Nat Hazards (2017) 86:S31–S54

DOI 10.1007/s11069-016-2647-8

ORIGINAL PAPER

Evaluation of the seismic risk of the unreinforced

masonry building stock in Antioquia, Colombia

Ana B. Acevedo1 • Juan D. Jaramillo1 • Catalina Yepes2 •

Vitor Silva2 • Fernando A. Osorio1 • Mabé Villar2

Received: 17 December 2015 / Accepted: 25 October 2016 / Published online: 2 November 2016
Ó Springer Science+Business Media Dordrecht 2016

Abstract This paper presents the development of an exposure model for the residential
building stock in Antioquia (the second most populated Department of Colombia), the
development of fragility functions for unreinforced masonry buildings, and estimation of
building damage for two possible seismic events. Both the exposure and fragility models
are publically available and can be used to calculate damage and losses due to single
events, or probabilistic seismic hazard. The exposure model includes information regarding
the total built-up area, number of buildings and inhabitants, building class, and replace-
ment cost. The methodology used for the creation of the exposure model was based on
available cadastral information, survey data, and expert judgment. Fragility functions were
derived using nonlinear time history analyses on single-degree-of-freedom oscillators, for
unreinforced masonry structures which represent more than 60% of the building stock in
the region. Both seismic scenarios indicate that an event corresponding to a return period
of 500 years located within the region of interest would cause slight or moderate damage to
nearly 95 thousand structures, and about 32 thousand would have severe damage or col-
lapse. This study was developed as part of the South America Risk Assessment project,
supported by the Global Earthquake Model and SwissRe Foundation.

Keywords Exposure  Seismic risk  Fragility functions  Masonry buildings

1 Introduction

Over the last four decades, two main seismic events have affected Colombia: the mag-
nitude 6.2 (Mw) Armenia earthquake of 1999 and the magnitude 5.0 (Mb) Popayán
earthquake of 1983. Both events caused considerable human losses with 1,185 and 287

& Ana B. Acevedo

aaceved14@eafit.edu.co
1
Department of Civil Engineering, Universidad EAFIT, Medellı́n, Colombia
2
GEM Foundation, Pavia, Italy

123
S32 Nat Hazards (2017) 86:S31–S54

deceased, respectively. The estimated economic losses from these events correspond to
1.88% of the gross domestic product (GDP) in 1999, and 1.5% in 1983 (AIS 2009; Cardona
et al. 2004). The experienced damage for relatively small magnitude events is a clear
indicator of the high vulnerability of some types of construction in Colombia. In addition,
more than 80% of the population is settled in areas of medium to high seismic hazard, thus
characterizing Colombia as a country with a high seismic risk.
For a developing country such as Colombia, the reduction in the economic impact of
earthquakes and the improvement in public safety are fundamental. As a first step, a
reliable seismic risk assessment is required, as this information can be used to support
earthquake disaster management and mitigation. An initiative from the Inter-American
Development Bank (IDB), among others, has estimated the seismic risk of some Latin-
American countries (including Colombia), through the development of Indicators of
Disaster Risk and Risk Management (http://idea.unalmzl.edu.co/). Some efforts have been
done in the main cities of Colombia such as Bogotá, Medellı́n, and Armenia, in which the
estimation of earthquake loss scenarios has been performed (Cardona et al. 1997; Salgado
et al. 2013; Salgado-Gálvez et al. 2014; Consorcio Microzonificación 2007).
Seismic hazard, exposure, and vulnerability information for the region of interest are
required in order to assess its seismic risk. The first seismic hazard study for Colombia
dates from 1972 (Atuesta 1972), and it has been updated several times. The results from
the latest seismic hazard assessment of the Colombian Association of Seismic Engineering
(AIS 2009) are featured in the current seismic regulation, the NSR-10 (AIS 2010). For
what concerns exposure, several models have been compiled for Bogotá, Medellı́n,
Manizales, Cali, and Bucaramanga, mainly for the purposes of microzonation, in which the
main structural characteristics of the building portfolio and soil conditions are described.
For what concerns structural vulnerability, a few models have been specifically developed
for Colombia, such as the one proposed by Bonnet (2003) for reinforced concrete frames
for the city of Manizales. Moreover, fragility/vulnerability curves for building classes such
as confined masonry, reinforced masonry, unreinforced masonry, adobe, earth, reinforced
concrete frames, pre-cast concrete, and dual frame-wall system have been used in the risk
assessments previously mentioned. Although the final results of the aforementioned studies
have been released to the public, the majority of the information such as hazard and
exposure datasets, and the statistical parameters that represent fragility/vulnerability curves
were kept private. The lack of data was one of the greatest challenges, as well as the
motivation, in the development of the current study.
The present manuscript describes an exposure model and fragility functions for unreinforced
masonry structures for the region of Antioquia. In order to explore the reliability and usefulness
of these models, two earthquake scenarios were performed. This research was carried out as part
of the South America Risk Assessment (SARA) project, supported by the Global Earthquake
Model (GEM) Foundation and funded by SwissRe Foundation. The exposure model and fra-
gility functions are accessible to the general public through the OpenQuake platform (https://
platform.openquake.org) and SARA wiki (https://sara.openquake.org/risk).

2 Seismic hazard in Antioquia

Colombia is located in the Ring of Fire, a zone where approximately a third of the global
seismicity takes place. The seismic activity in the country is the result of the interaction of
the Nazca, Caribbean and South America tectonic plates. This interaction generates regions
of low, medium, and high seismic hazard.

123
Nat Hazards (2017) 86:S31–S54 S33

Fig. 1 Geographical organization of the region of interest

Antioquia, the second most populated department of Colombia, is located in the

northeast of the country as can be observed in Fig. 1. Antioquia is divided in nine sub-
regions with 125 municipalities. The most populated area is the Aburrá Valley (Valle de
Aburrá), where more than 58% of the population of the Department is concentrated. The
Aburrá Valley has approximately 80 km length and constitutes the metropolitan area of
Medellı́n, the capital city. It includes the municipalities of Medellı́n, Caldas, La Estrella,
Sabaneta, Envigado, Itagüı́, Bello, Copacabana, Girardota and Barbosa.
In the case of Antioquia, the latest seismic hazard assessment of Colombia (AIS 2009)
indicates medium hazard for 62% of the municipalities and high hazard for the remaining
38%. Only two of the municipalities of the Aburrá Valley (Caldas and La Estrella) are
classified with high seismic hazard. The main sources of seismicity are crustal events from
seismogenic zones located in the North-West of Colombia, depth and shallow events from
the seismogenic zone of the ‘‘Eje Cafetero’’ (Middle-West), deep events from the sub-
duction of the Nazca Plate under the South American plate, and shallow seismicity from
the Romeral fault system. Romeral is the most active fault system of the country, with an
approximately total length of 700 km (Pulido 2003).
Figure 2 presents the distribution of both historical and instrumental seismic events for
magnitude larger than 4.0 (Mw) for Antioquia and its surroundings, as well as the active
faults near the Department. Instrumental events of Fig. 2a were extracted from the
Colombian Geological Survey, SGC (http://seisan.sgc.gov.co/RSNC/index.php/consultas/

Fig. 2 Seismic events of Antioquia. a Instrumental events. b Historical events

123
S34 Nat Hazards (2017) 86:S31–S54

consulexp). The historical events of Fig. 2b have been taken from the historical catalogue
developed by the SARA project.
Two main earthquakes have affected the region over the past 50 years: the crustal
Mistrató earthquake of 23 November 1979 with a magnitude of 7.9 (M0), in which several
buildings were damaged, and the Murindó earthquake of 18 October 1992 with a mag-
nitude of 7.3 (Ms), generated by the seismogenic zone of the ‘‘Eje Cafetero’’. In the latter
event, more than 240 buildings were heavily damaged (Martı́nez et al. 1994), despite the
low acceleration registered in Medellı́n (the epicentre was located about 150 km away).
Four historic events are reported by Ramı́rez (1975) with an epicentre around Medellı́n: the
13 April 1972 event, in which several houses were destroyed; and the events of 15
September 1868, 31 October 1928 and 11 October 1966. As paleoseismic studies are scarce
in the region, there is great uncertainty associated with the historical events.

