You are on page 1of 30

Bull Earthquake Eng (2011) 9:1067–1096

DOI 10.1007/s10518-010-9233-3

ORIGINAL RESEARCH PAPER

Seismic vulnerability and risk assessment: case study


of the historic city centre of Coimbra, Portugal

Romeu Vicente · Sonia Parodi · Sergio Lagomarsino ·


Humberto Varum · J. A. R. Mendes Silva

Received: 9 October 2009 / Accepted: 13 December 2010 / Published online: 31 December 2010
© Springer Science+Business Media B.V. 2010

Abstract Seismic risk evaluation of built-up areas involves analysis of the level of
earthquake hazard of the region, building vulnerability and exposure. Within this approach
that defines seismic risk, building vulnerability assessment assumes great importance, not
only because of the obvious physical consequences in the eventual occurrence of a seis-
mic event, but also because it is the one of the few potential aspects in which engineering
research can intervene. In fact, rigorous vulnerability assessment of existing buildings and
the implementation of appropriate retrofitting solutions can help to reduce the levels of phys-
ical damage, loss of life and the economic impact of future seismic events. Vulnerability
studies of urban centres should be developed with the aim of identifying building fragilities
and reducing seismic risk. As part of the rehabilitation of the historic city centre of Coimbra,
a complete identification and inspection survey of old masonry buildings has been carried
out. The main purpose of this research is to discuss vulnerability assessment methodologies,
particularly those of the first level, through the proposal and development of a method pre-
viously used to determine the level of vulnerability, in the assessment of physical damage
and its relationship with seismic intensity. Also presented and discussed are the strategy and

R. Vicente (B) · H. Varum


University of Aveiro, Aveiro, Portugal
e-mail: romvic@ua.pt
H. Varum
e-mail: hvarum@ua.pt

S. Parodi · S. Langomarsino
University of Genoa, Genoa, Italy
e-mail: sonia.parodi@unige.it
S. Langomarsino
e-mail: sergio.lagomarsino@unige.it

J. A. R. M. Silva
University of Coimbra, Coimbra, Portugal
e-mail: raimundo@dec.uc.pt

123
1068 Bull Earthquake Eng (2011) 9:1067–1096

proposed methodology adopted for the vulnerability assessment, damage and loss scenarios
for the city centre of Coimbra, Portugal, using a GIS mapping application.

Keywords Historic city centres · Traditional masonry buildings · Vulnerability ·


Seismic risk · Vulnerability index method · Damage scenarios · Loss estimation ·
GIS mapping

1 Introduction

The first attempts to characterise building vulnerability were made in the early 1980’s in the
United States and Central and Eastern Europe, particularly in seismically-prone countries
such Italy, Bulgaria, Romania and Greece. In the U.S., the Federal Emergency Management
Agency (FEMA) developed a tool (HAZUS 1999) for assessing earthquake losses in terms
of the built environment and population of urban areas. In Italy, the Servizio Sísmico Italiano
(GNDT 1994) has developed rapid screening methods which have been validated over time
by information gathered from post-seismic damage data.
Countries subject to low to moderate seismic hazard, such as Portugal, have suffered
strong earthquakes resulting in many casualties and considerable economic losses (Barata
2005). Older masonry constructions typically present several structural fragilities and are
therefore at significant risk if subjected to even moderate seismic activity. Evaluation of
potential damage and loss scenarios for future earthquake events is therefore essential in the
mitigation of seismic risk, even for regions with moderate to low hazard.
The conceptual and methodological aspects of seismic risk evaluation at an urban scale
define the nature of a holistic approach (Carreño et al. 2007). Within such an approach, vulner-
ability assessment of buildings is a complex yet essential process involving a variety of con-
siderations, including: (i) definition of the operational scale of the assessment: regional, urban
area, city block or single building; (ii) the financial investment and resources required; (iii)
the final objective of the assessment: urban planning, definition and optimisation of building
retrofitting strategies, emergency management planning; and, (iv) uniqueness of the assess-
ment methodology developed, driven by building surveys with different levels of inspection
and recording for a specific building type, construction technology and features, etc.
Stone masonry buildings represent the majority of the existing building stock in historic
city centres throughout Europe, including many classified as having World Heritage status by
UNESCO (United Nations Educational Scientific and Cultural Organization). However, such
buildings do not have adequate seismic capacity and consequently require special attention
due to their incalculable historical, cultural and architectural value. The amount of resources
spent on the vulnerability assessment and structural safety evaluation of these old masonry
buildings is justifiable, since not only does a first level assessment (Vicente 2008) include
building inspection, but also can help in the identification of buildings for which a more
detailed assessment is required, as well as the definition of priorities for both retrofitting and
in support of earthquake risk management.

2 Historic city centre of coimbra

2.1 Renovation and rehabilitation process

The renovation of the old city centre of Coimbra is a nationally unique urban rehabilitation
project. Other Portuguese cities have started renewal activities, many of which are ongoing,

123
Bull Earthquake Eng (2011) 9:1067–1096 1069

Fig. 1 Project perimeter of the renovation and rehabilitation of the old city centre of Coimbra

but none have undertaken the renewal process in such an organised, integrated and methodo-
logical manner. In order to survey and study the old city centre, the project perimeter (Fig. 1)
was divided into eight intervention zones. Each zone includes buildings that share architec-
tural, functional and occupational characteristics, with each zone studied and analysed as a
whole in the final stage of the rehabilitation process.
The renewal process started off as more of a heritage conservation concern, with the
aim of developing a building stock management platform leading to social and economic
improvement. However it was very soon understood that such a process could not be sus-
tainable without a vulnerability assessment and risk evaluation. Therefore the city council
invited the University to carry out a complete identification and inspection survey of old
masonry buildings (Vicente et al. 2005). The main goal was to develop a reliable first level
seismic vulnerability assessment approach that could be used to assess urban areas, specifi-
cally old masonry buildings. From the 1970’s onwards, several major cities in Portugal began
the renovation of their historic city centres, including Lisbon, Oporto, Faro and Guimarães.
Consequently assessment was just the first step, with the results used in the development
of damage distribution and loss estimation scenarios. The methodology developed here is
implemented within a GIS tool and is easily adaptable for eventual use in other case studies.

2.2 Inspection and appraisal: database

As part of the renovation and rehabilitation of this part of the city, a complete identification
and inspection survey of old masonry buildings was carried out on three different domains:
(a) architectural typologies and drawings; (b) structural and non-structural building features
and defects; and (c) socio-demographic characterisation. All data gathered was processed
and a database management system integrated into a GIS application developed to manage,
compare and analyse the information collected. The use of the information gathered in the
database is of particular importance to the seismic vulnerability assessment of old masonry
buildings at an urban scale, as will be further explained in Sect. 3.3.
This unique opportunity allowed the appraisal, diagnosis and inspection of building stock
dating from the XVII to the XX centuries, using detailed checklists. The information collected

123
1070 Bull Earthquake Eng (2011) 9:1067–1096

enabled the seismic vulnerability assessment of 679 buildings, with diverse construction fea-
tures. The development of such a building inventory is a key factor in assessing a wide
area.

2.3 Development and implementation of a GIS tool

Risk management of urban areas is in many cases undertaken without the use of a general
planning tool. A primary consequence of the absence of such a tool is that technicians and
decision makers (city councils or regional authorities) do not have a global view of the site
under analysis, which can lead to inadequate decisions being made in terms of rehabilitation
strategies and risk mitigation measures. This justifies the need for a multi-purpose tool con-
nected to a relational database and within a GIS environment, with which to apply a deeper
and integrated analysis of the project area and hence to manage data regarding historic build-
ing stock characteristics, conservation requirements, seismic vulnerability, damage and loss
scenarios, cost estimation and risk-impact assessment.
The GIS application software adopted in this study was ArcGis 9.3 (ESRI 2008), con-
nected to a relational database (see Fig. 2) of the structural building parameters that govern
the structural behaviour of old masonry buildings, with a particular focus on their seis-
mic vulnerability. In the GIS environment, graphical units (polygons) were mapped as
buildings registered from land surveys and associated with several features and attributes
allowing their visualisation, selection and multiple searches. The maps produced combine
geo-referenced data of building location as a graphical unit (in this case a polygon) linked to
building parameter information.
All routines were programmed and compiled in Visual Basic and in an ArcGis
9.3-compatible (ESRI 2008) programming language on a Windows platform. Various mod-
ules were developed via numerical algorithms for different tasks, including: visualisation of
general information and results by zones, vulnerability assessment, damage and loss esti-
mation for different earthquake intensities and information of the building parameters and
features used to estimate vulnerability. All database information associated with the GIS
application can be periodically updated and as such the latter is a valuable tool for the man-
agement of the old building stock, allowing data storage as well as having the capacity
for spatial analysis enabling the visualisation of data and results for different earthquake
scenarios.

