Vitor Silva - Seismic Vulnerability Course Notes

Seismic Vulnerability Assessment

Vitor Silva

May 2015
Seismic Vulnerability Assessment Course

Lecturer: Vitor Silva

Description:

Earthquake loss estimation can play a fundamental role in the sustainable development of a
given region, providing local governments and other decision makers with valuable
information necessary for the creation of risk mitigation actions. An important component
for this purpose is a vulnerability model that allows the estimation of losses from
structural/non-‐structural damage due to earthquakes, as a function of a set of ground motion
parameters. Structural vulnerability can thus be defined as the likelihood of a certain
element to suffer loss due to the effects of an earthquake. The influence of the vulnerability
of the exposed elements to seismic events is fundamental in the magnitude of losses. A
simple comparison between earthquakes that occurred in developing regions and developed
countries reveals the critical importance of vulnerability. For example, the Spitak (Armenia)
earthquake of December 1988 had a magnitude of Ms 6.7 and left a death toll of about 25000
casualties. Less than a year after, an earthquake with a greater magnitude (Ms 7.0) occurred
in Loma Prieta (California, USA) causing a number of human losses smaller than 70. This
way, the structural vulnerability assumes special importance in the estimation of seismic
risk, not only for reflecting directly the damage susceptibility of a structure, but also because
by intervening with appropriate strengthening solutions, it may be possible to significantly
reduce the vulnerability, and consequently the potential for human losses.

This course comprises a brief introduction to seismic hazard, risk and exposure modelling,
within the scope of vulnerability analyses. These notes leverage upon existing studies from a
number of sources that are cited and acknowledged throughout, and its structure and
contents are strongly based on the Seismic Risk and Vulnerability course from the
Understanding and Managing Extremes School at Pavia, Italy.

Course Schedule:
7th of May

-‐ Basic concepts of seismic hazard, risk and exposure modelling.
-‐ The importance of seismic vulnerability assessment
-‐ Overview of empirical, analytical and expert opinion methodologies
-‐ Questionnaire

8th of May
-‐ EMS-‐98 approach (Lagomarsino and Giovinazzi 2006)
-‐ Nonlinear static procedures in fragility assessment
o SPO2IDA (Vamvatsikos and Cornell 2006)
o Capacity Spectrum Method (HAZUS -‐ FEMA 2004)
o Displacement-‐based methodologies (Silva et al. 2013)
-‐ Nonlinear time history analysis (MDOF and SDOF)
-‐ Derivation of vulnerability functions
-‐ Exercises

11th/12th of May
-‐ Presentation of the Global Earthquake Model
-‐ Seismic Vulnerability Assessment in South America (SARA Project)
-‐ Examples of seismic vulnerability and risk assessment
-‐ Questionnaire
CHAPTER 1 INTRODUCTION ................................................................................................. 1
1.1. THE NEED FOR EARTHQUAKE LOSS MODELLING ............................................................ 1
1.2. THE IMPORTANCE OF SEISMIC VULNERABILITY .............................................................. 4
CHAPTER 2 BASIC CONCEPTS OF SEISMIC HAZARD ..................................................... 6

2.1. GROUND SHAKING INTENSITY MEASURES ....................................................................... 8
2.1.1. OBSERVATION-‐BASED INTENSITY ............................................................................................... 9
2.1.2. INSTRUMENTAL INTENSITY ........................................................................................................ 11
2.1.2.1. Peak Ground Acceleration ............................................................................................... 11
2.1.2.1. Arias Intensity ....................................................................................................................... 11
2.1.2.2. Significant Durations .......................................................................................................... 12
2.1.2.3. Pseudo-‐spectral acceleration ......................................................................................... 12
2.2. GROUND MOTION AND INTENSITY PREDICTION EQUATIONS ...................................... 13
2.3. SHAKEMAPS .................................................................................................................... 15
2.4. GROUND MOTION FIELDS .............................................................................................. 17
2.5. PROBABILISTIC SEISMIC HAZARD ASSESSMENT ........................................................... 18
CHAPTER 3 BASIC CONCEPTS OF EXPOSURE MODELLING ...................................... 21

3.1. GLOBAL EXPOSURE MODELLING ..................................................................................... 21
3.2. REGIONAL EXPOSURE MODELLING ................................................................................. 24
3.3. THE IMPORTANCE OF BUILDING TAXONOMIES ............................................................. 25
3.4. GEM BASIC BUILDING TAXONOMY ............................................................................... 26
CHAPTER 4 SEISMIC VULNERABILITY ASSESSMENT ................................................. 28

4.1. MAIN OUTPUTS IN VULNERABILITY ASSESSMENT ......................................................... 28
4.2. EXPERT OPINION VULNERABILITY ................................................................................ 30
4.3. EMPIRICAL VULNERABILITY .......................................................................................... 32
4.4. FRAGILITY ASSESSMENT ................................................................................................ 34
4.4.1. DAMAGE STATES .......................................................................................................................... 34
4.4.2. FRAGILITY FUNCTIONS AND DAMAGE PROBABILITY MATRICES ......................................... 35
4.4.3. EXPERT OPINION FRAGILITY ..................................................................................................... 36
4.4.4. EMPIRICAL FRAGILITY ................................................................................................................. 41
4.4.5. ANALYTICAL FRAGILITY .............................................................................................................. 44
4.4.6. NONLINEAR STATIC ANALYSIS ................................................................................................... 46
4.4.6.1. SPO2IDA (Vamvatsikos and Cornel, 2006) ............................................................... 50
4.4.6.2. The Capacity Spectrum Method (HAZUS -‐ FEMA, 2005) .................................... 52
4.4.6.1. DBELA (Silva et al. 2013) ................................................................................................. 56
4.4.7. NONLINEAR DYNAMIC ANALYSIS ............................................................................................... 62
4.5. CONSEQUENCE FUNCTIONS ............................................................................................ 64
CHAPTER 5 BASIC CONCEPTS OF SEISMIC RISK .......................................................... 68

5.1. SCENARIO RISK ASSESSMENT ........................................................................................ 68
5.2. SCENARIO DAMAGE ASSESSMENT .................................................................................. 69
5.3. PSHA AND EVENT-‐BASED RISK ASSESSMENT .............................................................. 70
5.4. RETROFITTING DECISION SUPPORT TOOL ................................................................... 72

Chapter 1
Introduction
1.1. The Need for Earthquake Loss Modelling
In 2011 the world celebrated the birth of the 7 billionth citizen. It has been
estimated that in 10 years the world population will reach 8.1 billions, and 9.4
billions in the year of 2050 (PRB, 2010). This uncontrolled growth of the population
has led to an increase of megacities (with a population greater than 2 million), often
located in areas prone to natural disasters, such as earthquakes. This peril has been
responsible for a death toll of over 60 thousand people per year in the last decades
and economic losses that can reach a great fraction of a country’s welfare. In the last
50 years in Central America, the earthquakes of Guatemala (1976), Nicaragua
(1972), and El Salvador (1986) caused economic losses of approximately 98%, 82%
and 40% of the nominal gross domestic product (GDP) of each country, respectively
(Daniell et al., 2010). In the Haiti earthquake of 2010, the economic losses were
above the nominal GDP (120%) and more than 300 thousand people are believed to
have perished. In addition to these direct consequences in the vicinity of the seismic
event, business disruption of multi-‐national enterprises can induce a negative
impact at a global scale.
The global economic losses and insured losses due to great natural disasters from
1980 to 2011 are illustrated in Figure 1.1.
Introduction

Figure 1.1. Overall losses and insured losses from 1980 until 2011 (MunichRe, 2012).
Earthquakes constitute on average 20% of the overall losses, but in some years,
this portion can be as high as 60% (e.g. 2010, 2011). Despite the great advances that
have been made in the last decades in the areas of probabilistic seismic hazard
assessment (e.g. Abrahamson 2006; Bommer and Abrahamson 2006), evaluation of
building seismic vulnerability (e.g. Calvi et al. 2006) and collection of information
regarding the elements exposed to the hazards (e.g. Gamba et al., 2012), an increase
in the trend of earthquake losses is still observed. The Global Assessment Report
(UNISDR, 2009) points out two probable causes for this tendency. Firstly, the
significant increase of population and capital stock in hazard prone areas that have a
direct impact in the associated catastrophe risk. Figure 1.2 illustrates the amount of
population and GDP exposed to hazard in the 20 countries with the highest values.

Figure 1.2. Absolute population and GDP exposed to natural disasters (adapted from UNISDR, 2009).
2
Chapter 1
Secondly, besides this growth in the property and property value, modern
societies strongly rely on inter-‐related systems (building stock, power, finance,
transport), which in case of a natural catastrophe might initiate a cascade effect,
where a disaster triggers another disaster. Modelling this system of systems might
be very challenging due to the required understanding of each component, and lack
of numerical tools to carry out such calculations.
The Great East Japan earthquake (2011) and tsunami sent a clear message that
both developed and developing countries are exposed to high risks. The year of
2011 was the most expensive year ever registered, far exceeding the 2005 economic
losses (hurricane Katrina), which yield the previous record. From the overall cost of
380 billion US$, the earthquake disasters in Japan and New Zeeland alone accounted
for absolute losses of 228 billion US$. Such losses can have a crippling effect in the
economy of countries whose governments have the legal liability to cover the full
costs of rebuilding. In the Kocaeli and Düzce earthquakes of 1999 (with a combined
fraction of GDP loss of approximately 8%), the Turkish government was faced with
an enormous financial burden due to its statuary obligation in covering the costs of
reconstruction. This situation propelled the creation of the Turkish Catastrophe
Insurance Pool (TCIP), which allowed transferring large parts of the financial
burden due to seismic losses to the world’s reinsurance market (Bommer et al.
2002). In this project, the creation of an earthquake loss model was fundamental to
the development of the economic model used to evaluate the impact of catastrophe
risk in the Turkish economy. Furthermore, earthquake loss modelling also serves as
the foundation to many other seismic risk mitigation actions. These may include
prioritization of zones within a country where the structural seismic vulnerability of
the building stock should be improved, planning of post-‐disaster emergency
response or definition of regulations to impose seismic-‐proof construction practices.
However, in less developed countries, the required resources, datasets and tools
might not exist in order to perform a comprehensive assessment of the seismic risk.
In fact, the evaluation of the annual costs of natural disasters from the last decades
between low/middle-‐income countries and high-‐income countries shows
considerably lower costs in the latter category, where catastrophe models are more
frequently available. This trend can be seen in Figure 1.3.
3
Introduction

Figure 1.3. Distribution of average annual cost of natural disaster in low-‐, middle-‐ and high-‐income
countries (adapted from Cummins and Mahul, 2009).
1.2. The importance of Seismic Vulnerability

Structural vulnerability can be defined as the likelihood of a certain element to
suffer loss due to the effects of an earthquake. The influence of the vulnerability of
the exposed elements to seismic events is fundamental in the magnitude of losses. A
simple comparison between earthquakes that occurred in developing regions and
developed countries reveals the critical importance of vulnerability. For example,
the Spitak (Armenia) earthquake of December 1988 had a magnitude of Ms 6.7 and
left a death toll of about 25000 casualties. Less than a year after, an earthquake with
a greater magnitude (Ms 7.0) occurred in Loma Prieta (California, USA) causing a
number of human losses smaller than 70 (Bommer, 2004). For this reason, the
structural vulnerability assumes special importance in the estimation of seismic
risk, not only for reflecting directly the damage susceptibility of a structure, but also
because by intervening with appropriate strengthening solutions, it may be possible
to significantly reduce the vulnerability, and consequently the seismic risk.
The recognition of the importance in understanding the vulnerability of the
exposed elements led to a rapid rise in demand for accurate and flexible
methodologies for its evaluation (Calvi et al. 2006). These may include empirical
methods that take advantage of post-‐earthquake damage data to derive fragility
functions (e.g. Colombi et al., 2008, Rota et al. 2008); simplified approaches where a
set of structural parameters are employed to derive a vulnerability index, which is
then used to calculate a curve relating levels of damage with a set of intensity
measure levels (Lagomarsino and Giovinazzi, 2006); or analytical methodologies
4
Chapter 1
that rely on numerical models to simulate the seismic performance of structures

against increasing levels of ground motion (e.g. Vamvatsikos and Cornell 2002;
Freeman 2004; Rossetto and Elnashai 2005). The latter category of methods has the
advantage of not depending on the availability of post-‐earthquake damage data, and
depending on the level of complexity of the numerical models, these approaches can
still incorporate fundamental structural characteristics in the analysis such as
vertical or plan irregularities or influence of higher modes of vibration (Chopra and
Goel 2002; Casarotti and Pinho 2007). Each analytical method considers different
simplifications, assumptions and algorithms, mainly in the way the structural
nonlinearity is handled. Consequently, the structural response will be dependent on
the chosen methodology, and discrepancies in the order of 2 can be observed
(Chopra and Goel, 2000; Lin et al, 2004). For these reasons, it is fundamental to fully
comprehend the limitations and strengths of each vulnerability methodology, and
ensure that the selected approach is capable of endorsing the accuracy and
reliability required by the end users.

5
Chapter 2
Basic concepts of seismic hazard
Seismic hazard can be described as the physical phenomenon induced by an
earthquake that may cause loss of life or injury, property damage, social and
economic disruption or environmental degradation. The physical phenomena
caused by earthquakes are summarized in Figure 2.1, and include tsunamis, surface
rupture, landslides, liquefaction and amplified ground shaking.
SURFACE FAULTING TSUNAMIS

RUPTURE
SEISMIC
ENERGY
RELEASE
LANDSLIDES LIQUEFACTION
TOPOGRAPHIC STRONG
SOIL DEPOSITS
RELIEF GROUND
MOTION
AMPLIFIED AMPLIFIED
GROUND GROUND
SHAKING SHAKING

Figure 2.1. The various natural phenomenon caused by the release of seismic energy.
The various phenomena caused by earthquakes will produce damage in different

manners in the surrounding building inventory, and it is fundamental to fully
comprehend their physical behavior in order to properly assess the seismic
Chapter 3
vulnerability of the built environment. The following figures demonstrate the effects
of these different phenomena.

Figure 2.2. Fault rupture.

Figure 2.3. Tsunamis.

Figure 2.4. Landslides.

Figure 2.5. Liquefaction.
7
The different natural phenomenon induced in earthquakes from 1989 to 2003 and
the associated damage was studied by Bird and Bommer (2004). The main objective
of the later study was to identify the main and secondary causes of damage to
buildings, transport systems (e.g. roads and railway lines) and utilities (e.g.
pipelines). Figure 2.6 shows the results considering only building damage.
Tsunami Shaking
Landslide
Landslide
None
Liquefaction
Tsunami
Shaking Fault rupture
Figure 2.6. The primary cause (left) and secondary cause of damage to buildings from 50 destructive
earthquakes from 1989-‐2003 (Bird and Bommer, 2004).
Bird and Bommer (2004) concluded that there were very few cases where ground
shaking did not dominate on a regional scale, but in smaller survey areas secondary
hazards were seen to also contribute significantly (e.g. in the Turkish town of
Adapazari, extensive liquefaction was observed durig the 1999 Kocaeli earthquake).
The majority of the existing vulnerability models thus focus on ground shaking.
2.1. Ground Shaking Intensity Measures

An earthquake can be defined as a sudden release of elastic strain energy in the
Earth’s crust as the result of a fault rupture (source). Energy radiates from the fault
rupture and from the surrounding volume of the crust in the form of elastic or
seismic waves (path). Strong ground shaking occurs when the seismic waves pass
through soil layers (which cause refraction and amplification of the seismic waves)
and reach the surface of the earth (site). The intensity of ground shaking from an
earthquake at a given site can be measured using a number of parameters. These
can be categorized into two types: intensity that is assigned based on observations
(also known as macroseismic intensity); and intensity that is measured from
instruments, such as accelerographs.
8
Chapter 3
2.1.1. Observation-‐based Intensity

Bolt (1999) defines intensity as “a measure of ground shaking obtained from
damage done to structures built by humans, changes in the Earth’s surface, and felt
reports”. Intensity is an index that reflects the strength of ground shaking at a
particular location during an earthquake. Therefore, it is not really a measure of the
size of the earthquake in the same way as moment or magnitude, but rather a
measure of ground motion. In order to make clear the index nature of intensity it
was historically represented by Roman numerals, but modern practice has moved
towards the use of Arabic numerals to ease their use in risk algorithms and
numerical methodologies.
Intensity scales were initially proposed by Rossi and Forel in 1883. The historical
evolution of intensity scales is described by Coburn and Spence (2002). Many of the
earlier scales have now been completely abandoned and the main scales are those of
the Modified Mercalli (MM) scale, used in the Americas, and the Medvedev-‐Karnik-‐
Sponheuer (MSK) scale used in most of Europe, except Italy where the Mercalli-‐
Cancani-‐Sieberg scale (MCS) is employed. The MSK scale has now generally been
replaced by the European Macroseismic Scale (Grünthal, 1998) (EMS-‐98). The MM,
MSK and EMS scales, as well as others, base the grades of intensity on (a) the
shaking felt by humans, (b) the movement of objects, (c) the damage to buildings,
and (d) changes in the ground. This last category is not a reliable indicator because
of the very high heterogeneity of soils and rocks and the strong influence of ground
water conditions and other parameters. It is generally advised not to use
observations of ground behavior other than for corroboration with the MM and MSK
scales. In fact, the EMS-‐98 removed the geotechnical and geological observations
from the descriptions of different grades of intensity.
The lower levels of intensity are defined primarily by how people feel the shaking,
but as seismic intensity becomes more severe, human perception of the movement
becomes progressively less important, and damage to buildings assumes a dominant
role. Intensity III is generally considered as the threshold of perceptibility, below
which the ground shaking is not felt by most people. Intensity VII can be thought of
as the threshold of appreciable building damage, although this might be intensity
VIII for engineered structures. Intensity XII is very rarely, if ever, encountered in
9
reality and therefore XI can be treated as the upper bound. In practice, X appears to
be an effective upper bound.
One final point must be made: the scales are neither continuous nor linear. The
variation from degree to degree is not gradual, each increase in one degree
representing a jump in the level of shaking. Furthermore, the jumps between
different degrees are not equal: the increase in the level of ground shaking from IV
to V is not the same as the increase from VII to VIII.
One of the most successful and well-‐established efforts to calculate and disseminate
observational ground shaking intensity shortly after the occurrence of an
earthquake is the Did You Feel it? initiative. This service has been supported by the
United States Geological Survey (USGS) to produce Community Internet Intensity
Maps (CCIM), which summarize the questionnaire responses provided by internet
users. “Communities” are defined as zip code, or postal code, regions. The form of
the questionnaire and the method for assignment of intensities are based on an
algorithm developed by Dengler and Dewey (1998) for determining a ”Community
Decimal Intensity". Figure 2.7 shows the CIIM for an earthquake in Northern Italy.

