Physics and Chemistry of the Earth xxx (2016) 1e9

Contents lists available at ScienceDirect

Physics and Chemistry of the Earth

journal homepage: www.elsevier.com/locate/pce

A new approach for computing a flood vulnerability index using

cluster analysis
Paulo Fernandez a, d, *, Sandra Mourato b, d, Madalena Moreira c, d, Luísa Pereira e, f
a
Instituto Polit

ecnico de Castelo Branco, Escola Superior Agra
ria, Portugal
b
School of Technology and Management, Polytechnic Institute of Leiria, Portugal
c

Universidade de Evora, Escola de Ci^
encias e Tecnologia, Portugal
d
ICAAM e Instituto de Ci^ encias Agra
rias e Ambientais Mediterra

^nicas, Universidade de Evora, Portugal
e
Universidade de Aveiro, Escola Superior de Tecnologia e Gesta

~o de Agueda, Portugal
f
Centro de Investigaça~o em Ci^encias Geo-Espaciais, Portugal

a r t i c l e i n f o a b s t r a c t

Article history: A Flood Vulnerability Index (FloodVI) was developed using Principal Component Analysis (PCA) and a
Received 12 February 2015 new aggregation method based on Cluster Analysis (CA). PCA simplifies a large number of variables into a
Received in revised form few uncorrelated factors representing the social, economic, physical and environmental dimensions of
1 April 2016
vulnerability. CA groups areas that have the same characteristics in terms of vulnerability into vulner-
Accepted 5 April 2016
Available online xxx
ability classes. The grouping of the areas determines their classification contrary to other aggregation
methods in which the areas' classification determines their grouping. While other aggregation methods
distribute the areas into classes, in an artificial manner, by imposing a certain probability for an area to
Keywords:
Aggregation methods
belong to a certain class, as determined by the assumption that the aggregation measure used is nor-
Cluster analysis mally distributed, CA does not constrain the distribution of the areas by the classes.
Flood vulnerability index FloodVI was designed at the neighbourhood level and was applied to the Portuguese municipality of
Principal components analysis Vila Nova de Gaia where several flood events have taken place in the recent past. The FloodVI sensitivity
was assessed using three different aggregation methods: the sum of component scores, the first
component score and the weighted sum of component scores.
The results highlight the sensitivity of the FloodVI to different aggregation methods. Both sum of
component scores and weighted sum of component scores have shown similar results. The first
component score aggregation method classifies almost all areas as having medium vulnerability and
finally the results obtained using the CA show a distinct differentiation of the vulnerability where hot
spots can be clearly identified.
The information provided by records of previous flood events corroborate the results obtained with
CA, because the inundated areas with greater damages are those that are identified as high and very high
vulnerability areas by CA. This supports the fact that CA provides a reliable FloodVI.
© 2016 Elsevier Ltd. All rights reserved.

1. Introduction climate models project an increase in rainfall intensity in warmer

From 2001 to 2010, hydrological disasters in Europe (flood and
mass movements) represented the largest share of total disaster
victims (55.1%) and millions of Euros worth of damages (Guha-
Sapir et al., 2012). Flood risk assessment entails understanding
vulnerability, which is an important issue at present, because

cnico de Castelo Branco, Escola Superior
* Corresponding author. Instituto Polite
Agr

aria, Portugal.
E-mail address: palex@ipcb.pt (P. Fernandez).
climates (Emori and Brown, 2005; Groisman et al., 2005; Santos
et al., 2015; Trigo and Palutikof, 2001) which will lead to an in-
crease in the frequency of flood events (Balica, 2012). Therefore,
vulnerability assessment is of paramount importance as a tool for
population safety and property protection.
In 2007, the Floods Directive (FD) created a Pan-European
framework to support the Member States in evaluating flood risk.
The FD is linked to the Water Framework Directive (WFD) and
should produce flood risk maps to help decision makers and au-
thorities take appropriate measures aimed at reducing flood risk in
an effective and sustainable manner (Mostert and Junier, 2009).

http://dx.doi.org/10.1016/j.pce.2016.04.003
1474-7065/© 2016 Elsevier Ltd. All rights reserved.

Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
2 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9
Furthermore, the development of techniques and assessment
methodologies as well as measures regarding the increase of
knowledge about flood vulnerability or flood risk can be of great
value for decision makers and can help reduce damage and
fatalities.
Risk may be defined as the probability that a particular level of
loss can be sustained by a given series of elements as a result of a
given level of hazard impact (Alexander, 2000). The exposed ele-
ments in flood risk are population, communities, buildings and
infrastructures as well as economic activities and the natural
environment, that are under threat in a given area.
Vulnerability is embedded into the concept of risk, as
Risk ¼ Hazard  Vulnerability (Wisner et al., 2004) and is under-
stood as “The characteristics and circumstances of a community,
system or asset that make it susceptible to the damaging effects of a
hazard” (UNISDR, 2009). The existing literature establishes several
definitions and conceptual frameworks of vulnerability were
established, according to the researchers' views. These can be
summarized in three classes (Adger, 2006; Fekete, 2009b; Tate,
2011): i) exposure to a natural event, risk stressor or shock; ii)
sensitivity, also described as susceptibility or resistance; and iii)
adaptative capacity, also expressed as recovery potential or
resilience.
There are usually four dimensions that need to be considered in
vulnerability assessment: i) the physical dimension that represents
the potential of physical impact on the built environment; ii) the
economic dimension that accounts for the potential impacts of
hazards on economic assets; iii) the social dimension that is related
to the presence of human beings, individuals or communities, and
their capacity to cope, resist and recover from hazard impacts; and
iv) the environmental dimension that refers to potential impacts on
the natural environment and the ability of ecosystems to cope and
recover from hazard impacts.
The complex structure of a vulnerability assessment framework
is described as a hierarchical model, a deductive model, or an
inductive model (Tate, 2012) and aggregate vulnerability indices
are computed using the mathematics of index construction
(Schmidtlein et al., 2008). Inductive methods were popularized by
the Social Vulnerability Index (Cutter et al., 2003) and are used by
Fig. 1. Study area location.
Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
the majority of the more recent vulnerability indices (Fekete,
2009b; Schmidtlein et al., 2011; Tate, 2012).
The main criticisms regarding indices construction methods are
the subjective process of both variable selection and weighting,
unavailability of certain variables, problems related to aggregation
at different scales, and difficulties in validating the results (Barnett
et al., 2008; Fekete, 2012; Jones and Andrey, 2007). Furthermore,
different combinations of the variables may produce diverse
vulnerability assessments (Chakraborty et al., 2005; Koks et al.,
2015). Nevertheless, the usefulness of indicators aimed at
reducing complexity, measuring progress, and establishing prior-
ities makes them an important tool for decision makers.
Jones and Andrey (2007) have argued that Principal Component
Analysis (PCA) offers an alternative to the otherwise subjective
variable selection by objectively simplifying a large number of
variables into a few uncorrelated factors that capture the variability
in the underlying data (Abdi and Williams, 2010). The PCA
approach increases flexibility regarding the choice and number of
variables, thereby allowing for a more robust and consistent set of
variables (Cutter et al., 2003) and provides several potential ad-
vantages with regard to aggregation of spatially explicit and
potentially incommensurable variables (Abson et al., 2012).
PCA is labelled as an inductive method and has been used by the
majority of the more recent vulnerability indices studies (Borden
et al., 2007; Cutter et al., 2003; Fekete, 2009a; Finch et al., 2010;
Rygel et al., 2006; Schmidtlein et al., 2011; Tate et al., 2010). The
vulnerability index is built as a function of principal components
(PC) and their subsequent aggregation. Aggregation of PC refers to
the procedure used to combine transformed, normalized, and
weighted indicators into a simpler measure, reducing the amount
and complexity of information that must be used during the pro-
cess of classifying the areas into vulnerability classes (Nguyen et al.,
2016). It should be emphasized that the existing aggregation
methods do not guarantee that those areas have similar charac-
teristics in terms of the variables of interest's values and thus of
vulnerability. Furthermore, and as stated before, the choice of the
aggregation method conditions the results, making the aggregation
of the PC a subjective decision in the index construction process
(Bo€ hringer and Jochem, 2007).
 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9
Commonly applied aggregation methods use addition and/or
multiplication of components. The aggregation based on the
addition of components using equal weights is used comprehen-
sively (Tate, 2012). In some recent vulnerability assessments,
eigenvector techniques similar to PCA have been used as robust
estimators of objective weights (Pacheco and Sanches Fernandes,
2013) and proven to be the most efficient among common
weighting techniques (Pacheco et al., 2015).
Cluster analysis (CA) appears as a serious alternative to these
aggregation methods. Although various studies, in distinct areas
other than flood risk assessment, have used CA to aggregate
different regions by their characteristics (Lee et al., 2016; Mourato
et al., 2010; Santos et al., 2014), a formal methodology to aggre-
gate principal components of similar neighbourhoods into the
same group and to perform their classification in a straightforward
manner is missing in the literature.
The partitional K-means clustering algorithm is a non-
hierarchical clustering technique which attempts to find k non-
overlapping clusters, by moving objects from cluster to cluster,
with the goal of minimizing the within-cluster variance and
maximizing the between-cluster variance (Jain and Dubes, 1988).
As a first step, k initial centroids are selected, where k is specified by
the user and indicates the desired number of clusters and then,
each additional iteration groups the observations, based on the
nearest Euclidean distance to the centre of the cluster. The centroid
of each cluster is then updated based on the points assigned to the
cluster. The process continues until cluster centres do not shift
more than a given cut-off value (Wu, 2012). By using CA, the areas
with similar characteristics, in terms of the variables of interest's
values, are gathered together. This is exactly what the other ag-
gregation methods lack.
In addition to the development of a new aggregation method,
we propose a vulnerability assessment addressing the physical,
economic, social, and environmental dimensions, whereas the
majority of studies use only the social dimension in a particular
area: United States (Cutter et al., 2003, 2013), United Kingdom
(Tapsell et al., 2002), Austria (Kienberger et al., 2009), Germany
(Fekete, 2009b), Norway (Holand and Lujala, 2013), Romania
(Armas and Gavris, 2013), China (Zhou et al., 2014), and Israel
(Felsenstein and Lichter, 2014). Therefore, the main differences
between this study and the previous ones on vulnerability assess-
ment are: the development and implementation of a new aggre-
gation method based on CA, which is objective, groups
neighbourhoods with similar characteristics into the same vulner-
ability class, incorporates a greater number of dimensions into the
assessment, and contains more detail terms of the in spatial scale
(neighbourhood level).
Thus, the main objective of this paper is to provide an auto-
mated framework for classifying vulnerability of neighbourhoods.
To achieve this objective, PCA and CA are used. PCA transforms a 15
dimensional space into a 4 dimensional one and determines scores
for each PC. This makes it possible to create and use a modified 4
dimensional space of the PC (by scores). CA groups neighbourhoods
with similar characteristics into 5 vulnerability classes. The index is
applied to the Vila Nova de Gaia municipality.
Although the input parameters of the Flood Vulnerability Index
(FloodVI) were selected according to the special focus of floods in
Portugal, they are not directly dependent on flood related infor-
mation. Therefore, this index can be applied to identify, quantify
and rank flood vulnerability not only in the study area but also in
any potential flood area.
The paper outline is as follows: Section 1 introduces the back-
ground and the aim of the study. Section 2 presents the study area
and describes the vulnerability assessment approach and the
sensitivity analysis of the vulnerability index. The results are
Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
3
presented and discussed in Section 3 and finally some conclusions
are presented in Section 4.
2. Materials and methods
2.1. Study site and data
Flood vulnerability is assessed for the municipality of Vila Nova
de Gaia (Fig. 1), situated in Northern Portugal, where several floods
have occurred (Branda ~o et al., 2014; Ze^zere et al., 2014). According
to the DISASTER Project (Ze ^zere et al., 2014), any flood event is
stored in the database if it led to casualties or injuries, and missing,
evacuated or homeless people, independent of the number of
people affected. Between 1865 and 2010, 57 floods were reported in
Vila Nova de Gaia municipality, accounting for a total of four deaths,
as well as evacuation of 123 and displacement of 2930. The mu-
nicipality ranks as the fourth in Portuguese ranking of flood di-
sasters. Between 1999 and 2009, 1275 flood events were reported
by the national civil protection service.
The municipality has an area of 168.46 km2 and is divided into
24 civil parishes and 3076 neighbourhoods. It is the third most
populous municipality in Portugal, with 302,295 inhabitants in
2011, approximately 180,000 of which are urban residents. From
2001 to 2011 the number of residents has increased by approxi-
mately 15,000. The population density is about 1795 inhabitants
per km2 and the building density is about 386 buildings per km2.
The predominant land uses are urban (43%) and forest (39%).
In this study two data sets were used: the Geographic Infor-
mation of Portuguese Statistics (2011 Census) provided by the
National Statistics Institute and the land use maps for 1990 and
2007 provided by the Portuguese Geographic Institute. The 2011
Census data are geo-referenced information based on small terri-
torial units. The neighbourhood is the territorial unit which iden-
tifies the smallest homogenous area, whether built-up or not, that
exists in the statistical section. It represents a block in urban areas.
2.2. Variable selection
A vulnerability study has to take into account several variables
within an area such as: age, gender, race, ethnicity, social class,
unemployment rate, immigrant status, density and quality of the
built environment, land use, housing tenancy and presence of
informal support networks (Borden et al., 2007; Burton and Cutter,
2008; Cutter et al., 2003; Cutter et al., 2000; Fekete, 2009a; Finch
et al., 2010; Lein and Abel, 2010; Masozera et al., 2007; Rygel
et al., 2006; Schmidtlein et al., 2011; Schneiderbauer and Ehrlich,
2006; Simpson and Katirai, 2006; Tapsell et al., 2002; Wisner
et al., 2004).
The variables for the social, economic and physical dimensions
were derived from the information available in the 2011 Census at
the neighbourhood level. Additionally, environmental variables
were added from the land use data. The criteria for variable se-
lection were taken from the literature and are presented in Table 1.
2.3. Flood vulnerability index design
PCA reduces the number of variables and determines some
components that summarise different vulnerability characteristics.
The proposed FloodVI is estimated according to the following steps,
as illustrated in Fig. 2:
1) Standardization of input variables to z-scores.
2) Calculation of the PCA with the standardized input variables.
The starting point for all factor analysis is the collinearity
4 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9