3 Exposure modelling of the building stock

As in many Latin-American countries, the building stock, infrastructure, and population

are mainly concentrated in a few regions. In Colombia, 59% of the population is located in
only 6 of the 32 departments. The second most populated department is Antioquia, with
more than 6 million inhabitants (14% of Colombia’s population) and an area of
63,612 km2. Antioquia is divided in 9 sub-regions with 125 municipalities. The sub-region
of the Aburrá Valley constitutes the metropolitan area of Medellı́n, the capital city; it
gathers ten municipalities and 58% of the Department’s population.
Although a modern seismic code is available for the country, an important number of
buildings in Antioquia does not comply with its minimum requirements. Formal con-
struction is usually located in the municipalities of Medellı́n, Envigado, Itagüı́ and Bello,
and in some developing areas within other municipalities (Consorcio Microzonificación
2007). However, a significant amount of informal construction can still be found in these
regions, due to the rapid increase in population in the last decades. In addition, many
residential buildings built before the year of 1984 did not include seismic provisions, as the
first seismic code was released in that year. Yepes et al. (2016) estimated that 65% of the
Colombian building stock has a lateral load resisting system not appropriated to sustain
seismic loads, while Mejı́a (2011) indicates that approximately 60% of the housing stock of
Antioquia is non-engineered unreinforced brick masonry buildings.
The exposure model for Antioquia developed in this study indicates a total of 147 km2
of built-up area, and 834 thousands of buildings. The Aburrá Valley sub-region has a total
of 109 km2 built-up area and 476 thousands of building; the city of Medellı́n has a total of
79 km2 of built-up area and 343 thousands of buildings. The total replacement cost for the
residential building stock of Antioquia was estimated as 135 9 106 million Colombian
Pesos—COP/43,038 million US dollars (values of replacement cost are given in COP as
the exchange rate to US dollars fluctuates severely for this currency; the average exchange
rate for the first semester of 2016 corresponds to 1 US dollar =3140 COP). For the Aburrá
Valley and Medellı́n, the total replacement cost was estimated as 97 9 106 million COP
(30,982 million US) and 76 9 106 million COP (24,327 million US), respectively.
Replacement cost refers to the cost of structural and non-structural components, and it is a
value associated with building rehabilitation. This value differs from the commercial value,
as land price is not included. It must be kept in mind that after a seismic event, a structure

123
Nat Hazards (2017) 86:S31–S54 S35

must be repaired according to modern seismic regulations, regardless of the structural

system.
Building classes included in the exposure model were defined according to the lateral
load resisting system, construction materials, and number of storeys. Differentiation
between ductile and non-ductile structures was specified only for reinforced concrete
structures. For Medellı́n and the Aburrá Valley, 80% of the reinforced concrete frame
buildings between one and five storeys were assumed as non-ductile as they were mainly
built before 1984; the remaining reinforced concrete frame buildings were considered as
ductile. On the other hand, even though the majority of the reinforced concrete frames
outside of the Aburrá Valley have been built in the last few decades when seismic design
was already mandatory, informal construction is still common in those municipalities. For
that reason, 50% of the reinforced concrete frames outside the Aburrá Valley were
assumed as ductile, while the remaining 50% were assumed as non-ductile. The GEM
taxonomy was used for the building classification (Brzev et al. 2013). Table 1 presents the
list of the resulting building classes. A total of 121 building classes were included in the
model. Figure 3 presents the replacement cost, built-up area, and building distribution for
Antioquia, the Aburrá Valley, and Medellı́n.
It can be observed from Fig. 3 that the majority of the built-up area of Antioquia
corresponds to unreinforced masonry structures, with a total of 78 km2, representing 508
thousands of buildings (53% of the total built-up area and 61% of the building stock). This
type of structures has been forbidden by all of the Colombian seismic codes, but it is still
common in the region. Confined masonry constitutes the second most common building
typology in Antioquia with a built-up area of 16 km2, representing 97 thousands of
buildings (11% of the total built-up area and 12% of the building stock).
An innovative methodology is described for the development of the exposure model of
Antioquia, which allows the calculations of the built-up area, number of buildings, number
of storeys, and structural system. Three exposure models have been developed, for each

Table 1 Building classes of the exposure model

Building class GEM taxonomy

Material and system Number of storeys

Reinforced concrete infilled frame, ductile CR/LFINF/DUC HEX:1 to HEX:24

Reinforced concrete infilled frame, non-ductile CR/LFINF/DNO HEX:1 to HEX:6
Reinforced concrete wall system CR/LWAL HEX:4 to HEX:39
Reinforced concrete dual frame-wall system CR/LDUAL HEX:11 to HEX:39
Confined masonry MCF HEX:1 to HEX:7
Reinforced masonry MR HEX:1 to HEX:7
Unreinforced masonry MUR HEX:1 to HEX:6
Unreinforced stone masonry MUR ? STRUB HEX:1
Reinforced rammed eartha ER ? ETR HEX:1 or HEX:2
Woodb W HEX:1 or HEX:2
Unknown or other typologies UNK HEX:1
a
This building class aggregates both rammed earth and wattle and daub buildings
b
Non-engineered wood

123
S36 Nat Hazards (2017) 86:S31–S54

Fig. 3 Replacement cost, built-up area, and building distribution for Antioquia (a), the Aburrá Valley (b),
and Medellı́n (c)

one of the three regions: the city of Medellı́n, the metropolitan area (Aburrá Valley sub-
region without Medellı́n), and the remaining 115 municipalities. Data availability was the
main reason for the development of three different models. Table 2 presents a summary of
the available information for each exposure model. A description of the assumptions made
in order to generate the missing data referred in Table 2 is presented in the next sec-
tion. The methodology developed for the exposure model definition can be applied to any
location, granted that the minimum information is available.
The exposure models include information about the location (neighbourhoods for the
model of Medellı́n, and municipalities for the other two models), the number of