Figure 2.7. Community Internet Intensity Map for the M5.2 earthquake in Northern Italy on 27th
January 2012
10
Chapter 3
2.1.2. Instrumental Intensity

The instruments that record the acceleration of the ground as a function of time
strong ground motion are called accelerographs (unlike seismographs which record
the displacement or velocity of the ground). The first accelerographs were
developed and installed in California in 1932, more than three decades after the first
seismographs came into operation. Accelerograms, the records obtained from
accelerographs, contain a wealth of information about the nature of the ground
shaking in strong earthquakes and also about the highly varied characteristics that
different earthquakes can produce at different locations.
2.1.2.1. Peak Ground Acceleration

As shown in Figure 2.8, peak ground acceleration (PGA) relates only to one isolated
peak within a record of ground motion. Therefore, it is a rather poor parameter for
characterizing the entire ground motion record. The numerical integration of the
accelerogram allows the velocity and displacement time-‐histories to be calculated as
well, from which peak ground velocity (PGV) and peak ground displacement (PGD)
can be obtained.
Figure 2.8. Representation of the peak ground acceleration (PGA) in a ground motion record.
2.1.2.1. Arias Intensity

Arias Intensity (AI) is another measure of the energy of the ground motion record
and it is defined as the time-‐integral of the square of the ground acceleration:
!!
𝜋
𝐼! = 𝑎(𝑡)! 𝑑𝑡
2𝑔 !
11
where g is acceleration due to gravity and Td is the duration of the accelerogram

above a given acceleration threshold.
Figure 2.9. Representation of cumulative Arias Intensity versus time.
2.1.2.2. Significant Durations

The significant duration is the time over which significant ground shaking occurs.
The Arias Intensity can be cumulatively calculated with time (as shown in the figure
above), and significant durations can be defined which describe the length of time
over which the Cumulative Arias Intensity is between 5% and 75% of IA, or between
5% and 95% of IA.
2.1.2.3. Pseudo-‐spectral acceleration

If a number of single degree of freedom (SDOF) systems (i.e. equivalent to a number
masses fixed on top of several sticks with distinct heights – see Figure 2.10) are
subjected to a given acceleration time-‐history at their base, they will respond
differently. If the peak response of each SDOF system (acceleration, velocity,
displacement) is registered, and plotted against its natural period of vibration, a
response spectrum is produced.
Figure 2.10. Representation of a number of single degree of freedom (SDOF) systems.
12
Chapter 3

Figure 2.11. Acceleration response spectra for three elastic damping levels.
2.2. Ground Motion and Intensity Prediction

Equations
Given a large number of records, one can calculate values for any of the parameters
described in the previous section and obtain a robust estimate of the correlation of
these values with any other parameter relevant to this suite of records, such as the
magnitude of the earthquake from which they came. This type of reasoning is the
basis for the development of empirical predictive equations for strong ground
motions. Usually, a relationship is sought between a suite of observed ground
motion parameters and an associated set of independent variables including a
measure of the size of the earthquake, a measure of distance from the source to the
site, some classification of the style-‐of-‐faulting involved and some description of the
geological and geotechnical conditions at the recording site. An empirical ground
motion prediction equation (GMPE) is simply a function of these independent
variables that provides an estimate of the expected value of the ground motion
parameter in consideration, as well as some measure of the distribution of values
about this expected value.
Thus far the development of empirical ground motion prediction equations has been
almost exclusively focused upon the prediction of peak ground-‐motions, particularly
PGA and, to a far lesser extent, PGV, and ordinates of 5% damped elastic acceleration
response spectra (Douglas, 2003). Regardless of the ground motion measure in
consideration, a GMPE can be represented as a generic function of predictor
variables, μ(M,R,θ) and a variance term, εσT, as in the following equation:
log 𝑦 = 𝜇 𝑀, 𝑅, 𝜃 + 𝜀𝜎!
13
For most ground motion measures, the values will increase with increasing
magnitude and decrease with increasing distance. These two scaling effects form the
backbone of prediction equations and many functional forms have been proposed to
capture the variation of motions with respect to these two predictors (Douglas,
2003). Different magnitude measures and different distance terms, such as
hypocentral distance and Joyner-‐Boore distance can be used in GMPEs, though the
most commonly used in recent studies are Moment Magnitude (MW) and Joyner-‐
Boore distance (rJB). For any particular ground motion record, the total variance
term given in the equation above may be partitioned into two components as shown
below:
log 𝑦!" = 𝜇 𝑚!" , 𝑟!" , 𝜃!" + 𝛿!" + 𝛿!,!!
The terms δe,i and δa,ij represent the inter-‐event and intra-‐event residuals,
respectively, and quantify how far away from the mean estimate of logyij the motions
from the ith event and the jth recording from the ith event are, respectively, as
illustrated in Figure 2.12.

Figure 2.12. The predicted median ground motion parameter varying with distance (bold line), the
actual recordings on which they are based for two different events (circle and square), and the
representation of inter-‐ and intra-‐event residuals (Bommer and Stafford 2008)
Equations to estimate the macroseismic intensity for a region given magnitude and
distance are also available, often being derived in regions where macroseismic data
is much more abundant than instrumental intensity (due to a lack of recording
14
Chapter 3
stations). These so-‐called Intensity Prediction Equations (IPEs) often have the same
format as that presented previously for instrumental ground motion parameters. A
review of various IPEs is available in Cua et al. (2010).

Figure 2.13. Variation in intensity predictions from various IPEs (Cua et al. 2010)
2.3. ShakeMaps
As defined by Wald et al. (1999), a shake map is a representation of ground shaking
produced by an earthquake. In order to generate rapid-‐response ground motion
maps it is necessary to determine the best format for reliable presentation of the
maps given the diverse audience, such as scientists, businesses, emergency response
agencies, media, and the general public. As discussed in Wald et al. (1999), the use of
intensity maps instead of peak ground acceleration and velocity maps could help to
simplify and maximize the flow of information to the public.
ShakeMap is a product of the United States Geological Survey (USGS) Earthquake

Hazards Program in union with regional seismic network operators. This project
makes available near-‐real-‐time maps of ground motion and shaking intensity
following significant earthquakes. The methodology applied for generating these
maps can be summarized as follows:
1. Gather the observed peak ground motions at recording stations in the region
of the earthquake (these are shown as triangles on the maps).
15
2. Determine the centroid (magnitude and location) of the event and assess
median ground motions on rock at sites far from recording stations (using
GMPEs).
3. Correct the recorded data to rock based on the site QTM (Quaternary Tertiary
Mesozoic) geology and interpolate the data plus the predicted ground motions
onto a fine rock grid.
4. Site amplify at each fine grid point based on its site QTM geology and
amplitude, fit a smooth function through all fine grid points, and contour.
It is important to note that most of the areas that could be affected by a large
earthquake do not have sufficient instrumentation for recording the ground motion
parameters associated with the seismic event. Hence, there are significant gaps in
the observed shaking distribution which makes it necessary to develop algorithms
to best describe the shaking in more remote areas by utilizing a variety of
seismological tools.
As described by Wald et al. (1999), to create the best composite map it is necessary
to combine information from individual stations, geological properties of the site
(representing site amplification), and ground motion attenuation for the distance to
the epicenter of the causative fault. The idea is to produce realistic estimates at grid
points located far from available data while preserving the detailed shaking
information existing for regions where there are stations nearby. Figure 2.14
illustrate an example ShakeMap for the magnitude 7.3 earthquake in Southern Peru.

Figure 2.14. ShakeMap for the September 25, 2013 magnitude 7.3 in Southern Peru.
16
Chapter 3
2.4. Ground Motion Fields

Whereas ShakeMaps provide a rapid, illustrative description of ground shaking after
an event has occurred, risk modelling requires the realistic prediction of ground
shaking for future possible events. The maps of ground shaking for a future scenario
earthquake (of which a set are produced to model the uncertainty in the ground
motion) are termed “ground motion fields” herein.
A value of a ground motion parameter (e.g. PGA) is calculated for a set of sites using
a GMPE or IPE for a given rupture model (with given geometry and magnitude). The
median ground motion can be calculated at each location, leading to a median
ground motion field as depicted in Figure 2.15.
−78˚ −76˚ −74˚ −72˚ −70˚
−6˚ −6˚
−8˚ −8˚
−10˚ −10˚
−12˚ −12˚
−14˚ −14˚
−16˚ −16˚
−18˚ −18˚
−78˚ −76˚ −74˚ −72˚ −70˚
PGA
0.0 0.1 0.2 0.3 0.4

Figure 2.15. Example of a median ground motion field, considering the 2007 Pisco earthquake in Peru.
However, as discussed previously, there is a large variability in the level of shaking
at a given site when this is predicted with an empirical equation. In order to account
for this variability, a number of ground motion fields should be generated, so that at
a given site the full range of possible levels of ground motion are modelled. Hence,
once the median (or logarthimic mean) ground motion at each site is estimated, an
estimate of the number of logarithmic standard deviations above or below the mean
(the residual) needs to be made by randomly drawing from a normal distribution.
The proximity of the sites can be accounted to consider that sites close to each other
will have similar residuals (known as spatial correlation). Examples of ground
motion fields with and without spatial correlation are illustrated in Figure 2.16b.
There are a number of spatial correlation models, but the most recent is that of
17
Jayaram and Baker (2009), which is based on spectral acceleration. Spatial

correlation models for macroseismic intensity have not been published in the past
but are the focus of recent research. A discussion of the modelling of uncertainty in
ground motion for the purposes of loss modelling is presented in Crowley et al.
(2008).
−78˚ −76˚ −74˚ −72˚ −70˚
−6˚ −6˚
−8˚ −8˚
−10˚ −10˚
−12˚ −12˚
−14˚ −14˚
−16˚ −16˚
−18˚ −18˚
−78˚ −76˚ −74˚ −72˚ −70˚
PGA
0.0 0.1 0.2 0.3 0.4

Figure 2.16. Ground motion field without spatial correlation (left) and with spatial correlation (right),
for the 2007 Pisco earthquake in Peru.
2.5. Probabilistic Seismic Hazard Assessment

From a probabilistic seismic hazard assessment (PSHA) it is possible to derive
hazard curves and maps. A hazard curve presents, at a given location, the
probabilities of exceedance within a given time period of a range of levels of ground
shaking. A hazard map presents the expected level of ground shaking for a given
probability of exceedance within a certain time interval throughout the region of
interest. Figure 2.17 depicts a hazard map for a probability of exceedance of 10% in
50 years for South America.
18
Chapter 3

Figure 2.17. Seismic hazard map in terms of peak ground acceleration with a 10% probability of
exceedance in 50 years for South America.
The four main steps to a PSHA are shown in Figure 2.18. The first step involves the
definition of seismic sources that are capable of generating future earthquakes. It is
generally assumed that any location within a source zone has an equal likelihood of
generating an event (i.e. the seismicity is uniform). Tectonics, observed seismicity
and a degree of judgement are all called upon in the determination of source zones.
Once the source zones have been defined, the temporal element of seismicity is
provided through recurrence relationships. Within each source zone, a recurrence
relationship can be derived through the extraction of the number of earthquakes of
different magnitude from the earthquake catalogue. A recurrence relationship
specifies the average rate at which an earthquake of some magnitude will be
exceeded, the first being proposed by Gutenberg and Richter (1994):
log 𝑁! = 𝑎 − 𝑏𝑀
where Nm is the mean annual rate of exceedance of the magnitude M, 10! is the
mean yearly rate of earthquakes of magnitude greater than or equal to zero, and b
describes the relative likelihood of large and small earthquakes. The a and b values
can be obtained through regression analysis of an earthquake catalogue, which is
pre-‐processed to remove fore-‐ and aftershocks. The third step involves the
definition of the ground motions at the site, using a GMPE as discussed previously.
The fourth step involves the calculation of a hazard curve at a given site by
integrating the exceedances of ground motions caused by earthquakes of different
sizes (magnitudes) occurring at varying locations in each source zone. The actual
details of the hazard calculations are not provided herein as this goes beyond the
19
scope of this course. Further details can be found in Reiter (1990) and McGuire
(2004).

Figure 2.18. The four-‐step process to a typical PSHA (adapted from Reiter, 1990) leading to the
calculation of a seismic hazard curve (bottom right)
Other intermediate products that can be obtained from a PSHA include stochastic
event sets and associated ground motion fields. Given a seismic source model, one
can generate a number of seismicity histories (i.e. randomly generated earthquake
catalogue), each one representing a possible realisation of the seismicity originating
from a seismic source model during a given time span. The ensemble of these
seismicity histories is called a Stochastic Event Set. Figure 2.19 shows the events
within a stochastic event set for Italy made up of two different seismicity histories of
50 years each.

Figure 2.19. Two different 50 year seismicity histories that make up a stochastic event set.
20
Chapter 3
Basic concepts of exposure
modelling
Exposure data as defined herein refers to the elements that are exposed to seismic
hazard, such as people, buildings and infrastructure.
3.1. Global exposure modelling

There are global datasets of gridded population with resolution of up to 30 arc
seconds (approximately 1km square at the ecuator), some of which are proprietary
(e.g. LandScanTM) and others that are open (e.g. Gridded Population of the World,
version 3, see Figure 3.1, and GRUMP v1). GRUMP (Global Urban and Rural Mapping
Project) also identifies whether a given grid cell is urban or rural. A review of global
datasets for population has been carried out by Salvatore et al. (2005), but some key
information on three main datasets is provided below.

