Content uploaded by Vitor Silva
Author content
All content in this area was uploaded by Vitor Silva on Sep 23, 2014
Content may be subject to copyright.
1
INTEGRATED RISK MODELING WITHIN THE GLOBAL
EARTHQUAKE MODEL (GEM): TEST CASE APPLICATION FOR
PORTUGAL
Christopher BURTON1 and Vitor SILVA2
ABSTRACT
At the forefront of the Global Earthquake Model (GEM) is the development of uniform
standards, datasets, and state-of-the-art modeling tools for the communication of earthquake risk. For a
more holistic assessment of the scale and consequences of earthquake impacts, spatially enabled and
open databases, methods, and Open Source software tools are being incorporated into the GEM
modeling framework to assess earthquake risk beyond the estimation of direct physical earthquake
impacts and loss of life. The latter is accomplished via the integration of estimates of physical risk (i.e.
estimates of human or economic loss) with quantified metrics that represent social and economic
characteristics of populations. This paper describes a test case/proof of concept for Portugal that was
developed to assess the total (or integrated) risk of the country. Integrated risk is described here as the
convolution of physical earthquake risk estimations with the social characteristics at a particular place.
The test case/proof of concept was constructed in order to inform the development of GEM’s
Integrated Risk Modelling Toolkit that is an Open Source Software tool that will allow users to
meaningfully integrate quantitative assessments of social and economic conditions with physical risk
estimates for earthquakes using GEM’s OpenQuake modeling suite. The results indicate that the
impacts from a damaging earthquake event in Portugal will not be random, but manifested from a set
of interacting conditions, some the result of geography and location, some the result of building
exposure, and some having to do with the social characteristics of populations.
INTRODUCTION
Earthquakes are a complex spatial phenomenon that vary greatly in magnitude and frequency, and
often result in the loss of lives, livelihoods, and property. There has been an exponential growth in the
losses from earthquakes throughout the world, and seismic disasters such as the Haitian Earthquake in
2010 and the Great East Japan Earthquake in 2011 provide reminders of the susceptibility of
communities to the loss of lives and property from damaging events. These disasters illustrated how
earthquakes impact people and communities, and despite sustained efforts to reduce earthquake risk, a
long history of development in seismically active areas has increased the susceptibility of populations
to earthquake impacts. This has stimulated great interest in understanding how to manage the
associated seismic risk.
While losses from earthquakes are the outcome most commonly associated with damaging
earthquakes, it is increasingly becoming clear that some populations are impacted differentially.
Additionally, the ability to prepare for, respond to, and recover from damaging events varies spatially.
1 Senior Scientist, Social Vulnerability and Integrated Risk Coordinator, GEM Foundation, Pavia, Italy,
christopher.burton@globalquakemodel.org
2 Senior Scientist, Physical Risk Coordinator, GEM Foundation, Pavia, Italy, vitor.silva@globalquakemodel.org
2
This is partially because the impacts from earthquakes are not the product of a singular source. Rather,
the impacts suffered are the result of interactions between the earth’s biophysical systems, the
engineered environment, and the social context at particular places. It is when damaging earthquakes
intersect with concentrations of populations and development that they become disasters.
To assess and communicate earthquake risk and the potential for disasters, the Global
Earthquake Model (GEM) addresses earthquake hazard, physical risk, and the differential
susceptibility of populations to adverse impacts. GEM is a foundation that was created to develop
state-of-the-art models, data, and open-source tools and software for understanding and
communicating earthquake risk (see Silva et al. 2013; Pagani et al. 2014). For a holistic evaluation of
the scale and consequences of earthquake impacts, GEM is developing a set of open-source software
tools, methods, and metrics to assess seismic risk and impact potential beyond the estimation of direct
physical impacts and loss of life. This is accomplished via the incorporation (or convolution) of
estimates of physical impacts from earthquakes with social characteristics at particular places, what we
refer to here as an integrated risk assessment. To calculate integrated risk, users will be able to draw
from estimates of physical earthquake risk (i.e. estimates of human or economic loss) and combine
those estimates with socio-economic indicators or computed measures of social vulnerability (i.e.
characteristics within social systems that create the potential for loss or harm). Within the GEM suite
of tools, the integrated evaluation of seismic risk will be accomplished using the Integrated Risk
Modelling Toolkit (Burton et al., in press) that will be publically available November 2014.