Fig. 2 Database and GIS framework

123
Bull Earthquake Eng (2011) 9:1067–1096 1071

3 Vulnerability assessment and risk evaluation

Part of the broader definition of risk provided by the United Nations Disaster Relief Office,
vulnerability may also be represented by a variable in a mathematical formulation used in
risk evaluation. It is agreed by most authors, such as Caicedo et al. (1994), Cardona (2001)),
Coburn and Spence (2002), McGuire (2004)) and Barbat et al. (2010), that absolute risk can
be expressed as the result of the mathematical relationship between hazard, vulnerability and
exposure:

Rie |T = |(Hi ⊗ Ve ) ⊗ E|T (1)

where R is the probability of exceedance of a certain level of loss of an exposed element e, as


a consequence of the occurrence of a seismic event of certain intensity i, H is the probability
of exceedance of a certain level of seismic activity of intensityi, during a specified recurrence
period T , V is the vulnerability, that is the intrinsic predisposition of a certain element e to
suffer damage resulting from a seismic event of intensity i, E is the exposure of the elements
at risk, reflecting the value of the elements exposed.
Vulnerability in this case is defined as the seismic vulnerability of buildings to suffer a
certain level of damage when subjected to a seismic event of defined intensity (Lang and
Bachmann 2003). On one hand it is used to indirectly measure reduction in building struc-
tural efficiency and on the other a building’s residual ability to guarantee its expected use
and function under normal conditions.
A number of methods have been developed during recent years to estimate damage for
a given level of ground shaking. Many of the ground motion-damage relationships used
were proposed independent of building vulnerability evaluation or classification, with most
established on the basis of post-seismic observations of damaged structures (ASER 2006).
However such observations are not available in regions of moderate to low hazard. Within the
framework of an inventorisation and inspection of building stock, the cost and time required
to carry out seismic vulnerability assessment is substantially reduced, since it is integrated
into the survey process (in this case involving a large number of buildings). All these reasons
provided motivation for the development and validation of such an approach for the renewal
process taking place here; the assessment of the old masonry buildings of the historic city
centre of Coimbra.

3.1 Vulnerability and risk assessment methodologies

When assessing the seismic vulnerability of buildings it is essential to first establish the project
objectives, before subsequently choosing the most appropriate strategy and tools necessary
for building assessment and fulfilment of these objectives. It is also extremely important to
understand the difference between the detailed approaches used for individual buildings and
those methods most efficient for larger scale analysis of groups of buildings. With respect
to the former, the use of a detailed methodology implies a very reliable evaluation with a
necessarily in-depth level of information regarding the analysed structure. However when
increasing the number of buildings and enlarging the area to be assessed, the resources and
quantity of information required are also increased and thus the use of less sophisticated and
onerous inspection and recording tools is more practical. Methodologies for vulnerability
assessment at the national scale should be based on few parameters, some of an empirical
nature based on knowledge of the effects of past earthquakes but also including statistical
analysis.

123
1072 Bull Earthquake Eng (2011) 9:1067–1096

Recently, European partnerships (RISK-UE 2004; LESSLOSS 2007) constituting various


workgroups on different aspects of vulnerability assessment and earthquake risk mitigation
have defined, particularly for the former, methodologies that are grouped into essentially three
categories in terms of their level of detail, scale of evaluation and use of data (first, second
and third level approaches). First level approaches use a considerable amount of qualitative
information and are ideal for the development of seismic vulnerability assessment for large
scale analysis. Second level approaches are based on mechanical models and rely on a higher
quality of information (geometrical and mechanical) regarding building stock. The third and
final level involves the use of numerical modelling techniques that require a complete and
rigorous survey of individual buildings. The definition and nature of the analysed criteria
(qualitative and quantitative) naturally condition the formulation of the methodologies and
their respective evaluation level, which can vary from the expedite evaluation of buildings
based on visual observation to the most complex numerical modelling of single buildings
(see Fig. 3).
The systemisation of these vulnerability assessment approaches has been developed by
many researchers and therefore differs due to varying levels of dependence of the following
factors: nature and objective of the assessment, quality and availability of information, char-
acteristics of the building stock inspected, scale of assessment, methodology criteria, degree
of reliability of the expected results and use by the end-user of the information produced. As
a result of these differences, the coherency of and consensus regarding this classification is
still a contentious issue.
In this context, we will now briefly describe the most important approaches (in terms of
the genesis of the methods and criteria used) of the many developed. Of the many different
classification systems developed over recent years, the first was by Corsanego and Petrini

B U ILD IN G AG G R E G ATE U R B A N A R EA

D etailed analysis m ethods (num erical and m echanical)

Experim ental
m ethods

Typoligical techniques

Indirect techniques

C onventional techniques

H ybrid techniques

Fig. 3 Analytical techniques used at different scales

123
Bull Earthquake Eng (2011) 9:1067–1096 1073

(1990) who proposed that the different methods be divided into four groups according to their
intended result: direct, indirect, conventional and hybrid techniques. Other proposed classifi-
cation systems are based either on analysis of the fundamental steps (data, methodology and
results) within each approach: statistic-based, mechanical and expert opinion-based methods;
or as a function of the quality of the source of information: empirical, analytical-theoretical
and experimental methods (Dolce et al. 1994).
The four different groups according to the most widely-recognised and complete clas-
sification system developed by Corsanego and Petrini (1990) are presented briefly in the
following section.

(i) Direct techniques use only one step to estimate the damage caused to a structure by
an earthquake, employing two types of methods; typological and mechanical:
Typological methods—classify buildings into classes depending on materials, con-
struction techniques, structural features and other factors influencing building
response. Vulnerability is defined as the probability of a structure to suffer a cer-
tain level of damage for a defined seismic intensity. Evaluation of damage probability
is based on observed and registered damage after previous earthquakes and also on
expert knowledge. Results obtained using this method must be considered in terms of
their statistical accuracy, since they are based on simple field investigation. In effect
the results are valid only for the area assessed, or for other areas of similar construction
typology and equal level of seismic hazard. Examples of this method are the vulnera-
bility functions or damage probability matrices (DPM) developed by Whitman et al.
(1974), in which a matrix for each building type or vulnerability class is defined that
directly correlates seismic intensity with probable level of damage suffered.
Mechanical methods—predict the seismic effect on the structure through the use of
an appropriate mechanical model, which may be more or less complex, of the whole
building or of an individual structural element. This type of method can be classified
into different groups depending on the type of model used to represent the structure,
i.e., analytical methods based on simple models and those based on more detailed
analysis. Methods based on simplified mechanical models are more suitable for the
analysis of a large number of buildings and require only a few input parameters. These
simple numerical procedures require an equivalently short time period for analysis
and can lead to reliable qualitative evaluations. A commonly used method belonging
to this group is the limit state method, based on limit state analysis (displacement
capacity and demand). Calvi (1999) applied this method to the analysis of the historic
city centre of Catania. The FAMIVE method (Speranza 2003) is another example of
a reliable mechanically-based method. The second group, including the more sophis-
ticated methods, are generally used to evaluate single structures at a higher level of
detail (in terms of building structure and construction) and are based on more refined
modelling techniques. Normally these methods are not applicable for seismic risk
assessment at an urban scale, which requires the evaluation of a great number of
buildings for several types of seismic action. The analytical procedure for this type of
method can involve the following: linear static, non-linear static, linear dynamic and
non-linear dynamic analysis. For example, the ATC-40 (1996), FEMA 273 (1996)
and Capacity Spectrum Method (CSM) Fajfar (1999) are non-linear static analysis
procedures.
(ii) Indirect techniques initially involve the determination of a vulnerability index, fol-
lowed by establishment of the relationships between damage and seismic inten-
sity, supported by statistical studies of post-earthquake damage data. This form of