Figure 3.1. Spatial distribution of population count according to LandScan 2012.
Basic concepts of seismic risk
LandScanTM uses a large number of data sources such as maps of major roads and
rail networks, drainage networks, utility systems, elevation contours, coastlines,
international boundaries and populated places, land cover data, high resolution
aerial photography and satellite imagery, and the US Bureau of Census data to
produce a 30 arc second grid of population. It is noted that the population is an
ambient population, as opposed to a residential population, and thus takes into
account the movements of individuals through a given area (or cell).
GPW version 3 is a 2.5 arc minutes (approximately 5km square at the equator)
spatially disaggregated population layer based on national statistical office
estimates of population, where the population is distributed uniformly across the
grid within a given administrative unit for which the census data is available.
Population data estimates are provided for 1990, 1995, and 2000, and projected to
2005, 2010, and 2015.
GRUMP v1 built upon GPW to construct a geo-‐referenced framework of urban and

rural areas by combining census data with satellite data. However, GRUMPv1
actually comprises three main data products: 1) a higher resolution gridded
population data product at 30 arc-‐seconds, or ~1km at the equator, for 1990, 1995,
and 2000; 2) urban extents dataset delineating urban areas based on NOAA’s night-‐
time lights data set and buffered settlement centroids (where night lights are not
sufficiently bright); 3) a points dataset of all urban areas with populations of greater
than 1,000 persons. These two datasets report residential as opposed to ambient
population.
The only currently available dataset related to building inventory on a global scale is
the USGS PAGER database v1.4 (Jaiswal and Wald, 2008) and the Global Exposure
Database of the Global Earthquake Model. Regarding the former dataset, a building
taxonomy (describing typologies of buildings and their characteristics) was
developed and datasets with residential housing building stock distributions
available for individual countries (from a variety of sources such as UN-‐HABITAT,
housing census, published literature, expert opinion) were mapped to these
construction types. Where no data was available for a given country, the distribution
of housing stock from selected neighbouring countries was used. Figure 3.2
represents the various sources of data used for each country in the World used by
PAGER.
22
Chapter 3
Figure 3.2. PAGER global building inventory data sources
The distributions of occupants in building types were divided between urban, rural,
residential and non-‐residential. Figure 3.3 provides an extract of the database for
Chile. Algorithms to further distribute these occupants during the day, night and
transit times were also developed. These distributions can be used in conjunction
with the aforementioned global population databases to estimate the occupants
within different building typologies at different times during the day.

Figure 3.3. Building distribution per type of area/use for Chile.
To improve on the aforementioned PAGER database, the Global Earthquake Model

has funded the development of the Global Exposure Database, that combines many
23
sources of data to produce a gridded distribution of buildings (counts, area,

replacement cost) and their occupants, distributed according to their structural
characteristics (following the GEM building taxonomy, as described in the following
section). Huyck et al. (2011) provide a detailed description of additional datasets
that are available around the world (at global and regional scales) to support the
modelling of global exposure data, such as administrative areas (boundaries of
countries, regions, provinces, municipalities etc.), population datasets, building data
(from census data and surveys), land use and land cover datasets.
3.2. Regional exposure modelling

The development of an exposure model capable of providing information about the
location, value and vulnerability classification of the exposed elements at a national
or regional scale might be a very challenging task, often only achievable through the
use of Census data (e.g. Erdik et al. 2003 (Turkey); Silva et al. 2014a (Portugal);
Crowley et al. 2008 (Italy)). The previously described approach can also be
employed, but the reliability of the final exposure model might not be sufficient for
regional or local risk analysis. These datasets may contain information about the
number of dwellings and/or building, material of the walls/roof/floor, number of
stories, type of dwelling and date of construction. With this information, it is
possible to associate each asset with a vulnerability class, and calculate a spatial
distribution of these categorized assets throughout the region of interest. Figure 3.4
illustrates a regional model that has been developed by the Global Earthquake
Model (SARA project) for the Andean countries following this approach.

Figure 3.4. Exposure modelling for the Andean countries.
24
Chapter 3
3.3. The importance of Building Taxonomies

A taxonomy is a classification scheme, and a number of schemes relevant to the
classification of buildings for seismic risk assessment have been proposed in the
past. These taxonomies have in general been developed to describe and classify
building structures in terms of seismic resistance and response. They have all been
developed since 1985 and can be expected to have improved upon earlier similar
taxonomies. The primary requirements of the proposed structural taxonomy are
listed below.
The criteria by which the existing taxonomies are assessed are listed below:
Distinguishes differences in seismic performance. The taxonomy distinguishes
earthquake-‐resistant versions of structural systems from non-‐earthquake resistant
version, including the " before" and "after" states of common seismic retrofits and
between ductile and non-‐ductile systems.
Observable. Two people examining the same structural system in the field or using
data obtained from the field should independently assign the same taxonomic group
based solely on the text definition of the taxonomic group.
Complete. The taxonomy must include all engineering features relevant to the global
seismic performance of a building structure. As mentioned above it is recognized
that there will be a need for additional taxonomies to capture all aspects of the
seismic performance and losses for an entire building, including building
dimensions and nonstructural components. The structural taxonomy must contain
sufficient attributes to meet the needs of the end users of GEM.
Simple and collapsible. The taxonomy should have as few groups as possible, while
still meeting the other requirements. It is also desirable to define common
combinations and relative quantities of structural systems so that fragility or
vulnerability functions can be created by aggregating the fragilities or vulnerabilities
of detailed components, while still distinguishing say, differences in ductility, design
or retrofit alternatives. A taxonomy is judged to be collapsible if taxonomic groups
can be combined and the resulting combinations still distinguish differences in
seismic performance.
Nearly exhaustive. Within practical limits almost every structural system can be
sensibly assigned to a taxonomic group.
25
Familiar to engineering practitioners and architects. It is desirable that engineers
and architects be familiar with the taxonomic system, particularly to readily and
accurately identify structural attributes. If the new taxonomic system corresponds
readily to an existing taxonomic system, it can give users access to existing data.
Engineers and architects should be familiar with the nomenclature to be defined to
avoid ambiguity.
Currently treats non-‐buildings. Built forms other than buildings need to be

included in the taxonomy sometime in the future. These include constructions like
dams, bridges and tunnels.
Extensible to other hazards. In the future parts of the GEM model may be used by
other groups working on other natural hazards such as floods, hurricanes and
volcanic eruptions.
User-‐friendly. The taxonomy should be straightforward, intuitive, and as easy to use
as possible by both those collecting data, those arranging for its analysis and the end
users.
International in scope. As far as possible the taxonomy should be appropriate for

any region of the world. It should not privilege any one region but be technically and
culturally acceptable to all regions.
3.4. GEM Basic Building Taxonomy

The GEM Basic Building Taxonomy relies on 8 key attributes, as reported in the table
below. See Brzev et al. 2013 for the full report on the taxonomy, and all of the
reference tables.
Table 3.1 GEM Basic Building Taxonomy summary table

Attribute Attribute levels Example
Material of the Lateral Material type (Level 1)

Load-‐Resisting Material technology (Level 2) Steel
System Material properties (Level 3)
Type of lateral load-‐resisting system (Level 1)
Lateral Load-‐Resisting
Braced frame
System Ductility (Level 2)

Roof material (Level 1) Masonry

Roof
Roof type (Level 2)

Floor material (Level 1)
Floor Concrete
Floor type (Level 2)
26

Chapter 3
Height Number of stories above ground 4

Date of Construction Construction completed (year) 1925
Type of irregularity (Level 1)
Irregularity Re-‐entrant corner
Irregularity description (Level 2)
Building occupancy class -‐ top level (Level 1)
Occupancy Residential
Building occupancy class -‐ detail (Level 2)

27
Chapter 4
Seismic vulnerability assessment
4.1. Main outputs in vulnerability assessment
Physical vulnerability is defined herein as the probability distribution of loss (or loss
ratio) due to structural or non-‐structural damage, given an intensity measure level
(see Figure 4.1). Loss might refer to repair costs or loss of life; loss ratio might refer
to the ratio of cost of repair to cost of replacement, or the ratio of number of
fatalities to the number of occupants. These vulnerability functions can be derived
directly, usually through empirical methods where the losses from past events at
given locations are related to the levels of intensity of ground motion at those
locations, or they can be derived indirectly by combining fragility functions and
consequence functions, as described in the following sections.
0.7
0.6
0.5
Loss ratio (%)
0.4
0.3
0.2
0.1
0
0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1 1.2
Spectral acceleration at T=0.30s (g)
Figure 4.1. Discrete vulnerability function
Chapter 4
Fragility functions describe the probability of exceeding a set of limit states, given an
intensity measure level; limit states describe the limits to performance levels, such
as damage or injury levels. Fragility functions can be derived by expert-‐opinion,
empirically (using observed data), or analytically, by explicitly modelling the
behaviour of a given asset typology when subjected to increasing levels of ground
motion. The relation between probability and ground shaking can be modelled by a
cumulative lognormal function (Figure 4.2), or in a discrete manner, in which each
intensity measure level is simply associated with a probability of exceedance per
damage state (Figure 4.3).

Figure 4.2. Continuous fragility functions
1
0.9
0.8
Probability of exceedance
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
5 6 7 8 9 10 11
MMI
Negligible Moderate Substantial Very Heavy Destruction

Figure 4.3. Discrete fragility functions
Consequence functions are used to derive the probability distribution of loss, given a
performance level, and are generally derived empirically or through expert opinion.
Figure 4.4 illustrates a consequence model for buildings in Greece.
29
1"
0.8"
Damage"ra>o"
0.6"
0.4"
0.2"
0"
Slight"
Moderate"
Collapse"
Heavy"
Very"heavy"

Figure 4.4. Consequence model showing mean ratios of cost of repair to cost of replacement, and bars
illustrating the uncertainty in these ratios (adapted from Kappos et al. 2006).
4.2. Expert Opinion Vulnerability

Expert opinion vulnerability assessment is needed in some situations, for example
due to the following reasons (Jaiswal, personal communication):
1. Observed damage and loss data exist only at places where large earthquakes
have occurred in the past;
2. Structural modelling and analysis techniques for estimating damage and loss
are not always suitable to model structural collapse behaviour of a large
number of structure types in different parts of the world;
3. It is not practical to perform hundreds of experiments/shake-‐table tests on
real-‐life structures as they are built today;
4. Structural design and construction practice may differ substantially over

time and space, and it is practically difficult to fully incorporate such changes
appropriately within available analytical models;
Vulnerability functions were produced using expert opinion as part of the ATC-‐13
project (ATC, 1985). More than 50 senior earthquake engineering experts were
asked to provide low, best and high estimates of the loss ratios (the ratio of repair
cost to replacement cost, expressed as a percentage, which they termed “damage
factor”) for Modified Mercalli Intensities (MMI) from VI to XII for 40 different
building classes. The low and high loss ratio estimates provided by the experts were
defined as the 90% probability bounds of a lognormal distribution, whilst the best
estimate was taken as the median loss ratio (Figure 4.5). Weighted means of the
experts’ estimates, based on the experience and confidence levels of the experts for
30
Chapter 4
each building class, were included in the averaging process, as described in

Appendix G of ATC-‐13 (ATC, 1985).
5% 5%
ML MB MH
4.00 10.60 18.20
Figure 4.5 Example of a lognormal distribution of the estimated loss ratio for a given macroseismic
intensity showing the mean low (ML), mean best (MB) and mean high (MH) estimates of the damage
factor (adapted from McCormack and Rad, 1997).
For each intensity measure level in the ATC-‐13 vulnerability model, a mean loss
ratio (so-‐called damage factor) and corresponding coefficient of variation is defined.
A lognormal distribution of the loss ratio for each intensity measure level is
assumed. Figure 4.6 presents the ATC-‐13 vulnerability functions (mean loss ratio
only) for moment-‐resisting concrete frame buildings.
The Global Earthquake Model funded an initiative that was carried out by the
physical vulnerability consortium to build upon the ATC-‐13 project to develop
expert opinion fragility/vulnerability assessment for a number of building types for
which empirical or analytical functions are not readily available. A new methodology
for weighting the experts’ opinions (Cooke’s method) was implemented. More
information on the results of this project are available on GEM Nexus.

Figure 4.6 ATC-‐13 vulnerability functions (mean loss ratio versus intensity measure level, IML) for
reinforced concrete moment-‐resisting frame buildings
31
It might be argued that expert opinion is close to empirical fragility/vulnerability

assessment as the experts are likely to be basing their opinion on their experience in
observing damage and loss in the field.
4.3. Empirical Vulnerability

Empirical vulnerability uses observations of loss and relates these to the observed
ground shaking to produce vulnerability functions.
An empirical global vulnerability model capable of predicting human losses

(fatalities) has been developed by Jaiswal et al. (2008), to rapidly and approximately
ascertain the impact in terms of fatalities immediately following a large earthquake
anywhere in the world. In order to develop this model, earthquake mortality rates
(attributed to damage from ground shaking) for more than 4,800 worldwide
earthquakes that occurred since 1973 were studied and a set of vulnerability
functions for specific countries or regions were built. This earthquake fatality rate
can be defined as the ratio of the total number of shaking-‐related fatalities to the
total population exposed at a given shaking intensity, using Modified Mercalli as the
intensity measure type.
The PAGER collection of vulnerability functions that return a fatality rate for a given
intensity measure level, can be expressed in terms of a two-‐parameter lognormal
cumulative distribution. For countries where at least 4 fatal seismic events have
occurred (earthquakes where casualties occurred) a specific vulnerability function
was computed. However, only a few countries have experienced >4 fatal
earthquakes since 1973 and therefore, it was necessary to aggregate some countries
in regions in order to have enough fatal earthquakes for this catalogue period. For
this reason, a new global regionalization scheme was proposed based on the
likelihood of suffering similar losses in future earthquakes. This spatial organization
was done based primarily on geography, building inventory and social-‐economic
similarities. Figure 4.7 presents the regionalization proposed by the aforementioned
methodology, where countries with the same colour all use the same empirical
vulnerability function. According to the proposed regionalization, the vulnerability
model is composed of 27 functions for specific countries and 23 functions for groups
of countries. These expressions, which give fatality rate (ν) (i.e. ratio of number of
32
Chapter 4
fatalities to exposed population) as a function of shaking intensity (S), are expressed
as a lognormal distribution function as follows:
1 𝐼
𝜈 𝑆 =𝜙 𝑙𝑛
𝛽 𝜃
where φ is the standard normal cumulative distribution function, I is the discrete

value of the intensity measure level and β and θ are the parameters of the
distribution. This vulnerability model is bounded between intensity V to IX (where
populations exposed to shaking intensity IX and above are collated at shaking
intensity IX). The two parameters of the distribution were computed in a way that
the residual error would be minimized for larger and smaller earthquakes
simultaneously.

Figure 4.7 Regionalisation scheme for PAGER empirical casualty vulnerability model.
Jaiswal and Wald (2011) extended the aforementioned methodology to develop

empirical vulnerability functions in terms of economic loss. In order to estimate
economic losses, an assessment of the economic exposure at various levels of
shaking intensity was required. The economic value of all the physical assets
exposed at different locations in a given area is generally not known and extremely
difficult to compile at a global scale. In the absence of such a dataset, the total Gross
Domestic Product (GDP) exposed at each shaking intensity was estimated by
multiplying the per-‐capita GDP of the country by the total population exposed at that
33
shaking intensity level. The total GDP estimated at each intensity was then scaled by
an exposure correction factor, which is a multiplying factor calculated for each
country and which aims to account for the disparity between wealth and/or
economic assets to the annual GDP. The economic exposure obtained using this
procedure is thus a proxy estimate for the economic value of the actual inventory
that is exposed to the earthquake. The economic loss ratio, defined in terms of a
country-‐specific lognormal cumulative distribution function of shaking intensity, has
been derived and calibrated against the losses from past earthquakes. Economic loss
data available for global earthquakes from 1980 to 2007 from the MunichRe
NatCatSERVICE databasehas been used for this purpose. The same regionalization
scheme as proposed for the fatality model (Figure 4.7) is used with the economic
vulnerability model.
4.4. Fragility Assessment
4.4.1. Damage States

The two previous sections dealt with the direct estimation of vulnerability (i.e.
distribution of loss ratio given intensity), whereas as described in the introduction,
vulnerability can be estimated through damage, by first estimating the distribution
of damage (using fragility functions) and then relating that damage distribution to
loss (using consequence functions).
There are many structural damage scales that consist of discrete damage state
descriptions that are meant to represent different levels of building performance,
from no damage to collapse. Damage scales can be used to assign damage to
buildings for post-‐earthquake damage and safety assessment, or for the definition of
performance of buildings in analytical vulnerability assessment. Existing damage
scales thus vary greatly in the number of damage scales adopted (seen to vary
between 3 and 12 in the literature), details of the damage state descriptions, damage
criteria used and the amount and type of information needed for the damage
description. An example damage scale for Reinforced Concrete Moment Resisting
Frames (so-‐called C1) according to HAZUS (FEMA, 2003) is provided in Table 4.1.
34
Chapter 4
Table 4.1 Damage scale from HAZUS (FEMA, 2003) for Reinforced Concrete Moment Resisting Frames
(C1)
Damage State Damage Description
Flexural or shear type hairline cracks in some beams and
Slight Structural Damage
columns near joints or within joints.
Most beams and columns exhibit hairline cracks. In ductile
frames some of the frame elements have reached yield
Moderate Structural Damage capacity indicated by larger flexural cracks and some
concrete spalling. Non-‐ductile frames may exhibit larger
shear cracks and spalling.
Some of the frame elements have reached their ultimate
capacity indicated in ductile frames by large flexural
cracks, spalled concrete and buckled main reinforcement;
Extensive Structural Damage non-‐ductile frame elements may have suffered shear
failures or bond failures at reinforcement splices, or
broken ties or buckled main reinforcement in columns
which may result in partial collapse.
Structure is collapsed or in imminent danger of collapse
due to brittle failure of non-‐ductile frame elements or loss
of frame stability. Approximately 13%(low-‐rise),
Complete Structural Damage
10%(mid-‐rise) or 5%(high-‐rise) of the total area of C1
buildings with Complete damage is expected to be
collapsed.