The purpose of this paper is to describe a test case/proof of concept for Portugal that was
developed to assess the integrated risk of that country using OpenQuake. The test case/proof of
concept for developing an integrated risk assessment was carried out in order to inform the
development of GEM’s Integrated Risk Modeling Toolkit. The test case was also carried out to
demonstrate the context in which the methods, metrics, and software under development for the
integrated assessment of risk within OpenQuake may be used for decision-making.
BACKGROUND
Perhaps the first attempts to combine assessments of risk with social characteristics were the product
of multihazard analysis where in a seminal work Hewitt and Burton (1971) coupled event magnitude
and frequency with a measure of human impact potential to better understand natural hazard impacts.
Building upon this work, Cutter (1996) and Cutter et al. (2000) formulated the hazards-of-place
approach to vulnerability analysis. The hazards-of-place approach constitutes a detailed delineation of
exposure for a particular study area and an investigation of population sensitivities at levels of analysis
appropriate for both regional and local investigations. The approach is prevalent within the literature
in the United States and has been applied to case studies in South Carolina (Cutter et al. 2000),
California (Burton and Cutter 2008), and Oregon (Wood et al. 2010). International applications of the
hazards-of-place approach include New Zealand (Montz 2000), Europe (Kumpulainen 2006), and
Barbados and St. Vincent (Boruff and Cutter 2007).
In earthquake engineering, the Earthquake Disaster Risk Index (EDRI) (Davidson 1997)
provides an early example of an integrated risk assessment framework. The EDRI includes measures
of hazard, exposure, vulnerability, external contexts, and emergency response and recovery capacities.
Building upon this work, the Urban Disaster Risk Index (UDRI) (Carreño et al. 2007; Carreño et al.
2012) describes risk using a weighted combination of indices. The UDRI was initially applied to
Barcelona (Spain) and Bogota (Colombia) using a wieghted combination of indices aimed at
measuring physical risk and social fragilities within communities. Similar approaches include the
application of physical risk and social indicators and indices in Metro Manila (Fernandez et al. 2007),
Istanbul (Khazai et al. 2008) and Mumbai (Khazai and Bendimerad, 2011). It is within this context
that the term index (or the plural form indices) designates the manipulation of individual variables to
produce an aggregate measure of a phenomenon (e.g. social vulnerability). An indicator is a
quantitative or qualitative measure derived from observed facts that simplify and communicate the
reality of a complex situation (Freudenberg 2003). Indicators are pieces of information that summarize
the characteristics of a system or highlight what is happening in a system. The mathematical
combination of a set of indicators is a composite index.
C. Burton and V. Silva 3
OVERARCHING INTEGRATED RISK MODELLING FRAMEWORK FOR OPENQUAKE
Although a multitude of individual assessment frameworks are applicable, Figure 1 represents the
overarching framework for an integrated risk assessment in OpenQuake. The framework was inspired
by theoretical constructs designed to guide the convolution of assessments of a natural hazard threat,
potential economic losses, and social vulnerability (see Cutter 1996; Cardona 2005). The starting point
of the integrated risk-modeling framework is the modeling of seismic hazard that may be
accomplished for a particular study area in GEM’s OpenQuake-Hazard Engine. The OpenQuake-
Hazard Engine is a Python-based module that uses OpenSHA-lite for modeling earthquake ruptures
and calculating hazard results such as stochastic event sets and ground motion fields. Here, the seismic
hazard is combined with exposure and physical vulnerability from which estimates of physical
earthquake risk are derived using a number of possible calculation workflows. These modules include
1) a deterministic scenario (i.e. a single event) calculator which estimates loss and damage for a
collection of exposed assets such as buildings; 2) a probabilistic event-based risk calculator to estimate
the probability of exceedance of certain levels of loss in a given time span; and 3), a classical PSHA-
based risk calculator to compute the probability of losses for single assets (Silva et al. 2013).