Table 1
Considerations for variable selection.

Variables Considerations

Building density In urban areas with high population density, the rescue process is often rather complicated. In some cases, high population density is related to a
relative number of lower income families (Masozera et al., 2007). Buildings' density is a factor which influences vulnerability in inundated areas
because of potential increase of building's exposure to flooding (Cardona, 2005; Tapsell et al., 2002).
Number of floors The upper floors of buildings may be used to protect people and their belongings (Schneiderbauer, 2007).
Construction period The more recent building constructions are based on structure safety regulations. Therefore, these buildings are often more resistant. Older
neighbourhoods also have older sewage systems which may be more susceptible to flooding (Simonovic et al., 2007).
Building structure The main floor and wall construction materials determine the building's physical fragility to a flood event and indicate resistance to damage, as well
as the social status of the residents (Müller et al., 2011; Schneiderbauer, 2007).
Housing occupancy Landlords are more likely to pursue construction changes in their buildings and to have insurance or increase their insurance coverage than tenants
(Tapsell et al., 2002). People that rent a house usually do not have the financial resources for home ownership and often do not have access to
information about financial support during recovery (Cutter et al., 2003; Fekete, 2009a).
Gender Women have a higher perception of risk and are better prepared for action (Fekete, 2009a). However, women can have more difficulty during
recovery, often due to lower income and greater family care responsibilities (Cutter et al., 2003; Fekete, 2009a; Hewitt, 1997).
Education level The level of education and illiteracy rate are clear factors of socioeconomic vulnerability, because there is a direct relationship between these and
economic capacity, social status and job opportunities (Cutter et al., 2003; Fekete, 2009a).
Age The elderly have limited mobility and physical difficulties in evacuations. They are more reluctant to leave their homes, have health-related
problems and longer recovery time (Rygel et al., 2006). The very young also have high physical fragility and dependency (Cutter et al., 2003; Fekete,
2009a; Hewitt, 1997; Kuhlicke et al., 2011).
Unemployment The unemployed comprise a special group that is more dependent on other family members and on the government (Fekete, 2009a). The
unemployed potentially have lower financial assets, so their houses are of lower quality and are most probably not insured (Balica, 2012).
Household Larger families often have to share their income and have more dependents such as children and elderly to evacuate (Fekete, 2009a). Frequently,
composition they have limited financial resources to outsource care for dependents and difficulties in reconciling work responsibilities with care for family
members (Cutter et al., 2003; Fekete, 2009a; Hewitt, 1997).
Economic activity The services sector is an expression of areas with high economic development areas. (Cutter et al., 2009).
sector
Land use In agricultural areas there are also economic losses resulting from temporary inundation of crops, which can be frequent due to the increase in
extreme events resulting from climatic changes (Morris and Brewin, 2013). Damages are lower in inundated forest areas. Some species are flood
tolerant (survived more than 150 days), and others can withstand up to 50 days of flooding (Whitlow and Harris, 1979). Floods in urban areas
encumber emergency management and result in large economic losses, because the temporary water coverage affects economic activities and a
large number of people (Balica, 2012).
Urban growth Rapid urban growth combined with wrong planning policies may result in poor quality housing which makes residents more vulnerable. Urban land
use increases impervious areas and hence runoff (Balica and Wright, 2010).