123
Nat Hazards (2017) 86:S31–S54 S37

Table 2 Available and required data for exposure model development

Required data Available information for each region

Medellı́n Aburrá valley Other

municipalities

Built-up area Cadastral map of year From homogenous areas defined Cadastral map of
2012 (building in the microzonation study of year 2012 (for
resolution) Medellı́n and the Aburrá some
Valley municipalities)
Number of floors Cadastral map of year Distribution from previous Cadastral map of
2012 (building studies year 2012 (for
resolution) some
municipalities)
Structural system N/A N/A N/A
distribution
Socio-economic strata Information for each Information for each N/A
neighbourhood municipality
Population Information for each Information for each Information for
neighbourhood municipality each
municipality
Replacement cost N/A N/A N/A

inhabitants, socio-economic strata, total built-up area, and replacement cost for the dif-
ferent building classes. Socio-economic strata refer to the hierarchical economical dif-
ference between dwellings. Colombia’ social strata are divided in a scale from one to six,
in which strata 1–3 refer to dwellings which are occupied by the lowest income inhabitants,
while strata 4–6 correspond to dwellings where the highest income inhabitants live. Strata
1 and 2 can be associated with uncontrolled urban development in marginal areas, which
with time is included into the city limits. Although replacement cost was not used in the
damage scenario presented in Sect. 6, it is still presented in this study as it might be useful
for future studies.
As results in the models are given in terms of built-up area, additional calculations were
required to compute the number of buildings and dwellings. The number of buildings for
each region was computed by defining an average number of dwellings and average
dwelling area for each building class. The average number of dwellings was defined as a
function of the number of storeys, while average dwelling area was defined as a function of
the socio-economic strata (dwellings in the highest strata have larger area). These
parameters were defined based on expert judgement and visual inspections. Furthermore, in
order to assess realistic values for the number of dwellings and average dwelling area for
unreinforced masonry structures (as this type of construction constitutes more than half of
the building stock), 151 real structures from Medellı́n and surrounding municipalities were
surveyed in order to relate the aforementioned parameters. This exercise allowed obtaining
information regarding the socio-economic strata, built-up area, number of storeys, number
of dwellings per building, and dwelling area.
Results of the number of dwellings of the exposure models were compared with
available information. Values of Medellı́n and the metropolitan area (Aburrá Valley) were
compared to values from the ‘‘Life quality survey’’ of the year 2011 (DANE 2011). Data
from the 2005 General Census (DANE 2005) were used for the comparison of the number
of dwellings of the municipalities outside the Aburrá Valley. The exposure model of

123
S38 Nat Hazards (2017) 86:S31–S54

Medellı́n indicates a total of 713 thousands of dwellings when all of the neighbourhoods
are considered, which equals the reference value. If only urban dwellings are considered,
the exposure model of Medellı́n indicates 660 thousands of dwellings, while the reference
reports 569 thousand of dwellings (ratio of 0.86). For the remaining municipalities of the
Aburrá Valley, the ratio of the number of dwellings of the reference to the number of the
dwellings of the exposure model ranges between 0.81 and 0.97. The small variation of the
number of dwellings of the exposure model of Medellı́n and the Aburrá Valley to the
reference numbers indicates a reasonable reliability of the model. The ratio of the number
of dwellings of the exposure model for the municipalities outside the Aburrá Valley to the
reference number ranges between 0.5 and 2.0 for 86% of the municipalities. These results
indicate large differences, but it is not clear to the authors of this study what was the
criterion used in the 2005 General Census for the definition of the urban area of each
municipality. Moreover, a larger number of dwellings than the value reported in the Census
is expected as the population have increased significantly in the last decade.

4 Methodology for development of the exposure model

As can be observed from Table 2, an important amount of data required for the devel-
opment of the exposure model was missing. A brief description of the assumptions made in
order to generate the missing data is explained in this section for the three regions:
Medellı́n, Aburrá Valley and municipalities outside the Aburrá Valley.

4.1 Medellı́n exposure model

The city of Medellı́n has the most detailed information of all of the municipalities of
Antioquia. Cadastral information of the year 2012 (map of plan built areas and number of
storeys) was available. Total built-up area was computed by the multiplication of the plan
area by the number of storeys. Information regarding the population and socio-economic
strata was also available for each of the 350 neighbourhoods of Medellı́n. The cadastral
information did not include information concerning the building class, which is a key
aspect for the development of an exposure model for seismic risk analyses.
The structural system was defined as a function of the socio-economic strata and the
number of storeys, relying on expert judgment and data collected from surveys compiled
for the microzonation study of the Aburrá Valley (Consorcio Microzonificación 2007).
Data from the General Census of 2005 were not used as it only differentiates floor and
roof materials, and it is not possible to differentiate from this information a structural
system such as reinforced concrete, steel, or masonry (reinforced, unreinforced and
confine).
Figure 4 presents the structural system distribution defined for Medellı́n for the lowest
income strata (socio-economic strata 1) and the highest income strata (socio-economic
strata 6). The distribution for the remaining strata can be found in the SARA wiki. From this
figure, it can be observed that the building class distribution differs greatly between
buildings with one to ten storeys according to the socio-economic strata. Unreinforced
masonry structures, which constitute more than half of the built-up area of Antioquia, are
more common in socio-economic strata from one to four. Buildings of this type in high-
income zones usually correspond to structures built before the endorsement of seismic
codes in Colombia. Reinforced concrete buildings have mainly five to ten storeys and low-

123
Nat Hazards (2017) 86:S31–S54 S39

Fig. 4 Structural system distribution for Medellı́n as a function of the number of storeys and socio-
economic strata. a Socio-economic strata 1, b socio-economic strata 6

rise reinforced concrete structures are commonly found in high-income strata. High-rise
buildings of more than ten storeys are reinforced concrete frames, wall systems, or dual
frame-wall systems. High-rise buildings have been built mainly over the last decade, for
either the lowest income or the highest income zones. Earth structures are not commonly
found in the city, as they have been replaced by other structural systems.
The replacement cost was defined as a function of the area and socio-economic strata.
Several construction companies were consulted in order to estimate a reasonable cost.
Values in Colombian Pesos/US dollars (COP/US) per square meter were defined as
500,000 COP/159 US for strata 1 and 2; 740,000 COP/236 US for strata 3; 1,150,000 COP/
366 US for strata 4; 1,450,000 COP/462 US for strata 5; and 1,750,000 COP/557 US for
strata 6.

4.2 Aburrá Valley exposure model

The Aburrá Valley sub-region has ten municipalities: Medellı́n, Caldas, La Estrella,
Sabaneta, Envigado, Itagüı́, Bello, Copacabana, Girardota, and Barbosa. Cadastral infor-
mation (map of plan built areas and number of storeys) was only available for Giradota
(year of 2007). Homogenous areas defined in the microzonation study of Medellı́n and the
Aburrá Valley (Consorcio Microzonificación 2007) were used in order to obtain built-plan
area of the municipalities without cadastral information. Homogenous areas refer to res-
idential areas within each municipality where buildings have similar number of storeys.
Five types of residential areas were defined in the microzonation study based on building’s
most common number of storeys as follows: zone 1 (1 and 2 storeys), zone 2 (3–5 storeys),
zone 3A (6–10 storeys), zone 3B (11–15 storeys), and zone 3C (16 and more storeys).
Figure 5 shows the number of storeys distribution of the homogeneous areas zone 1 and
zone 2 of the municipality of Bello. It can be observed from the figure that the most
common number of storeys for zone 1 is one and two; in the case of zone 2 the most
common number of storeys is three to five. As plan areas defined in the microzonation
study included green areas, streets, and other open spaces, only a percentage of each area
was considered in this study. As building shapes are easy to identify from aerial pictures, a
comparison of such pictures with the areas defined in the microzonation study allowed for
the percentage definition, leading to values between 35 and 50%.
Survey data specifically collected for the microzonation study were used to define the
distribution of number of storeys per building for each homogeneous zone, with the

