Risk Targeting in Seismic Design Codes:

The State of the Art, Outstanding Issues

and Possible Paths Forward

John Douglas and Athanasios Gkimprixis

Abstract Over the past decade there have been various studies on the development
of seismic design maps using the principle of “risk-targeting”. The basis of these
studies is the calculation of the seismic risk by convolution of a seismic hazard
curve for a given location (derived using probabilistic seismic hazard analysis) with
a fragility curve for a code-designed structure (ideally derived from structural
modelling). The ground-motion level that the structure is designed for is chosen so
that the structure has a pre-defined probability of achieving a certain performance
level (e.g. non-collapse). At present, seismic design maps developed using this
approach are only widely applied in practice in the US but studies have also been
conducted on a national basis for France, Romania, Canada and Indonesia, as well
as for the whole of Europe using the European Seismic Hazard Model. This short
article presents a review of the state of the art of this technique, highlighting efforts
to constrain better some of the input parameters. In addition, we discuss the diffi-
culties of applying this method in practice as well as possible paths forward,
including an empirical method to estimate an upper bound for the acceptable col-
lapse and yield risk.

Keywords Seismic hazard Earthquake engineering
Fragility curves



Risk targeting Design Acceptable risk

J. Douglas (&)
A. Gkimprixis

Department of Civil and Environmental Engineering,
University of Strathclyde, Glasgow, UK
e-mail: john.douglas@strath.ac.uk
A. Gkimprixis
e-mail: athanasios.gkimprixis@strath.ac.uk

© Springer International Publishing AG, part of Springer Nature 2018 211

R. Vacareanu and C. Ionescu (eds.), Seismic Hazard and Risk Assessment,
Springer Natural Hazards, https://doi.org/10.1007/978-3-319-74724-8_14
212 J. Douglas and A. Gkimprixis

1 Introduction

Current seismic building codes (e.g. Eurocode 8), based on results from a proba-
bilistic seismic hazard analysis (PSHA), generally adopt a constant hazard approach
to define the ground motions used for design. In other words, the peak ground
acceleration (PGA, or other intensity measure, IM, e.g. spectral acceleration) used
for design in one location has the same probability of being exceeded in a given
year as the design PGA in another location. Often this annual probability is
1/475 = 0.0021 (equivalent to 10% in 50 years or a return period of 475 years
assuming a Poisson process). Ten years ago, Luco et al. (2007) proposed a new
approach that targets a constant risk level across a territory. This has three principal
advantages over the use of design levels defined in the traditional way: trans-
parency, a uniform risk level across a territory and the ability to compare (and
ideally control) risk for different types of hazard (e.g. earthquake and wind). It does
come, however, with the disadvantage of making more choices explicit, rather than
implicitly assumed because of convention (e.g. the choice of 475 years as the
design return period).
The procedure of Luco et al. (2007), although often using different input
parameters (see below), has been applied to France (Douglas et al. 2013), Romania
(Vacareanu et al. 2017), Indonesia (SNI 2012), Canada (Allen et al. 2015) and at a
European scale (Silva et al. 2016), as well as forming the basis of the current US
seismic design code (ASCE 2010). Despite its numerous attractions (see above) and
the fact that it is a relatively simple procedure to implement, there are a number of
outstanding issues. For example, Douglas et al. (2013) note that the collapse
probabilities targeted by Luco et al. (2007) appear to be at least an order of mag-
nitude too high when compared with observed damage in previous earthquakes
(also see Sect. 3.1 of this article).
The next section presents an overview of the risk targeting approach and dis-
cusses previous choices of the critical input parameters. Section 3 highlights the
outstanding problems and some potential solutions, which are currently being
investigated by the authors.

2 Method and Required Inputs to Risk Targeting

The risk of collapse (or other level of structural damage) of a building at a given site
from earthquake shaking can be estimated by convolving the seismic hazard curve,
expressing the probability of different levels of ground motion, with the fragility
curve, expressing the probability of collapse given these ground motions (e.g.
Kennedy 2011). This so-called “risk integral” forms the basis of the risk-targeting
approach. For this approach there needs to be a link between the design acceleration
and the fragility curve used to compute the risk of collapse. For a standard fragility
curve based on the lognormal distribution, a single point on this curve (if the
Risk Targeting in Seismic Design Codes … 213

standard deviation is fixed) is required to define the building’s fragility completely.