Hill and Rossetto (2008a) provide a detailed review and comparison of damage
scales for different building types (Figure 4.8).

Figure 4.8 Comparison of damage scales for reinforced concrete with masonry infill (adapted from Hill
and Rossetto, 2008a)
4.4.2. Fragility Functions and Damage Probability Matrices

Fragility functions can be discrete or continuous. A continuous fragility function was
shown previously in Figure 4.2. When macroseismic intensity is the intensity
measure, the functions are generally discrete (due to the integer scale of
macroseismic intensity), such as the example shown in Figure 4.9.
35

Figure 4.9. Discrete fragility functions; the distribution of damage at MMI = 8.5 is indicated.
This fragility function can also be represented in terms of a damage probability

matrix, where for a given intensity measure level (e.g. MMI = 8) the percentage of
buildings within each damage state is reported (obtained by probability of attaining
or exceeding that damage state minus the probability of attaining or exceedance the
next damage state).
Table 4.2 Damage probability matrix (after Whitman et al. 1973)
Non-‐ Intensity of Earthquake

Damage Structural
Structural
State Damage
Damage V VI VII VIII IX
0 None None 10.4 -‐ -‐ -‐ -‐

1 None Minor 16.4 0.5 -‐ -‐ -‐
2 None Localised 40.0 22.5 -‐ -‐ -‐
3 Not noticeable Widespread 20.0 30.0 2.7 -‐ -‐
4 Minor Substantial 13.2 47.1 92.3 58.8 14.7
5 Substantial Extensive -‐ 0.2 5.0 41.2 83.0
6 Major Nearly total -‐ -‐ -‐ -‐ 2.3
7 Building condemned -‐ -‐ -‐ -‐ -‐
8 Collapse -‐ -‐ -‐ -‐ -‐

4.4.3. Expert Opinion Fragility

A method for fragility assessment that can be placed in the expert opinion category
is that proposed by Lagomarsino and Giovinazzi (2006), which is based on the EMS-‐
98 macroseismic scale. The outputs of this method include damage probability
matrices, and mean damage factor functions, which can then be combined with
consequence functions to produce vulnerability functions.
36
Chapter 4
According to the European Macroseismic Scale (Grunthal, 1998), intensity can be

evaluated either by direct field observations or by questionnaires sent to the
affected areas, although the latter is really only suitable for lower intensity values.
The “modal” intensity is assigned to a given area based on the observations for the
following three categories: (a) descriptions of people’s reactions, (b) the movement
of objects and (c) the proportions of each building category expected to be within
each damage state.
Building damage begins at intensity grade V, whereas the degrees of intensity VII,
VIII and IX on the EMS scale describe the most important degrees of shaking in many
ways, from an engineering point, since they are frequently encountered and
potentially damaging. Intensity is sometimes thought of as an indication of building
damage, but this is really a misperception. Different types of building are treated
differently and levels of damage are carefully defined precisely to deconvolute the
ground motion from the response of buildings. Figure 4.10 shows how EMS-‐98
classifies the vulnerability class of different types of building. Figure 4.11 illustrates
the different grades of damage for a given building typology. Buildings are assigned
to a vulnerability class, and the damage in buildings within a given vulnerability
class is observed and recorded (few, many or most assigned to a given damage
grade), and then damage probability matrices are used to assign a macroseismic
intensity for the area.
37

Figure 4.10. Relationship between building types and vulnerability classes.

Figure 4.11. Damage grade descriptions for a given building type
38
Chapter 4
Table 4.3 Damage Probability Matrix (DPM) for vulnerability class A.
DAMAGE GRADE
CLASS A
0 1 2 3 4 5
V Few
VI Many Few
INTENSITY
VII Many Few
DEGREE VIII Many Few
IX Many Few
X Many
XI Most
XII All

As described previously, the EMS-‐98 scale defines qualitative descriptions of “Few”,
“Many” and “Most” for five damage grades for levels of intensity ranging from V to
XII for six different classes of decreasing vulnerability (from A to F). Damage
matrices containing a qualitative description of the proportion of buildings that
belong to each damage grade for various levels of intensity were presented
previously in Table 4.3 for vulnerability class A.
The problems related to the vagueness of the matrices (i.e. they are described
qualitatively) and the incompleteness of the matrices (i.e. the lack of information for
all damage grades for a given level of intensity) has been tackled by Giovinazzi and
Lagomarsino (2004; 2006) by applying Fuzzy Set Theory and by assuming a beta
damage distribution, respectively. In classical set theory, the membership of
elements in a set is assessed in binary terms -‐ an element either belongs or does not
belong to the set. By contrast, fuzzy set theory permits the gradual assessment of the
membership of elements in a set; this is described with the aid of a membership
function valued in the real unit interval [0, 1], where a value of 0 means the
parameter does not belong to the set, a value of 1 indicates the degree of belonging
is almost sure whilst anything between 0 and 1 means rare but possible.
Membership functions have been applied to the percentage ranges of few, many and
most described in EMS-‐98, as shown in Figure 4.12.
39

Figure 4.12 Membership functions of the percentage ranges used for the quantitative terms few, many
and most in EMS-‐98 (Giovinazzi and Lagomarsino, 2004; 2006).
Possible and plausible upper and lower bounds are assigned to each linguistic term
(e.g. for many: the possible lower bound is 10, the plausible lower bound is 20, the
possible upper bound is 50, the plausible upper bound is 60). The discrete beta
distribution is assumed for the damage distribution for a given intensity level and
the parameters of the distribution (where the mean relates the mean damage grade
varying between 0 and 5), which leads to the matching of the values of the plausible
and possible upper and lower values of the linguistic terms. This leads to four curves
(mean damage grade versus EMS-‐98 intensity) for each vulnerability class, as shown
in Figure 4.13.

Figure 4.13 Plausibility and possibility curves for vulnerability class B and C
To treat the belonging of a given building class to a vulnerability class, which is also
treated in linguistic terms in EMS-‐98 (see Figure 4.10), Fuzzy Set Theory was also
used. Each vulnerability class is related to a vulnerability index, which depends on
the building typology, the characteristics of the building stock (e.g. number of floors,
40
Chapter 4
irregularity etc.) and the regional construction practices. The vulnerability index
membership function is show in Figure 4.14. The plausible and possible upper and
lower vulnerability index bounds of each vulnerability class are then related to the
four functions of mean damage grade shown in Figure 4.13, and an analytical
function is fit to the curves as a function of macroseismic intensity and Vulnerability
Index as follows:
𝐼 + 6.25𝑉! − 13.1
𝜇! = 2.5 1 + 𝑡𝑎𝑛ℎ
2.3

Figure 4.14 Vulnerability index membership functions for EMS-‐98 vulnerability classes (Giovinazzi and
Lagomarsino, 2006)
4.4.4. Empirical Fragility

Empirical fragility assessment requires the observation of damage to different
building typologies under different levels of ground shaking. The method relies on
the collection of damage data for different building types following an earthquake.
This is often done for a variety of reasons: rapid structure evaluations are carried
out to ascertain the safety of the buildings in the immediate aftermath of an
earthquake; detailed damage evaluations are carried out in certain regions, for a
subset of the total number of buildings, perhaps following a request from the
building owner; detailed post-‐earthquake surveys are undertaken by
reconnaissance teams on small samples of the buildings. The ideal source of data
should include all of the buildings in a given area (including the non-‐damaged
buildings), each building should be georeferenced, and the data should be collected
by structural engineers in areas close to ground motion recording stations. Such
data is, however, hard to come by and often assumptions are needed (e.g. to
41
estimate the number of undamaged buildings, to estimate the ground shaking levels
to which the buildings were subjected).
The choice of the damage scale in empirical fragility functions is often limited to the
damage scale used in the post-‐earthquake damage survey, though when a number of
different surveys need to be combined (perhaps from different events in different
countries, or from surveys using different data collection forms), then
harmonisation to a common damage scale is necessary and studies such as the one
of Hill and Rossetto (2008a) can be particularly useful.
The choice of the building taxonomy can also be limited by the one used on the post-‐
earthquake damage survey, though as for damage scales, harmonisation to a
common taxonomy might be necessary when different surveys are combined in the
generation of empirical fragility functions.
The geographical unit for which damage statistics are developed should be the
smallest area possible (with uniform soil conditions) that will provide a statistically
representative sample. When the data is not available on a building-‐by-‐building
basis, this choice of geographical unit is imposed by the resolution of the data, and
might refer to an administrative area. The minimum number of observations per
dataset and the minimum number of data points and their distribution with respect
to the intensity measure levels is an open issue that has been investigated by GEM’s
Global Vulnerability Consortium (Rossetto et al. 2015). Fragility functions have been
developed in the past with as few as 3 data points.
The levels of ground motion to which the damaged buildings were subjected is
usually the most difficult data to obtain, due to a lack of observations/recordings at
the locations where the buildings were damaged. Recent post-‐earthquake damage
surveys are concentrating their reconnaissance efforts close to ground motion
recordings, but this does not completely remove the uncertainty in the ground
motion (as local site conditions can vary the ground shaking even over distances of
100s of metres) and a lot of the existing data occurs in areas for which recorded
ground motion is not available. In this case, there are two choices that can be made:
1) the damage can be related to the observed macroseismic intensity estimates, 2)
GMPEs can be used to estimate the ground motion. In the former case, the problem
becomes circular as the same damage data is used to define the intensity measure
level and the damage distribution; an alternative could be to use Internet-‐based
42
Chapter 4
intensity such as Did You Feel it?, or ShakeMaps. When GMPEs are used, the choice of
the intensity measure (e.g. PGA, PGV, SA) and the choice of the predictive equation
will influence the results, and the uncertainty and correlation in the ground shaking
should be included in the estimation of the ground motion, though due to its
complexity this is often ignored. Considering the large uncertainty in the conversion
between observed and instrumental intensity measure, it is important that the
intensity measure type selected for the derivation of the fragility functions is one for
which prediction equations are available, such that the functions can be used in the
forward modelling of damage and loss.
Once the building damage data has been prepared, and the ground motion levels
have been estimated for each geographical unit in the dataset, regression analysis is
needed to fit a curve to the data (Figure 4.15). A number of regression methods (e.g.
linear least squares, nonlinear least squares) can be applied, and the best practice in
the case of instrumental intensity is to trial different intensity measure types (from
different GMPEs) and different regression methods to find which combination
provides the best fit to the data.

Figure 4.15 Empirical fragility functions for RC MRF (Orsini, 1999)
There are thus a number of drawbacks to empirical fragility assessment:
o When macroseismic intensity is used for the intensity measure type, there is
circularity in the method;
o To cover a wide range of intensity measure levels and building typologies, it is
often necessary to combine damage surveys which might rely on different
damage scales and different building taxonomies;
43
o The data is often clustered at the low damage/ground motion area, which can
influence the regression analysis;
o Only building types that have experienced damage can be modelled (and thus
these methods are not suitable for evaluating new building typologies, or
retrofitted buildings, as discussed further in Chapter 6);
o The level of ground shaking often needs to be predicted (thus the functions are
not purely “empirical” or “observational”), and the uncertainty and correlation
in the latter accounted for;
o To use intensity measure types that correlate better with structural response
(such as spectral acceleration at the fundamental period) an estimation of the
period of vibration of the buildings is needed, which introduces further
uncertainty.
Nevertheless, when the data is adequately processed and uncertainties are

acknowledged (even if not explicitly modelled), then empirical fragility assessment
has the advantage of providing a realistic estimate of the expected damage.
The following publications are recent attempts to model empirical fragility

functions: Rossetto and Elnashai (2003), Liel and Lynch (2009), Rota et al. (2008),
Sarabandi et al. (2004); Colombi et al. (2008), Goretti and Di Pasquale (2004), Di
Pasquale et al. (2005), Karababa and Pomonis (2010), Amiri et al. (2007).
4.4.5. Analytical Fragility

The fragility of classes of buildings can be modelled through numerical models,
especially in the case of reinforced concrete buildings where extensive research and
development in this field has taken place over the past 30 to 40 years.
Numerical models attempt to represent the nonlinear behaviour of buildings under
different loading conditions. There are two main camps of nonlinear structural
analysis: plastic hinge modelling (“concentrated inelasticity”) and fibre element
modelling (“distributed inelasticity”). In the former method, the inelastic behaviour
is concentrated in springs at the ends of the elements of the building (the beams and
columns), which is where the inelastic response is normally concentrated, and the
hysteretic behaviour of these springs has to be calibrated before the analysis. In the
latter method, the constitutive behaviour of each material type is explicitly modelled
44
Chapter 4
(through appropriate stress-‐strain relationships) and the cross-‐section of the

element is divided into a number of fibres and the individual inelastic behaviour of
each fibre is integrated across the section. The member is divided into a number of
elements, thus allowing the reproduction of plastic hinges and the spread of
inelasticity along the member length (see Figure 4.16).
G aus s node B
S ec tion b
G aus s
S ec tion a
node A
σ
B
ε
L /2 3 L /2
ε
R C S ection Unc onfined C onfined S teel F ibres
C oncrete F ibres C oncrete F ibres

Figure 4.16 Discretisation in fibre modelling of a typical reinforced concrete member
A larger number of a priori assumptions are required with plastic hinge modelling,
and thus for fibre element modelling is recommended for less experienced
modellers.
For fragility assessment of classes of buildings there are a number of issues that
need to be addressed when creating a numerical model:
o Should the building typology be modelled in 3D or simplified to a 2D model? In

general 2D analyses are adopted, though in this case torsional behaviour will be
ignored and thus the typology should be regular in plan (see Figure 4.17).
o How can the variability in the geometrical and material properties amongst the
buildings found within a given typology be modelled? Once the typical range of
values for these properties is obtained, Monte Carlo simulation can be used
such that a number of random buildings can be generated, each with a given
beam length, beam depth, column length, column depth, wall thickness, infill
panel thickness, concrete compressive strength, yield strength (and so on)
obtained from the relevant distribution.
45
o How many fibres should be employed across the section, how many elements
should the members be divided into? This usually requires sensitivity analysis to
understand the influence that these modelling choices have on the response of
the structure.
CR F
R CM
Torsional Vibration

Figure 4.17 Unfavourable plan layouts inducing torsional response in buildings
Once the geometry, material properties, and loading conditions have been assigned
in the numerical model a choice needs to be made on the type of nonlinear analysis
to be undertaken. There are two main types of analysis of interest for fragility
assessment: nonlinear static and nonlinear dynamic.
4.4.6. Nonlinear static analysis

Nonlinear static analysis involves the introduction of forces or displacements at each
floor of the building or frame to laterally displace the structure, following which the
internal forces and deformations are calculated and balanced. The aim of a nonlinear
static analysis is to provide an envelope to the outputs of nonlinear dynamic
analysis under increasing levels of intensity (Figure 4.18). The output of such
analysis is a pushover curve, which relates the shear force at the base of the building
(i.e. the force that is counteracting the lateral actions being applied to the building)
to the displacement of a given node of the building, generally taken as the roof.
46
Chapter 4
120
Max Ba se She ar (kN)

100
80
60
40
20
0
0 0.5 1 1.5 2 2.5 3
drift (%)
7000
base shear (kN)
6000
5000
4000
3000 dynamic
DAP
uniform
2000
triangular
1000
0
0% 1% 2% 3%
total drift
Figure 4.18 Maximum base shear versus top displacement obtained by running many dynamic analyses
within increasing intensity (top), pushover curves with different lateral load assumptions compared
with the dynamic output (bottom).
Nonlinear static analyses are less time consuming, and more numerically stable than
dynamic analysis. However, as shown in Figure 4.18, the results can vary as a
function of the assumptions made on the application of the lateral actions. However,
the definition of global damage states in terms of displacement limits is easier with
pushover curves. There are methods to identify the yield displacement (passing
through 60-‐70% of the ultimate base shear), as well as the ultimate displacement
(when the base shear drops to 80% of the maximum value after the peak). These
possible thresholds are presented in Figure 4.19. However, a check on the response
of the members at a given global limit state to damage should be undertaken (for
example to check that the assumed ultimate displacement really does correspond to
the formation of plastic hinges in the majority of the columns in a single storey, as
shown in Figure 4.20b).
47
150
100
Base shear (kN)