With respect to the distribution of potential losses in an area, both exposure and physical
vulnerability interact with the underlying social fabric of a particular area of analysis. The social fabric
includes socioeconomic characteristics and measures of the overall capacity of a population to respond
to an event (Cutter et al. 2000). It is the inherent characteristics of populations or communities that
help to redistribute risk before an event and after an event in the distribution of losses, and it is the
underlying social fabric of a place that creates a community’s social vulnerability, which when
measured can be viewed as a factor that aggravates or attenuates risk.
Seismic Hazard
Earthquake Scenarios;
Probabilistic Hazard
Geographic
Context
Exposure;
Physical Vulnerability
Physical Risk
Estimates: Human &
Economic Loss Potential
Social Fabric
Place Specific;
Context Specific
Social & Economic
Vulnerability/
Resilience
Integrated
(Place-Based) Risk
Figure 1. Framework for integrated risk assessment in OpenQuake
PHYSICAL SEISMIC RISK MODEL
The evaluation of the physical seismic risk model arises from the convolution of three main
components: seismic hazard, physical vulnerability and exposure data. The probabilistic seismic
4
hazard for Portugal was derived based on an existing source model proposed by Vilanova and Fonseca
(2007) and a set of ground motion prediction equations compatible with the tectonic environment of
the region (Vilanova et al. 2012). For what concerns the physical vulnerability of the exposed
elements, a recently proposed set of vulnerability functions was employed for the reinforced concrete
building typologies (Silva et al. 2014a), whilst for the masonry building stock an existing model
developed by Carvalho et al. (2002) was adopted. Finally, with respect to the location and economic
value of the residential building portfolio, information from the Portuguese Building Census Survey of
2011 was utilized to derive an exposure model with a spatial resolution at the level of the counties.
The seismic risk calculations were carried out using the Classical PSHA-based Risk calculator from
the OpenQuake-engine (Silva et al. 2013). The results included loss exceedance curves for each
county, as well as average annual losses, as illustrated in Figure 3 (Silva et al. 2014b).
SOCIAL VULNERABILITY MODEL
Models serve an important role in helping to understand earthquake risk. The social analog to a
quantitative physical risk model for earthquakes is a social vulnerability index. Building an index
involves a number of steps that are articulated in the literature, including indicator selection, variable
transformation, scaling, weighting, and aggregation. Typically, social vulnerability indices include
deductive, hierarchical, and inductive modelling arrangements where: 1) deductive models contain
typically fewer than 10 indicators which are normalized and aggregated into an index; 2) hierarchical
models employ roughly 10 to 20 indicators that are separated into groups (or sub-indices) that share
the same dimension in which individual indicators are aggregated into sub-indices, and sub-indices are
aggregated; and 3) inductive approaches begin with a large set of indicators which are reduced to a
smaller set of uncorrelated factors using principle components analysis (PCA) (Tate 2012). All model
types were constructed to help inform the development of the workflow of the Integrated Risk
Modelling Toolkit. Only the hierarchical model is reported here due to space constraints.
Because there is no definitive set of indicators for measuring social vulnerability, the selection
of variables was subjective. The index was developed using data for 278 counties in Portugal that were
culled from the country’s census in order to characterize the broader dimensions of social vulnerability
in which there is a consensus in the literature. Initially data to compute 95 variables were collected and
categorized into five basic themes (population, economy, education, infrastructure, and governance
and institutional capacity). Each theme was treated separately for variable selection and aggregation (a
procedure discussed in more detail below), yet when all subcomponents are combined it is intended
that the coupling of constituent parts represent the social vulnerability concept as a whole.