A convenient choice is to use the design IM and the corresponding probability of a
building attaining the considered damage state when subjected to that IM.
The general procedure for finding the design value for the considered IM (e.g.
PGA) is shown in Fig. 1. The key input parameters, using the nomenclature of
Douglas et al. (2013), are: b (the standard deviation of the fragility curve assuming
a lognormal distribution), X (the probability of collapse at the design IM) and Y
(the targeted annual probability of collapse). Seismic design codes generally do not
report these values and hence assessing them has been the focus of considerable
efforts over the past decade (see following section).
There is a closed-form solution for the risk given a hazard curve expressed as a
power law and a lognormal fragility curve (e.g. Kennedy 2011), which can be used
to understand the influence of different parameters on the design IM. For typical
hazard curves, however, an iterative technique is required to determine the design
IM (Fig. 1). We have found that a bisection method, which bounds the targeted risk
from above and below until convergence to a given tolerance, is the best approach.
To compute the convolution, numerical integration of the derivative of the fragility
curve with the hazard curve using the trapezium rule works well.
The seismic hazard curve used within the calculation needs to be defined down
to potentially very low probabilities of exceedance because the probabilities of
collapse (or another damage state) defined by the fragility function are often far
from one for large accelerations. This means that the hazard curve may need
extrapolation, for which the power-law expression (based on the IMs for the
smallest calculated probabilities) works well for most examples tested to date. We
validated our algorithm by comparing our results with those from the Risk-Targeted
Ground Motion Calculator available on the USGS website (https://earthquake.usgs.
gov/hazards/designmaps/rtgm.php).

Fig. 1 Flowchart of the method to find risk-targeted design parameters. l is the mean of the
fragility curve assuming a lognormal distribution
214 J. Douglas and A. Gkimprixis

2.1 The State of the Art

A summary from the literature of the three key inputs to the procedure is given in
this section (we are assuming that the seismic hazard curve has been correctly
defined by a recent PSHA for a wide range of probabilities of exceedance and that
lognormal fragility curves are used).
ASCE (2005) provides design criteria for nuclear power plants and critical
facilities based on a risk-targeting framework. The targeted risk ranges from 10−5 to
10−4, while for lower design categories values up to 10−3 are given in Braverman
et al. (2007). To comply with the code, a nuclear plant must have a smaller than 1%
chance of unacceptable performance for the design IM and less than 10% for 1.5
times the design IM.
Kennedy (2011) comments on the ASCE (2005) approach and implements it for 28
US nuclear plants. With a b in the range 0.3–0.6, the Seismic Core Damage
Frequencies (SCDFs) are between 6  10−7 and 6  10−6. Based on this, the United
States Nuclear Regulatory Commission (2007) has set a target SCDF of 10−5.
Luco et al. (2007) use generic fragility curves with b = 0.8. They assume a
probability of collapse of 10% under the ground motion corresponding to a
2475-year return period. The ATC-63 project initially proposed this value after
analysing buildings designed with ASCE 7-05. Luco et al. (2007) find that the 2003
NEHRP design ground motions led to non-uniform risk across US territory. By
targeting the average collapse probability in 50 years, which they found to be
around 1%, Luco et al. (2007) calculated new design ground motions with ratios of
0.7–1.15 of the 2003 NEHRP design ground motions.
The results of that report were considered for the ASCE 7-10 and FEMA (2009)
NEHRP Provisions. The latter propose the convolution of a hazard curve with a
fragility curve with b = 0.8 (decreased to 0.6 in ASCE 7-10), X = 0.1 and a target
probability of collapse of 1% in 50 years. Liel et al. (2015) comment on ASCE 7-10
and suggest modifications to consider subduction earthquakes and near-fault effects.
With the same X and b as in the regulation, the probability of collapse in 50 years
varied from 0.21 to 0.62%, for areas affected by subduction earthquakes, while at
near-fault sites, the risk was much higher, reaching in some cases 6%.
Allen et al. (2015) comment on the implementation of risk-targeting in Canada
for future versions of its National Building Code (NBCC). For comparison, ground
motion values derived from the risk-targeting approach of Luco et al. (2007) are
divided with those proposed in 2015 NBCC (for a 2%-in-50-year exceedance
probability). The resulting risk coefficients for spectral accelerations at 1.0 s are
between 0.84 and 1.00 in southwest Canada.
In Goulet et al. (2007) the collapse probabilities for a four-storey
reinforced-concrete (RC) frame under the 2%-in-50-year ground motion were cal-
culated as between 0 and 2%; after considering structural uncertainties the values
increased to 2–7%. They also found a variation of Y between 0.1  10−4 and
0.5  10−4 and between 0.4  10−4 and 1.4  10−4, when they considered struc-
tural uncertainties.
Risk Targeting in Seismic Design Codes … 215