50
Capacity curve
1st limit state
2nd limit state
3rd limit state
0
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14
Displacement at the top (m)
Figure 4.19 Identification of global damage states from a pushover curve
Δp Δp
θp θp

Figure 4.20 Collapse mechanism (a) beam-‐sway, (b) column-‐sway, (c) mixed mechanism
Simplified procedures have emerged in recent years to replace the need to

undertake complicated nonlinear analyses of 2D frames, which attempt to represent
pushover curves through a number of simple formulae based on structural
mechanics principles, and calibrated using nonlinear analyses (Calvi, 1999; Glaister
and Pinho, 2003; Crowley et al. 2004; Bal et al. 2010). These equations include the
period of vibration as a function of the height (which can be used to obtain the initial
stiffness), and the displacements at different performance levels as a function of the
geometrical and material properties and the mechanism that is assumed to form in
the building (Figure 4.20). The collapse mechanisms shown in Figure 4.20a and c are
less brittle than that shown in Figure 4.20b and occur at higher levels of lateral
deformation. Older buildings or informal construction, that has not been adequately
designed to withstand seismic actions, is more likely to fail with a column-‐sway,
brittle mechanism. A methodology for fragility derivation based on this approach
(Silva et al. 2013) is presented in the following section.
48
Chapter 4
Once pushover curves have been obtained (either through complex nonlinear
modelling or simplified equations), the response of the building of frame to
increasing levels of seismic intensity needs to be estimated. There are a number of
methods to estimate the nonlinear response, often termed Nonlinear Static
Procedures (NSPs), which the main ones being N2 (Fajfar and Fischinger, 1988;
Fajfar and Gaspersic, 1996), Capacity Spectrum Method, with the modifications of
FEMA 440 (ATC, 2005), Displacement Coefficient Method, with the modifications of
FEMA 440 (ATC, 2005). The key steps of all of these methods include:
1. The bilinearisation of the pushover curve;
2. The representation of the structure as a single-‐degree-‐of-‐freedom system (as

opposed to a multi-‐degree-‐of-‐freedom structure);
3. The calculation of the nonlinear response of the SDOF system using an

elastic response spectrum.
Once the response of the structure to a given accelerogram (represented in terms of

a response spectrum) is found with one of the aforementioned methods, the damage
state of the structure can be found. Fragility functions can be generated as follows:
1. Random generation of a population of 2D frames through Monte Carlo

simulation;
2. Pushover curve for each frame, and representation as the curve for a SDOF
system;
3. Estimate nonlinear response for each frame, using a large selection of

records and a given Nonlinear Static Procedure;
4. Global damage state identification based on the nonlinear response;
5. For each intensity measure level, plot the cumulative percentage of buildings
in each global damage state;
6. Regression analysis to calculate the parameters (mean and standard

deviation) of the fragility functions (can usually be assumed to be a
lognormal distribution);
The results of fragility functions obtained with nonlinear static and nonlinear
dynamic methods are likely to be different, mainly due to the uncertainty in the NSP,
which is not accounted for in the method described above. In order to include the
49
latter uncertainty in the analysis, a number of nonlinear dynamic analyses are

required in order to estimate the uncertainty in the dynamic response, given the
estimated static response (see Figure 4.21). This uncertainty (logarithmic standard
deviation) can then be combined with the standard deviation calculated in point 6
above.

Figure 4.21 Comparison of nonlinear static response estimated with the FEMA 440 method and with
nonlinear dynamic analysis for 7 reinforced concrete frames
For this course, three nonlinear static procedures will be described in depth. These
methods vary considerably in terms of manner in which the uncertainties and
ground shaking are handles, and thus each one of its own limitations and
advantages.
4.4.6.1. SPO2IDA (Vamvatsikos and Cornel, 2006)

The Static PushOver to Incremental Dynamic Analysis (SPO2IDA) is a simplified
methodology, which combines the results of incremental dynamic analysis (IDA)
with pushover analysis. The procedure was originally developed by Vamvatsikos
and Cornell (2006) and has been recommended in ATC-‐58 (FEMA P-‐58, 2012).
SPO2IDA uses empirical relationships from a large database of incremental dynamic
analysis results to convert static pushover curves into probability distributions for
building collapse as function of ground shaking intensity. The process of SPO2IDA
uses only one ground motion record for a given site, scaled up or down to reflect the
full range of demands for different return periods, of acceleration or displacement
50
Chapter 4
amplitudes. An Excel workbook application tool of SPO2IDA is available for

download from the ATC-‐58 project website1.
It should be noted that this procedure is best suited to low-‐rise buildings of regular
configuration. The reliability of median values derived from this procedure is
dependent upon the quality of analyses performed and the number of intensities
evaluated. The calculation of median capacity in terms of spectral acceleration,
accounting for record-‐to-‐record variability, can be conducted with the SPO2IDA tool
(Excel Workbook).
SPO2IDA is applicable to any form of capacity curve, either bilinear or multilinear

elasto-‐plastic. Given the idealized capacity curve, the SPO2IDA tool uses an implicit
R-‐µ-‐T relation to correlate nonlinear displacement expressed in terms of ductility
(𝜇!"! = ∆!"! /∆! , where ∆! = ∆!"! ), to the corresponding medians capacity in terms
of spectral acceleration, Ŝ a ,dsi (T1 ), as shown in Figure 4.22. These medians capacity
are defined as the values at which a given damage threshold has been attained or
overcome by 50% of the analysis conducted with the ground motion pairs. The
record-‐to-‐record dispersion 𝛽! is estimated by considering the number of records
that have been used in the SPO2IDA spreadsheet (𝛽! is estimated based on a large
empirical database of incremental dynamic analysis results).
Figure 4.22 Sample SPO2IDA results of normalized spectral acceleration at different damage state
thresholds
The steps required to calculate the medians capacity using the SPO2IDA procedure
are described below:
1
https://www.atcouncil.org/Projects/atc-58-project.html
51
1. Develop an appropriate mathematical model of the building for nonlinear static

analysis. Identify primary and secondary elements or components; define your
nonstructural elements; foundation flexibility; gravity loads; and P-‐Delta effects.
2. Run a static pushover analysis; and make sure that the analysis is extended to a
deformation such that collapse is judged to occur. Construct a force-‐
displacement relationship considering different damage states thresholds
(capacity curve).
3. Provide additional input information: building weight, building height, and

fundamental period.
4. Fit the force-‐displacement relationship into an appropriate idealized

representation by supplying the following reference control points: 1-‐ the yield
point; 2-‐ point of peak strength; 3-‐ the onset of residual strength response; 4-‐
the ultimate deformation at collapse.
5. Convert the fitted force-‐displacement relationship to normalized spectral

acceleration and displacement/ductility parameters to allow selection and
calculation of SPO2IDA input.
6. Execute SPO2IDA and extract the median capacity associated to each damage
threshold. These medians capacity are defined as the values corresponding to
the attainment or overcome of the threshold value by 50% of the analysis
conducted with ground motion pairs (see Figure 4.24).
A fragility model developed using the methodology was developed by in Chaulagain
et al. 2015 for reinforced concrete structures, as depicted in Figure 4.23.

1 .0
0 .8
P roba bility of exc eedenc e
0 .6

0 .4
m ode ra te
e x te ns iv e
0 .2
da m a g e
0 .0
0 .0 0 .2 0 .4 0 .6 0 .8 1 .0 1 .2 1 .4 1 .6 1 .8 2 .0 2 .2 2 .4 2 .6
S pe c tra l a c c e le ra tion (g ) T = 0 .2 9 (s e c )

Figure 4.23 Fragility model for reinforced concrete buildings in Nepal (Chaulagain et al. 2015).
4.4.6.2. The Capacity Spectrum Method (HAZUS -‐ FEMA, 2005)
52
Chapter 4
The capacity spectrum method (CSM) was initially proposed by Freeman et al.
(1975), and it represents a simplified methodology for many purposes such as the
evaluation of a large inventory of buildings, assessment of new or existing structures
or to identify the correlation between damage states and levels of ground motion.
ATC-‐40 (1996) proposes three different procedures (A, B and C) for the application
of the Capacity Spectrum Method. However, procedure B adopts some
simplifications that might not always be valid and procedure C has a very strong
graphical component, making it difficult for systematic applications. Hence,
procedure A, which is characterized by its intuitiveness and simplicity, is adopted
herein. This procedure iteratively compares the capacity and the demands of a
structure, using a capacity curve (for the simplified SDOF) and a damped response
spectrum, respectively. The ground motion spectrum is computed for a level of
equivalent viscous damping that is estimated as a function of the displacement at
which the response spectrum crosses the capacity curve, in order to take into
account the inelastic behaviour of the structure. Iterations are needed until there is
a match between the equivalent viscous damping of the structure and the damping
applied to the spectrum. The final intersection of these two curves approximates the
displacement response of the structure.
The initial proposal of this method was heavily criticized due to its tendency to
underestimate the deformation of the structures. Thus, in FEMA-‐440 (2005), some
modifications were proposed mainly regarding the calculation of the equivalent
viscous damping. Another aspect worth further investigation in this method is how
well the bilinear curves represent the yielding point for different response spectra,
as this value is crucial for the calculation of the equivalent viscous damping.
ATC-‐40 defines the slope of the first segment of the bilinear curve based on the
initial stiffness, which in order to respect the equal energy dissipated rule (i.e. the
area under the capacity curve needs to be equal to the area under the bilinear
curve), a yielding point located in the elastic portion of the capacity curve might be
obtained. This misrepresentation of the yielding occurs mostly when combining this
method with response spectra with low levels of ground motion. An alternative for
this procedure has been proposed in FEMA-‐273 (1997), in which the slope of the
first segment of the bilinear is computed based on the effective stiffness, allowing a
more realistic shape of the bilinear curves. This aspect is shown in Figure 4.24 and
Figure 4.25, where the same capacity curve was used against a weak (PGA=0.74
53
m/s2) and a strong (PGA=1.42 m/s2) accelerograms, respectively, for each

procedure. 0.16
Sd =0.0226 m Sd =0.0393 m
0.14 y y
Say=0.0625 g Say=0.1028 g
0.16
Spectral acceleration (g)

0.16
0.16
1.6 0.16
0.16
1.6 0.12
Sd =0.0226 m Sd =0.0393 m Sd =0.0353 m
0.14 y 2 y y 2
0.14
0.14 Say=0.6252
=0.0625 m/s
g 0.14Sa =0.1028 g
0.14
1.4 0.1 Sa =0.9271
=0.0927 m/s
g
1.4 y y y
(m/s2)
0.12
(m/s2)
Sdy=0.0353
=0.0373 m
(g)
(g)
m Sdy=0.0373 m

0.12
0.12
1.2 0.12
0.12
1.2 0.08
Sd =0.0373 m Sd =0.0373 m
Say=0.0927
y g y
acceleration
acceleration
0.1 Sa =0.0999
Say=0.0999 gg Sa=0.0999
Sa =0.0999 gg
acceleration
0.1 0.1
acceleration
0.1
1 y 0.1
1 0.06 yy
0.08
0.08
0.08
0.8 0.08
0.08
0.8 0.04
0.06 Capacity curve
Spectral
Spectral
0.06
0.06 0.06
0.06
0.6 0.02 Bilinear curve
0.6
Spectral
Spectral
0.04 Response spectrum

0.04
0.04
0.4 Capacity
0.04
0.04
0.4 curve 0
Capacity
Capacity curve
curve 0.02 Bilinear curve 0 Capacity
Capacity 0.05
curve
curve 0.1 0.15
0.02 Bilinear
Bilinear curve
curve Response
0.02 spectrum Bilinear Spectral displacement (m)
curve
0.02
0.2 0.02
0.2 Bilinear curve
0 Response
Response spectrum
spectrum Response
Response spectrum
spectrum
0 0.05 0.1 0.15
0 00 0.05 0.1 0.15 0 00 0.05 0.1 0.15
0 0.05 0.1 Spectral 0.15
displacement 0(m) 0.05 0.1 0.15
0.15
Spectraldisplacement
Spectral displacement(m)
displacement (m)
(m) Spectraldisplacement
displacement (m)
(m)
Figure 4.24 Bilinear curves according to ATC-40 (left) and FEMA-273 (right) for a “weak”
response spectrum.
0.16
0.14
0.16
0.16 0.16
(g)
0.16
1.6
0.12 0.16
1.6
Sd =0.0373
=0.0373 mm Sdy=0.0226 m Sdy=0.0393 m
y
Spectral acceleration
=0.9991 gm/s2 0.14 y

0.14 Say=0.0999 0.14 Say=0.0625 g Sa =1.0287 m/s2
1.4
0.14
0.1 y 0.14
1.4 y =0.1028 g
y
(g)(g)
0.12
(m/s2)
(m/s2)
Sdy=0.0353 m
(g)
(g)
Sdy=0.0373 m Sdy=0.0373 m
(g)
0.12
0.12
1.2
0.08 0.12
0.12
acceleration
Sd =0.0373 m 1.2 Sd =0.0373 m

y Say=0.0927 g y
acceleration
acceleration
acceleration
Spectral acceleration
=0.0999 gg
Sa=0.0999 0.1 =0.0999 gg
Sa=0.0999
Sa Sa
acceleration
0.1 0.1
acceleration
0.1
1
0.06 yy 0.1
1 yy
0.08
0.08
0.08
0.8
0.04 0.08
0.08
0.8
Spectral
Capacity curve 0.06

Spectral
Spectral
0.06 0.06
Spectral
0.06
0.6
0.02 Bilinear curve 0.06
0.6
Spectral
Spectral
Response spectrum 0.04

0.04
0.04
0.4
0 0.04
0.04
0.4 Capacity curve
0 0.05 Capacity
0.1
Capacity curve
curve 0.15 0.02 Capacity
Bilinear
Capacity curve
curve
curve
0.02
0.02
0.2 Spectral displacementBilinear
Bilinear curve
(m) curve 0.02
0.02 Bilinear
Response
Bilinear curve
spectrum
curve
0.2
Response
Response spectrum
spectrum 0 Response
Response spectrum
spectrum
0 0.05 0.1 0.15
00 0 00 0.05 displacement 0.1 0.15
00 0.05
0.05 0.1
0.1 0.15
0.15 0 Spectral
0.05 0.1(m) 0.15
Spectraldisplacement
displacement (m)
(m) Spectral displacement
Spectral displacement (m)(m)

Figure 4.25 Bilinear curves according to ATC-40 (left) and FEMA-273 (right) for a “strong”
response spectrum.
A significant discrepancy is observed in the bilinear curves for the “weaker”

response spectrum. In fact, in the case where the ATC-‐40 guidelines were followed,
an equivalent viscous damping equal to 8.94% was obtained, whilst in the second
case where the FEMA-‐273 was used, a lower equivalent damping of 5.91% was
attained. With regards to the second situation in which a stronger spectrum is
employed, both methods seem to provide reasonable results (equivalent damping of
13.21% and 12.30% for ATC-‐40 and FEMA-‐273, respectively). To summarize,
according to what is prescribed in ATC-‐40, the Capacity Spectrum Method can be
applied taking the following steps:
1. Create a numerical model of the structure and apply lateral forces to the
structure, in order to assess the relation between base shear and a reference
node displacement (i.e. pushover curve);
54
Chapter 4
2. Convert the obtained curve for the MDOF, to the equivalent SDOF (in terms of
spectral acceleration (Sa) versus spectral displacement (Sd));
3. Construct a bilinear representation of the capacity curve, using one of the

aforementioned methods;
4. Calculate the response spectra (code-‐based or derived from a ground motion
record) for an equivalent 5% elastic damping;
5. Intersect capacity spectrum and reduced demand spectrum, build new

bilinear capacity spectrum for that point and update effective viscous
damping;
6. If the updated damping acceptably matches the previously

determined/assumed one, the Performance Point (PP) of the structure is
found, otherwise, iteration is carried out until convergence is reached.
This performance point is equivalent to what would be obtained by subjecting the

equivalent single degree of freedom to a nonlinear time history analysis. This
response (i.e. performance point) can be used to allocate the structure in a damage
state, based on a pre-‐established set of displacement thresholds.
This process can be repeated several times considering other ground motion
records, as well as structures (i.e. building class). The result will be a distribution of
buildings in different damage states, according to a set of ground motion records
(analytically derived damage probability matrix). These results can then be used to
derive a fragility model, as described in the preceding sections.
An alternative method for modelling the fragility of structures with nonlinear static
methods is to calculate the nonlinear response of the median building for a building
class on-‐the-‐fly for a given earthquake scenario. The pushover curve of the SDOF
system for the median building is thus needed. Pre-‐computed fragility functions
which relate the nonlinear response to the probability of exceedance of different
damage states are then used to obtain the damage distribution; the uncertainty in
the aforementioned fragility functions combines the uncertainty in the capacity and
the uncertainty in the response. This procedure is shown in Figure 4.26. However, it
is noted that HAZUS also includes a list of fragility functions in terms of peak ground
acceleration that can be used directly to calculate the damage distribution as a
function of the ground shaking (using GMPEs).
55