123
S40 Nat Hazards (2017) 86:S31–S54

Fig. 5 Number of storeys distribution for homogeneous zones 1 and 2 of Bello

exception of Girardota. For the latter municipality, a map with the number of storeys was
already available.
As in the case of Medellı́n, structural system distribution was defined as a function of
the socio-economic strata and the number of storeys. Data collected from the micro-
zonation study, as well as judgement of local experts, were used to define the structural
system distribution. Three main groups of municipalities were defined according to the
expected common building classes and economic development of each municipality:
Group 1—Itagüı́, Bello and Copacabana; Group 2—La Estrella, Girardota, Caldas and
Barbosa; and Group 3—Envigado and Sabaneta. A structural system distribution was used
for each group, and the same replacement cost used in Medellı́n was assumed for the entire
Aburrá Valley.

4.3 Exposure model of municipalities outside the Aburrá Valley

The exposure model for the 115 municipalities outside the Aburrá Valley incorporates
many uncertainties, due to the several assumptions required to compensate for the lack of
data (see Table 2). These municipalities constitute 41.6% of Antioquia’s population, and
are organized in eight sub-regions (Bajo Cauca, Magdalena Medio, Nordeste, Norte,
Occidente, Oriente, Suroeste and Urabá). The 2005 General Census Survey (population,
number of dwellings and distribution of wall and floor material) was available for all of the
municipalities. Cadastral information of the year 2012 (map of plan built areas and number
of built storeys) was available for the urban area of 80 municipalities.
For the municipalities in which cadastral information was not available, the built-plan
area was defined from the analysis of aerial imagery. Figure 6 presents an example of the
built area definition for the municipality of Apartadó, in which, based on buildings plan
shapes, general residential areas were drawn (black polygons). A reduction factor was
applied to these areas in order to exclude non-residential areas such streets and parks. The
reduction factor was computed by the comparison of several general areas (black poly-
gons) to their real built-plan areas (white rectangles). Inaccuracy in the resultant built area
is accepted as the definition from aerial imagery of all of the building shapes in a given
municipality is an extremely time-consuming task.
For the municipalities with cadastral information, the distribution of number of storeys
was obtained directly from the cadastral map. For the remaining municipalities, the dis-
tribution of number of storeys was defined as: (a) distribution of a similar municipality

123
Nat Hazards (2017) 86:S31–S54 S41

Fig. 6 Definition of built-plan area for Apartadó

with cadastral information (similar population and/or area), or (b) if there was not a similar
municipality with cadastral information, a virtual survey was performed to obtain the
typical number of floors. Virtual surveys refer to the collection of information about storey
height and structural system from randomly selected buildings, using imagery available in
the Google Street View application.
As no information was available about the structural system, virtual surveys were
performed. This process was comprised of four steps: (1) the municipality with the greatest
population within each sub-region was selected; (2) data from the virtual survey was used
for the determination of the building classes; (3) the building class distribution of the
selected municipality was applied to all of the municipalities of the same sub-region; and
(4) adjustments were done in each municipality as data of the 2005 General Census
(DANE 2005) identifies the percentage of earth and wooden dwellings. Figure 7 presents
the number of storeys and the structural system distribution for Apartadó.
The structural system distribution was defined only according to the number of storeys,
as socio-economic strata information was not available. Regardless of the structural sys-
tem, a unique replacement cost of 1,000,000 Colombian Pesos (318 US) per meter square
was assumed for all of the municipalities. This value was defined based on the judgement
of several construction companies from the region, and taking into consideration that the
majority of the structures outside the metropolitan area do not comply with code regula-
tion, and therefore the replacement cost will be rather uniform considering that a damaged
building must be replaced by a structure that complies with code regulations.

Fig. 7 Number of storeys and structural system distribution for Apartadó. a Number of storeys distribution.
b Structural system distribution

123
S42 Nat Hazards (2017) 86:S31–S54

5 Vulnerability of unreinforced masonry structures

As shown in Fig. 3, near 60% of the residential building stock in Antioquia is unreinforced
masonry structures (508 thousands of buildings), which unfortunately are highly vulnerable
to seismic hazard. Colombian unreinforced masonry structures showed a poor performance
during the Armenia earthquake of 1999 (Mw 6.2); nearly 50,000 dwellings were reported as
affected by the event, from which 17,551 urban dwellings were reported as destroyed or
non-habitable (Cardona et al., 2004). Although data of the building typology of the affected
dwellings are not available, it is well known that many of those structures were unrein-
forced masonry buildings. Figure 8 shows images of typical unreinforced masonry struc-
tures in Medellı́n. The majority of these buildings in Antioquia are characterized by
informal construction at low-income neighbourhoods, ignoring seismic provisions, as
presented in Fig. 8a. On the other hand, Fig. 8b illustrates pre-code unreinforced masonry
structures (before the year 1984) at medium-income neighbourhoods, but with better
construction practice.
Figure 9 presents the distribution of unreinforced masonry buildings according to the
number of storeys for Medellı́n, Aburrá Valley, and Antioquia. It can be observed that
about 80% of these buildings have one or two storeys and nearly 10% of the buildings have
three storeys. A very small percentage of buildings are structures of four or more storeys
(2, 4 and 3% for Medellı́n, the Aburrá Valley and Antioquia, respectively). Six storeys
buildings are only found in the Aburrá Valley and Medellı́n and represent 0.10% of the
total number of unreinforced masonry buildings.
In order to perform an earthquake loss scenario, fragility and/or vulnerability functions
are required. Fragility functions relate the probability of exceeding a set of damage states

Fig. 8 Examples of unreinforced masonry structures. a Lower socio-economic strata. b Medium socio-
economic strata

123
Nat Hazards (2017) 86:S31–S54 S43

Fig. 9 Number of storeys distribution for unreinforced masonry buildings in Medellı́n, Aburrá Valley, and
Antioquia

conditional on a given ground motion level, while vulnerability functions relate the
probability of loss ratio (ratio of repair cost to replacement cost) for a set of ground motion
levels. Fragility functions can be converted into vulnerability functions using a damage-to-
loss model.
In this study, analytical fragility functions were derived for in-plane response of
unreinforced masonry structures based on the simplified pushover-based earthquake loss
assessment method, SP-BELA (Borzi et al. 2008). A simplified procedure was selected in
order to reduce the computational and modelling effort. The SP-BELA methodology for in-
plane failure mechanism is partially based on the MeBaSe procedure (Restrepo-Vélez and
Magenes 2004), which has been proven to be suitable for application to Colombian
unreinforced masonry structures. Six building classes were considered according to the
number of storeys. In order to derive the fragility functions, the structural capacity of each
building class was represented by a set of 100 equivalent single-degree-of-freedom
(SDOF) oscillators, while the demand was represented by 300 ground motion records, thus
allowing the propagation of the building-to-building and the record-to-record variabilities
(see Sect. 5.3). In order to obtain the response of each SDOF oscillator under each
accelerogram, a nonlinear time history analysis (NLTHA) procedure was used (see
Sect. 5.4). The following sections present in detail the steps involved in the fragility curves
generation.