123
1074 Bull Earthquake Eng (2011) 9:1067–1096

evaluation is used extensively in the analysis of vulnerability on a wide scale. Of the


various techniques currently available, the methodology developed by GNDT in Italy
is still widely applied. Created in the 1980’s, the GNDT method was initially applied
in the analysis of several historic city centres in various regions of Italy; for example
Catania in 1999 and Molise in 2001 (GNDT 1994). The method involves the deter-
mination of a building vulnerability classification system (vulnerability index) via
observation of physical construction and structural characteristics, based on simple
structural calculations and on the identification of factors responsible for the control
of building seismic response. Each building is classified in terms of a vulnerability
index related to a damage grade determined via the use of vulnerability functions.
These functions enable the formulation, in a fast and simple manner, of the damage
suffered by buildings for each level of seismic intensity (or peak ground acceleration,
PGA) and vulnerability index. These types of methods use extensive databases of
building characteristics (typological and mechanical properties) and rely on observed
damage after previous earthquakes to classify vulnerability, based on a score assign-
ment. The rapid screening ATC-21 technique (1988) is extensively used in the U.S. to
obtain such a vulnerability score.
(iii) Conventional techniques are essentially heuristic, introducing a vulnerability index
independently of the prediction of the level of damage. They are used to compare
different constructions of the same typology in a certain region, analysing features
that predominantly affect seismic resistance and calibrated by expert opinion. These
indexes provide a relative measure of vulnerability, since the obtained results are
difficult to relate to other constructive typologies due to the difference between the
considered features. There are essentially two types of approach: those that qualify the
different physical characteristics of structures empirically and those based on the crite-
ria defined in seismic design standards for structures, evaluating the capacity-demand
relationship of buildings. ATC-13 (1985) is one of the most well-known approaches
of this type. Here damage probability matrices were built for 78 classes of structure,
40 of which refer to buildings. Starting from the basis of expert judgement, ATC-13
estimates the expected percentage of damage a certain structure can suffer if subjected
to a seismic event of certain intensity. The degree of uncertainty is assumed as a limi-
tation, treated through probabilistic processing. Although this method does not allow
the introduction of new data or its application to other buildings and regions, it has
been used widely in many studies of seismic vulnerability in recent decades. Another
example belonging to this category of technique is the HAZUS (1999) methodology of
the National Institute of Building Science (NIBS), which characterises seismic action
through acceleration and displacement spectra. This method employs the relative sub-
jectivity of expert opinion for the classification of damage for 36 types of structural
model, based on the classification made in FEMA 178 (1992). For each construction
type and level of earthquake-resistant design, the capacity of the structure, spectral
displacement and inter-story drift limit are defined for different levels of damage.
(iv) Hybrid techniques combine features of the methods described previously, such as
vulnerability functions based on observed vulnerability and expert judgment, in which
vulnerability is based on the vulnerability classes defined in the European Macroseis-
mic Scale, EMS-98 (Grünthal 1998). This is the case in the Macroseimic Method
devised by Lagomarsino and Giovinazzi (2006), which combines the characteris-
tics of typological and indirect methods using the vulnerability classes defined in
the EMS-98 scale and a vulnerability index improved by the use of modification
factors.

123
Bull Earthquake Eng (2011) 9:1067–1096 1075

Table 1 Vulnerability index (Iv )

Parameters Class Cvi Weight Vulnerability index


A B C D pi

1. Structural building system


14
P1 Type of resisting system 0 5 20 50 0.75 I V∗ = i=1 C V i × pi
P2 Quality of the resisting system 0 5 20 50 1.00
P3 Conventional strength 0 5 20 50 1.50
P4 Maximum distance between walls 0 5 20 50 0.50
P5 Number of floors 0 5 20 50 1.50
P6 Location and soil conditions 0 5 20 50 0.75
2. Irregularities and interaction
P7 Aggregate position and interaction 0 5 20 50 1.50 0 ≤ I V∗ ≤ 650
P8 Plan configuration 0 5 20 50 0.75
P9 Regularity in height 0 5 20 50 0.75
P10 Wall façade openings and alignments 0 5 20 50 0.50
3. Floor slabs and roofs
P11 Horizontal diaphragms 0 5 20 50 1.00
P12 Roofing system 0 5 20 50 1.00
4. Conservation status and other elements
P13 Fragilities and conservation state 0 5 20 50 1.00 Normalised index
P14 Non-structural elements 0 5 20 50 0.50 0 ≤ I V ≤ 100

From analysis of these approaches, their applicability in terms of project scale (note that
the range defined varies from case to case) is shown schematically in Fig. 3.

3.2 Proposed methodology for the vulnerability assessment of the historic city centre of
Coimbra

The methodology applied here in the evaluation of the vulnerability of the entire historic city
centre of Coimbra can be considered a hybrid technique in respect to criteria presented in
the previous section. The vulnerability index formulation proposed is based essentially on
the GNDT II level approach (GNDT 1994), for the vulnerability assessment of residential
masonry buildings. This methodology is based on post-seismic damage observation and sur-
vey data covering a vast area, focussing on the most important parameters affecting building
damage which must be surveyed individually.
Overall vulnerability is calculated as the weighted sum of 14 parameters (see Table 1)
used in the formulation of the seismic vulnerability index. These parameters are related to
4 classes of increasing vulnerability: A, B, C and D. Each parameter represents a building
feature influencing building response to seismic activity. A weight pi is assigned to each
parameter, ranging from 0.50 for the less important parameters (in terms of structural vulner-
ability) up to 1.5 for the most important (for example parameter P3 represents conventional
strength) as shown in Table 1. The vulnerability index obtained as the weighted sum of the 14
parameters initially ranges between 0 and 650, with the value then normalised to fall within
the range 0 < I v < 100. The calculated vulnerability index can then be used to estimate
building damage after a specified intensity of seismic event, as will be discussed in Sect. 4.2.

123
1076 Bull Earthquake Eng (2011) 9:1067–1096

This procedure has been used in Italy for the last 25 years, but has been adapted here for use
with Portuguese masonry buildings and improved by: (i) introducing a more detailed analysis,
derived from the good level of building stock information; (ii) the discussion and redefinition
of the criteria of some of the most important parameters; and (iii) the introduction of new
parameters that take into account the interaction between buildings (structural aggregates)
and other overlooked building features. The insertion of parameters P5, P7 and P10 introduces
the following features: the height of the building (P5); the interaction between contiguous
buildings (P7)—a very important feature when assessing buildings in urban areas; and the
alignment of wall façade openings which affects the load path and load bearing capacity
(P10).
The 14 parameters are arranged into four groups, as shown in Table 1, in order to empha-
sise their differences and relative importance (see Vicente 2008). The first group includes
parameters (P1, P2) characterising the building resisting system, the type and quality of
masonry, through the material (size, shape and stone type), masonry fabric, arrangement and
quality of connections between walls. Parameter P3 roughly estimates the shear strength
capacity of the building, parameter P4 evaluates the potential out-of-plane collapse risk,
while parameters P5 and P6 evaluate the height and soil foundation conditions of the build-
ings. The second group of parameters is mainly focused on the relative location of a build-
ing in the area as a whole and on its interaction with other buildings (parameter P7). This
feature is not considered in other methodologies but we consider it extremely important,
since the seismic response of a group of buildings is rather different to the response of
a single building. Parameters P8 and P9 evaluate irregularity in plan and height, while
parameter P10 identifies the relative location of openings, which is important in terms
of the load path. The third group of parameters, which includes P11 and P12, evaluates
horizontal structural systems, namely the type of connection of the timber floors and the
impulsive nature of pitched roofing systems. Finally, parameter P13 evaluates structural
fragilities and conservation level of the building, while parameter P14 measures the neg-
ative influence of non-structural elements with poor connections to the main structural
system.
The definition of each parameter weight is a major source of uncertainty, despite being
based on expert opinion. Parameters were grouped to emphasize the differences and rela-
tive importance amongst them (see Vicente 2008) Consequently in order for the results to
be accurately interpreted statistically, upper and lower bounds of the vulnerability index, Iv
were defined.
The method proposed here is considered robust, taking into account that the inspection of
the majority of buildings was carried out in detail and accurate geometrical information was
available. Therefore, uncertainty in the assignment of vulnerability classes to each parameter
is considered low. A confidence level indicator was associated with each parameter, so that
the vulnerability index is also coupled to a confidence level rating (see Vicente 2008).
Figure 4 illustrates the distribution of the most important parameters in terms of their over-
all influence in the vulnerability assessment, as well as the vulnerability class distribution
of each parameter evaluated in the detailed vulnerability assessment of over 410 masonry
buildings.
Among all parameters evaluated, P3, P5 and P7 (see Fig. 4) have the highest weight val-
ues ( pi ) in the vulnerability index. However in the case study presented here, parameters
P4, P9 and P10 with weights ranging from 0.5 to 0.75, also have a significant influence on
the final vulnerability index, due to their low classification (mostly C to D). Parameters P2,
P11, P12 and P13, with weights ( pi ) of 1.00, are those parameters whose increase could be
defined as representing a strengthening action (masonry consolidation, timber floor stiffening,

123
Bull Earthquake Eng (2011) 9:1067–1096 1077

Parameter 3 Parameter 5 Parameter 8


500 500 500

400 400 400

Frequency
Frequency
Frequency

300 300 300

200 200 200

100 100 100

0 0 0
A B C D A B C D A B C D
Vulnerability Class Vulnerability Class Vulnerability Class

Parameter 9 Parameter 12 Parameter 13


500 500 500

400 400 400


Frequency

Frequency

Frequency
300 300 300

200 200 200

100 100 100

0 0 0
A B C D A B C D A B C D
Vulnerability Class Vulnerability Class Vulnerability Class

100%

90%

80%
D
70%
C
60%
Influence

B
50%
A
40%

30%

20%

10%

0%
P1 P2 P3 P4 P5 P6 P7 P8 P9 P10 P11 P12 P13 P14
Vulnerability parameters

Fig. 4 Parameter distribution and its influence on the vulnerability index for buildings assessed in Coimbra

retrofitting of trussed roofing systems, effective connection between horizontal and vertical
structural elements and building maintenance strategy).