Figure 4.26 HAZUS methodology for damage distribution assessment (FEMA, 2005)
4.4.6.1. DBELA (Silva et al. 2013)

This section presents a procedure to derive fragility functions for populations of
buildings that relies on the displacement-‐based earthquake loss assessment
(DBELA) methodology. This method builds upon the urban assessment methodology
proposed by Calvi (1999), in which the principles of structural mechanics and
seismic response of buildings were used to estimate the seismic vulnerability of
classes of buildings. In this method, the displacement capacity and demand for a
number of limit states needs to be calculated. Each limit state marks the threshold
between the levels of damage that a building might withstand, usually described by
a reduction in strength or by exceedance of certain displacement/drift levels. Once
these parameters are obtained, the displacement capacity of the first limit state is
compared with the respective demand. If the demand exceeds the capacity, the next
limit states need to be checked successively, until the demand no longer exceeds the
capacity and the building damage state can be defined. If the demand also exceeds
the capacity of the last limit state, the building is assumed to have collapsed. This
56
Chapter 4
procedure is schematically depicted in Figure 4.27, in which the capacities for three
limit states are represented by Δi and the associated demand by Sdi.
Spectral
Displacement
Sd,1 > Δ1 Moderate damage reached
Sd,2 > Δ2 Heavy damage reached Disp. spectrum for moderate damage LS
Sd,3 < Δ3 Collapse not reached
Disp. spectrum for heavy damage LS
Δ3
Disp. spectrum for collapse LS
Sd,3
Sd,2
Δ2
Sd,1 Δ1
Period
T1 T2 T3
Figure 4.27 -‐ Comparison between limit state capacity and the associated demand (adapted from Bal et
al., 2010).
In this example, the demand exceeds the capacity in the first and second limit state
but not in the third limit state, thus allocating the building to the third damage state.
The demand in this methodology is represented by a displacement spectrum which
can be described as the expected displacement induced by an earthquake on a
single-‐degree-‐of-‐freedom (SDOF) oscillator of given period and damping. Therefore,
the displacement capacity equations that are derived must describe the capacity of a
SDOF equivalent structure, and hence must give the displacement capacity at a given
limit state (which could be structural or non-‐structural).
When considering structural limit states, the displacement at the height of the
centre of seismic force of the original structure (HCSF) can be estimated by
multiplying the base rotation by the height of the equivalent SDOF structure (HSDOF),
which is obtained by multiplying the total height of the actual structure (HT) by an
effective height ratio (efh) (see Figure 4.28).
57

Figure 4.28 – Definition of effective height coefficient (Glaister and Pinho, 2003).
Pinho et al. (2002) and Glaister and Pinho (2003) proposed formulae for estimating
the effective height coefficient for different response mechanisms. For what
concerns the beam sway mechanism (or distributed plasticity mechanism, as shown
in Figure 4.29), a ratio of 0.64 is proposed for structures with 4 or less storeys, and
0.44 for structures with 20 or more storeys. For any structures that might fall within
these limits, linear interpolation should be employed. With regards to the column-‐
sway mechanism (or concentrated plasticity mechanism, as shown in Figure 4.29),
the deformed shapes vary from a linear profile (pre-‐yield) to a non-‐linear profile
(post-‐yield). As described in Glaister and Pinho (2003), a coefficient of 0.67 is
assumed for the pre-‐yield response and the following simplified formula can be
applied post-‐yield (to attempt to account for the ductility dependence of the
effective height post-‐yield coefficient):
𝜀! !"! !!!
𝑒𝑓! = 0.67 − 0.17 ∙
𝜀! !"!
Plastic
hinges

Figure 4.29 – Deformed profiles for beam-‐sway (left) and column-‐sway (right) mechanisms (adapted
from Paulay and Priestley, 1992).
58
Chapter 4
The displacement capacity at different limit states (either at yield (Δy) or post-‐yield
(ΔLSi)) for bare frame structures can be computed using simplified formulae, which
are distinct if the structure is expected to exhibit a beam-‐ or column-‐sway failure
mechanism. These formulae can be found in Bal et al. (2010) or Silva et al. (2013).
In order to estimate whether a given frame will respond with a beam-‐ or a column-‐
sway mechanism it is necessary to evaluate the properties of the storey. A
deformation-‐based index has been proposed by Abo El Ezz (2008) which reflects the
relation between the stiffness of the beams and columns. This index can be
computed using the following formula:
ℎ!
𝑙!
𝑅=
ℎ!
𝑙!
where lc stands for the column length. Abo El Ezz (2008) proposed some limits for
this index applicable to bare and fully infilled frame structures, as described in Table
4.4.
Table 4.4 -‐ Limits for the deformation-‐based sway index.

Failure mechanism
Building typology
Beam sway Column Sway
Bare frames R ≤ 1.5 R > 1.5
Fully infilled frames R ≤ 1.0 R > 1.0

The displacement demand is initially represented by the 5% damped spectra. The

displacement is computed for the period at each limit state and modified by a
correction factor (η), representative of the equivalent viscous damping and limit
state ductility. In EC8 (CEN, 2004), the following equation is proposed for the
calculation of the correction factor:
10
𝜂=
5 + 𝜉!"
where ξeq stands for the equivalent viscous damping, which can be estimated using
the formula proposed by Priestley et al. (2007) for RC structures with a non negative
post-‐yield ratio:
where μLSi stands for the ductility at the considered limit state (assumed as the ratio
between ΔLSi and Δy). More accurate approaches have recently been proposed to
estimate the correction factors (η), considering additional parameters, such as the
59
magnitude or source-‐to-‐site distance (Rezaeian et al., 2012). The integration of such

methods in the calculation of the displacement demand is in the future development
plans of the proposed methodology.
𝜇!"! !!
𝜉!" = 0.05 + 0.565
𝜇!"! 𝜋
With regards to the calculation of the yield period (Ty in seconds) for bare frame
structures, Crowley and Pinho (2004) and Crowley et al. (2008) proposed a
relationship between the period and the total height of 0.10HT and 0.07HT for
structures without and with lateral load design, respectively. For infilled frames, a
relation equal to 0.06HT has been recommended by Crowley and Pinho (2006) for
structures without lateral load design. The elongated period of vibration for any of
the limit states (TLSi) can be computed using the following formula:
𝜇!"!
𝑇!"! = 𝑇!
1 + 𝛼 ∙ 𝜇!"! − 𝛼
where α stands for the post-‐yield stiffness ratio. In cases where this ratio can be
assumed as zero, the relation between TLSi and Ty will depend purely on the limit
state ductility as follows:
𝑇!"! = 𝑇! 𝜇!"!
In this methodology to derive fragility functions, a randomly generated population

of buildings is produced, according to the probabilistic distribution of a set of
material and geometrical properties (e.g. Bal et al. (2008), Silva et al. 2014b). Using
the set of synthetic buildings, the displacement capacity Δi is calculated using the
equations described above. Similarly, the displacement demand Sdi is also computed
for each limit state period using damped spectra at a level of equivalent viscous
damping, from a set of accelerograms. However, if a user wishes to avoid this
additional computational effort, the displacement spectra can also be provided
directly. The displacement demand for each limit state is then computed by
modifying the elastic displacement spectrum by a correction factor ηi,
representative of the equivalent viscous damping and limit state ductility.
Once the capacity and demand displacements for the whole group of synthetic
buildings are computed, both sets of displacements are compared, and used to
allocate each building in a certain damage state. Thus, for each ground motion
60
Chapter 4
record, percentages of buildings in each damage state can be obtained and fragility
curves can be extrapolated. In Figure 4.30, the building damage distributions for 4
records with different levels of spectral acceleration at the fundamental period are
presented.
Record 1 – Sa 0.10 g Record 2 – Sa 0.20 g Record 3 – Sa 0.30 g Record 4 – Sa 0.45 g

Figure 4.30 -‐ Derivation of fragility curves based on building damage distribution.
The logarithmic mean, λ, and logarithmic standard deviation, ζ, that are estimated
for each fragility curve will have some uncertainty, due to the scatter of the results.
Hence, a sampling method can be used to properly evaluate the uncertainty on the
statistics. This method can be a continuous bootstrap sampling with replacement
from the original dataset (Wasserman, 2004). Figure 4.31 shows an example of the
estimated limit state exceedance probability from 250 records.

Figure 4.31 - Statistical treatment of the parameters defining the curve.
61
This methodology to derive fragility functions is summarized in Figure 4.32.

Figure 4.32 -‐ Workflow of the DBELA fragility functions calculator.
4.4.7. Nonlinear dynamic analysis

Nonlinear dynamic analysis involves the introduction of accelerograms at the base of
the model and the nonlinear structural package can be used to compute the
response of the building or frame in terms of a number of parameters such as: the
floor displacements or accelerations or the relative displacement between floors or
the member forces and deformations, at any given time in the analysis. Nonlinear
analyses can be numerically unstable and it can be difficult to derive meaningful
results without significant modelling knowledge. For fragility assessment, a large
number of accelerograms with increasing intensity need to be run through the
model to capture the record-‐to-‐record variability. For fragility assessment, the
response of the structure (in terms of forces and deformations as mentioned
previously) needs to be related to the performance in terms of a global damage
state. Experimental tests of members can be useful for calibration of the
aforementioned relationship between response and damage, but the main difficulty
lies in defining how many members should reach a given level of response for the
62
Chapter 4
definition of the global damage state. Attempts have been made to combine the
member responses into a single global damage index e.g. Park and Ang, 1985).
12,000
10,000
8,000
6,000
4,000
Acceleration
2,000
-2,000
-4,000
-6,000
-8,000
0 5 10 15
Time
12,000
10,000
8,000
6,000
4,000
2,000
Acceleration
0
-2,000
-4,000
-6,000
-8,000
-10,000
-12,000
-14,000
0 5 10 15
Time
Figure 4.33 Numerical model of a RC MRF for nonlinear dynamic analysis
The following steps can thus be followed to produce fragility functions with
nonlinear dynamic analysis:
1. Random generation of population of 2D frames through Monte Carlo

simulation;
2. Nonlinear dynamic analysis for each frame using a large selection of ground
motion records;
3. Identification of response and associated global damage state for each ground
motion records;
4. For each intensity measure level, plot the cumulative percentage of buildings
in each global damage state;
5. Apply regression analysis to calculate the parameters (mean and standard

deviation) of the fragility functions (usually assumed to follow a cumulative
lognormal distribution).
Figure 4.34 illustrates a fragility model developed by Silva et al. (2014b) for
moment-‐frame reinforced concrete structures using nonlinear dynamic analyses.
63
0.9
0.8
Probability of exceedance
0.7
0.6
0.5
0.4
0.3
0.2 Slight
Moderate
0.1 Extensive
Collapse
0
0 0.2 0.4 0.6 0.8 1 1.2 1.4 1.6

Figure 4.34 Fragility function obtained from nonlinear dynamic analyses (Silva et al. 2014b)
The following studies have used nonlinear dynamic analysis to produce fragility
functions: Singhal and Kiremidjian (1996), Dumova-‐Jovanoska (2004) and Silva et
al. (2014b).
There are important drawbacks in analytical fragility function assessment which

need to be recognized:
o To reduce the complexity of analyses, buildings are often simplified as 2D

regular models and thus ignore the irregularities, structural deficiencies and
construction defects that cause the majority of structural collapses;
o The relationship between the modelled response and the actual damage of
structures is not straightforward and often requires experimental testing for
improved calibration.
Nevertheless, analytical fragility assessment allows for more control on the

uncertainties being modelled, easier interpretation of the results, models for new
and retrofitted buildings can be developed, and models for structures which have
not experienced damage can be produced.
4.5. Consequence Functions

As previously mentioned, consequence functions (also known as damage-‐to-‐loss
conversion equations), allow fragility functions to be transformed into vulnerability
functions. These functions are generally based on observations (empirical) or expert
opinion.
In order to transform fragility functions into vulnerability functions in terms of

repair cost, so-‐called damage factors (ratios of cost of repair to cost of replacement)
64
Chapter 4
are needed for the considered building typology for each damage state considered in
the fragility function. A number of studies have proposed damage factors for
different damage states (generally based on damage cost data from previous
earthquakes), though only a few have looked at the uncertainty in these damage
factors for a given damage state. Masi et al. (2002) represents this variation as a
Beta distribution, based on post-‐earthquake survey data from Italy, whilst Goretti
and Di Pasquale (2004) assume normal distributions for the damage factors.
Construction practices and costs, in particular those applied in the repair and
reconstruction phase, vary significantly from one country to another, and thus
damage factors should vary accordingly. Assigning repair types and associated costs
to damage states depends on a number of factors from failure mechanism, to size
and geometry of buildings in the building class, to engineering practice of the
country. Some examples of attempts to calculate damage factors include Bal et al.
(2008) and Hill and Rossetto (2008b, c).
For the method proposed by Giovinazzi and Lagomarsino (2004, 2006), a

correlation between the mean damage grade μD and what is referred to as the
damage index DI (i.e. the ratio between the repair and the reconstruction cost) has
been proposed, which has been fit to data present in the literature (though it is not
clear whether this is purely empirical or a mixture of expert opinion and observed
data):
µD = 5 DI0.57
Hence, the curves presented in Figure 4.13 can be converted from mean damage
grade to loss ratio (ratio of cost of repair to cost of replacement), and thus
vulnerability functions can be produced.
The definition of replacement cost varies in the aforementioned studies, from “new”
build cost (Bal et al. 2008) to market value of the property before the earthquake
(Whitman et al. 1973), to construction replacement cost (FEMA, 2003). It is noted
that whatever is the definition, it should be compatible with the definition of
replacement cost in the exposure model.
There are thus a number of issues that should be considered when using existing
damage factors:
o Are the damage states equivalent to those of the fragility functions?
65
o What is the meaning of the replacement cost used in their derivation?
o Are they for the same country, or is the engineering and construction practice
similar?
Coburn and Spence (2002) discuss the many earthquake-‐induced causes of death,
which include fires following earthquakes, tsunamis, rockfalls and landslides.
However, it is clear that in most large-‐scale earthquake disasters the main cause of
death is the collapse of buildings (see Figure 4.35). Over the last century about 75%
of fatalities attributed to earthquakes have been caused by the collapse of buildings
(Coburn and Spence, 2002), and the greatest proportion of victims die in the
collapse of masonry buildings. Unfortunately these buildings are very common in
seismic areas and make up a large proportion of the building stock around the
world. It should be noted, however, that in some areas of the world the buildings
constructed in reinforced concrete, which has become the material of choice for new
construction, are actually highly vulnerable and when they do collapse can be more
lethal and cause the death of a higher percentage of their occupants than masonry
buildings. Figure 4.35 shows how there has been an increase in fatalities caused by
the collapse of reinforced concrete buildings in the second half of the last century.