Fajfar and Dolšek (2012) examined a three-storey RC building with no provi-

sions for earthquake resistance. They found an annual probability of collapse equal
to 0.65  10−2. When designed with Eurocode 8 (CEN 2004b), this risk reduced to
2.22  10−4 and to 2.7  10−4 when they accounted for epistemic uncertainties.
Ramirez et al. (2012) examine 30 buildings designed using the 2003
International Building Code (IBC) together with the ASCE provisions. They find
the probabilities of collapse at the design PGA to be in the range 0.4–4.2% for these
buildings.
Douglas et al. (2013) try to avoid significant changes in the existing design
ground motions for France, which correspond to a 475-year return period.
Considering previous studies, as well as the results of some sensitivity analyses,
they finally choose b = 0.5 and X = Y = 10−5. Under these assumptions, the
estimated risk-targeted design PGAs are not very different from the values proposed
by the current French code.
Ulrich et al. (2014a) present fragility curves for three-storey RC buildings
designed according to Eurocodes 2 and 8 (CEN 2004a, b) for different levels of
design PGA, ag. They find yielding probabilities between 0.14 (for ag = 0.07 g) and
0.85 (for ag = 0.3 g) and in the range 1.7  10−7 (ag = 0.07 g) to 1.0  10−5
(ag = 0.3 g) for the probabilities of collapse at the design PGA.
In Martins et al. (2015) fragility curves are also derived, firstly considering the
spectral acceleration at the fundamental period of vibration Sa(T1). This results in b
ranging from 0.35 to 0.45, which increased up to 0.8 when they used PGA instead.
For a three-storey RC building, they found the X to be equal to 5.20  10−3 and
2.21  10−2 for design accelerations of 0.2 and 0.4 g, respectively, which change
to 3.95  10−3 and 5.57  10−2 for a five-storey RC building. Then, using Sa(T1)
as the IM, the annual probabilities of collapse for the three-storey buildings were
9.50  10−5, 1.67  10−5 and 1.07  10−5, and for the five-storey buildings
1.78  10−4, 7.34  10−5 and 2.97  10−5, for design accelerations of 0.0, 0.2
and 0.4 g, respectively.
Likewise, the annual rates of collapse for the Italian territory are investigated by
Iervolino et al. (2017). The current version of the Italian seismic code was used for
the design of regular three-, six-, and nine-story RC buildings for two soil condi-
tions. The results of their analyses indicate that annual collapse risk for
code-conforming RC structures is between 10−5 and 2.02  10−4. For the case of
one-storey precast RC structures (also designed with the seismic code) Y values are
found between 10−5 and 6.35  10−3. The reader is referred to RELUIS and
EUCENTRE (2014–2018) for details of this ongoing project.
Silva et al. (2016) study the fragility curves derived by the European project
Syner-G and find an average value of b = 0.5, which they consider as a lower bound.
Assuming a 10% probability of collapse at the 2475-year ground motion and the
hazard curves derived by the European project SHARE (www.share-eu.org), for two
values of b (0.6 and 0.8) they find a probability of collapse at the 475-year design
acceleration ranging from 10−3 to 10−2. Choosing then a value of X = 10−3 and Y =
5.0  10−5 (by relating the risk of collapse to human losses), they propose new
design maps for Europe.
216 J. Douglas and A. Gkimprixis