analysis and the elimination of redundant data. This procedure normalized PC using the Euclidean distance measure. The K-
measures if the correlations are appropriate to carry out PCA. means is a partition clustering approach by which each point is
3) Analysis of PCA output results. The KaisereMeyereOlkin test assigned to the cluster with the closest centre, with a pre-defined
(KMO) measures the sampling adequacy and shows the extent number of clusters. The K-means algorithm aims at minimizing
to which the data fits factor analysis, thus determining the level an objective function (equation (1)), in this case a squared error
of confidence that can be expected when using factor analysis function.
(Hair et al., 2009). KMO values above 0.6 indicate an acceptable
Xk Xn   ðjÞ
2

level and above 0.8 a good compatibility level of variables J¼ x  c  (1)
j¼1 i¼1 i j
(Hutcheson and Sofroniou, 1999). Those components whose
eigenvalues are greater than one are selected (Kaiser, 1960).
where
 kis the number of clusters, n is the number of cases and
Communalities measure the extent to which the variance of the  ðjÞ 2
xi  cj  is a distance function between case i and the centre of
original variables is accounted for by the observed components.
cluster j (cj) (Jain and Dubes, 1988).
The communalities values should all be greater than 0.5.
The cluster centre is defined as a parameter set that has the
4) Analysis of variance explained. The number of components
minimum average Euclidean distance to each of the members in
needed in order to account for a pre-specified amount of orig-
the cluster. This distance is weighted by the explained variance
inal data variation should be retained. The smallest number of
obtained for the PC. In this way, the dimensions that are considered
components is chosen such that at least 80% of the original data
more important will have more impact on the clustering process.
variation is explained.
The number of clusters used in clustering is five according to the
5) Rotation of the initial PCA solution using the Varimax rotation.
five vulnerability classes (very low, low, medium, high, and very
This is a popular orthogonal factor rotation method and the
high).
factors are extracted so that their axes are maintained at 90 .
Each cluster is then characterized by a value, to reduce the 4
This generally simplifies the relationships among the variables
dimensional space to one dimension, during the classification
and clarifies the interpretation of the factors.
process. This value is the mean value of the coordinates of the
6) Calculation of the component scores for each neighbourhood.
cluster centre. The cluster with the smallest value will be classified
The component score is a composite measure created for each
as being the least vulnerable and the one with the highest value as
observation of each extracted factor in the factor analysis. The
he most vulnerable.
component scores are standardized to a z-score.
To assess FloodVI sensitivity three other aggregation methods
7) Aggregation method, assigning a vulnerability class to each
are considered:
neighbourhood.
Sum of components (Aggregation 1): this is a simple approach
that adds the component scores (CS), assigning equal weight to
The novel aggregation method proposed here is based on CA,
each component of the index (equation (2)).
which is regarded as the most practical method of establishing
regions with similar characteristics from large data sets (Hosking Xn
Aggregation 1 ¼ CSi (2)
and Wallis, 1997). A K-means clustering method is applied to the i¼1

Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9 5

Fig. 2. FloodVI construction flowchart.

First component (Aggregation 2): the first extracted compo- Weighted sum of components (Aggregation 3): This is a
nent is the linear combination of variables that explain the largest compromise between the first two methods, where each compo-
amount of variation in the original data. Therefore, selecting only nent's weight (vi) is the proportion between the explainable vari-
the first component (CS1) will give the mathematically optimal ance and the total variation (equation (4)).
value that summarizes all the input variables in a single combina-
tion (equation (3)). Xn
Aggregation 3 ¼ v
i¼1 i
 CSi (4)
Aggregation 2 ¼ CS1 (3)
For each aggregation method the resulting score is standardized

Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
6 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9

to z-scores. These z-score values are then used to classify the cor- loads of the four components after Varimax rotation. This pro-
responding neighbourhood into a vulnerability class. These classes vides a clear depiction of the four PC and the variables that form
correspond to an interval of dimensions that is determined by the it.
standard deviation of the z-score values (that has a value of 1 since
the z-scores are normalized). Hence, the corresponding neigh- 3.2. Flood vulnerability assessment
bourhoods are classified respectively as having very high vulnera-
bility for z-score values > 1.5, high vulnerability for 1.5  z-score The four aggregation methods were applied (2.3) to compute
values 0.5, medium vulnerability for 0.5 <z-score values < 0.5, the FloodVI. The first three aggregation methods consider that
low vulnerability for 1.5  z-score values  0.5, and finally very among the 3076 neighbourhoods, 205 have very low vulnerability,
low vulnerability for z-score values < 1.5. Selection of these values 744 have low vulnerability, 1178 have medium vulnerability, 744
is supported by the existing literature (Cutter et al., 2003; Dunning have high vulnerability and 205 have very high vulnerability.
and Durden, 2013; Schmidtlein et al., 2008) and assumes that the z- Noteworthy is the artificial distribution of the number of neigh-
scores have a normal distribution. In this way, and considering the bourhoods per class due to the classification procedure. Indeed, the
5 vulnerability classes, the probability of having a neighbourhood commonly used thresholds (2.3) force, for example, the medium
with medium vulnerability class is 38.292% against 24.173% for the vulnerability class to have a greater number of neighbourhoods.
low and high vulnerability classes and 6.681% for the other two The method using CA provides the following results: 1354 neigh-
classes. This form of classification forces the neighbourhoods to bourhoods have very low vulnerability, 369 have low vulnerability,
pertain to a certain class, depending on the thresholds used. On 215 have medium vulnerability, 861 have high vulnerability and
the contrary, and being more accurate, CA does not impose any kind 277 have very high vulnerability. The spatial distribution of these
of constraint on the distribution of the areas to be classified. classified neighbourhoods per aggregation method is illustrated in
Henceforth, cluster aggregation method will be referred to as Ag- Fig. 3.
gregation 4. Fig. 3 shows that the spatial distribution of the 5 vulnerability
classes produced by the aggregations methods 1 and 2 are very
different, especially in the southeast part of the municipality. This
3. Results and discussion indicates that it is important to also take into account components
two, three and four, which represent 41.7% of the variance of the
3.1. Component scores variables involved. Fig. 3 also shows that the results produced with
aggregation method 3 are very similar with those produced with
After PCA, a KMO test value of 0.882 was achieved, which can be method 1 since the aggregation measures used by both methods
considered good. The relationship between the selected variables are very similar. As the first component is valued more in aggre-
can be described by 4 components which explain 86.1% of the gation method 3 than in method 1, the results obtained with
variance. The first component explains 44.4% of the variance and methods 3 and 2 are less dissimilar than those obtained with
can be related to the social and economic dimensions of vulnera- methods 1 and 2, with most of dissimilarities also located in
bility. It includes the following variables: male inhabitants; female southeast part of the municipality. The major differences are be-
inhabitants; age; unemployment; education level; economic ac- tween the results of aggregation methods 1, 2 and 3 and those of
tivity sector; housing occupancy and household composition. The method 4, where there is a predominance of areas with high
second component that explains 24.0% of the variance addresses vulnerability.
building features and includes: building density; number of floors; Correlation of the results produced by the four different aggre-
construction period and building structure. The third and fourth gation methods confirms the aforesaid (Table 3).
components include the environmental dimension where compo- Analysing the percentage of the areas classified according to
nent three (which explains 11.2%) is related to the urban aspects: each aggregation method (Fig. 4a) it can be concluded that both
urban land use and urban land use change (1990 and 2007) and aggregation methods 1 and 3 present similar results. According to
component four (which explains 6.5%) comprises the rural vari- aggregation method 2, 52.5% of the area is classified as having
ables: agricultural land use and forest land use. medium vulnerability, which strongly disagrees with the 7.1%
The rotated component matrix (Table 2) shows the variable (lowest percentage) obtained by aggregation method 4, which
classifies 45.6% of the areas as having high vulnerability.
Although we have endorsed 5 vulnerability classes, the sensi-
Table 2
tivity analysis is more straightforward if we combine Low and Very
Variable loads in the rotated component matrix.
low vulnerability classes into one single class. The same is true for
Variables Components the High and Very high vulnerability classes (Fig. 4b). From Fig. 4b
1 2 3 4 one can conclude that both aggregation methods 1 and 3 provide
Gender 0.957 0.271 0.019 0.001 similar results, although the latter is sounder, from the conceptual
Age 0.899 0.380 0.017 0.011 point of view. Considering aggregation method 2, the medium and
Unemployment 0.878 0.277 0.030 0.012 low vulnerability classes still have the highest percentage of area,
Education level 0.741 0.530 0.060 0.007 which can be considered conservative since the physical and the
Economic activity sector 0.961 0.091 0.17 0.001
Housing occupancy 0.944 0.137 0.004 0.001
environmental dimensions are not accounted for. Aggregation
Household composition 0.956 0.250 0.002 0.004 method 4 now classifies 92.9% of the area as having high and low
Building density 0.393 0.899 0.043 0.027 vulnerability.
Number of floors 0.214 0.946 0.072 0.042 Intersecting the results obtained with the four aggregation
Construction period 0.166 0.836 0.002 0.046
methods, it is noted that 36 neighbourhoods maintain their clas-
Building structure 0.412 0.821 0.030 0.011
Urban land use 0.058 0.071 0.906 0.043 sification as high or very high, none keep the medium vulnerability
Urban growth 0.007 0.012 0.672 0.157 classification and 49 neighbourhoods maintain their classification
Agriculture land use 0.046 0.092 0.355 0.787 of low or very low vulnerability. Aggregation methods 1, 2 and 3
Forestry land use 0.066 0.002 0.615 0.624 have been used and divulged in the literature and no attention has
The bold values correspond to variables load on which components. been paid to their somewhat discordant results.

Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9 7

Fig. 3. Flood vulnerability index at Vila Nova de Gaia municipality according to different aggregation methods.

4. Conclusions

The developed and implemented FloodVI describes how social

Table 3
Correlation values between aggregation methods. and economic characteristics of the population, building features
and environmental issues behave in terms of resistance and resil-
Aggregation Aggregation
ience to flood impact. This study integrates mathematical analysis
1 2 3 4 (PCA and CA) and Geographic Information System (GIS) techniques
1 1 e e e to estimate several vulnerability dimensions. It is also a contribu-
2 0.46 1 e e tion to flood risk assessment.
3 0.79 0.74 1 e FloodVI was been proven to be aggregation model sensitive with
4 0.19 0.19 0.07 1
both sum of components and weighted sum of components

Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
8 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9
Fig. 4. Percentage of area for each vulnerability classes according to the four different aggregation methods: a) five classes b) three classes.
aggregation methods showing similar results. The percentage
values of the areas classified as having low, medium and high
vulnerability are almost equal. The first component aggregation
method gives conservative results classifying the majority of the
areas as having medium and low vulnerability, and the CA aggre-
gation method provides completely different results that present
the majority of the areas as having high and low vulnerability.
Considering that it is not possible to completely validate which
aggregation method is the most accurate, and bearing in mind that
methods involving the addition of components are very similar, the
novel approach suggested here has several advantages: i) mathe-
matical background supporting the ability to bring together similar
areas; ii) providing the decision maker with an option, since the
results clearly classify the neighbourhoods in high or low vulner-
ability; and iii) not imposing an artificial distribution of neigh-
bourhoods per class of vulnerability. In addition, it is possible to
state that CA produces a reliable FloodVI since overlaying of inun-
dated areas, where damages are recorded to have occurred in
previous flood events, are in strong agreement with high and very
high vulnerability areas as classified by the proposed FloodVI.
The potential presented by FloodVI to clearly identify an area as
belonging to an extreme class of vulnerability is an important asset
for decision makers. This index makes it possible to easily detect
the hot spots, and improves interventions in the more vulnerable
areas. Thus FloodVI is a potentially powerful tool because it sum-
marizes complexity, provides quantitative metrics to compare
places and track progress, and it is relatively easy for non-experts to
interpret.
References
Abdi, H., Williams, L.J., 2010. Principal component analysis. Wiley Interdiscip. Rev.
Comput. Stat. 2, 433e459.
Abson, D.J., Dougill, A.J., Stringer, L.C., 2012. Using principal component analysis for
information-rich socio-ecological vulnerability mapping in southern Africa.
Appl. Geogr. 35, 515e524.
Adger, W.N., 2006. Vulnerability. Global Environ. Change 16, 268e281.
Alexander, D., 2000. Confronting Catastrophe. Oxford University Press, New York.
Armas, I., Gavris, A., 2013. Social vulnerability assessment using spatial multi-
criteria analysis (SEVI model) and the social vulnerability index (SoVI model)
e a case study for Bucharest, Romania. Nat. Hazards Earth Syst. Sci. 13,
1481e1499.
Balica, S., Wright, N.G., 2010. Reducing the complexity of the flood vulnerability
index. Environ. Hazards 9, 321e339.
Balica, S.F., 2012. Applying the Flood Vulnerability Index as a Knowledge Base for
Flood Risk Assessment. Delft University of Technology and Academic Board of
Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
the UNESCO-IHE, Delft, Netherlands, p. 152.
Barnett, J., Lambert, S., Fry, I., 2008. The Hazards of Indicators:Insights from the
Environmental Vulnerability Index. Ann. Assoc. Am. Geogr 98, 102e119.
Borden, K.A., Schmidtlein, M.C., Emrich, C.T., Piegorsch, W.W., Cutter, S.L., 2007.
Vulnerability of U.S. Cities to Environmental Hazards. J. Homel. Secur. Emerg.
Manag 4, 1547e7355.
€
Bohringer, C., Jochem, P., 2007. Measuring the immeasurable e a survey of sus-
tainability indices. Ecol. Econ. 63 (1), 1e8.
Branda ~o, C., Saramago, M.M., Ferreira, T., Cunha, S., Costa, S., Alvarez, T.,
Carvalho, F.F.d., Silva, M., Duarte, C., Braunschweig, F., Brito, D., Fernandes, L.,
Jauch, E., Silva, R.P., 2014. Elaboraça ~o de Cartografia Específica sobre Risco de
Inundaça ~o para Portugal Continental. Relato
rio Final, Volume 1-Memo
ria
Descritiva. Age ^ncia Portuguesa do Ambiente, Lisbon, Portugal, p. 260.
Burton, C., Cutter, S., 2008. Levee Failures and Social Vulnerability in the Sacra-
mento-San Joaquin Delta Area, California. Nat. Hazards Rev 9, 136e149.