Figure 4.35 Breakdown of earthquake-‐induced fatalities by cause (Coburn and Spence, 2002)
Coburn and Spence (2002) provide a plot of the total number of people killed versus
the total number of heavily damaged buildings (Figure 4.36), where it can be readily
seen that deaths can be correlated to the destruction caused by earthquakes.
However, it should be noted that the casualty values are highly dispersed when less
than 5000 buildings were damaged.
66
Chapter 4

Figure 4.36 Relationship between the number of fatalities and the number of buildings damaged
heavily in earthquake (Coburn and Spence, 2002)
In order to transform fragility functions into vulnerability functions in terms of

casualties and fatalities, casualty loss ratios are needed which relate the number of
casualties to the number of occupants of buildings of a given typology damage to a
given damage state. Detailed post-‐earthquake consequences data is needed in order
to anchor casualty loss ratios to the damage and building type, and this information
is often not available. Furthermore, it is difficult to extract the death and injury
counts that are due to ground shaking alone. Despite these difficulties, Spence
(2007) and So (2009) have attempted to produce casualty loss ratios as described
previously (Figure 4.4) and recent earthquakes have proved to have extremely
detailed casualty information (e.g. 2011 New Zealand earthquake: So, personal
communication).
Table 4.5 Injury distributions for specific building types (Spence, 2007). UI=uninjured; I1=slight
injuries; I2=moderate injuries; I3=serious injuries; I4=critical injuries; I5=deaths
Complete Damage State
UI I1 I2 I3 I4 I5
Masonry (1F) 23.6% 50.0% 12.0% 8.0% 0.4% 6.0%
Masonry (2-‐3F) 16.5% 50.0% 15.0% 10.0% 0.5% 8.0%
Masonry (≥4F) 9.4% 50.0% 18.0% 12.0% 0.6% 10.0%
RC (1F) 32.9% 30.0% 19.0% 3.0% 0.2% 15.0%
RC (2-‐3F) 20.8% 30.0% 23.0% 4.0% 0.2% 22.0%
RC (≥4F) 9.7% 30.0% 27.0% 5.0% 0.3% 28.0%

67
Chapter 5
This Chapter describes the convolution of seismic hazard, exposure and
vulnerability for the calculation of physical risk. The following sections focus on the
methods that have been implemented within the open-‐source software OpenQuake
(Silva et al. 2014c) that is being developed by the Global Earthquake Model.
5.1. Scenario Risk Assessment

Scenario risk assessment involves the calculation of losses and loss statistics due to
a single, deterministic earthquake (such as a large damaging historical earthquake),
for a collection of assets. Such analyses are of importance, for example, for
understanding the validity of a given model, for emergency management planning,
for raising societal awareness of risk, or for rapid post-‐earthquake loss assessment.
This workflow requires the definition of a finite rupture (rupture scenario model),
an exposure model and a physical vulnerability model. The characteristics of the
rupture are used to generate ground motion fields (i.e. estimates of the intensity of
ground shaking at each location in the model), considering the aleatory variability in
the GMPE or IPE with or without spatial correlation. For a given ground motion field,
the intensity measure level at a given site is used to obtain a random sample of the
loss ratio (considering the mean and standard deviation of the loss ratio at the given
intensity measure level from the vulnerability model) for each asset contained in the
exposure model. Using the sampled loss ratios for each asset, the mean and standard
deviation of the loss ratios across all ground motion fields can be calculated. Loss
ratios are converted into losses by multiplying by the value of the asset given in the
Chapter 5
exposure model. It is furthermore possible to sum the losses throughout the region
and to compute the mean and standard deviation of the total loss, and to plot
statistics such as the mean loss on maps, as presented in Figure 5.1 for the Pisco
(Peru) earthquake of 2007.

Figure 5.1 Mean economic loss map for the Pisco (Peru) earthquake of August 2007.
5.2. Scenario Damage Assessment

Scenario damage assessment leads to the calculation of the distribution of buildings
in various damage states (e.g. no damage, moderate, collapse), given a single
earthquake. This module requires the definition of a finite rupture, an exposure
model and a fragility model. The main results of general interest are the damage
distribution per asset, total damage distribution and collapse maps. This calculator
is very similar to the Scenario Risk calculator, however instead of a vulnerability
model, a fragility model is required. Once the ground motion field have been
generated, the damage distribution for each asset in each location is obtained from
the appropriate fragility model (as a function of the building typology). Using the
number or area of each building typology in each location, the damage distributions
can be weighted and an aggregated damage distribution for the whole region is
obtained. The mean and standard deviation of the aggregated damage distribution
across the ground motion fields can also be obtained. This type of results is
illustrated in Figure 5.2.
69
Number#of#buildings#
Collapse#
Extensive#
Moderate#
Slight#
Adobe#
Wooden# No#damage#
Masonry#
Stone#
RC#

Figure 5.2 Distribution of damage per building typology (left) and mean collapse map (right) for the
Pisco (Peru) earthquake of August 2007.
5.3. PSHA and Event-‐based Risk Assessment

The aforementioned types of risk analysis use single, pre-‐defined events and are
thus conditional on that event occurring. In order to consider the risk considering
any possible event that could occur within a given region, the probability of future
events needs to be incorporated into the analysis. There are two ways in which this
can be done:
o By using the output of Probabilistic Seismic Hazard Assessment (PSHA) in

terms of hazard maps and curves (known as classical PSHA-‐based risk
assessment herein).
o By using stochastic event sets, generated from seismic hazard input models
(known as probabilistic event-‐based risk assessment herein).
When the risk assessment only needs to provide mean outputs (e.g. average annual
losses, AAL) or relative estimates of risk, then the former approach is appropriate.
This method is commonly used for ranking countries in terms of risk or for
producing risk maps of AAL. When the full range of probabilities of losses need to be
robustly estimated (through loss exceedance curves), then the latter method is more
appropriate.
As with the scenario-‐based risk calculator, the main input models are exposure and
vulnerability. Hazard curves are the output of the hazard module that are then input
70
Chapter 5
into the risk calculator. These hazard curves can be combined with a vulnerability
and exposure model to derive asset-‐specific loss exceedance curves. Due to the
employment of logic trees, it is possible to have multiple hazard curves at each
location (one hazard curve per logic tree branch). Thus, various loss exceedance
curves are also calculated for each asset, and a user has the possibility of extracting
results from a specific logic tree branch, the mean loss exceedance curves, or loss
curves corresponding to a specific percentile. These loss exceedance curves can be
used to derive risk maps, containing the expected loss for a given return period, as
presented in Figure 5.3.
Figure 5.3 Economic losses for a probability of exceedance of 10% (left) and 2% (right) in 50 years.
The same exposure and vulnerability models mentioned previously are also valid for
the probabilistic event-‐based risk assessment approach. As with the scenario risk
calculator, ground motion fields (GMFs) are used in the risk calculations, but a much
larger number of GMFs are used; one per event in the stochastic event set. For each
ground motion field, the intensity measure level at a given site is combined with a
vulnerability function, from which a loss ratio is randomly sampled for each asset.
Spatial correlation of the ground motion residuals and vulnerability correlation in
the loss ratios within assets of the same taxonomy can be accounted, as described in
the Scenario Risk Calculator. The distribution of loss per asset across all ground
motion fields is used to estimate the rate of exceeding a number of loss thresholds,
which can then be converted into a loss exceedance curve assuming a Poissionian
distribution. Once again, a logic tree structure can be employed in the calculation of
71
the ground motion fields, which leads to a number of loss exceedance curves per
asset equal to the number of branches considered in the hazard calculations. These
loss curves can be for a single asset, or aggregated for all assets in the exposure
model. Once again, loss maps can also be generated, as illustrated in in Figure 5.3.
5.4. Retrofitting Decision Support Tool

Seismic retrofitting involves structural interventions that aim to improve the
performance of buildings in earthquakes. Conventional intervention techniques for
retrofitting can be subdivided into global and member intervention types. Examples
of the former include addition of RC structural walls, steel bracing and buttresses
whilst the latter can take the former of member jacketing or injection of epoxy resin.
Figure 5.4 shows four different global techniques that could be applied to retrofit a
building with an open ground floor storey. Each method has its pros and cons, which
need to be considered:
o What impact will the scheme have on the foundations?
o How much impact will it have on the functioning of the building?
o How much does it cost?
o How much is the performance of the building improved?

Figure 5.4 Four global retrofitting schemes for a building with a soft storey.
72
Chapter 5
Benefit-‐cost analysis (BCA) is a systemic procedure for evaluating decisions that

have an impact on society and that can be applied in the case of retrofitting to find
the most attractive solution that maximises the net present value.
The direct costs of the various options (such as those shown in Figure 5.4) need to
be estimated, considering also the impact that the time that the building will be
unavailable might have on the cost. The different retrofitting schemes that can be
applied need to be modelled. This is where analytical fragility assessment is
fundamental in risk assessment, such that a robust assessment of the impact of the
different retrofitting schemes on the response of the building can be estimated. The
previously described methods can be applied for models with different retrofitting
schemes and vulnerability functions of the original and retrofitted building derived
(Figure 5.5).

Figure 5.5 Discrete vulnerability functions (uncertainty not plotted) for original and retrofitted
buildings
The next step involves assessing the benefits of the different retrofitting schemes in
terms of reduced losses. The losses should be economical in order to be compared
with the cost of retrofitting, but this does not mean that only repair cost can be
accounted for. Although a difficult and somewhat controversial subject, a value of life
can be assigned to the number of fatalities as well. Economists have used several
estimation techniques for estimating the value of life. These range from hypothetical
surveys where people are asked how much they must be paid to accept certain risks,
to examining the wage premium people working in hazardous jobs are given to
compensate them for the additional risks they are incurring.
73
Loss exceedance curves need to be produced using the classical PSHA-‐based risk
workflow, considering the original vulnerability functions, and then again with the
retrofitted vulnerability model.

Figure 5.6 Loss exceedance curves for original and retrofitted buildings
The Benefit-‐Cost Ratio (BCR) can then be calculated as follows:
(𝐴𝐴𝐿! − 𝐴𝐴𝐿! ) (1 − 𝑒 !!" )

𝐵𝐶𝑅 = .
𝐶 𝑟
where AALC is the average annual loss of the current (original) building, AALR is the
average annual loss of the retrofitted building, r is the discount rate (to bring the
losses to the future present monetary value), t is the design life of the building (e.g.
50 years) and C is the cost of retrofitting (that needs to be included in the exposure
model). The difference in average annual loss with and without retrofitting provides
the average avoided loss per year should retrofitting be employed. This is then
multiplied by a factor that brings the avoided losses over the life of the building to
the present value. For a given design life, as the discount/interest rate increases, the
future benefits decrease as the losses in the future are higher, hence the average
saving in loss over t years (the benefit) is lower and the BCR decreases. Discount
rates are generally assumed to be between 3 and 6% (FEMA, 1992).
The decision on whether to retrofit and which retrofitting scheme to adopt can be
based on the scheme with the highest BCR (above 1).
74
REFERENCES
Abo El Ezz, A. (2008). “Deformation and strength based assessment of seismic failure
mechanisms for existing RC frame buildings”. MSc Thesis, ROSE School, Pavia, Italy.

Abrahamson, N. A. (2006). “Seismic hazard: Problems with current practice and future
developments”. Proceedings of the 1st European Conference on Earthquake Engineering
and Seismology, Geneva, Switzerland.

Amiri, G.G., Jalalian, M., Amrei, S.A.R. (2007) “Derivation of vulnerability functions based on
observational data for Iran,” Proceedings of International Symposium on Innovation
and Sustainability of Structures in Civil Engineering, Tongji University, China.

Applied Technology Council (ATC) (1985) ATC-‐13, Earthquake Damage Evaluation Data for
California, Applied Technology Council, Redwood City, CA, 492 pp.

Applied Technology Council (ATC) (1996). ATC-‐40 Seismic Evaluation and Retrofit of
Concrete Buildings, Volumes 1 and 2, Report No. ATC-‐40, Applied Technology
Council, Redwood City, California, USA.

Bal, I., Crowley, H., Pinho, R., Gulay, F. (2008b). “Detailed assessment of structural
characteristics of Turkish RC building stock for loss assessment models”, Soil
Dynamics and Earthquake Engineering, 28:914-‐932.

Bal I., Crowley, H. and Pinho, R. (2010) Displacement-‐Based Earthquake Loss Assessment:
Method Development and Application to Turkish Building Stock, ROSE Research Report
2010/02, IUSS Press, Italy.

Bird, J.F. and Bommer, J.J. (2004) “Earthquake losses due to ground failure,” Engineering
Geology, 75(2), pp. 147–79.

Bolt, B.A. (1999) Earthquakes, Fourth edition, W.H. Freeman.

Bommer, J. J., Spence, R., Erdik, M., Tabuchi, S., Aydinoglu, N., Booth, E., Re, D. D.,
Pterken, D. (2002). “Development of an Earthquake Loss Model for Turkish
Catastrophe Insurance”. Journal of Seismology, 6:431-‐446.

Bommer, J.J. (2004) Engineering Seismology, Class notes for the ROSE School, IUSS, Pavia,
Italy

Bommer, J. J., Abrahamson, N. A. (2006). "Why do Modern Probabilistic Seismic
Hazard Analyses Often Lead to Increased Hazard Estimates?". Bulletin of the
Seismological Society of America, 96:1967-‐1977.

Bommer, J.J. and Stafford, P.J. (2008). Seismic hazard and earthquake actions, In: Elghazouli,
A.Y., Seismic Design of Buildings to Eurocode 8, Taylor & Francis

Brzev S., C. Scawthorn, A.W. Charleson, L. Allen, M. Greene, K. Jaiswal, and V. Silva (2013),
GEM Building Taxonomy Version 2.0, GEM Technical Report 2013-‐02 V1.0.0, 188 pp.,
GEM Foundation, Pavia, Italy, doi:10.13117/GEM.EXP-‐MOD.TR2013.02.

Calvi, G.M. (1999) “A displacement-‐based approach for vulnerability evaluation of classes of
buildings,” Journal of Earthquake Engineering, Vol. 3, No. 3, pp. 411-‐438.

Calvi, M., Pinho, R., Magenes, G., Bommer, J., Restrepo-‐Vélez, L., Crowley, H. (2006).
“Development of Seismic Vulnerability Assessment Methodologies over the
past 30 years“, ISET Journal of Earthquake Technology, 43(3):75-‐104.

Casarotti, C., Pinho, R. (2007). “An adaptive capacity spectrum method for
assessment of bridges subjected to earthquake action”. Bulletin of Earthquake
Engineering; 5(3):377–390.

CEN (2004). Eurocode 8: Design of Structures for Earthquake Resistance -‐ Part 1:
General rules, seismic actions and rules for buildings. European Committee for
Standardization, Brussels, Belgium.

Chaulagain, H., Rodrigues, H., Silva, V., Spacone, E., Varum, H. (2014) "Earthquake
Loss Estimation for the Kathmandu Valley", Bulletin of Earthquake
Engineering.

Chopra, A. K., Goel, R. K., (2000). “Evaluation of NSP to estimate seismic deformation:
SDF systems”. Journal of Structural Engineering, 126(4):482-‐490.

Chopra, A. K., Goel, R. K. (2002). “A modal pushover analysis procedure for
estimating seismic demands for buildings”. Earthquake Engineering &
Structural Dynamics; 31(3):561–582.

Coburn, A. W. and Spence, R. (2002) Earthquake Protection, John Wiley & Sons.

Colombi, M., Borzi, B., Crowley, H., Onida, M., Meroni, F., Pinho, R. (2008) “Deriving
vulnerability curves using Italian earthquake damage data,” Bulletin of Earthquake
Engineering, 6, pp. 485-‐504

Crowley, H. and Pinho, R. (2004) “Period-‐height relationship for existing European
reinforced concrete buildings,” Journal of Earthquake Engineering, Vol. 8, Special Issue
1., pp.93-‐119

Crowley H., Pinho R. and Bommer J.J. (2004) “A probabilistic Displacement-‐based
Vulnerability Assessment Procedure for Earthquake Loss Estimation,” Bulletin of
Earthquake Engineering, Vol. 2, pp. 173-‐219.

Crowley, H., Bommer, J.J. and Stafford, P.J. (2008) “Recent Developments in the Treatment of
Ground-‐Motion Variability in Earthquake Loss Models, “ Journal of Earthquake
76

Engineering, Vol. 12 (S2), pp. 71-‐80.

Crowley, H., Pinho, R. (2006). “Simplified equations for estimating the period of
vibration of existing buildings”. Proceedings of the 1st European Conference on
Earthquake Engineering and Seismology, Geneva, Switzerland.

Crowley, H., Miriam, C., Borzi, B., Faravelli, M., Onida, M., Lopez, M., Polli, D., Meroni,
F. (2009). "A comparison of seismic risk maps for Italy". Bulletin of Earthquake
Engineering, 7(1):149-‐180.

Cua et al. (2010) Available from URL:
http://www.globalquakemodel.org/system/files/doc/GEM-‐TechnicalReport_2010-‐
4.pdf

Cummins, J. D., Mahul, O. (2009). Catastrophe Risk Financing in Developing
Countries: Principles for Public Intervention. The World Bank, Washington
D.C., USA.

Daniell, J. E., Wenzel, F., Khazai, B. (2010). "The Cost of Historic Earthquakes Today –
Economic Analysis since 1900 through the use of CATDAT". Proceedings of the
Australian Earthquake Engineering Society Conference, Perth, Australia.

Dengler, L.A. and Dewey, J.W. (1998) “An Intensity Survey of Households Affected by the
Northridge, California, Earthquake of 17 January, 1994,” Bulletin of the Seismological
Society of America, Vol. 88, p. 441-‐462.

Di Pasquale, G., Orsini, G. and Romeo, R.W. (2005) “New developments in seismic risk
assessment in Italy”, Bulletin of Earthquake Engineering, Vol. 3, No. 1, pp. 101-‐128.