The quality of composite indices and the soundness of the messages they convey depend not
only on the methods used in the construction process, but also on the internal consistency of the
variables selected (i.e. how well the variables may measure the underlying concept). A series of
multivariate analysis were conducted to select an internally consistent and parsimonious set of metrics.
As a first step, the raw data was transformed into comparable scales using percentage, per capita, and
density functions. The data was then standardized using a Min-Max rescaling scheme to create a set of
indicators on the same measurement scale. Min-Max rescaling rescales each variable into an identical
range between 0 and 1 (a score of 0 being the worst rank for an indicator score and 1 being the best
rank). Min-Max rescaling was chosen due the simplicity of the scaling algorithm and interpreting the
resulting indicator ranks for the proof of concept. Further work is being conducted using alternate data
transformation methods to better understand the extent to which data transformation contributes to
sensitivities and uncertainties within the final model output.
A correlation analysis comprised the second step. Preliminary testing of the data revealed a
number of non-parametric and non-linear relationships that were non-transformable. Thus, a non-
linear/non-parametric correlation analyses was applied to assess associations between the data. Highly
correlated variables (Spearman’s R>0.700) were eliminated from further consideration to avoid
subjectively choosing one variable over another for inclusion in subsequent analyses.
In addition to correlation, multidimensional scaling was used to gauge the internal consistency
of the variables to discriminate relevant data from potentially irrelevant data. Multidimensional scaling
is a technique that is often considered a non-parametric alternative to Factor Analysis (FA). Given a
matrix of variables, the procedure represents the data as points in a Euclidian plane in a manner in
C. Burton and V. Silva 5
which two points are closer together when the respective variables are similar in terms of their
distances. The Euclidean plane of points was evaluated under the premise that points spaced closer
together may be internally consistent, e.g. appropriate for measuring the same underlying phenomenon
which is the 5 data themes referred to above. Using this procedure, variables mapped at great distances
from clusters of similar points were scrutinized to understand their source for being an outlier and all
were removed from subsequent analyses.
The multivariate analysis procedures were useful in reducing the data from n=95 to n=27
variables. The remaining 27 variables (Table 1) were considered internally consistent and appropriate
for social vulnerability modeling. The method of aggregation that we employed represents the
summation of equally weighted sub-index scores. In other words, variable scores in each sub-index
(e.g. population, economy, etc.) were averaged to reduce the influence of the number of variables in
each sub-index. Each sub-component was then summed to derive a final composite score. Since there
are five sub-components, the summed score of the composite index ranges between 0 and 5 (0 being
the least socially vulnerable and 5 being the most). As a subsequent step, the composite social
vulnerability scores were rescaled using Min-Max rescaling to produce a final composite score
between zero and one (0 being the least socially vulnerable and 1 being the most vulnerable). An
aggregation method using equal weights was applied due to the lack of theoretical justification for
weighting one variable over another for use in this proof of concept.