Cardona, O.D., 2005. Indicators of Disaster Risk and Risk Management: Summary
Report. Inter-American Development Bank, Washington, D.C.
Chakraborty, J., Tobin, G., Montz, B., 2005. Population evacuation: assessing spatial
variability in geophysical risk and social vulnerability to natural hazards. Nat.
Hazards Rev. 6, 23e33.
Cutter, S.L., Boruff, B.J., Shirley, W.L., 2003. Social vulnerability to environmental
hazards. Soc. Sci. Q. 84, 242e261.
Cutter, S.L., Emrich, C.T., Morath, D.P., Dunning, C.M., 2013. Integrating social
vulnerability into federal flood risk management planning. J. Flood Risk Manag.
6, 332e344.
Cutter, S.L., Emrich, C.T., Webb, J.J., Morath, D., 2009. Social Vulnerability to Climate
Variability Hazards: a Review of the Literature. Hazards and Vulnerability
Research Institute, Department of Geography e University of South Carolina,
Columbia.
Cutter, S.L., Mitchell, J.T., Scott, M.S., 2000. Revealing the vulnerability of people and
places: A case study of Georgetown County, South Carolina. Ann. Assoc. Am.
Geogr 90 (4), 713e737.
Dunning, C.M., Durden, S., 2013. Social Vulnerability Analysis: A Comparison of
Tools. U.S. Army Corps of Engineers Institute for Water Resources (IWR),
Alexandria, VA.
Emori, S., Brown, S.J., 2005. Dynamic and Thermodynamic Changes in Mean and
Extreme Precipitation Under Changed Climate. Geophysical Research Letters
32.
Fekete, A., 2009a. Assessment of Social Vulnerability for River-floods in Germany.
Institute for Environment and Human Security. United Nations University,
Bonn.
Fekete, A., 2009b. Validation of a social vulnerability index in context to river-floods
in Germany. Nat. Hazards Earth Syst. Sci. 9, 393e403.
Fekete, A., 2012. Spatial disaster vulnerability and risk assessments: challenges in
their quality and acceptance. Nat. Hazards 61, 1161e1178.
Felsenstein, D., Lichter, M., 2014. Social and economic vulnerability of coastal
communities to sea-level rise and extreme flooding. Nat. Hazards 71, 463e491.
Finch, C., Emrich, C., Cutter, S., 2010. Disaster disparities and differential recovery in
New Orleans. Popul. Environ 31, 179e202.
Groisman, P.Y., Knight, R.W., Easterling, D.R., Karl, T.R., Hegerl, G.C., Razuvaev, V.N.,
2005. Trends in intense precipitation in the climate record. J. Clim. 18,
1326e1350.
Guha-Sapir, D., Vos, F., Below, R., Ponserre, S., 2012. Annual Disaster Statistical Re-
view 2011: the Numbers and Trends. CRED, Brussels.
Hair, J.F., Black, W.C., Babin, B.J., Anderson, R.E., 2009. Multivariate Data Analysis: a
Global Perspective. Pearson Education, London.
 P. Fernandez et al. / Physics and Chemistry of the Earth xxx (2016) 1e9
Hewitt, K., 1997. Regions of Risk: a Geographical Introduction to Disasters. Themes
in Resource Management Addison Wesley Longman, Essex, England.
Holand, I.S., Lujala, P., 2013. Replicating and adapting an index of social vulnerability
to a new context: a comparison study for Norway. Prof. Geogr. 65, 312e328.
Hosking, J.R., Wallis, J.R., 1997. Regional Frequency Analysis: an Approach Based on
L-moments. Cambridge University Press, Cambridge, UK.
Hutcheson, G., Sofroniou, N., 1999. The Multivariate Social Scientist: Introductory
Statistics Using Generalized Linear Models. SAGE Publications Ltd.
Jain, A.K., Dubes, R.C., 1988. Algorithms for Clustering Data. Prentice-Hall, Inc.
Jones, B., Andrey, J., 2007. Vulnerability index construction: methodological choices
and their influence on identifying vulnerable neighbourhoods. Int. J. Emerg.
Manag. 4, 269e295.
Kaiser, H.F., 1960. The application of electronic computers to factor analysis. Educ.
Psychol. Meas. 20, 141e151.
Kienberger, S., Lang, S., Zeil, P., 2009. Spatial vulnerability units e expert-based
spatial modelling of socio-economic vulnerability in the Salzach catchment,
Austria. Nat. Hazards Earth Syst. Sci. 9, 767e778.
Kuhlicke, C., Scolobig, A., Tapsell, S., Steinführer, A., Marchi, B., 2011. Contextualizing
social vulnerability: findings from case studies across Europe. Nat. Hazards 58,
789e810.
Koks, E.E., Jongman, B., Husby, T.G., Botzen, W.J.W., 2015. Combining hazard,
exposure and social vulnerability to provide lessons for flood risk management.
Environ. Sci. Policy 47, 42e52.
Lein, J.K., Abel, L.E., 2010. Hazard vulnerability assessment: How well does nature
follow our rules? Environ. Hazards 9, 147e166.
Lee, C.T., Guzman, D., Ponath, C., Tieu, L., Riley, E., Kushel, M., 2016. Residential
patterns in older homeless adults: Results of a cluster analysis. Soc. Sci. Med.
Masozera, M., Bailey, M., Kerchner, C., 2007. Distribution of impacts of natural di-
sasters across income groups: a case study of New Orleans. Ecol. Econ. 63,
299e306.
Morris, J., Brewin, P., 2013. The impact of seasonal flooding on agriculture: the
spring 2012 floods in Somerset, England. J. Flood Risk Manag. 7 (2), 128e140.
Mourato, S., Moreira, M., Corte-Real, J., 2010. Interannual variability of precipitation
distribution patterns in Southern Portugal. Int. J. Climatol 30, 1784e1794.