3.3 Vulnerability assessment completion

The methodology proposed here requires accurate knowledge of building characteristics,


which can only be obtained via thorough and detailed inspection. In accordance with this,
evaluation of vulnerability was undertaken in two phases. In the first phase, an evaluation
of vulnerability index, Iv , was made for those buildings for which detailed information was
available (410 buildings out of 679). In the second phase a more expeditious approach to the
assessment of the remaining buildings was adopted in function of the mean values obtained
from the detailed analysis of the first group of buildings (Sect. 3.2), considering that the
masonry building characteristics are homogeneous in this region.

123
1078 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 5 Vulnerability modifier factors and scores

As a lack of information meant carrying out a more detailed study of these 269 buildings
was not possible, a simpler approach (Giovinazzi and Lagomarsino 2004) was used. Since
masonry building characteristics tend to be homogeneous in this region of Portugal, the vul-
nerability index values of the second group of buildings were taken as a function of the mean
values obtained from the detailed analysis of the first group.
The mean vulnerability index value obtained for all masonry buildings in the first detailed
evaluation was used as a typological vulnerability index (average value) that can be affected
by modifiers of the mean vulnerability index for each building. Classification of these modi-
fiers can influence the final vulnerability index as the sum of the scores for all modifiers. The
modifiers are some of the vulnerability index definition parameters shown in Table 1.
Figure 5 presents the seven modifier parameters and their scores in relation to the average
vulnerability value for each parameter. The vulnerability index, Iv , is defined according to
the sum of the modifier parameter scores for each non-detailed assessment. The scores for
each parameter are defined with respect to the average value of that parameter used in the
average vulnerability index value and the weight of each parameter in the overall definition.
For example, as the mean vulnerability class value for parameter P8 (plan configuration)
derived from the detailed vulnerability assessment of the 410 buildings is taken as that of
class C, the modifier scores are therefore only considered if there is a shift in parameter
vulnerability class that increases or reduces the final vulnerability index (see Fig. 5).
The final vulnerability index is defined as

I V = I¯V + I V (2)

where I V is the final vulnerability


 index, I¯v is the average vulnerability index from the
detailed assessment and Iv is the sum of the modifier scores. This strategy is only valid
if a reliable detailed assessment of a large number of buildings in the study area is initially
obtained, and also only when applied to a single building typology (in this case masonry
buildings). All results from both approaches can be found in Sect. 4.1.
Using the vulnerability functions developed by Giovinazzi (2005) it is also possible to
correlate vulnerability with damage and loss estimation, as will be explained in Section 3.5.

3.4 Validation and implementation of the methodology

3.4.1 GNDT II level approach

The method proposed here is based on the original GNDT II level approach (GNDT 1994)—
although with some modifications (as discussed in Sect. 3.2). However the ‘backbone’ param-
eters, shared by every vulnerability assessment according to Combescure et al. (2005), are

123
Bull Earthquake Eng (2011) 9:1067–1096 1079

essentially the same. Considering this fact, the similarity in terms of the definition of the
vulnerability index of the two methodologies (original and improved) enables the use of the
same vulnerability functions relating vulnerability to a damage index (Benedetti and Petrini
1984)
In the GNDT II level methodology, the vulnerability index is used as an intermediate step
to estimate damage suffered by a building after seismic activity of specified intensity. For
each value of the vulnerability index, a deterministic correlation between the seismic event
(expressed in terms of PGA) and damage (expressed as an economic index d ranging from
0 to 1) is defined (see Fig. 6). The index d represents the cost of returning a building to
its original undamaged condition, taking into account the actual value of the building. In
terms of the damage index, a d value ranging from 0.8 to 1.0 is conventionally considered
as representing building collapse.
To simplify the procedure a trilinear function is assumed for each vulnerability level, with
the value of (yi) corresponding to the initial acceleration that induces damage and that of (yc)
to collapse. Figure 6 illustrates the relationship between damage and seismic action (in terms
of intensity or ground acceleration) for different values of building vulnerability.
The choice of acceleration as a parameter indicative of the severity of the seismic event
is derived from the idea of using a continuous parameter—even if the data set is expressed
in terms of macroseismic intensity (MCS). This study adopted the following expression
proposed by Guagenti and Petrini (1989) relating seismic intensity with acceleration:
ln(y) = a × IMCS − b with a = 0.602, b = 7.073 (3)
where y represents the PGA, IMCS the intensity according to the MCS scale and a and b are
constants given by the values referred to above.

3.4.2 Macroseismic method

Since this study adopted the analytical vulnerability curves of the Macroseismic Method
(Giovinazzi and Lagomarsino 2004), it is essential to establish the correspondence between
the Macroseismic Method and the GNDT II level approach.
The former method makes reference to the EMS-98 Macroseismic Scale (Grünthal 1998),
which implicitly contains a model of vulnerability. On the basis of the definition of damage

Fig. 6 Vulnerability functions: (a) Trilinear function (Grimaz et al. 1996); (b) Vulnerability curves proposed
by Benedetti and Petrini (1984)

123
1080 Bull Earthquake Eng (2011) 9:1067–1096

described in the EMS-98 scale (Grünthal 1998), it is possible to derive damage probability
matrices for each of the defined vulnerability classes (A to F). Through use of the linguistic
definitions (Few, Many and Most) and their respective numerical interpretation, complete
Damage Probability Matrices (DPM) for every vulnerability class may be obtained. Having
solved the incompleteness using probability theory, the ambiguity and overlap of the lin-
guistic definitions is then tackled using fuzzy set theory (Giovinazzi 2005), in which upper
and lower boundary limits for the correlation between the macroseismic intensity and mean
damage grade of the distribution are defined and derived for each building typology and
vulnerability class (Grünthal 1998).
A conventional vulnerability index, V , is then introduced in which each building belongs
to a vulnerability class. The numerical values of this index are arbitrary, being only scores
with which to quantify building behaviour in a conventional manner. The vulnerability
index (ranging from 0 to 1) is initially defined on the basis of typological information
(building type), but is subsequently affected by modification factors such as maintenance
condition, materials and construction quality, number of floors, plan and height irregu-
larity and the level of earthquake resistant design. These behaviour modifiers were cho-
sen on an empirical basis, involving the observation of typical damage patterns and
suggestions in documents and other studies [e.g. ATC-21 1988; Benedetti and Petrini 1984;
UNDP/UNIDO 1985]. Modifying scores are attributed through expert judgment, with fur-
ther calibration recommended based on damage and vulnerability data collected after future
earthquakes.
For the operational implementation of the methodology, an analytical expression was
proposed (Bernardini et al. 2007) which correlates hazard with the mean damage grade
(0 < μ D < 5) of the damage distribution (discrete beta distribution) in terms of the vulner-
ability value, as shown in Eq. 4.
 
I + 6.25 × V − 12.7
μ D = 2.5 + 3 × tanh × f (V, I ) 0 ≤ μ D ≤ 5 (4)
Q

where I is the seismic hazard described in terms of macroseismic intensity, V the vulner-
ability index, Q a ductility factor and f (V, I ) is a function of the vulnerability index and
intensity. The latter is introduced in order to understand the trend of numerical vulnerability
curves derived from EMS-98 DPMs for lower values of the intensity grades (I = V and VI)
where:

e V /2×(I −7) I ≤ 7
f (V, I ) = 0 ≤ μD ≤ 5 (5)
1 I >7

This analytical expression derives from the interpolation of vulnerability curves calculated
from the completed DPMs, as suggested in the EMS-98 scale. Used to estimate physical
damage, this mathematical formulation is based on work previously proposed by Sandi and
Floricel (1995).
The vulnerability index, V , determines the position of the curve, while the ductility fac-
tor, Q, determines the slope of the vulnerability function (rate of damage increases with
rising intensity). In this study a ductility factor of 3.0 was adopted, a value suggested as
providing the best fit for the comparison between the GNDT curves and EMS-98 functions.
The regression analysis and parametric studies performed by Sandi and Floricel (1995)
also lead to a mean value of 3.0 being suggested for masonry buildings of fairly ductile
behaviour.