5.1 Capacity curves

Capacity curves for in-plane failure mechanism were calculated using a simplified push-
over analysis (Borzi et al. 2008). The pushover curve was defined as a bilinear curve (see
Fig. 10) using three parameters: yield displacement (Dy), ultimate displacement (Du), and
collapse multiplier (k).
The yielding and ultimate displacements can be calculated using the following
formulae:
Dy ¼ k1 hT dy ð1Þ
 
Du ¼ k1 hT dy þ k2 du  dy hp ð2Þ

where hT and hp are the total building height and inter-storey height, and dy and du are the
inter-storey drift corresponding to the drift limit at yield and the collapse drift, respectively.

123
S44 Nat Hazards (2017) 86:S31–S54

Fig. 10 Capacity curve for

elastic perfectly plastic structural
behaviour for in-plane
mechanism (adapted from Borzi
et al. 2008)

Parameters k1 and k2 are coefficients required to obtain the equivalent height of the single-
degree-of-freedom system, as defined by Restrepo-Vélez and Magenes (2004) for buildings
with masses uniformly distributed along the building height and for walls with a mass
equal to 30% of the floor mass.
The collapse multiplier, k, relates to the lateral load required to produce the structure
collapse, F, by the structure mass, M, and the gravitational acceleration, g, as follows:
F ¼ kMg ð3Þ

The structure collapse multiplier will be the smallest among the collapse multiplier of
each floor (ki). The collapse multiplier for the ith floor (ki), was computed using formula
(4), as proposed by Benedetti and Petrini (1984).
 Pn 1=2
1 k¼i Wk
ki ¼ P n Ai ski 1 þ ð4Þ
hk Wk 1:5ski Ai ð1 þ cAB Þ
WT Pk¼i
n
j¼1
hj Wj

where WT stands for the total weight of the building (including weight of walls, floor, roof,
dead loads and 30% of live loads), Wi refers to the weight of the floor i, ski represents the
shear resistance of the masonry at the floor i, Ai refers to the total area of the resisting walls
at level i in the direction of application of the seismic loads, cAB represents the ratio
between Ai and Bi, with Bi being the maximum area between the area of wall in the loaded
direction and the orthogonal direction, and n is the number of storeys. Masonry shear
resistance was conservatively taken as 200 kPa based on experimental results (López et al.
2012) and expert judgment. The building collapse multiplier is the smallest among all of
the calculated collapse multipliers.
Typically, unreinforced buildings in the metropolitan area of Medellı́n have similar
floor weights. Thus, the weakest floor (smallest collapse multiplier) is usually located in
the lower floor. In addition, it was assumed equal values of Ai and Bi, floor weight, storey
height, and shear resistance throughout the building height. Equation (4) is rewritten taken
into consideration these assumptions and leading to the following formula:
 1=2
1 qwt
k¼ q sk 1 þ ð5Þ
qwt wA 1:5sk qwA ð1 þ cAB Þ
where qwA is the ratio between area of walls in a floor to floor area (wall density), qwt
stands for the ratio of total building weight to floor area (weight density), cAB is the ratio
between qwA and qwB, with wB being the maximum wall density between the wall density
in the direction of the applied seismic loads and the orthogonal direction.

123
Nat Hazards (2017) 86:S31–S54 S45

Torsional effects were taken into account by the incorporation of the correction coef-
ficient /c suggested by Restrepo-Vélez and Magenes (2004) and expressed in formula (6).
sk
/c ¼ 5:5   þ 0:5 ð6Þ
Lw=
LT

The correction coefficient is function of the shear resistance of the masonry, sk, and the
ratio of the total length of the perimeter walls (including openings and non-structural
walls), in the direction of the applied seismic loads, LT, to the sum of the lengths of the
resistant perimeter walls in the same direction, Lw. The coefficient was derived by
Restrepo-Vélez and Magenes (2004) from the comparison of the structural behaviour of 3D
models versus the simplified 2D models of five different buildings with two to five storeys
and different shear strength of masonry. Formula (7) presents the final equation used for
the calculation of the collapse multiplier.
(  1=2 )
1 1 qwt
k¼  q sk 1 þ ð7Þ
5:5 L ski þ 0:5 qwt wA 1:5sk qwA ð1 þ cAB Þ
w=
LT

Forty unreinforced masonry buildings were selected from the 151 structures initially
surveyed (see Sect. 3) in order to gather data required for the generation of the capacities
curves. Plan views of the lower floor of the forty buildings were analysed in order to
calculate wall densities in both directions. Figure 11 shows plan views of two of the
surveyed buildings; it can be observed that the typical dimensions of this building typology
(15 m length, 6 m width, and inter-storey height of 2.5 m) reported by Mejı́a (2011) are in
agreement with the surveyed buildings. Floor weights were estimated from the observed
type and thickness of both floor and roof. Values of dead and live loads from current
Colombian seismic code (NSR-10) were used (AIS 2010). As stated before, value of
masonry shear resistance was based on experimental results. The collected data displayed
no significant correlation between the structural parameters and the number of storeys.
Therefore, data from the entire sample (regardless the number of storeys) were used to
generate the statistics (best-suited distribution, mean and variance) for the wall density,
slab weight, floor area and inter-storey height, as describe in Table 3. The values presented
in the table were used to generate the capacity curves for buildings between one and six
floors.
In addition to the parameters presented in Table 3, yielding and ultimate displacements
are required for the calculation of the capacity curves. As previously depicted in Fig. 10,
the yielding and ultimate displacements depend on the total height of the building (ht) and
the inter-storey height (hp). For the purposes of this study, the same inter-storey height was
assumed throughout each building, and thus the total height was estimated by multiplying
the inter-storey height by the number of storeys.
For the definition of the collapse drift and the ratio between the collapse and yielding
drifts, experimental results of unreinforced masonry walls reported by Magenes and Calvi
(1997) were used. These results indicated a uniform distribution of the du and a large
scatter in the ratio du/dy. Lower and upper limits for both du and du/dy were defined as plus/
minus one standard deviation, as described in Table 4.
One hundred capacity curves were randomly generated for each building class (defined
by the number of storeys). Figure 12 shows the capacity curves for one, three and six
storeys buildings. Darker lines indicate the mean value of each set of capacity curves.