Mostert, E., Junier, S.J., 2009. The European flood risk directive: challenges for
research. Hydrol. Earth Syst. Sci. Discuss. 6, 4961e4988.
Müller, A., Reiter, J., Weiland, U., 2011. Assessment of urban vulnerability towards
floods using an indicator-based approach e a case study for Santiago de Chile.
Nat. Hazards Earth Syst. Sci. 11, 2107e2123.
Nguyen, T.T.X., Bonetti, J., Rogers, K., Woodroffe, C.D., 2016. Review: indicator-based
assessment of climate-change impacts on coasts: a review of concepts, meth-
odological approaches and vulnerability indices. Ocean Coast. Manag. 123,
18e43.
Pacheco, F.A.L., Pires, L.M.G.R., Santos, R.M.B., Sanches Fernandes, L.F., 2015. Factor
weighting in DRASTIC modeling. Sci. Total Environ. 505, 474e486.
Pacheco, F.A.L., Sanches Fernandes, L.F., 2013. The multivariate statistical structure
of DRASTIC model. J. Hydrol. 476, 442e459.
Rygel, L., O’sullivan, D., Yarnal, B., 2006. A method for constructing a social
vulnerability index: an application to hurricane storm surges in a developed
country. Mitig. Adapt Strat. Glob. Change 11, 741e764.
^zere, J.L., 2014. Risk analysis for local management from
Santos, P.P., Tavares, A.O., Ze
hydro-geomorphologic disaster databases. Environ. Scie. Policy 40, 85e100.
Please cite this article in press as: Fernandez, P., et al., A new approach for computing a flood vulnerability index using cluster analysis, Physics
and Chemistry of the Earth (2016), http://dx.doi.org/10.1016/j.pce.2016.04.003
9
Santos, R.M.B., Sanches Fernandes, L.F., Varandas, S.G.P., Pereira, M.G., Sousa, R.,
Teixeira, A., Lopes-Lima, M., Cortes, R.M.V., Pacheco, F.A.L., 2015. Impacts of
climate change and land-use scenarios on Margaritifera margaritifera, an
environmental indicator and endangered species. Sci. Total Environ. 511,
477e488.
Schmidtlein, M.C., Shafer, J.M., Berry, M., Cutter, S.L., 2011. Modeled earthquake
losses and social vulnerability in Charleston. South Carolina. Appl. Geogr 31,
269e281.
Schmidtlein, M.C., Deutsch, R.C., Piegorsch, W.W., Cutter, S.L., 2008. A sensitivity
analysis of the social vulnerability index. Risk Anal. Off. Publ. Soc. Risk Anal. 28,
1099e1114.
Schneiderbauer, S., Ehrlich, D., 2006. Social Levels and Hazard (In)dependence in
Determining Vulnerability. In Measuring Vulnerability to Natural Hazards: To-
wards Disaster Resilient Societies. United Nations University Press, Tokyo.
Schneiderbauer, S., 2007. Risk and Vulnerability to Natural Disasters e from Broad
View to Focused Perspective. Department Geosciences. University Berlin, Berlin
(Germany).
Simonovic, S.P., McBean, G., Prodanovic, P., Burn, D.H., Wey, K., Kay, P., Emerson, A.,
Mortsch, L., Hebb, A., Goldt, R., Helsten, M., Davidge, D., Wood, M., 2007. Floods:
Mapping Vulnerability in the Upper Thames Watershed under a Changing
Climate. University of Western Ontario, University of Waterloo, Environment
Canada and Upper Thames River Conservation Authority.
Simpson, D.M., Katirai, M., 2006. Indicator Issues and Proposed Framework for a
Disaster Preparedness Index (DPi), Working Paper 06e03. Center for Hazards
Research and Policy Development, University of Louisville, p. 49.
Tapsell, S.M., Penning-Rowsell, E.C., Tunstall, S.M., Wilson, T.L., 2002. Vulnerability
to flooding: health and social dimensions. Philosophical Trans. R. Soc. Lond. Ser.
A Math. Phys. Eng. Sci. 360, 1511e1525.
Tate, E., 2012. Social vulnerability indices: a comparative assessment using uncer-
tainty and sensitivity analysis. Nat. Hazards 63, 325e347.
Tate, E., Cutter, S.L., Berry, M., 2010. Integrated multihazard mapping. Environ. Plan.
B Plann. Design 37, 646e663.
Tate, E.C., 2011. Indices of Social Vulnerability to Hazards: Model Uncertainty and
Sensitivity, College of Arts and Sciences. University of South Carolina.
Trigo, R.M., Palutikof, J.P., 2001. Precipitation scenarios over Iberia: a comparison
between direct GCM output and different downscaling techniques. J. Clim. 14,
4422e4446.
UNISDR, 2009. Terminology on Disaster Risk Reduction. United Nations, Geneva,
Switzerland.
Whitlow, T.H., Harris, R.W., 1979. Flood Tolerance in Plants: a State-of-the-art Re-
view. U.S.Army Engineer Waterways, Vicksburg, Mississippi.
Wisner, B., Blaikie, P., Cannon, T., Davis, I., 2004. At Risk e Natural Hazards, People's
Vulnerability and Disasters, second ed. Routledge, London.
Wu, J., 2012. Cluster analysis and K-means clustering: an introduction. In: Wu, J.
(Ed.), Advances in K-means Clustering: a Data Mining Thinking. Springer Berlin
Heidelberg, Berlin, Heidelberg, pp. 1e16.
^zere, J.L., Pereira, S., Tavares, A.O., Bateira, C., Trigo, R.M., Quaresma, I., Santos, P.P.,
Ze
Santos, M., Verde, J., 2014. DISASTER: a GIS database on hydro-geomorphologic
disasters in Portugal. Nat. Hazards 72, 503e532.
Zhou, Y., Li, N., Wu, W., Wu, J., Shi, P., 2014. Local spatial and temporal factors
influencing population and societal vulnerability to natural disasters. Risk Anal.
34, 614e639.