123
Bull Earthquake Eng (2011) 9:1067–1096 1081

3.4.3 Comparison of methodologies

As the final vulnerability index is obtained differently in each method, a final comparison
between the GNDT II level approach and the Macroseismic method is made through the use
of a mean damage grade. By correlating the two measures of damage, it is then possible to
define the relationship between the measures of vulnerability used by the two methodolo-
gies. With reference to Eq. 2, ground acceleration, y (PGA) can be converted into intensity
according to the MCS scale. With the purpose of equalising intensity values with the EMS-98
intensity scale (Grünthal 1998) in order to carry out a coherent comparison, the acceleration-
intensity relationships developed by Margottini et al. (1992) were employed, in which one is
defined in terms of I (MCS) and the other in terms of I (MSK). Using the equivalence of the
MSK and the EMS-98 scales (IMSK = IEMS-98 ), it is then possible to obtain the correlation
between the two measures of intensity, as defined below:

IMSK = 0.734 + 0.814 × IMCS (6)

In order to make an accurate comparison between the two methods, it is also necessary to
relate the economic damage index adopted in the GNDT II level approach (1994) to the
physical damage grades defined in the Macroseismic method. A number of different authors
have proposed correlations between these two measures of damage (see Table 2).
The mean damage grade, μ D , given by the Macroseismic Method represents the mean
value that is used to define a discrete damage distribution and is expressed as


5
μD = pk × D k (7)
k=0

where pk is the probability associated with a specific damage grade Dk , with k  [0–5]
and where the mean damage grade, μ D , is the barycentric value of the discrete damage
distribution.
Transformation of the economic damage index in terms of structural damage is achieved
using the correlation proposed by FEMA-NIBS (HAZUS 1999). In order to define the corre-
lation between the economic damage index, de , and the mean damage value, μ D , a probabi-
listic distribution of damage is assumed, derived from the discretisation of a beta distribution
defined in the interval [0–5]. For easier implementation of this procedure, the approximation
given by the following analytical function was used:

μ D = 4 × de0.45 (8)

Table 2 Correlation between damage grade and economic damage index

Damage grade, Dk 0 1 2 3 4 5
Level of damage No damage Slight Moderate Severe Very severe Destruction

Economic damage index, de


ATC-13 (1985) 0.000 0.050 0.200 0.550 0.900 1.000
Bramerini et al. (1995) 0.000 0.010 0.100 0.350 0.750 1.000
HAZUS (1999) 0.000 0.020 0.100 0.500 1.000 1.000
Dolce et al. (2000) 0.000 0.035 0.145 0.305 0.800 1.000

123
1082 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 7 Vulnerability curves obtained using the GNDT II level and Macroseismic method

Table 3 Correlation between the vulnerability indexes and the vulnerability classes defined in terms of the
EMS-98 scale
Macroseismic method Class A (V = 0.88) Class B (V = 0.72) Class C (V = 0.56)

GNDT II level Iv = 50 Iv = 25 Iv = 0

Once the transformations of acceleration (y) into intensity (IEMS-98 ) and the economic
damage index, d, into the mean damage grade, μ D , have been carried out, it is then possible
to compare the vulnerability curves of both methods in a IEMS-98 -μ D format as shown in
Fig. 7.
By comparing the two types of vulnerability curve with respect to a central mean damage
value (μ D = 2.5), the correlation presented in Table 3 was achieved.
Based on the comparison presented above, the following analytical correlation was derived
between the vulnerability indexes indices of the two methods:
V = 0.56 + 0.0064 × I V (9)
Via this relationship, the vulnerability index, Iv , can be transformed into the vulnerability
index, V (used in the Macroseismic Method), enabling the calculation of the mean damage
grade through Eq. 4 and subsequently the estimation of damage and loss. For those buildings
where detailed evaluation was not carried out, the mean vulnerability index can be defined
as a function of the vulnerability classes defined in terms of the EMS-98 scale. In this case,
the modifier parameters can also be expressed in vulnerability index V format, taking into
account the Iv values re-defined in Eq. 9.

4 Vulnerability assessment results, damage and loss scenarios

4.1 Seismic vulnerability assessment

The entire traditional limestone masonry building stock of the city centre of Coimbra was
assessed, with each building assigned a vulnerability index, Iv , value. The mean Iv value
lay between 38 and 39, corresponding to a building typology of field and simple stone and
vulnerability classes A to B by adjusting the construction description available in the EMS-
98 scale (Grünthal 1998). Table 4 shows the correspondence between various vulnerability

123
Bull Earthquake Eng (2011) 9:1067–1096 1083

Table 4 Mean vulnerability index (Iv ), vulnerability class and typology according to the EMS-98 scale

I V (Vulnerability index) EMS-98 Vulnerability class EMS-98 Typology

38.13 A and B (most probable) Field stone and Simple stone

Fig. 8 Vulnerability index: (a) histogram distribution; (b) adjusted normal distribution

classes and the calculated value of mean vulnerability. The mean value obtained suggests that
the building stock of Coimbra belongs to building typologies between M1 and M3 according
to the Macroseismic Method (Giovinazzi 2005).
While detailed evaluation of over 410 buildings resulted in a mean vulnerability index
value of Iv = 38.13, the use of the proposed complementary approach for the remaining
269 buildings led to very little change in this mean value, shifting it to Iv = 38.38 (see
Fig. 8). Around 39% of the building stock have a vulnerability index value of over 40, and
20% over 45 (equivalent to vulnerability class A). Only 1% of buildings have an Iv of below
20 (equivalent to vulnerability class B). The maximum and minimum Iv values obtained for
the whole building stock are 60.58 and 12.12, respectively. Figure 8 shows the vulnerability
index distribution histograms for the detailed approach and for the overall assessment using
the completion strategy presented in Sect. 3.3.
The standard deviations (σ I v ) of vulnerability index data obtained via the detailed anal-
ysis and the complete analysis incorporating the simpler vulnerability assessment for the
entire study area differ very little, being 9.12 and 7.86 respectively, representing a variance
of 14%. This completion strategy therefore seems reliable for the analysis of buildings of
similar typology, especially where the quality and reliability of the detailed assessment data
is high.
Seismic vulnerability was mapped using GIS software connected to a relational database
of structural building characteristics. Proper storage and management of data requires the
use of a GIS in order to obtain a spatial and organic view of the inter-connected problems
associated with a given study area. GIS tools were used in this project to provide a good
global overview of the vulnerability assessment and risk scenarios (including potential dam-
age mapping and loss scenarios for events of different seismic intensity, as will be presented
in a later section).
Figure 9 shows the spatial distribution of building stock seismic vulnerability. Such seis-
mic vulnerability maps enable the identification of more vulnerable areas and buildings,
which can be very useful for the planning of urban management and protection strategies.
From analysis of Fig. 9, the more vulnerable buildings are found in the northeast of the study
area, correlating with the extremely poor conservation status of the building stock in zones

123
1084 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 9 Building stock vulnerability maps: (a) Global vulnerability distribution; (b) Identification of buildings
with Iv >40

Z1 and Z2. It is also possible to conclude that the corner and end buildings of street blocks
are in general more vulnerable due to their location and interaction with other buildings. It
is recommended that masonry buildings with values of Iv over 45 should undergo a more
detailed assessment.

4.2 Building stock physical damage estimation scenarios

Once vulnerability has been defined, the mean damage grade, μ D , can be calculated for
different macroseismic intensities, using Eq. 4. Figure 10 shows vulnerability curves for the
mean value of the vulnerability index, Iv,mean , as well as for the upper and lower bound
ranges (Iv,mean − 2σ I v ; Iv,mean − 1σ I v ; Iv,mean + 1σ I v ; Iv,mean + 2σ I v ).
From these mean damage grade values, μ D , different damage distribution histograms for
events of varying seismic intensity and their respective vulnerability index values can be
defined, using a probabilistic approach. The most commonly-applied methods are based on
the binomial probability mass function and the beta probability density function.
n!
P M F : pk = × d k × (1 − d)n−k n ≥ 0 ; 0≤ p≤1 (10)
k!(n − k)!
The damage distribution adopted was fitted to a beta distribution function, where t and r
are geometric parameters associated with the damage distribution. Research carried out by
Giovinazzi (2005) has shown that the beta distribution is the most versatile, as by controlling
the shape of the distribution via the parameters t and r , it enables the fitting of both very
narrow and broad damage distributions. This continuous beta probability density function in
which  is the known gamma function is expressed as the following:

123
Bull Earthquake Eng (2011) 9:1067–1096 1085

Fig. 10 Vulnerability curves for the building stock of the historic city centre of Coimbra