Table 1: Variable selection for social vulnerability index
Category
Indicator description
Justification
Population
Percent of the population that is female
Cutter et al. 2003
Population
Percent of the population living in statistical cities
Cutter et al. 2003
Population
Number of recent foreign in-migrations per 1000 population
Cutter et al. 2003
Population
Number of recent in-migrations from another municipality per
1000 Population
Fekete 2009
Population
Percent of the population of foreign nationality
Cutter et al. 2003
Population
Percent of the population under 5 years of age and over 65 years of
age
Cutter et al. 2003
Population
Population density
Mendes 2009
Population
Number of persons per housing unit
Mendes 2009
Population
Percent of female headed households
Mendes 2009
Population
Percent renter occupied housing units
Mendes 2009
Population
Percent of the population with a disability
Mendes 2009
Population
Percent population receiving social integration income of social
security
Fekete 2009
Economy
Percent of the working aged population that is unemployed
Cutter et al. 2003
Economy
Percent female labor force participation
Cutter et al. 2003
Economy
Percent of the labor force working in secondary sector employment
Mendes 2009
Economy
Percent of the labor force that is employed in service industries
Mendes 2009
Economy
Percent of the labor force employed in non-skilled elementary
occupations
Mendes 2009
Economy
Per capita purchasing power
Khazai et al. 2013
Infrastructure
Percent of buildings in need of large and very large repairs
Mendes 2009
Infrastructure
Completed buildings in new constructions per 1000 population
Mendes 2009
Infrastructure
Percentage of the population not served by public water supply
systems
Schneiderbauer and Ehrlich
2006
Infrastructure
Percentage of the population not served by wastewater treatment
Mendes 2009
Education
Percentage of the population without a complete level of education
Mendes 2009
Education
Percent of the population with tertiary education completed
Mendes 2009
Governance
Abstention rate in election for presidency
Cutter et al. 2003
Governance
Abstention rate in municipal elections
Cutter et al. 2003
Governance
Crime rate
Khazai et al. 2013
SIMPLE INTEGRATED RISK MODEL
The evaluation of integrated risk (i.e. the combination of estimated losses with the social vulnerability
index outlined above) required the modeling of expected economic losses for each county (or first
6
order impacts) and the modeling of conditions within social systems that create the potential for harm
and loss. The modeled losses in the form of estimates of average annual loss for each county were also
rescaled using the Min-Max rescaling method to render them commensurate to the social vulnerability
index. To derive an estimate of integrated risk, a total risk index was constructed via the convolution
of the social vulnerability index with the estimates of physical earthquake risk. Carreño et al. (2007;
2012) provide the aggregation method that was adopted for this work due to its mathematical
simplicity. In this method, the direct potential impact of an earthquake (in a general sense) is denoted
as
( )
FRR FT += 1
where
T
R
is a total risk index,
F
R
is a physical earthquake risk index which is an
average annual loss estimate for Portugal derived utilizing the physical risk model outlined above, and
F
is the composite social vulnerability index which may be described as an aggrevating coefficient of
the estimated loss.
RESULTS
The probabilistic seismic hazard for mainland Portugal was calculated using the Classical PSHA-
based hazard calculator (Pagani et al. 2014). A large number of seismic hazard curves were derived
following a 0.01x0.01 decimal degrees spatial resolution considering a wide spectrum of epistemic
uncertainties (e.g. various ground motion prediction models, seismic zonations, magnitude-frequency
distributions), through the employment of a logic tree structure. Using the set of hazard curves at each
location, a mean seismic hazard map for a probability of exceedance of 10% in 50 years was
calculated, as depicted in Figure 2.
Figure 2. Hazard map PGA (g) for Portugal counties
The appraisal of the spatial distribution of hazard in Portugal indicates an expected higher peak
ground acceleration on rock for the Lower Tagus Valley (region around Lisbon) and the south of
Portugal, thus highlighting regions where a combination of seismically vulnerable structures and high
population concentration could lead to significant earthquake losses. The hazard results were
combined with the exposure and vulnerability models to derive loss exceedance curves for each
county. These curves were converted into annual rate of exceedance, and numerically integrated in
order to obtain the average annual economic loss, as illustrated in Figure 3.