Following a different approach, Tsang and Wenzel (2016) firstly define an

acceptable fatality risk. Based on a literature review they choose a value of 10−6 for
the acceptable annual fatality rate. They then estimate the corresponding limit for
the targeted annual risk of complete structural damage as roughly 10−4. For a
different limit state, the annual risk of “Real collapse” falls within the range
6  10−6 to 8  10−6.
Tsang et al. (2017) investigate buildings with precast RC columns designed
using the risk-targeted ground motions (MCER) proposed in IBC-2012 (ICC IBC
2012) and ASCE/SEI 7-10. Theoretically, a 10% probability of collapse under the
MCER in 50 years is expected. They estimate, however, probabilities that are in
every case lower than 10% (a maximum of 6.1% was found). In addition, the
average annual collapse risk was estimated as 2.5  10−6 (with a maximum of
1.6  10−4), while the regulation imposes 2  10−4. Also, the value of 0.25% in a
design life of 50 years (5  10−5 annually), as proposed in Silva et al. (2016), was
in some cases exceeded. Judd and Charney (2014) state that “the assumed ASCE
7-10 fragility curve is conservative” and that “the conditional probability of col-
lapse may exceed 10%”.
In Vacareanu et al. (2017) the proposed values of Luco et al. (2007) and Silva
et al. (2016) are tested for Romania. Compared to previous uniform hazard maps,
the risk distribution changes significantly when targeting uniform risk. Vacareanu
et al. (2017) calculate the ratios between the design PGA for a mean return period
of 475 years and those resulting from risk targeting for Y = 2  10−4, b = 0.8 and
X = 0.1 and 0.001. For the lower X, the ratios were below 0.6 whilst for the higher
X they were larger than 1.0. Finally, considering two mean return periods, 475 and
2475 years, new risk maps are derived using X = 0.1 and 0.001. Based on the
distribution of risk over the country, they conclude that using X = 0.001, as pro-
posed in Silva et al. (2016), leads to a less realistic distribution of Y.
Following US practice, the philosophy of risk targeting has been followed by the
new Indonesian Earthquake Resistance Building Code SNI 1726-2012 (SNI 2012).
By setting a target of 1% for the probability of collapse in 50 years, this code maps
risk-targeted spectral response accelerations for 0.2 and 1.0 s. The generic fragility
curves follow a lognormal distribution with X = 0.1. Considering the material
properties and human-related parameters representative of the broader area of
Indonesia, Sengara et al. (2016) report values of b around 0.7, which is adopted by
SNI 1726-2012.

3 Outstanding Issues and Possible Paths Forward

As can be seen by Sect. 2 there is a rapidly increasing number of studies that have
attempted to apply the risk-targeting approach to different areas or sought to con-
strain the various inputs upon which this technique relies. Despite these many
studies a number of problems remain, which we highlight in this section along with
a brief discussion of potential solutions.
Risk Targeting in Seismic Design Codes … 217

As noted above, when evaluating the risk integral, ground motions for very low
probabilities of exceedance sometimes need to be estimated. A power-law
extrapolation appears to work well for most examples we have studied to date.
Further research on the technicalities of the calculations, however, needs to be
conducted to define a stable procedure that works for all possible inputs.
Targeting a non-zero value of collapse risk accepts that some buildings will
collapse (potentially leading to human casualties) in earthquakes, even when they
are designed in compliance with the building code. From a moral point of view this
is problematic and it leads to the difficulty of trying to define what risk is “ac-
ceptable”. One potential solution to this is to estimate an upper bound on the risk
that has been “accepted” historically based on the levels of observed damage in
previous earthquakes, as attempted by Labbé (2010). This “risk” is an upper bound,
as generally after every damaging earthquake the population lament the damage
that occurred. In the following Sect. (3.1), some preliminary assessments using this
technique are presented. Another potential solution to this problem is to target a
damage state that is less severe than collapse, for example structural yielding.
Targeting this limit state is less morally problematic and the level of acceptable risk
in this context could be defined using, for example, cost-benefit analysis based on
the cost of reducing the risk further (perhaps adopting an ALARP “as low as
reasonably practicable” philosophy). This choice also has other benefits, e.g. it is
generally easier to assess in numerical modelling when a structure yields rather than
when it collapses.
Once the targeted risk is chosen, unless great changes to the accelerations cur-
rently used for design are accepted by practicing engineers, the probability of
collapse at the design acceleration is automatically implied, as shown by Fig. 3 of
Douglas et al. (2013). Here there is a trilemma: any two out of the three input
parameters, design IM, X and Y, can be chosen independently but not all three.
To make this point clearer, we can use the work of Žižmond and Dolšek (2017).
In their article, they follow a risk-targeting approach for the derivation of the q
factor for use in force-based design using Eurocode 8 (CEN 2004b). In Eq. 4 of that
paper an analytical formula is presented for the median acceleration of the
risk-targeted lognormal fragility curve, l. Adapting it to our nomenclature and
recalling that our design acceleration ag implies a probability of collapse of X, we
have the following equations, which show that the various input parameters are
connected:

k2
b2
!1k
k0
e 2
l¼ ð1Þ
Y

ag ¼ lognormal1 ðX; l; bÞ ð2Þ

where ag is the risk-targeted design acceleration and k0, k are parameters that define
the idealized hazard curve. The solution to this potential problem is to check all
three values are physically reasonable. This trilemma may be the reason for the
218 J. Douglas and A. Gkimprixis

apparently high target collapse probabilities (Y) used by Luco et al. (2007), as they
were forced to adopt them once X had been defined and they did not want the
design IM to change greatly from the previous code.
There is a need to derive fragility curves for a wide variety of code-designed
structures with different geometries and materials. Previous studies have adopted
generic fragility curves that scale constantly with the design acceleration so that the
iterative procedure used to converge to the targeted risk is simple. As shown by
Fig. 3 of Ulrich et al. (2014b), however, this desired feature appears not to be true
following current design practice, as they were not created with risk targeting in
mind. It is possible, however, to generate a suite of fragility curves for all potential
design accelerations and then to use the appropriate ones when iterating to find the
actual design acceleration for a location.
Two issues concerning the used fragility functions are: what value of b to use
and, indeed, whether the lognormal distribution should continue to be used at all?
Fragility curves that are used within the risk targeting calculations can imply very
strong buildings (e.g. Table 1 of Douglas et al. 2013) as well as non-negligible
chances of collapse for very low ground accelerations. This is a consequence of the
high values of b that need to be used to account for different types of structures of
varying geometries. Two solutions to this problem are: (a) adopt a different func-
tional form for the fragility curve that equals zero for low ground accelerations and
unity for very high accelerations, or (b) move from generic fragility curves for all
types and geometries to a curve covering only a small set of structures. The second
of these changes would mean a change of philosophy of design codes from being
associated with a single map giving the design accelerations for all structures to
potentially many maps giving design accelerations for different structural types and
geometries. This additional complexity, however, appears to be necessary if the
risk-targeting approach is to imply physically realistic buildings and levels of risk.

3.1 Assessing Risk Using Earthquake Damage Databases

In this section, we use field observations of building damage from recent earth-
quakes and an empirical Fermi-type approach to estimate an upper bound on the
“acceptable” probability of collapse (Y) discussed above. In particular, the number
of RC buildings that collapsed due to earthquakes during recent time periods in
Italy and Greece are used to estimate the observed annual collapse rate.
The technique relies on the ratio of the total number of observed cases of a given
damage level in a given period of time to the total number of buildings that could
have been affected by earthquakes, which is then normalized to give an annual rate.
Estimating both of these totals for a given country requires: (a) post-earthquake
field mission reports that provide the number of cases of damage and (b) building
census data indicating how many structures were present when the earthquake
occurred.
Risk Targeting in Seismic Design Codes … 219

There are a number of difficulties and uncertainties in applying this approach.