Douglas, J. (2003) “Earthquake ground motion estimation using strong-‐motion records: a
review of equations for the estimation of peak ground acceleration and response
spectral ordinates,” Earth-‐Science Reviews, 61, pp. 43-‐104.

Dumova-‐Jovanoska, E. (2004) “Fragility curves for RC structures in Skopje region,”
Proceedings of 13th World Conference on Earthquake Engineering, Vancouver, Canada.

Erdik, M., Aydinoglu, N., Fahjan, Y., Sesetyan, K., Demircioglu, M., Siyahi, B., Durukal,
E., Ozbey, C., Biro, Y., Akman, H., Yuzugullu O. (2003). “Earthquake Risk
Assessment for Istanbul Metropolitan Area”, Earthquake Engineering and
Engineering Vibration, 2(1).

FEMA (1992) A benefit/cost model for the seismic rehabilitation of buildings (FEMA 227), Vols
1, 2. VSP Associates, Sacramento, California.

FEMA-‐273 (1997). NEHRP Guidelines for the Seismic Rehabilitation of Buildings.
Report No. FEMA 273. Federal Emergency Management Agency, Washington
D.C., USA.

77
FEMA (2003) HAZUS-‐MH MR4 Technical Manual, Federal Emergency Management Agency.
http://www.fema.gov/plan/prevent/hazus/hz_manuals.shtm

FEMA-‐440 (2005). Impovement of Nonlinear Static Seismic Analysis Procedures.
Federal Emergency Management Agency, Washington D.C., USA.

Gamba, P., Cavalca, D. Jaiswal, K., Huyck, C., Crowley, H. (2012). "The GED4GEM
Project: Development of a Global Exposure Database for the Global
Earthquake Model Initiative". Proceedings of the 15th WCEE, World Conference
on Earthquake Engineering, Lisbon, Portugal.

Freeman, S., Nicoletti, J., Tyrell, J. (1975). “Evaluation of Existing Buildings for
seismic risk -‐ A case study of Puget Sound Naval Shipyard, Bremerton,
Washington”. Proceedings of the 1st U.S. National Conference on Earthquake
Engineering, Berkley, USA.

Freeman, S., (2004). “Review of the development of the capacity spectrum
method”. ISET Journal of Earthquake Technology, 41:1-‐13.

Glaister, S. and Pinho, R. (2003) “Development of a Simplified Deformation-‐Based Method for
Seismic Vulnerability Assessment,” Journal of Earthquake Engineering, Vol. 7, pag 107-‐
140.

Goretti, A. and Di Pasquale, G (2004) “Building inspection and damage data for the 2002
Molise, Italy earthquake,” Earthquake Spectra, Vol. 20, pp. S167-‐S190

Grunthal, G. (ed.), (1998) European Macroseismic Scale 1998 (EMS-‐98). Cahiers du Centre
Europeen de Geodynamique et de Seismologie 15, Centre Europeen de Geodynamique
et de Seismologie, Luxembourg.

Hill, M. and Rossetto, T. (2008a) “Comparison of building damage scales and damage
descriptions for use in earthquake loss modelling in Europe,” Bulletin of Earthquake
Engineering, Vol. 6, pp. 335-‐365.

Hill, M. and Rossetto, T. (2008b) “Improving seismic loss estimation for Europe through
enhanced relationships between building damage and repair costs,” Proceedings of
14th World Conference on Earthquake Engineering, Beijing, China, Paper S01-‐02-‐009.

Hill, M. and Rossetto, T. (2008c) “Development of parameters for use in seismic recovery
estimation of residential buildings in Lima, Peru,” Proceedings of 14th World Conference
on Earthquake Engineering, Beijing, China, Paper 09-‐01-‐0027.

Huyck, C., Esquivias, G., Gamba, P., Hussain M., Odhiambo, O., Jaiswal, K., Chen, B., Becker, M.,
Yetman, G., Crowley, H. (2011) D2.2 Survey of available input databases for GED.
Available from URL: http://www.nexus.globalquakemodel.org/ged4gem/posts

Jaiswal, K.S., Wald, D.J. and Hearne, M. (2008) Estimating Casualties for Large Earthquakes
Worldwide using an Empirical Approach, U.S. Geological Survey Open-‐File Report
78

2009-‐1136, 78 pp.

Jaiswal, K.S., and Wald, D.J. (2008) Creating a Global Building Inventory for Earthquake Loss
Assessment and Risk Management, U.S. Geological Survey Open-‐File Report 2008-‐1160,
103 pp.

Jaiswal, K.S. and Wald, D.J. (2011) Rapid Estimation of the Economic Consequences of Global
Earthquakes, U.S. Geological Survey Open-‐File Report 2011-‐1116, 47 pp.

Jayaram, N. and Baker, J. W. (2009) “Correlation model for spatially distributed ground-‐
motion intensities,” Earthquake Engineering and Structural Dynamics, 38(15), pp.
1687–1708.

Kappos, A., Panagopoulos, G., Panagiotopoulos, C., Penelis, G. (2006). "A hybrid
method for the vulnerability assessment of R/C and URM buildings". Bulletin
of Earthquake Engineering, 4(4):391-‐413.

Karababa, F.S. and Pomonis, A. (2010) “Damage data analysis and vulnerability estimation
following the August 14, 2003 Lefkada Island, Greece earthquake,” Bulletin of
Earthquake Engineering, Vol. 9, No. 4

Lagomarsino S., Giovinazzi, S. (2006). “Macroseismic and Mechanical Models for the
Vulnerability Assessment of Current Buildings”. Bulletin of Earthquake
Engineering, Special Issue “Risk-‐Ue Project”, 4(4).

Liel, A.B. and Lynch, K.P. (2009) “Vulnerability of reinforced concrete frame buildings and
their occupants in the 2009 L’Aquila, Italy earthquake,” Natural Hazards Review, in
press

Lin, Y. Y., Chang, K. C., Wang, Y. L. (2004). “Comparison of displacement coefficient
method and capacity spectrum method with experimental results of RC
columns”. Earthquake Engineering & Structural Dynamics, 33:35-‐48.

Masi, A., Dolce, M. and Vona, M. (2002) A procedure to estimate economic losses due to
damage of residential buildings (in Italian), DiSGG (Dipartimento di Strutture,
Geotecnica, Geologica Applicata all’Ingegneria) Proceedings 5, University of Basilicata,
Potenza, Italy.

McCormack, T.C. and Rad, F.N. (1997). “An earthquake loss estimation methodology for
buildings based on ATC-‐13 and ATC-‐21”, Earthquake Spectra, Vol. 13, No. 4, pp. 605-‐
621.

McGuire, R.K. (2004) Seismic Hazard and Risk Analysis. Earthquake Engineering Research
Institute.

MunichRe (2012). Topics Geo: Natural catastrophes 2011 Analyses, assessments,
positions. 2012 Issue, Munich, Germany.

79
Orsini G. (1999) “A model for buildings’ vulnerability assessment using the parameterless
scale of seismic intensity (PSI),” Earthquake Spectra, 15(3), pp. 463–83

Park, Y.J. and Ang, A.H.S. (1985). “Mechanistic seismic damage model for reinforced
concrete”, Journal of Structural Engineering, ASCE, Vol. 111, No. 4, pp. 722-‐739.

Paulay, T., Priestley, M. J. N. (1992). Seismic Design of Reinforced Concrete and
Masonry Buildings. John Wiley and Sons, Inc., New York, USA.

Pinho, R., Bommer, J. J., Glaister, S. (2002). “A simplified approach to displacement-‐
based earthquake loss estimation analysis”. Proceedings of the 12th European
Conference on Earthquake Engineering, London, England.

PRB (2010). World population data sheet. Population Reference Bureau,
Washington D.C., USA.

Priestley, M. J. N., Calvi, G. M., Kowalsky, M. J. (2007). Displacement-‐based Seismic
Design of Structures. IUSS Press, Pavia, Italy.

Reiter, L. (1990) Earthquake Hazard Analysis: Issues and Insights. Columbia University Press,
New York.

Rezaeian, S., Bozorgnia, Y., Idriss, I. M., Campbell, K., Abrahamson, N., Silva, W.
(2012). Spectral Damping Scaling Factors for Shallow Crustal Earthquakes in
Active Tectonic Regions. PEER Report no 2012/01, Pacific Earthquake
Engineering Research Center, California, USA.

Rossetto, T. Elnashai, A. (2003) “Derivation of vulnerability functions for European-‐type RC
structures based on observational data,” Engineering Structures, Vol. 25, pp. 1241-‐
1263.

Rossetto, T., Elnashai, A. (2005) “A new analytical procedure for the derivation of
displacement-‐based vulnerability curves for populations of RC structures”.
Engineering Structures 27(3):397-‐409

Rota, M., Penna, A., Strobbia, C.L. (2008) “Processing Italian damage data to derive
typological fragility curves,” Soil Dynamics and Earthquake Engineering, Vol. 28, No.
10-‐11, pp. 933-‐947

Salvatore, M., Pozzi, F., Ataman, E., Huddleston, B. and Bloise, M. (2005) “Mapping global
urban and rural population distributions,” Food and Agriculture Organization of the
United Nations, Rome, 2005. Available from URL:
http://www.fao.org/docrep/009/a0310e/A0310E00.htm#TOC

Sarabandi, P., Pachakis, D., King, S., Kiremidjian, A. (2004) “Empirical fragility functions from
recent earthquakes,” Proceedings of 13th World Conference on Earthquake Engineering,
Vancouver, Canada.

80

Silva, V., Crowley, H., Pinho, R., Varum, H. (2013) “Extending Displacement-‐Based Earthquake
Loss Assessment (DBELA) for the Computation of Fragility Curves”. Engineering
Structures, 56:343-‐356.

Silva, V., Crowley, H., Pinho, R., Varum, H. (2014) “Seismic Risk Assessment for mainland
Portugal”. Bulletin of Earthquake Engineering, 13(2):429-‐457.

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”. Bulletin of Earthquake Engineering, DOI 10.1007/s10518-‐014-‐9669-‐y.

Silva, V., Crowley, H., Pagani, M., Monelli, D., Pinho, R. (2014) “Development of the
OpenQuake engine, the Global Earthquake Model’s open-‐source software for seismic
risk assessment”. Natural Hazards, 72(3):1409-‐1427.

Singhal, A. and Kiremidjian, A.S. (1996) “Method for probabilistic evaluation of seismic
structural damage,” Journal of Structural Engineering, ASCE, Vol. 122, No. 12, pp. 1459-‐
1467

So, E. (2009) The assessment of casualties for earthquake loss estimation, PhD Thesis,
University of Cambridge, UK.

Spence, R. Ed (2007) Earthquake disaster scenario predictions and loss modelling for urban
areas, LESSLOSS Report 7, IUSS Press, Pavia, Italy.

UNISDR (2009). “Global Assessment Report on Disaster Risk Reduction 2009”.
Report, United Nations International Strategy for Disaster Reduction
Secretariat, Geneva, Switzerland.

Vamvatsikos, D., Cornell, A. C. (2002). “Incremental dynamic analysis”. Earthquake
Engineering & Structural Dynamics, 31(3):491-‐514.

Vamvatsikos D, Cornell CA (2006) Direct estimation of the seismic demand and capacity of
oscillators with multilinear static pushovers through Incremental Dynamic Analysis,
earthquake Engineering and Structural Dynamics 35(9):1097–1117.

Wald, D., Quitoriano, V., Heaton, T., Kanamori, H., Scrivner, C., Worden, B. (1999) "TriNet
ShakeMaps: Rapid generation of peak ground motion and intensity maps for
earthquakes in southern California," Earthquake Spectra, Vol. 15 (N° 3), pp. 537-‐555.

Wasserman, L. (2004) All of Statistics: A Concise Course on Statistical Inference.
Springer, New York, USA.

Whitman, R.V., Reed, J.W. and Hong, S.T. (1973) “Earthquake Damage Probability Matrices,”
Proceedings of 5th World Conference on Earthquake Engineering.
81

Vitor Silva - Seismic Vulnerability Course Notes

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Vitor Silva - Seismic Vulnerability Course Notes

Uploaded by

Copyright:

Available Formats

Seismic Vulnerability Assessment

7th of May

CHAPTER 2 BASIC CONCEPTS OF SEISMIC HAZARD ..................................................... 6

CHAPTER 3 BASIC CONCEPTS OF EXPOSURE MODELLING ...................................... 21

CHAPTER 4 SEISMIC VULNERABILITY ASSESSMENT ................................................. 28

CHAPTER 5 BASIC CONCEPTS OF SEISMIC RISK .......................................................... 68

1.2. The importance of Seismic Vulnerability

that rely on numerical models to simulate the seismic performance of structures

SURFACE FAULTING TSUNAMIS

The various phenomena caused by earthquakes will produce damage in different

2.1. Ground Shaking Intensity Measures

2.1.1. Observation-­‐based Intensity

2.1.2. Instrumental Intensity

2.1.2.1. Peak Ground Acceleration

2.1.2.1. Arias Intensity

where g is acceleration due to gravity and Td is the duration of the accelerogram

Figure 2.9. Representation of cumulative Arias Intensity versus time.

2.1.2.2. Significant Durations

2.1.2.3. Pseudo-­‐spectral acceleration

2.2. Ground Motion and Intensity Prediction

log 𝑦!" = 𝜇 𝑚!" , 𝑟!" , 𝜃!" + 𝛿!" + 𝛿!,!!

ShakeMap is a product of the United States Geological Survey (USGS) Earthquake

2.4. Ground Motion Fields

−78˚ −76˚ −74˚ −72˚ −70˚

0.0 0.1 0.2 0.3 0.4

Jayaram and Baker (2009), which is based on spectral acceleration. Spatial

−78˚ −76˚ −74˚ −72˚ −70˚

0.0 0.1 0.2 0.3 0.4

2.5. Probabilistic Seismic Hazard Assessment

3.1. Global exposure modelling

GRUMP v1 built upon GPW to construct a geo-­‐referenced framework of urban and

Figure 3.2. PAGER global building inventory data sources

To improve on the aforementioned PAGER database, the Global Earthquake Model

sources of data to produce a gridded distribution of buildings (counts, area,

3.2. Regional exposure modelling

3.3. The importance of Building Taxonomies

Currently treats non-­‐buildings. Built forms other than buildings need to be

International in scope. As far as possible the taxonomy should be appropriate for

3.4. GEM Basic Building Taxonomy

Table 3.1 GEM Basic Building Taxonomy summary table

Material of the Lateral Material type (Level 1)

Roof material (Level 1) Masonry

Height Number of stories above ground 4

Negligible Moderate Substantial Very Heavy Destruction

4.2. Expert Opinion Vulnerability

4. Structural design and construction practice may differ substantially over

each building class, were included in the averaging process, as described in

It might be argued that expert opinion is close to empirical fragility/vulnerability

4.3. Empirical Vulnerability

An empirical global vulnerability model capable of predicting human losses

where φ is the standard normal cumulative distribution function, I is the discrete

Jaiswal and Wald (2011) extended the aforementioned methodology to develop

4.4. Fragility Assessment

4.4.1. Damage States

Damage State Damage Description

4.4.2. Fragility Functions and Damage Probability Matrices

This fragility function can also be represented in terms of a damage probability

Non-­‐ Intensity of Earthquake

0 None None 10.4 -­‐ -­‐ -­‐ -­‐

4.4.3. Expert Opinion Fragility

According to the European Macroseismic Scale (Grunthal, 1998), intensity can be

4.4.4. Empirical Fragility

Nevertheless, when the data is adequately processed and uncertainties are

The following publications are recent attempts to model empirical fragility

2.1.1. Observation-‐based Intensity

2.1.2.3. Pseudo-‐spectral acceleration

GRUMP v1 built upon GPW to construct a geo-‐referenced framework of urban and

Currently treats non-‐buildings. Built forms other than buildings need to be

Non-‐ Intensity of Earthquake

0 None None 10.4 -‐ -‐ -‐ -‐

(through appropriate stress-‐strain relationships) and the cross-‐section of the

2. The representation of the structure as a single-‐degree-‐of-‐freedom system (as

4. Fit the force-‐displacement relationship into an appropriate idealized

5. Convert the fitted force-‐displacement relationship to normalized spectral

4.4.6.2. The Capacity Spectrum Method (HAZUS -‐ FEMA, 2005)

Table 4.4 -‐ Limits for the deformation-‐based sway index.

magnitude or source-‐to-‐site distance (Rezaeian et al., 2012). The integration of such

5.3. PSHA and Event-‐based Risk Assessment

Benefit-‐cost analysis (BCA) is a systemic procedure for evaluating decisions that

Engineering, Vol. 12 (S2), pp. 71-‐80.

2009-‐1136, 78 pp.