 (t)
P D F : pβ (x) = (x − a)r −1 (b − x)t−r −1 ; a ≤ x ≤ b ; a = 0; b = 5
 (r )  (t − r )
(11)
Assuming that a = 0 and b = 5, it can then be simplified to
pβ (x) = k(t, r ) × x r −1 × (5 − x)t−r −1 (12)
where for a continuous variable x, the variance (σx2 )
and the mean value (μx ) of the values
are related to t and r as defined below:
μx (5 − μx )
t= −1 (13)
σx2
μx
r =t· (14)
5
The aim of this study is to obtain a discrete distribution in which the probability associated
with each damage grade, Dk , with k [0, 5], is defined as:
0.5
P(D0 ) = p(0) = k(t, r ) · x r −1 (5 − x)t−r −1 d x
0

k+0.5

P(Dk ) = p(k) = k(t, r ) · x r −1 (5 − x)t−r −1 d x (15)


k−0.5
5
P(D5 ) = p(5) = k(t, r ) · x r −1 (5 − x)t−r −1 d x
4.5

For the definition of parameters t and r in the beta discrete distribution, the numerical damage
distributions derived from the EMS-98 scale (Bernardini et al. 2007) are used. The reduced
variation obtained for parametert in the numerical damage distributions justifies the adoption
of a unique value of t (equal to 8) with which to represent the variance of all possible damage
distributions. Based on this assumption, it is then possible to define the damage distributions

123
1086 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 11 Discrete damage distribution histograms for Iv = 38.38: (a) I (EMS-98) = VIII; (b) I (EMS-98) = IX

exclusively through use of the average value μ D , characterised by variance coherent with
that found via completion of the EMS-98 DPM’s.
μD
r =8× (16)
5
The mean value, μ D , is obtained as a function of the vulnerability index (Iv ) and inten-
sity (I ), as expressed in Eq. 4, while the variance of the distributions is defined using the
previously assumed value of 8 for parameter t. Figure 11 presents examples of damage dis-
tributions obtained through use of the beta probability distribution (t = 8; a = 0; b = 5) for
events of different seismic intensity and the mean value of the building vulnerability index
(Iv = 38.38).
Another method of representing damage using damage distribution histograms involves the
use of fragility curves. Here the probability of exceeding a certain damage grade or state, Dk
(k [0; 5]) is obtained directly from the physical building damage distributions derived from
the beta probability function for a determined building typology. Just like the vulnerability
curves, fragility curves define the relationship between earthquake intensity and damage in
terms of the conditional cumulative probability of reaching a certain damage state.
Probability histograms of a certain damage grade, P(Dk = d), are derived from the difference
of cumulative probabilities:

P(Dk = d) = PD [Dk ≥ d] − PD Dk+1 ≥ d (17)
Fragility curves are influenced by the parameters of the beta distribution function and allow
for the estimation of damage as a continuous probability function. Figure 12 shows fra-
gility curves corresponding to the damage distribution histograms of the mean vulnerabil-
ity index value (Iv = 38.38) as well as of the mean value plus one standard deviation
(Iv + σ I v = 47.50).
Figures 13 and 14 present the damage scenarios mean damage grade assessment, com-
bining building vulnerability and seismic hazard data for earthquake intensities of VIII and
IX. These two seismic scenarios correspond to the strongest historically-felt earthquake ever
experienced in the district of Coimbra (in 1755). Damage estimation ranges from 1.3 to 3.6 for
the earthquake scenario using I (EMS-98) = VIII and from 2.49 to 4.4 for I (EMS-98) = IX.
Maps in this GIS format enable the identification of urban areas with higher exposure and
at greater risk of damage. However damage assessment is just the first step in the estimation
of seismic loss, with the potential number of victims (deaths and severely injured), number of
unusable buildings, homelessness and the economic impact of an earthquake also considered.
These aspects are discussed in the following section.

123
Bull Earthquake Eng (2011) 9:1067–1096 1087

Fig. 12 Examples of fragility curves: (a) Iv = 38.38; (b) Iv = 47.50

Fig. 13 Damage distribution for


I (EMS-98) = VIII

4.3 Seismic loss assessment

The development of GIS tools in recent years has highlighted their potential for the man-
agement of data. The example developed here represents the implementation of a workable
and progressive platform, integrating all seismic risk evaluation, from building characteris-
tics to estimation of economic loss. Building vulnerability assessment and loss evaluation
algorithms (mathematical and probability functions) were also integrated into the GIS tool,

123
1088 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 14 Damage distribution for


I (EMS-98) = IX

Fig. 15 GIS tool environment

enhancing the whole analytical process. Such a procedure enabled rapid data layering and
editing of building information, inter-correlation of building feature vulnerability, physi-
cal risk and loss estimation results, supporting the making of responsible risk management
actions and decisions. Figure 15 shows a screenshot of the GIS tool environment.
This GIS tool not only enables the visualisation of results, but also the handling of data for
sensibility analysis, scenario modelling and simulation—for instance in order to foresee the
impact of retrofitting strategies in the reduction of vulnerability and consequent economic
losses.

123
Bull Earthquake Eng (2011) 9:1067–1096 1089

Loss evaluation can be presented in two modes:

(i) Construction of damage scenarios using representative values of the vulnerability index
for the 679 buildings evaluated
(Iv,mean − 2σ I v ; Iv,mean − 1σ I v ; Iv,mean ; Iv,mean +1σ I v ; Iv,mean + 2σ I v );
(ii) Using GIS to develop loss scenarios for the whole area evaluated and for each building.
In this way it is possible to inter-correlate and evaluate the spatial distribution of results.

4.3.1 Collapsed and unusable buildings

The expected mean damage grade μ D was evaluated as a function of macroseismic inten-
sity I (EMS-98) and the vulnerability index Iv obtained from the building assessment. A beta
function was employed to obtain the damage distribution, the latter being defined as the prob-
ability of reaching each damage grade, Dk (k [0, 5]) for the assessed mean damage grade
μ D . The loss estimation models adopted in this analysis are also based on damage grades
and involve correlating the probability of the occurrence of a certain damage level with the
probability of building collapse and loss of functionality. The most frequently employed
approaches are those based on observed damage data, such as the HAZUS (1999) proposal
and that of the Italian National Seismic Survey. The latter was based on work by Bramerini
et al. (1995) which involved the analysis of data associated with the probability of buildings
to be deemed unusable after minor and moderate earthquakes. Although such events produce
lower levels of structural and non-structural damage, higher mean damage values may occur
which are associated with a higher probability of building collapse.
The probability of the occurrence of each damage grade is multiplied by a factor. These range
from 0 to 1 and differ from proposal to proposal, but all involve statistical correlation which
validates the algorithms used. In Italy, data processing undertaken by Bramerini et al. (1995)
enabled the establishment of these weighted factors and respective expressions for their use
in the estimation of building loss. For the analysis of collapsed and unusable buildings the
following equations have been derived:

Pcollapse = P(D5 ) (18)


Punusable buildings = P(D3 ) × Wub,3 + P(D4 ) × Wub,4 (19)

where P(Dk ) is the probability of the occurrence of a certain level of damage (D1 to D5 )
and Wub,3 , Wub,4 are weights indicating the percentage of buildings associated with the dam-
age level Dk , that have suffered collapse or that are considered unusable. The values of the
weighting factors presented in the SSN (Bramerini et al. 1995) and HAZUS (1999) propos-
als are slightly different. For the analysis of Coimbra the following weights were applied:
Wub,3 = 0.4; Wub,4 = 0.6.
Figure 16 shows the results of building collapse and unusable building estimations for the
mean value of the vulnerability index (Iv ) for the entire study area, as well as for other values
of vulnerability, namely: Iv,mean − 2σ I v ; Iv,mean − 1σ I v ; Iv,mean + 1σ I v ; Iv,mean + 2σ I v .
For moderate to strong seismic activity, corresponding to intensities of XIII and IX on the
EMS-98 scale, the overall results listed in Table 5 are concerning.
Using the GIS tool at an urban scale it is possible to define the probability of collapse of
each single building. As this information is also linked to all building data collected, different
data layers can be combined. Figure 17 shows examples of layered plot results produced by
the GIS tool.

