each of them according to the pairwise comparisons made by a panel of experts. As a result, the method supplies a pairwise comparison matrix, where the higher the weight, the more important the corresponding indicator is. In addition, the AHP method allows checking the consistency of the evaluations to reduce bias in the decision-making process. This is essential when many pairwise comparisons are carried out, as some inconsistencies may arise.

To develop the method, the respondent is asked to:

  1. (1)

    Order the indicators according to their importance (1 the most important) giving the same value to those that they consider equal.

  2. (2)

    Provide their opinion regarding the weight of the selected indicators, comparing them two by two in a table and assigning to each relationship the values detailed in Table 4.

where j and k are indicators and ajk represents the importance of the jth indicator relative to the kth indicatorTo facilitate their assignment, an example is presented below in which the following indicators are compared (Table 5):

  • J is the percentage of homes affected by the sheet of water considering the height at which they are located

  • kis the percentage of old and poorly maintained homes affected by the sheet of water

  • h is the percentage of roads affected by the sheet of water

Table 5 Example of values assigned by an expert applying the AHP method
Full size table

Based on the values assigned by all experts, it is possible to assign weights to each physical vulnerability indicator. The final weights are shown in Table 6.

Table 6 Indicator weights proposed to estimate the exposure level (Tascón-González 2017)
Full size table

Subsequently, the three indicator raster maps were mathematically overlaid with their corresponding weights and the result generated the flood exposure map of the physical vulnerability index.

The following step was to calculate the physical vulnerability index map. Based on Eq. 1, the exposure level should be divided by the resistance value, 0.1, after considering the low resistance that Ponferrada now shows concerning flood hazards without any planning. The division of the exposure value map by such a low resistance value map assumed that the PVI values were higher than the 0–1 range. Consequently, the results were divided by ten to normalize the PVI. For this reason, with this method, where the resistance value is almost null, the PVI map presents the same values as the exposure map.

4 The physical vulnerability map

The application of the previous method allowed the generation of the physical vulnerability map of Ponferrada where the closeness to the infrastructure breach has appeared as an aggravation of risk in several cases (Creach et al. 2016). At the same time, a distributed map of the PVI (Fig. 9) has been achieved. This index is grouped into 5 classes: very low, low, medium, high and very high considering the relationship of values between the degree of exposure and resistance (Müller et al. 2011; Baky et al. 2020). The analysis of the vulnerability data by districts shows their internal variability. The minimum value in all of the districts is 0.059. This value partly corresponds to those urban parcels without buildings, where the PVI value is due to the harmed road indicator of the overall municipality. For the following analysis, these parcels without buildings are out of scope. The rest of the parcels that show this low vulnerability value correspond with buildings located close to the water sheet perimeter, where water depth is so low that buildings are not affected by the flood.

Fig. 9
figure 9

Map of the physical vulnerability index

Full size image

Overall, of the 6441 studied urban parcels in Ponferrada, 261 parcels (4%) showed values higher than 0.5, which were considered very highly vulnerable. In addition, 1203 urban parcels (18.7%) were included in the interval 0.4–0.5, which was considered highly vulnerable. The rest of the parcels showed low values. The mean PVI value overall in the study area is 0.26, which is considered a medium–low value in terms of vulnerability.

From a district point of view, district number 4 presents the lowest PVI value (0.188). This district only has a census section affected by the flood, so its PVI values are quite homogenous. The maximum PVI values reached inside all of the districts are over 0.45 (Table 7). The only exception is district 4, where the maximum value is 0.378. This low minimum value matches the lowest average PVI value of the district. Another district with homogeneous PVI values is number 2. In this case, it is in the right margin of the river where the floodwater reaches the greatest depth in the census section number 214 (Table 8 and Fig. 10).