C. Burton and V. Silva 7
Figure 3. Map of predicted average annual loss for Portugal counties
In addition to hazard and physical risk, understanding the distribution of the social vulnerability
of populations is an integral part of disaster management, planning, and mitigation. Figure 4 depicts
the spatial variation in social vulnerability for Portugal’s counties. The classification scheme was
simplified (i.e. high, moderate, low) for presentation purposes. The counties delineated in the darker
shades of red along the classification continuum exhibit higher levels of social vulnerability. While the
spatial pattern is not uniform throughout, there are significant pockets of social vulnerability that could
warrant management concern given the seismic threat. Of special interest is the clustering of moderate
to high and high levels of social vulnerability in the southwestern portion of the country. Here,
counties that host cities such as Santarem, Setubal, and Faro have populations with high levels of
social vulnerability. These areas are also in zones of high seismicity (see Figure 2) and zones of high
physical earthquake risk (see Figure 3). In addition to these clusters of counties with high social
vulnerability and physical risk, there is another large section in Portugal’s northern portion with high
levels of social vulnerability. These are in largely rural areas outside of the high-risk zones. However,
exceptions in these areas exist where counties are exposed to considerable risk of loss and that also
contain highly socially vulnerable populations. Populations within these counties may not only suffer
greater impacts due to ground shaking and building damage, they may also lack the ability to
adequately mitigate, prepare for, and recover from a damaging earthquake event. The spatial
distribution of risk within these counties is compounded when viewed as an integrated risk map
(Figure 5).
8
Figure 4. Map of social vulnerability for Portugal counties
Figure 5. Map of integrated risk for Portugal counties
CONCLUSION
The impacts from an earthquake will be expressed differentially across communities. To be effective,
governments, disaster planners, and managers must not only understand the physical agents of
earthquake risk, but also the social characteristics that give rise to vulnerabilities within the
communities they protect. This paper presented a method, workflow, and analysis conducted within
GEM’s OpenQuake that consists of a spatial delineation of physical earthquake risk combined with an
index of social vulnerability. The overall approach leads to an encompassing perspective on risk
C. Burton and V. Silva 9
assessment that considers loss and damage as part of a dynamic system, and our findings suggest that
there are spatial differences in physical earthquake risk, social vulnerability, and integrated risk within
Portugal. Disaster mitigation and planning under such circumstances may require special attention
where different aspects of social vulnerability affect the way in which communities may prepare for
and respond to the seismic threat. In sum, the approach mainstreams risk and social vulnerability into
policy discussions on earthquake loss and damage reduction, makes it possible to use risk estimates in
benchmarking exercises to monitor changes in loss potential over time, and recognizes that both the
causes and solutions for earthquake loss are found in human, environmental, and built-environment
relationships.
REFERENCES
Boruff BJ and Cutter SL (2007) “The Environmental Vulnerability of Caribbean Island Nations,”
Geographical Review, 97(1): 932-942
Burton, CG and Cutter SL (2008) “Levee Failures and Social Vulnerability in the Sacramento-San
Joaquin Delta Area, California, “Natural Hazards Review”, 9(3): 136-149
Burton CG, Khazai B, Silva V, (in press) “Social Vulnerability and Integrated Risk Assessment
within the Global Earthquake Model”, Proceedings of 10th National Conference in Earthquake
Engineering, Earthquake Engineering Research Institute, Anchorage, Alaska
Cardona, OD (2005) Indicators of Disaster Risk and Risk Management, IADB-UNC/IDEA Program
for Latin America and the Caribbean, Universidad Nacional de Colombia, Instituto de Estudios