123
1090 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 16 Estimative of the collapsed and unusable buildings for different vulnerability values

Table 5 Estimation of the number of collapsed and unusable buildings

Total number of buildings: 679 Intensity, I (EMS-98)

VII VIII IX X

Collapsed buildings 0 1 (0.14%) 39 (5.74%) 334 (49.1%)


Unusable buildings 11 (1.62%) 158 (23.2%) 468 (68.9%) 324 (47.7%)

Fig. 17 Mapping results for the evaluation of collapse probability and unusable building assessment

123
Bull Earthquake Eng (2011) 9:1067–1096 1091

4.3.2 Human casualties and homeless

One of the most serious consequences of an earthquake is the loss of human life and thus
one of the major goals of all risk mitigation strategies is ensuring human safety. Over the
last hundred years the world has been struck by more than 1,250 strong quakes and over
1.5 million people have died as a consequence (Coburn and Spence 2002). However official
numbers are not always accurate and the actual totals may be much higher. Of the various
casualty rate analyses and correlation laws found in the literature, those developed by Coburn
et al. (1992), Tiedemann (1989), the HAZUS model (1999) and Bramerini et al. (1995) are
the most frequently cited.
Once again the Italian proposal (Bramerini et al. 1995) was used here to guarantee the
typological consistency of the loss assessment procedure. The rate of dead and severely
injured is projected as being 30% of the residents living in collapsed and unusable buildings,
with the survivors assumed to require short term shelter. Casualty (dead and severely injured)
and homelessness rates are determined via Eqs. 20 and 21, respectively.

Pdead and severely injured = 0.3 × P(D5 ) (20)


Phomeless = P(D3 ) × Wub,3 + P(D4 ) × Wub,4 + P(D5 ) × 0.7 (21)

These two indicators are of great interest for risk management. Figure 18 shows an esti-
mation of the numbers of dead, severely injured and homeless for the mean value of the
vulnerability index (Iv ), as well as for other vulnerability values (Iv,mean − 2σ I v ; Iv,mean −
1σ I v ; Iv,mean ; Iv,mean + 1σ I v ; Iv,mean + 2σ I v ).
For events of moderate to high seismic activity, corresponding to intensities of VIII and
IX on the EMS-98 scale, results obtained for the entire study area are displayed in Table 6.
Representation of the results in map form is an extremely useful method of presenting a
global overview of the potential effects of an earthquake, enabling the identification of the

Fig. 18 Estimation of homeless and casualty rate for different values of vulnerability

Table 6 Estimation of the number of homeless, dead and severely injured

Population: 1800 Intensity, I (EMS-98)

VII VIII (%) IX (%) X

Homeless 30 (1.67%) 419 (23.3%) 1313 (72.9%) 1478 (82.1%)


Dead and severely injured 0 0 31 (1.72%) 265 (14.7%)

123
1092 Bull Earthquake Eng (2011) 9:1067–1096

Fig. 19 Mapping results for the evaluation of death rate and shelter requirements for different seismic sce-
narios

worst affected areas and helping the development of rescue plans and post-event intervention.
Figure 19 shows layered results for part of the urban area under study.

4.3.3 Economic losses and repair costs

The estimated damage grade can either be interpreted economically, as defined by Benedetti
and Petrini (1984), or as an economic damage index, i.e. the ratio between the repair cost
and the replacement cost (building value). The correlation between damage grades and the
repair and rebuilding costs are obtained by the processing of post-earthquake damage data.
As shown in Table 7, a variety of correlations have been derived in earlier studies.
The most reasonable relationship, as confirmed by the post-seismic investigation of
Pasquale and Goretti (2001), is that which assumes a similar value of the damage index
for damage grade 4 and 5 and a greater difference between the damage index for the lower

Table 7 Correlation between damage levels and damage index

Damage grade, Dk 0 1 2 3 4 5

Bramerini et al. (1995) 0.000 0.010 0.100 0.350 0.750 1.000


ATC-13 (1985) 0.000 0.05 0.200 0.550 0.900 1.000
Dolce et al. (2005) 0.005 0.035 0.145 0.305 0.800 0.950

123
Bull Earthquake Eng (2011) 9:1067–1096 1093

Fig. 20 Estimation of repair costs

damage grades of 1 and 2. The values obtained by ATC-13 (1985) and Dolce et al. (2005)
are in agreement with these criteria.
The correlation adopted here was that established by Dolce et al. (2005). The statistical
values obtained by these authors were derived from analysis of the data collected, using the
GNDT-SSN procedure, after the Umbria-Marche (1997) and the Pollino (1998) earthquakes
(Dolce et al. 2005), and based on the estimated cost of typical repairs for more than 50,000
buildings.
The probabilities of the repair costs are defined as the product of the following two prob-
abilities: The conditional probability of the repair cost for each damage level, P[R|Dk ],
expressed by the values presented in Table 7, and the known conditional probability of the
damage condition for each level of building vulnerability and seismic intensity, P[Dk |v, I ],
given by:


5 
100
Pr ob [R|I ] = Pr ob [R|Dk ] × Pr ob [Dk |I V , I ] (22)
Dk =1 I V =0

These values were calculated for both the mean vulnerability index value and the lower and
upper bound values (Iv,mean −2σ I v ; Iv,mean −1σ I v ; Iv,mean ; Iv,mean +1σ I v ; Iv,mean +2σ I v ).
Estimated cost of repair for the building stock of the historic city centre of Coimbra relative
to total value of buildings (in terms of unit area as 800e/m2 ) are shown in Fig. 20.
For seismic events of intensity in the range of V to IX, the variation between estimated
minimum and maximum repair cost is significant. For higher earthquake intensities, the
difference is much smaller as a result of the high damage levels caused by severe seismic
events. Taking into account the regional hazard posed to Coimbra, the repair cost estima-
tions for earthquake intensities VII, VIII and IX are the most representative. The high repair
costs for events of intensity VIII to X confirm the rather high vulnerability of the building
stock. It should also be noted that the total repair cost in terms of building area ranges from
3.4% (IEMS-98 = VII) to 35.7% (IEMS-98 = I X ) of the cost of replacing the entire build-
ing stock. For a seismic event intensity of VIII and assuming a mean vulnerability value
(Iv = 38.38), total repair costs amount to around 60% of the 2007 annual budget of Coimbra
city council.

123
1094 Bull Earthquake Eng (2011) 9:1067–1096

5 Final comments

The vulnerability assessment method developed here has been proven to be extremely useful
and reliable for the analysis of building construction characteristics and as a consequence so
are the results obtained from its use. Integration of this vulnerability assessment technique
into a Macroseismic method has enabled the development of damage and loss scenarios for
urban risk mitigation and management. The proposed vulnerability assessment method and
risk scenario mapping technique can easily be adapted for specific building features and
adopted by other regions and old city centres.
Application of this methodology was only possible since it was integrated within the reha-
bilitation process along with other goals, taking advantage of available resources. However,
it has revealed itself to be a first level approach which provided a highly reliable evaluation
result for most of the parameters classified.
The GIS application and database management system enables the storage of building fea-
ture and survey information, assessment of seismic vulnerability and damage and risk scenario
prediction, as well as allowing the upgrading and improvement of data. This integrated tool
can be helpful for the development of strengthening strategies, cost-benefit analyses, civil
protection and emergency planning.
With respect to the damage scenarios studied here, the results obtained correlate well with
the observed building construction features and general fragilities of the built-up environ-
ment. Even though the city of Coimbra is located in a low to moderate seismic hazard region,
the level of building vulnerability is such that it presents a considerable global seismic risk
for the building stock and historic area. The seismic vulnerability of the old city centre of
Coimbra is relatively high and therefore interventions focussing on the improvement of seis-
mic performance and conservation of old buildings, assisted by materials’ characterisation
and mechanical modelling, are required for seismic risk reduction.
Methods of vulnerability assessment based on statistical approaches and damage obser-
vation are far more suitable for large scale analysis, essentially for two reasons: they require
less information and fewer resources and the currently available simplified mechanical mod-
els still require experimental testing validation. However, the uncertainties associated with
the empirical vulnerability curves and the quality of vulnerability classification data are still
issues that must be studied further with respect to post-seismic data collection. For risk
mitigation, a reduction in building vulnerability is a priority and therefore the development
of more reliable vulnerability assessment models which combine statistical and mechanical
methods should lead to better results and bidirectional validation through the adjustment of
criteria.
This research could enable the following steps to be taken in the future: Development of
the framework of a comprehensive database and guidance tool for local authorities respon-
sible for rehabilitation and renewal in urban areas; integration of this database within a GIS
environment for risk management of buildings at a urban scale; rapid creation of earthquake
scenarios validated by decision makers and technicians for urban areas, enabling the esti-
mation and forecasting of direct and indirect consequences of their economic and physical
impact; establishment and validation of a modular approach for the creation of building
vulnerability databases, as well as damage and cost evaluation algorithms for other historic
urban areas.

Acknowledgments We gratefully acknowledge the financial support given by the Portuguese Foundation
for Science and Technology to carry out this research, the City Council of Coimbra for their cooperation
throughout the inspection and appraisal of buildings and the Portuguese Geographical Institute for granting
the use of the Orthophotomaps.