Table 7 Basic statistics of PVI values by district
Full size table
Table 8 Basic statistics of PVI values by census section
Full size table
Fig. 10
figure 10

Mean PVI values and their standard deviation distributed by census section

Full size image

The rest of the districts show more heterogeneity in PVI values. District 1 corresponds to the old town area, partly located on a hill. These elevated areas may be affected by the flood but with low water depths. Nevertheless, other urban parcels in this district are at lower altitudes. This is the reason why some PVI values are higher than 0.5. Therefore, District 1 displays a wide range of PVI values. District 5 presents low PVI values in general. It is a district located close to the perimeter of the floodwater sheet, with relatively new buildings. The exception is Section 504, where water reaches a high-water depth together with bad building conditions. Finally, District 3 developed along the river with a low concentration of urban areas. It shows the highest internal PVI heterogeneity values because it is an extended district with some houses close to the river and others close to the perimeter of the floodwater sheet. The highest PVI values are reached in census Section 307, where there are many one-floor buildings in bad condition.

All these PVI values are conditioned by the distribution of the different indicators. The most influential indicator in Ponferrada is the building percentage covered by water (Fig. 9). It is important to note that the largest water depth reaches 20.25 m in Ponferrada, so the number of buildings highly affected by an elevated water depth is high (Table 9). The comparison of these values with the building heights (Table 10) makes it easy to understand the high vulnerability of the urban parcels.

Table 9 Simulated water depths in the flood area
Full size table
Table 10 Building heights
Full size table

Indeed, the main map (Fig. 9) shows many buildings (2786 urban parcels) affected 100% by the water depth. Again, those 1879 urban parcels without buildings appear with a 0-value indicator. The high level of vulnerability index present in the urban parcels located in district 3 is relevant because of the type of building. Most of them are one-floor buildings, so although the water depths are not as high as in other districts, the floodwater easily covers the houses.

Despite the high importance of the water depth versus building height, the panel of experts gave their maximum weight value to the building condition indicator. The analysis of the results showed that 1810 buildings had the minimum value because of their good building condition. All of them were built after 1970 and are spread over all of the districts of the town. The highest exposure values with this indicator are in districts 3 and 5 and match with constructions built in the period 1900–1920 (Table 11). None of the other districts has buildings of this period.

Table 11 Number of buildings according to their building condition and district (source: INE 2011)
Full size table

A very small area of district 4 is affected by the flood. It is only a small part of the census Section 401, and it is not representative of the overall district. For this reason, assigning the building condition values of the district 4 in general would include errors. That is why the same building conditions of the neighbouring 103 census section were assigned to those buildings in Section 401.

The general good building conditions, regardless of age, are responsible for not reaching the maximum potential vulnerability value in the final map.

In the case of harmed roads, the importance of this indicator in the analysis is very low due to the low number of affected kilometres and the low AHP weight value. From the 884 km of roads that are in the municipality, only 292 km would be harmed by the flood. This represents 33% of the total studied roads.

5 Discussion

At this moment, the open-access data available in several countries allow an improve in the analysis scale without an increase in time and money. Ten years ago, this study would have presented just a unique indicators’ value for the entire municipality. Instead, the new studies may show the spatial variability of the indicators inside the municipality, and it is possible to differentiate the flood vulnerability of each urban parcel. Therefore, they are basic to increasing societal security and improve the decision-making process (Abdelrahman 2021). For instance, in the studied case of a dam break, a large area will be flooded due to its proximity to the dam, but different risk degrees have been detected. As a result, the re-use of the open-access data for specific applications like the one presented in this study supports that these data create value for the society and its economic development (Cowan et al. 2014). In cases where cadastral mapping is unavailable in the study area, an alternative approach for generating a vector map with urban parcels is through satellite image segmentation (Chhor et al. 2017; Feng et al. 2020). So, parcels will be distinguished according to their different type of the rooftops. However, when utilizing this method, detailed information such as building age, number of floors, and other characteristics may not be attainable. Then, conducting a survey among the local population would be the only viable option to gather comprehensive data on these aspects. But this option only is possible when the towns are small.