Ambientales (IDEA), Manizales
Carreño M-L, Cardona OD, Barbat AH (2007) “Urban Seismic Risk Evaluation: A Holistic
Approach,” Natural Hazards, 40: 137-172
Carreño M-L, Cardona OD, Barbat AH (2012) “New Methodology for Urban Seismic Risk
Assessment from a Holistic Perspective,” Bulletin of Earthquake Engineering, 40: 137-172
Carvalho EC, Coelho E, Campos Costa A, Sousa ML, Candeias P (2002) “Vulnerability Evaluation of
Residential Buildings in Portugal,” Proceedings of the 12th European Conference on
Earthquake Engineering, London, United Kingdom
Cutter, SL (1996) “Vulnerability to Environmental Hazards,” Progress in Human Geography, 20:
529-539
Cutter SL, Boruff BJ, Shirley WL (2003) “Social Vulnerability to Environmental Hazards,” Social
Science Quarterly, 84: 242-261
Cutter SL, Mitchell J, Scott M (2000) “Revealing the Vulnerability of People and Places: A Case
Study of Georgetown County, South Carolina,” Annals of the Association of American
Geographers, 90 (4): 713-737
Davidson, R (1997) “A Multidisciplinary Urban Earthquake Disaster Risk Index,” Earthquake
Spectra, 13(2): 211-223
Fekete, A (2009) “Validation of a Social Vulnerability Index in Context to River-floods in Germany,”
Natural Hazards and Earth Systems Sciences, 9: 393-403
Fernandez J, Mattingly S, Bendimerad F, Cardona OD (2006) Application of Indicators in Urban and
Megacities Disaster Risk Management, A Case Study of Metro Manila, EMI Topical Report
TR-07-01
Freudenberg M (2008) Composite Indicators of Country Performance: A Critical Assessment, OECD,
Paris
Hewitt K and Burton I (1971) The Hazardousness of a Place: A Regional Ecology of Damaging
Events, University of Toronto Press, Toronto
Khazai B and Bendimerad F. (2011) Risk and Resiliency Indicators, EMI Topical Report TR-1 1-03
Khazai B, Kilic O, Basmaci A, Konukcu B, Sungay B, Zeidan A, Wenzel F (2008) Megacity
Indicators System for Disaster Risk Management - Implementation in Istanbul, in M Erdik,
10
Zschau J. Megacity Istanbul Project Reports. Municipality Disaster Management Center
(AKOM), Istanbul
Khazai B, Merz M, Schulz C, Borst D (2013). “An Integrated Indicator Framework for Spatial
Assessment of Industrial and Social Vulnerability to Indirect Disaster Losses,” Natural
Hazards, 67: 145-167
Kumpulainen S (2006) Vulnerability Concepts in Hazard and Risk Assessment, in Schmidt-Thomé P,
Natural and Technological Hazards and Risks Affecting the Spatial Development of European
Regions, Geological Survey of Finland, Helsinki
Mendes, JM (2009) “Social Vulnerability Indexes as Planning Tools: Beyond the Preparedness
Paradigm,” Journal of Risk Research, 12: 43-58
Montz BE (2000) “The Hazardousness of a Place: Risk from Multiple Natural Hazards. Papers and
Proceedings of the Applied Geography Conferences, 23: 331-339
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,” Seismological Research Letters, 85(3): 1-13
Schneiderbauer S and Ehrlich D (2006) Social Levels and Hazard (In) dependence in Determining
Vulnerability, in Birkmann J, Measuring Vulnerability to Natural Hazards: Towards Disaster
Resilient Cities, United Nations Press, Tokyo
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,” Natural
Hazards, DOI 10.1007/s11069-013-0618-x
Silva V, Crowley H, Pinho R, Varum H (2014a) “Development of a Vulnerability Model for the
Portuguese RC Building Stock,” Bulletin of Earthquake Engineering, under review
Silva V, Crowley H, Pinho R, Varum H (2014b) “Seismic Risk Assessment for Mainland Portugal,”
Bulletin of Earthquake Engineering, under review
Tate, EC (2012) “Social Vulnerability Indices: A Comparative Assessment using Uncertainty and
Sensitivity Analysis,” Natural Hazards, 63: 325-347
Vilanova S and Fonseca J (2007) “Probabilistic Seismic-hazard Assessment for Portugal,” Bulletin of
the Seismological Society of America, 97: 1702-1717
Vilanova S, Fonseca J, Oliveira C (2012) “Ground-motion for Seismic Hazard Assessment in Western
Iberia,” Bulletin of the Seismological Society of America, 102: 169-184
Wood NJ, Burton CG, Cutter SL (1991) “Community Variations in Social Vulnerability to Cascadia-
related Tsunamis in the U.S. Pacific Northwest,” Natural Hazards, 52: 369-389