123
S46 Nat Hazards (2017) 86:S31–S54

Fig. 11 Examples of plan views of the surveyed unreinforced masonry buildings

Table 3 Statistical parameters from gathered data

Parameter Mean Variance Statistical
distribution

Wall density in the longitudinal direction, qwx (m2/m2) 0.05 1.02E-04 Log-normal
Wall density in the orthogonal direction, qwy (m2/m2) 0.0745 1.50E-04 Log-normal
Ratio between slab weight and floor area, qslab (MN/m2) 0.0033 1.45E-07 Log-normal
Dead load (MN/m2) 0.001 – –
30% of live load (MN/m2) 0.00054 – –
Inter-storey height (m) 2.4 5.55E-02 Log-normal
Masonry shear resistance (MN/m2) 0.2 0.0025 Log-normal

5.2 Damage states

For each SDOF, the observed maximum spectral displacement due to each ground motion
record was estimated and compared with the corresponding damage state thresholds in
order to allocate the structure within a damage state. Damage limit states were defined as a

123
Nat Hazards (2017) 86:S31–S54 S47

Table 4 Results for the collapse drift and ratio between the collapse and yielding drifts
Parameter Mean C.O.V. (%) SD Lower limit Upper limit

du 0.005 11 0.00055 0.00445 0.0055

du/dy 5.0 46 2.3 2.7 7.3

Fig. 12 Capacity curves for in-plane failure mechanism of unreinforced masonry buildings of one, three
and six storeys. a 1 storey. b 3 storeys. c 6 storeys

Table 5 Damage state criterion

Limit state Displacement

Slight 0.7dy
Moderate 0.25  (3dy ? du)
Extensive 0.5  (dy ? du)
Collapse du

function of the yielding and ultimate displacements (dy and du, respectively), as described
in Table 5 and suggested by several studies (Lagomarsino and Giovinazzi 2006; Erberik
2007; Silva et al. 2014). Comparison of the displacements to damage state thresholds
allowed building a damage matrix containing the number of SDOF in each damage state
per ground motion record.

5.3 Ground motion records selection

The ground motion records used in this study were selected from the PEER (Pacific
Earthquake Engineering Research) database, in such a way that they would be compatible
with the local tectonic environment and seismicity, as previously described in Sect. 2. The
selected set was composed of ground motion records with moment magnitudes between 5
and 9, depending on the distance. For long distances (i.e. over 50 km), only records with a
moment magnitude above 7 were considered, while for shorter distances magnitudes
between 5 and 7 were selected. A rock soil type was assumed and no near-fault effects
were considered. A maximum scaling factor of 2 was considered (Watson-Lamprey and
Abrahamson 2006). Two intensity measure types (IMT) were defined as representative of
the dynamic characteristics of the building classes: peak ground acceleration (PGA) for the
stiff buildings (one to three storeys) and spectral acceleration (Sa) at 0.4 s for buildings
with four or more storeys (0.4 s was defined based on the elongated period, in order to
account for the effect of structural damage in the dynamic properties of the building stock).

123
S48 Nat Hazards (2017) 86:S31–S54

For each IMT, 10 levels of intensity were defined and 30 records were scaled, leading to
two sets of 300 ground motion records.

5.4 Fragility functions

To develop the fragility functions equivalent SDOF oscillators were generated based on the
capacity curves for each building class. Nonlinear time history analyses (NLTHA) were
performed using the GEM’s Risk Modeller’s Toolkit (Silva et al. 2015). This module uses
the open-source software for nonlinear structural analysis OpenSEES (McKenna et al.
2000) to perform the nonlinear dynamic analyses. The hysteretic behaviour of each SDOF
was defined based on its capacity curve and following the ‘‘Pinching4 Material’’ model
with degradation, as defined by McKenna et al. (2000).
For each SDOF, the maximum spectral displacement due to each record was estimated
and compared with the corresponding damage state thresholds (see Sect. 5.2) in order to
allocate the structure within a damage state. This process allowed building a damage
matrix containing the number of SDOF in each damage state. Finally, each fragility
function was modelled using a cumulative lognormal distribution, whose parameters (i.e.
logarithmic mean and logarithmic standard deviation) were derived through a statistical
regression (least square method).
Figure 13 presents the resulting fragility functions for the six building classes, and
Table 6 contains the statistical parameters for each function.

6 Seismic damage scenarios

In order to explore the seismic risk of the unreinforced masonry building stock, a scenario
damage assessment was conducted considering the six building classes (from one to six
storeys). The calculations were performed using the OpenQuake engine, the open-source
software for seismic hazard and risk analysis developed by the Global Earthquake Model
(Silva et al. 2013; Pagani et al. 2014).
Two scenarios were selected for the damage assessment corresponding to an active
shallow event of magnitude 5.9 at a depth of 10 km in the south-west of the Aburrá Valley
(6.21 N, 75.64 W) and a subduction intraslab event with magnitude 7.1 at a depth of
112 km (epicentre at 6.08 N, 75.70 W), as shown in Fig. 14. Both scenarios were selected
based on probabilistic seismic risk analysis, in which events that led to a number of
collapsed buildings with a return period close to 500 years in the metropolitan area were
identified. These calculations were performed using the recently proposed probabilistic
seismic hazard model for South America (https://sara.openquake.org/hazard) and consid-
ering an investigation time of 100,000 years. Four equally weighted ground motion
attenuation equations for active shallow events (Akkar et al. 2014; Bindi et al. 2014; Boore
et al. 2014; Sadigh et al. 1997) and two equally weighted ground motion attenuation
equations for subduction intraslab events (Abrahamson et al. 2016; Montalva et al. 2016)
were used to account for the epistemic uncertainty in the selection of the ground motion
model. Site effects were taken into account through the shear wave velocity in the top 30-m
layer (Vs30), as indicated in the microzonation study for the region (Consorcio Micro-
zonificación 2007). Figure 14 presents the mean ground motion field for PGA for both
scenarios.

123
Nat Hazards (2017) 86:S31–S54 S49

Fig. 13 Fragility curves for in-plane failure mechanism of masonry buildings between one and six floors.
a 1 storey. b 2 storeys. c 3 storeys. d 4 storeys. e 5 storeys. f 6 storeys

To assess the damage distribution, one thousand ground motion fields were generated
and combined with the exposure and fragility functions to calculate the mean and standard
deviation of the number of buildings in each damage state (e.g. Ansal et al. 2009; Fiorini
et al. 2012). Figure 15 presents the spatial distribution of the mean number of collapsed
MUR buildings in each municipality. A total of 21 thousand and 16 thousand buildings are
expected to collapse in the active shallow and subduction scenario, respectively. In both
scenarios the majority of collapsed buildings are located in the Aburrá Valley (6% of the
exposed assets for the active shallow scenario; 4% for the subduction scenario). Collapsed
buildings from the shallow event concentrate in the city of Medellı́n and its neighbourhood
municipalities; on the other hand, collapsed buildings from the subduction event are
widespread in the department.