Overlaying maps of different scales can introduce inconsistencies and inaccuracies into the results (Bailey 1988 and Chrisman 1987). In this study different map scales with different elementary units, such as the urban parcel, the census section, the district area, the municipality and the 5 m × 5 m pixel, were used (Table 12). The choice of the unit type for each variable was made based on the intrinsic nature of the indicator. For example, the number of road km does not make sense by urban parcel, it is a variable for the overall municipality. Nevertheless, the overlay of these layers does not involve any inconsistency nor inaccuracy since the polygons of the larger scale are completely inside of one polygon of the layer with smaller scale. There cannot be an urban parcel between two census section nor two census section between two municipalities. In this way, the information of the smallest scale has been extracted towards the spatial units of the largest scale layer (the cadastre parcel). Furthermore, adopting the urban parcel as the unit of analysis for most indicators enables the application of this method in towns of any scale, enhancing the adaptability and versatility of the proposed methodology.

Table 12 Elementary units used in the index analysis
Full size table

Another advantage of the proposed method is that the results may help to define the emergency planning of the town. Water might reach the town of Ponferrada in a very short time (15 min). In case that some citizens cannot reach out the water perimeter, the index maps can become a useful tool to identify those buildings located inside the water sheet that fulfil saved characteristics as meeting points. Therefore, this type of analysis represents a good improve in the flood planning for those scenarios originated by rainfall or snow melting. But the emergency planning might present some problems when trying to define meeting points in buildings when floods come from a dam break. In this case, it is not possible to model the exact behaviour of the building structures when the water wave reaches them. This means that some of them might fall because of the water impact, although other buildings that are located inside the flood extent map might remain safe in the uppermost floors. The uncertainties of engineering studies in this aspect are still very high (Paquier and Goutal 2016). For this reason, it is necessary to improve the wave impact and sediment transport modelling before buildings inside the flood area are designated as meeting points. Despite the presence of these uncertainties, the awareness of their existence may facilitate the design of a flexible emergency planning to anticipate and adjust to future changes. This approach is consistent with the resilience-enhancing strategies supported by Berkes (2007).

At the same time, there are other uncertainties related to the number of citizens at each urban parcel that would help to better define the meeting points. In this case, the data exists, but the present Spanish legislation does not allow its use out of the Administration’s studies. Despite all the advantages, the use of open data may lead to claims for individual rights due to conflict with the personal data protection laws. So, while everybody defends that public data belong to the citizens, it is important that the access to these governmental data follows a legal framework to guarantee everybody’s rights (Abdelrahman 2021). Otherwise, identity can be inferred from data and the privacy of individuals can be affected (Cowan et al. 2014). Anyhow, although these uncertainties, the proposed PVI analysis based on cadastre parcels represents an important new data source to be included in emergency planning.

The time interval between flood hydraulic modelling and emergency planning may lead to other inconsistencies, in this case related to altitude data. The introduction of LiDAR data in altitude mapping in recent years has improved the accuracy of height data but has resulted in problems when combining with older data. For this reason, it is necessary to adapt the water levels of former modelling to LiDAR altitudes, considering the perimeter points of the water sheet extension. Again, GIS is an extremely helpful tool in this process and allows mapping the percentage of each building that is covered by the floodwater. With this method, water depth is considered spatially variable and diminishes the overestimations of a unique water depth value in the entire town, as noted by Breilh et al. (2013). Nevertheless, this spatial analysis would have been avoided if data from the previous dam-break study would have been open-accessed and instead of few points with water height data the complete modelled raster map would have been left as open-access data. It is what Abdelrahman (2021) calls prevalence of closed government culture and lack of open government data policy. Reports of this type of studies are often available after requesting personally for them, but the Administration is reticent to give spatial data of this kind, maybe due to fear to data misinterpretation and misuse (Zuiderwijk and Janssen 2014).