123
Bull Earthquake Eng (2011) 9:1067–1096 1095

References

ArcGis 9.3 (2008) Geographic information systems, 310 New York Street, Redlands, 92373-8100, USA,
ESRI
ASER (2006) Assessing and managing earthquake risk. In: Goula X, Oliveira CS, Roca A (eds) Geo-scientific
and engineering knowledge for earthquake risk Mitigation: development, tools, techniques. Springer,
ISBN 1-4020-3524-1
ATC-13 (1985) Earthquake damage evaluation data for California. Applied Technology Council, Redwood
City, California
ATC-21 (1988) Rapid visual screening of building for potential seismic hazards: a handbook. Applied Tech-
nology Council, FEMA 145, Redwood City, California
ATC-40 (1996) Seismic evaluation and retrofit of concrete buildings, Technical report, ATC-40. Applied
Technology Council, Redwood City, California
Barata MRT (2005) The Lisbon earthquake of 1st November 1755—a historical perspectives approach. In:
Proceedings of the international conference 250th anniversary of the 1755 Lisbon earthquake, 1st–4th
November, pp 31–41
Barbat AH, Carreño ML, Pujades LG, Lantada N, Cardona OD, Murulanda MC (2010) Seismic vulnerability
and risk evaluation methods for urban areas. A review with application to a pilot area. Struct Infrastruct
Eng 6(1–2):17–38
Benedetti D, Petrini V (1984) On seismic vulnerability of masonry buildings: proposal of an evaluation pro-
cedure. Ind Constr 18:66–78
Bernardini A, Giovinazzi S, Lagomarsino S, Parodi S (2007) Vulnerabilità e previsione di danno a scala terri-
toriale secondo una metodologia macrosismica coerente con la scala EMS-98. ANIDIS, XII Convegno
Nazionale l’ingegneria sismica in Italia, 10 a 14 Giugno, Pisa
Bramerini F, Di Pasquale G, Orsini A, Pugliese A, Romeo R, Sabetta F (1995) Rischio sismico del territorio
italiano. Proposta per una metodologia e risultati preliminari. Rapporto tecnico del Servizio Sismico
Nazionale, SSN/RT/95/01, Roma (in Italian)
Carreño ML, Cardona OD, Barbat AH (2007) Urban seismic risk evaluation: a holistic approach. Nat Hazards
40:137–172
Caicedo C, Barbat AH, Canas JA, Aguiar R (1994) Vulnerabilidad sísmica de edificios, Monografía CIMNE
IS-6. Barcelona
Cardona, O.D. (2001) Estimación holística del riesgo sísmico utilizando sistemas dinámicos complejos.
Universidade Politécnica de Cataluña. PhD Thesis, Barcelona, Spain
Coburn AW, Spence R (2002) Earthquake protection. Wiley, Chichester, England
Coburn AW, Spence R, Pomodis A (1992) Factors determining human casualty levels in earthquakes: mortality
prediction in building collapse; 10th WCEE. Madrid, pp 5989–5994
Corsanego A, Petrini V (1990) Seismic vulnerability of buildings. In: Proceedings of the SEISMED 3. Trieste,
Italy
Calvi GM (1999) A displacement-based approach for vulnerability evaluation of classes of buildings. J Earthq
Eng 3(3):411–438
Combescure D, Guéguen P, Lebrun B (2005) Vulnérabilité sismique du bâti existant: approche d’ensemble.
Cahier technique AFPS 25:121
Dolce M, Kappos A, Zuccaro G, Coburn AW (1994) Report of the EAEE working group 3: vulnerability and
risk analysis. Technical Report 10th european conference on earthquake engineering. Vienna, No. 4, pp
3049–3077
Dolce M, Marino M, Masi A, Vona M (2000) Seismic vulnerability analysis and damage scenarios of Potenza.
International workshop on seismic risk and earthquake damage scenarios of potenza
Dolce M, Kappos A, Masi A, Penelis G, Vona M (2005) Vulnerability assessment and earthquake damage
scenarios of the building stock of Potenza (southern Italy) using Italian and Greek methodologies. Eng
Struct 28:357–371
Fajfar P (1999) Capacity spectrum method based on inelastic demand spectra. Earthq Eng Struct Dyn 28:
979–993
FEMA 273 (1996) NEHRP guidelines for the seismic rehabilitation of buildings. Federal Emergency
Management Agency, Washington DC
FEMA 178 (1992) NEHRP handbook for the seismic evaluation of existing buildings. Federal Emergency
Management Agency, Washington, DC
Gruppo Nazionale per la Difesa dai Terremoti–GNDT-SSN (1994) Scheda di esposizione e vulnerabilità e di
rilevamento danni di primo livello e secondo livello (muratura e cemento armato). Gruppo Nazionale
per la Difesa dai Terremoti: Roma

123
1096 Bull Earthquake Eng (2011) 9:1067–1096

Grünthal G (1998) European macroseismic scale. Centre Européen de Géodynamique et de Séismologie:


Luxembourg. vol 15
Giovinazzi S, Lagomarsino S (2004) A macroseismic model for the vulnerability assessment of buildings. In:
Proceedings of 13th world conference on earthquake engineering. Vancouver: Canada
Giovinazzi S (2005) The vulnerability assessment and damage scenario in seismic risk analysis. PhD Thesis,
International doctorate, University of Florence, Technical University of Carolo-Wilhelmina
Grimaz S, Meroni F, Petrini V, Tomasoni R, Zonno G (1996) Il ruolo dei dati di danneggiamento del terrem-
oto del Friuli nello studio di modelli di vulnerabilità sísmica degli edifici in muratura; La Scienza e i
Terremoti. Analisi e prospettive dal’ esperienza del Friuli, 1976-1996. Atti del Convegno, Udine, 14–15
Novembro
Guagenti E, Petrini V (1989) Il caso delle vecchie costruzioni: verso una nuova legge danni-intensità. In:
Proceedings of the 4th Italian conference on earthquake engineering, Milano I, pp 145–153
HAZUS MH (1999) Earthquake loss estimation methodology—technical and user manuals. Federal Emer-
gency Management Agency, Washington, D.C
Lang K, Bachmann H (2003) On the seismic vulnerability of existing unreinforced masonry buildings. J Earthq
Eng 7(3):407–426
LESSLOSS (2007) European research project for risk mitigation for earthquakes and landslides (2004–2007);
Website: http://www.lessloss.org/main/index.php (last viewed December 2007)
Lagomarsino S, Giovinazzi S (2006) Macroseismic and mechanical models for the vulnerability assessment
of current buildings. Bull Earthquake Eng. doi:10.1007/s10518-006-9024-z
Margottini C, Molin D, Narcisi B, Serva L (1992) Intensity versus ground motion: a new approach using
Italian data. Eng Geolog 33:45–48
McGuire RK (2004) Seismic hazard and risk analysis, EERI Monography. Earthquake Engineering Research
Institute, Oakland, CA
Di Pasquale G, Goretti A (2001) Economic and functional vulnerability of residential buildings stricken by
Italian recent seismic events. In: Proceedings of the 10th Italian national conference on earthquake
engineering (in Italian)
RISK-UE Project (2004). An advanced approach to earthquake risk scenarios, with applications to different
European cities; Website: http://www.risk-ue.net (last viewed December 2007)
Speranza E (2003) An integrated method for the assessment of the seismic vulnerability of historic buildings.
PhD Thesis. University of Bath
Sandi H, Floricel I (1995) Analysis of seismic risk affecting the existing IX building stock. In: Proceedings
of the 10th European conference on earthquake engineering, vol 3, pp 1105–1110
Tiedemann H (1989) Casualties as a function of building quality and earthquake intensity. In: Proceedings of
the international workshop on earthquake injury epidemiology for mitigation and response; 10–12 July,
Baltimore, Maryland. Johns Hopkins University, Baltimore, pp 420–434
UNDP/UNIDO (1985) Project RER/79/015. Post-earthquake damage evaluation and Strength Assessment of
Buildings under Seismic Condition. UNDP, Vienna
Vicente RS (2008) Strategies and methodologies for urban rehabilitation interventions. The vulnerability
assessment and risk evaluation of the old city centre of Coimbra. PhD Thesis, University of Aveiro (in
Portuguese)
Vicente RS, Varum H, Mendes da Silva JAR (2005) Seismic vulnerability assessment of buildings in the old
city centre of Coimbra. In: Proceedings of the international conference 250th anniversary of the 1755
Lisbon earthquake, 1st–4th November, pp 206–213
Whitman RV, Reed JW, Hong ST (1974) Earthquake damage probability matrices. In: Proceedings of the 5th
world conference on earthquake engineering, p 2531, Rome

123

You might also like