123
S50 Nat Hazards (2017) 86:S31–S54

Table 6 Statistical parameters of the fragility functions

Unreinforced Intensity Slight Moderate Extensive Collapse
masonry measure type
Mean SD Mean SD Mean SD Mean SD

1 Storey PGA [g] 0.606 0.288 1.390 0.599 1.547 0.616 1.783 0.656
2 Storeys PGA [g] 0.207 0.057 0.424 0.119 0.524 0.139 0.702 0.183
3 Storeys PGA [g] 0.324 0.095 0.450 0.150 0.594 0.224 0.778 0.243
4 Storeys Sa (0.4 s) [g] 0.278 0.081 0.331 0.067 0.418 0.134 0.563 0.141
5 Storeys Sa (0.4 s) [g] 0.215 0.037 0.286 0.085 0.346 0.105 0.451 0.113
6 Storeys Sa (0.4 s) [g] 0.204 0.057 0.239 0.067 0.280 0.089 0.359 0.103

Fig. 14 Earthquake scenarios in Antioquia: epicentre and mean ground motion. a Active shallow.
b Subduction intraslab

Results of the shallow event demonstrate that for a moderate earthquake southwest of
the Aburrá Valley, almost 133 thousand buildings can be significantly affected (26% of
Antioquia’s UMR building stock). From these affected buildings, 72% will experience
slight and moderate damage, and the remaining 28% will undergo extensive damage or
collapse. Similar results are produced by the subduction event assessment, in which 129
thousand buildings would be affected. 25% of those affected buildings would experience
extensive damage or collapse.
Figure 16a shows mean and standard deviation of the fraction of assets in each
damage state for each of the ground motion attenuation equation considered in both
scenarios. Figure 16b shows the distribution of damage buildings per class. Both events
indicate that the most vulnerable building classes are the unreinforced masonry damage
structures between four and six storeys (around 65% of the buildings suffered extensive
damage or collapsed). These results highlight the importance of accurate vulnerability
and risk studies and the need to develop appropriate risk mitigation actions for this type
of construction.

123
Nat Hazards (2017) 86:S31–S54 S51

Fig. 15 Spatial distribution of number of collapse buildings in the Aburrá Valley: active shallow event
(a) and subduction event (b)

Fig. 16 Damage assessment for a shallow M5.9 event in southwest Aburrá Valley and a subduction M7.1
event. a Total damage distribution. b Damage distribution per building class

7 Conclusions

An exposure model for the residential building stock of the Department of Antioquia, as
well as fragility functions for unreinforced masonry buildings (the most common type of
construction) were developed for the purposes of assessing seismic risk.

123
S52 Nat Hazards (2017) 86:S31–S54

The development of an ideal exposure model requires detailed information that might
not be accessible. Nevertheless, as long as aerial imagery and images of the building stock
(as those provided by Google Street View) are available, an approximated exposure
model—in a reasonable amount of time—can be generated with the methodology pre-
sented in this work. Three different regions were considered due to data availability:
Medellı́n (Antioquia’s capital), Aburrá Valley (metropolitan area of Medellı́n) and
municipalities outside the Aburrá Valley. The methodology used for the development of
these models has been thoroughly described and could be extended to other regions, even
outside of Colombia. The exposure model includes data on built-up area, number of
buildings and inhabitants, building class, and replacement cost. Exposure models devel-
oped with the proposed methodology would be as precise as the quality of the input data.
Comparison of results of the presented exposure models with available data on number of
dwellings shows a good estimation of the building stock for the exposure models devel-
oped with good quality input data: the city of Medellı́n and the municipalities of the
metropolitan area (Aburrá Valley). Nevertheless, comparisons about the structural typol-
ogy distribution of the building stock could not be performed as references were not
available. The exposure model developed for the municipalities outside the Aburrá Valley
incorporates many uncertainties as input data was scarce. However, the presented model
constitutes an important contribution that can be used for an initial estimation of the
seismic risk in the region. The models can be easily improved as new data becomes
available.
The exposure models indicate unreinforced masonry structures as the main typology in
Antioquia, with 53% of the built-up area, representing a total of 508 thousands buildings
(61% of the building stock). Confined masonry constitutes the second most common
building typology with a total of 97 thousand buildings (11% of the built-up area and 12%
of the building stock).
Fragility functions for in-plane failure mechanism were derived for unreinforced
masonry structures based on the local construction practice. Forty unreinforced masonry
buildings were inspected in order to generate statistics of the parameters that define the
structure capacity. One hundred capacity curves were generated for six different building
classes (from one to six storeys) in order to consider building-to-building variability.
Seismic demand was considered by the use 300 hundred real ground motion records to take
into account record-to-record variability.
As both fragility and exposure model could be used for seismic assessment, two seismic
scenarios for unreinforced masonry structures were performed. The scenarios considered
are plausible events that could take place in the region of interest. The earthquake scenarios
were performed using the OpenQuake engine (Silva et al. 2014; Pagani et al. 2014).
Results from both scenarios indicate that about 20% of the total building stock of unre-
inforced masonry structures would undergo extensive damage to collapse, being the
structures between four to six storeys the most vulnerable building classes. Future studies
should focus on the development of fragility functions for the additional building
typologies of the building stock in order to evaluate seismic risk for all building types.
Results of the considered damage scenarios clearly demonstrate the high vulnerability
of Colombian unreinforced masonry buildings. Even though unreinforced masonry is
forbidden by the Colombian code, this building typology represents 61% of the building
stock. It becomes a necessity to endorse retrofitting campaigns for this type of buildings.
Local researchers such as Vega and Torres (2015), López et al. (2012), and Bastidas et al.
(2002), have addressed the issue and have proposed methods for the improvement of the
seismic capacity of this type of construction. Nonetheless, it is important to understand that

123
Nat Hazards (2017) 86:S31–S54 S53

the majority of the unreinforced masonry buildings are inhabited by low-income residents,
and thus it is unlikely that the residents will take initiative. Governmental policies should
encourage the enhancement of this type of construction, such as the reduction in taxes or
development of financial mechanisms.

Acknowledgements The authors would like to express their gratitude to the cadastral offices of the
municipalities of Medellı́n and Girardota for the data provided. Valuable information used for the micro-
zonation of the metropolitan area of Medellı́n was available to the authors of this work, which is highly
appreciated. The authors would also like to acknowledge the contribution of local experts for the definition
of some parameters required for the exposure model development. Funding was provided by Swiss Re
Foundation.