The building condition map is another tool to help diminish the vulnerability of a society. The objective of this variable is to consider the potential of roof collapses, wall failures or foundations undermined, all of which are part of the building elements most often damaged by a flood (Kelman and Spence 2004). Although the open government data do not offer extensive details about building conditions, cross referencing with other data related to the age of the infrastructure (or the age of the last extension or rehabilitation of the building) allows improvement of building characteristics. It also provides a first global view of the building capability of resistance. In the case of Ponferrada, this factor is the critical when analysing the town’s resilience. The majority of the buildings shows good conditions which enhances their resistance when facing the water wave generated by the potential dam break. The main issues are concentrated in districts 3 and 5, located downtown, in the oldest area of the city. Nevertheless, some of these buildings may have undergone various types of internal rehabilitation (not recorded by the Cadastre Administration) that could have affected their structural integrity. In this case, the concept of data collaboratives would represent a benefit for citizens (Susha et al. 2019), considering the insurance companies as the collaborative entities. Sanderson et al. (2022) remarked the importance of the interactions among different entities to improve resilience planning. The incorporation of this collaborative data could facilitate the implementation of local policies focused on improving the resistance of the building materials or the urban planning to relocate potentially affected population in the event of flooding, among others. As a consequence, this investment would help to reach some of the 4 R’s of resilience (Kameshwar et al. 2019): robustness (ability to absorb damage) and rapidity (ability to rapidly recover). While there is enough data about building conditions in the national open-access database, there is a lack of available data related to infrastructures. The physical vulnerability map of this study only includes roads. Other infrastructure networks, such as electricity, water, telecommunications and internet, should also be included in future studies to improve the physical vulnerability results. At this moment, almost without exception, societies are highly dependent on these infrastructures for their current way of life. They are critical infrastructures, and without them, the goods and services that protect life would be at risk. This type of data faces two problems. The first is that most of these data belong to private enterprises, so their access is difficult to achieve. Data exchange among companies and government is still difficult due to legal barriers related to proprietary nature of data (Susha et al. 2019). The second is the lack of methodology to establish indicators that represent the vulnerability of these variables. Until now, few studies have focused on estimating the damage after the event has occurred (Dall’Osso et al. 2009; Guillard-Goncalves 2016; Azmeri and Isa 2018). Although these difficulties for working with infrastructure data, we consider that their inclusion in future PVI will represent an important improve in the vulnerability evaluation of the society.

This study has focused on physical vulnerability. However, the analysis of social and economic factors is fundamental to enhance the resilience of the society, where disaster data such as the inclusion of rescue, evacuation or hospitalization operations during and immediately after the flood event might be included (Lin et al. 2017).

6 Conclusion

The proposed method has shown a high adaptability and versatility to data sources with different units of analysis after being tested in a medium-sized population. At the same time, the spatial analysis has once again showed being a powerful tool to estimate the flood vulnerability, displaying the geographical diversification of each indicator and the final PVI. Of all indicators, the most relevant ones have been, firstly, the percentage of affected buildings based on the depth and, secondly, the age of the buildings. The AHP approach to define the weight of each indicator gives excellent results, in case besides the weight matrix the evaluators add an ordered list of the indicators according to their importance. The overall resulting maps represent an important progress when developing prevention programs, as well as emergency and evacuation plans, by delimiting the most vulnerable areas and pointing potential meeting points to increase the resilience against floods. Future local policies should focus in these areas to achieve a resilient society, improving the quality of the buildings and developing new urban planning to avoid increasing the population on them. However, none of these measures should be considered without combining the results of the present study with a socio-economic approach to the vulnerability of Ponferrada.

On the other hand, in this paper, we have argued that the new open-access data facilitates the estimation of the physical vulnerability index in flood risk analysis from a potential dam break. Nevertheless, though these new open sources have increased the amount of available data for this type of studies, there is still a lack of data related to material construction and the infrastructure networks considered critical for the present standard way of life. Efforts should be focused in the following years on making open access this type of data that is currently in private hands. The access to these data should follow a legal framework to guarantee everybody’s rights, but, at least, it is necessary to share this cartography with the local administration. In this way, risk analysis studies ordered by municipalities will be able to include other indicators related to infrastructure, such as telephone, electricity or water networks, improving the resilience of the society and decreasing the vulnerability.