References
Abrahamson N, Gregor N, Addo K (2016) BC hydro ground motion prediction equations for subduction
earthquake. Earthq Spectra 32(1):23–44
Asociación Colombiana de Ingenierı́a Sı́smica. AIS (2010) Reglamento Colombiano de Construcción Sismo
Resistente NSR-10, Bogotá, DC
Akkar S, Sandikkaya MA, Bommer JJ (2014) Empirical ground-motion models for point- and extended-
source crustal earthquake scenarios in Europe and the Middle East. Bull Earthq Eng 12(1):359–387
Ansal A, Akinci A, Cultrera G, Erdik M, Pessina V, Tonuk G, Ameri G (2009) Loss estimation in Istanbul
based on deterministic earthquake scenarios of the Marmara Sea region (Turkey). Soil Dyn Earthq Eng
29:699–709
Asociación Colombiana de Ingenierı́a Sı́smica—AIS (2009) Estudio General de Amenaza Sı́smica de
Colombia. Comité AIS-300, Bogotá, DC
Atuesta JA (1972) Evaluación del riesgo sı́smico para Colombia. Proyecto de grado. Universidad de Los
Andes, Bogotá
Bastidas ED, Molina LF, Yamin L, Sarria A, Ortega JP (2002) Non-structural masonry walls strenthened
with GFRP laminates. Revista de Ingenierı́a. Universidad de los Andes, Bogotá
Benedetti D, Petrini V (1984) Sulla vulnerabilità sı́smica di edifici in muratura: proposta su un método di
valutazione. L’industria delle Costruzioni 149:66–74
Bindi D, Massa M, Luzi L, Ameri G, Pacor F, Puglia R, Augliera P (2014) Pan-European ground motion
prediction equations for the average horizontal component of PGA, PGV and 5%-damped PSA at
spectral periods of up to 3.0 s using the RESORCE dataset. Bull Earthq Eng 12(1):391–430
Bonnet R (2003) Vulnerabilidad y riesgo sı́smico de edificios. Aplicación a entornos urbanos en zonas de
amenaza alta y moderada. Tesis Doctoral, Universidad Politécnica de Cataluña, Barcelona
Boore DM, Stewart JP, Seyhan E, Atkinson G (2014) NGA-West2 equations for predicting PGA, PGV, and
5% damped PGA for shallow crustal earthquake. Earthq Spectra 30(3):1057–1085
Borzi B, Crowley H, Pinho R (2008) Simplified pushover-based earthquake loss assessment (SP-BELA)
method for masonry buildings. Int J Archit Herit Conserv Anal Restor 2:353–376
Brzev S, Scawthorn C, Charleson AW, Allen L, Greene M, Jaiswal K, Silva V (2013) GEM Building
Taxonomy Version 2.0, GEM Technical Report 2013-02 V1.0.0, GEM Foundation, Pavia, Italy.
doi:10.13117/GEM.EXP-MOD.TR2013.02
Cardona OD, Eeri M, Yamı́n LE (1997) Seismic microzonation and estimation of earthquake loss scenarios:
integrated risk mitigation project of Bogotá, Colombia. Earthq Spectra 13(4):795–814
Cardona OD, Wilches-Chaux G, Garcı́a X, Mansilla E, Ramı́rez F, Marulanda MC (2004) Estudio sobre
desastres ocurridos en Colombia: Estimación de pérdidas y cuantificación de costos, Departamento
Nacional de Planeación, ACCI, Banco Mundial
Consorcio Microzonificación 2006, Área Metropolitana del Valle de Aburrá (2007) Microzonificación y
Evaluación del Riesgo Sı́smico del Valle de Aburrá, Medellı́n
Departamento Administrativo Nacional de Estadı́stica. DANE (2011) Encuesta de Calidad de Vida-2011
Departamento Administrativo Nacional de Estadı́stica—DANE (2005) Censo General 2005—República de
Colombia, Bogotá, DC
Erberik MA (2007) Fragility-based assessment of typical mid-rise and low-rise RC buildings in Turkey. Eng
Struct 30(5):1360–1374
Fiorini E, Borzi B, Iaccino R (2012) Real time damage scenario: case study for the L’Aquila earthquake. In:
Proceedings of the 15th world conference on earthquake engineering, Lisbon, Portugal

123
S54 Nat Hazards (2017) 86:S31–S54

Lagomarsino S, Giovinazzi S (2006) Macroseismic and mechanical models for the vulnerability and damage
assessment of current buildings. Bull Earthq Eng 4(4):415–443
López S, Quiroga PN, Torres N (2012) Evaluation of unreinforced masonry walls covered with reinforced
mortars. In Proccedings of the XXXV Jornadas Sul Americanas de Engenharia Estructural, Rio de
Janiero, Brasil
Magenes G, Calvi GM (1997) In-plane seismic response of brick masonry walls. Earthq Eng Struct Dyn
26:1091–1112
Martı́nez JM, Parra E, Parı́s G, Forero C, Bustamente M, Cardona OD, Jaramillo JD (1994) Los sismos del
Atrato Medio 17 y 18 de octubre de 1992 Noroccidente de Colombia. Rev Ing 4:35–76
McKenna F, Fenves G, Scott M, Jeremic B (2000) Open system for earthquake engineering simulation
(OpenSees). Pacific Earthquake Engineering Research Center. Technical report. University of Cali-
fornia, Berkeley, CA
Mejı́a LG (2011) Non-engineered unreinforced brick masonry building. Housing Report, World Housing
Encyclopedia. Earthquake Engineering Research Institute (EERI) and International Association of
Earthquake Engineering (IAEE)
Montalva G, Bastias N, Rodriguez-Marek A (2016) Ground motion prediction equation for the chilean
subduction zone. Seismol Res Lett (under review)
Pagani M, Monelli D, Weatherill G, Danciu L, Crowley H, Silva V, Henshaw P, Butler L, Nastasi M,
Panzeri L, Simionato M, Vigano D (2014) OpenQuake-engine: an open hazard (and risk) software for
the global earthquake model. Seismol Res Lett 85(3):692–702. doi:10.1785/0220130087
Pulido N (2003) Seismotectonics of the Northern Andes (Colombia) and the development of seismic
networks. Bulletin of the International Institute of Seismology and Earthquake Engineering, Special
Edition, pp 69–76
Ramı́rez JE (1975) Historia de los Terremotos en Colombia. Documentación Geográfica, Instituto Geo-
gráfico Agustı́n Codazzi, Bogotá
Restrepo-Vélez, LF, Magenes G (2004) Simplified procedure for the seismic risck assessment of unrein-
forced masonry buildings. In: Proceedings of the thirteenth world conference on earhtquake engi-
neering. Vancouver, Canada, paper no. 2561
Sadigh K, Chang C-Y, Egan JA, Makdisi F, Youngs RR (1997) Attenuation relationships for shallow crustal
earthquakes based on California strong motion data. Seismol Res Lett 68(1):180–189
Salgado MA, Zuloaga D, Cardona OD (2013) Evaluación probabilista del riesgo sı́smico de Bogotá y
Manizales con y sin la influencia de la Caldas Tear. Revista de Ingenierı́a, enero-junio:6–13.
Universidad de los Andes, Bogotá DC
Salgado-Gálvez MA, Zuloaga-Romero D, Bernal GA, Mora MG, Cardona OD (2014) Fully probabilistic
seismic risk assessment considering local site effects for the portfolio of buildings in Medellı́n,
Colombia. Bull Earthq Eng 12:671–695. doi:10.1007/s10518-013-9550-4
Silva V, Crowley H, Pagani M, Monelli D, Pinho R (2013) Development of the OpenQuake engine, the
Global Earthquake Model’s open-source software for seismic risk assessment. Nat Hazard. doi:10.
1007/s11069-013-0618-x
Silva V, Crowley H, Pinho R, Varum H (2014) Investigation of the characteristics of Portuguese regular
moment-frame RC buildings and development of a vulnerability model. Bull Earthq Eng
13(5):1455–1490
Silva V, Casotto C, Rao A, Villar M, Crowley H, Vambatsikos D (2015) Openquake risk modeller’s
toolkit—user guide. Global Earthquake Model (GEM). Technical Report 2015-06. doi:10.13117/GEM.
OPENQUAKE.MAN.RMTK.1.0/01
Vega C, Torres N (2015) Comportamiento dinámico de muros de mamposterı́a no estructural reforzados
mediante polı́meros reforzados con fibra. Memorias del VII Congreso Nacional de Ingenierı́a Sı́smica,
Universidad de los Andes y Asociación Colombiana de Ingenierı́a Sı́smica. Bogotá, mayo de 2015
Watson-Lamprey J, Abrahamson N (2006) Selection of ground motion time series and limits on scaling. Soil
Dyn Earthq Eng 26:477–482
Yepes C, Silva V, Valcárcel J, Acevedo A, Hube M, Coronel G (2016) A uniform residential building
inventory for South America. Earthq Spectra (under review)

123