For example, in some reports the number of collapses of RC buildings is combined
with the number of collapses of other structural types, e.g. masonry. In others only a
general “damage” category is used without specifying, for example, the number of
collapses or partial collapses. Sometimes the focus of post-earthquake field mis-
sions is on estimating the number of casualties and not on collecting accurate
building damage statistics. Also damage records may be incomplete, either by not
including all damaging earthquakes during a time period or by not totalling all
incidences of damage for a given earthquake (e.g. damage within cities but not rural
areas). For example, Colombi et al. (2008), when deriving empirical vulnerability
curves from Italian data, state that roughly half of the available observations cannot
be used because of a lack of information on the structural type and level of damage.
In addition, when combining data from different databases it is necessary to assume
equivalence between various damage scales (or convert between scales). Finally
(and probably most important), the sample sizes available are small, both in terms
of the number of collapsed buildings and the small number of possible earthquake
scenarios (locations and magnitudes) that have been sampled during the time period
considered, which is relatively short given the recurrence interval of large earth-
quakes. Therefore, the annual collapse rates estimated here should be considered
very rough.
We conduct a preliminary analysis for Italy and Greece, for which the earth-
quake damage databases are roughly complete in recent years and easily accessible,
and for a combined “total or partial collapse” damage state for RC buildings. It
would be relatively simple to conduct such an analysis to other developed countries,
such as the USA and Japan.
The Cambridge Earthquake Impact Database (CEQID) (http://www.ceqid.org/
CEQID/Home.aspx) reports the number of RC building collapses in Italian earth-
quakes from 1980 to 2009 inclusive (30 years) as: Irpinia 1980 (58 buildings),
Eastern Sicily 1990 (3), Umbria-Marche 1997 (49), Umbria-Marche 1998 (0),
Pollino 1998 (9), Molise 2002 (29) and L’Aquila 2009 (57), giving a total of 205
total or partial collapses. Colombi et al. (2008) gives an estimate of 842 collapses
for almost the same period range, thereby demonstrating large uncertainty in this
total.
Next it is necessary to estimate the total number of RC buildings in Italy.
EPISCOPE (2014) estimates that 24.7% of Italian residential buildings are of RC
and Istat (2011, http://dati-censimentopopolazione.istat.it) states that there were
14,452,680 buildings in Italy of which 12,187,698 are residential. Therefore, it can
be estimated that there are roughly 3.5 million RC buildings in total of which about
3 million are residential. We assume that this total has not changed over the
30 years covered by the collapse database, which given the slow population growth
of Italy could be thought a reasonable assumption.
Dividing the total number of collapsed buildings by the building population and
the number of years gives an estimate of the collapse rate. Using the two estimates
of 205 and 842 in 30 and 23 years, respectively, and 3 or 3.5 million buildings
(depending on whether the database covers all buildings or only residential) gives
220 J. Douglas and A. Gkimprixis

annual collapse rates of between 2  10−6 and 1  10−5, which are much lower
than some of the probabilities assumed in previous risk-targeting exercises (e.g.
Luco et al. 2007).
For the period 1978–2003 inclusive the following RC building collapses due to
earthquakes in Greece have been reported: Thessaloniki 1978 (4 buildings)
(CEQID), Gulf of Corinth 1981 (15) (Carydis et al. 1982), Kalamata 1986 (5),
Aigion 1995 (1), Athens 1999 (69) (CEQID) and Lefkada 2003 (0) (Karababa and
Pomonis 2011), giving a total of 94 total or partial collapses. Using information
from the Greek statistical service (http://www.statistics.gr) and the TABULA pro-
ject (www.building-typology.eu) for the total number of RC buildings in Greece,
and assuming that the building total remained constant, leads to estimates of the
collapse rate between 1  10−6 and 2  10−6, again much lower than some
assumed probabilities for the targeted risk.
The accuracy of these estimates could be improved by using a more complete
database both for the numerator (the damage) and the denominator (the number of
buildings at risk) and by extending the time period covered so that more potential
earthquake scenarios are considered. Nevertheless, such an analysis is constrained
by the long recurrence intervals of damaging earthquakes, which means that the
assessed rates will always be associated with large uncertainties. In addition,
variations due to building type (e.g. masonry versus reinforced concrete), location
(e.g. developed versus less-developed countries) and damage level (e.g. collapse
versus yield) could be studied as well.
The same procedure was repeated for the same two countries considering a
different damage state. A broader class was defined to represent ‘yield’, as con-
sidered by Ulrich et al. (2014a, b), which includes the total number of RC buildings
that are neither not-damaged nor collapsed (partially or in total). The same limi-
tations discussed before are also the case here. In addition, complete data were
available for a shorter period of 18 years for Greece (but still 30 years for Italy).
With these assumptions, the annual percentage of yield is found equal to 3  10−5
for Italy and 1  10−4 for Greece. Apart from providing useful constraints on the
inputs to risk-targeting this type of analysis could have other benefits as well.
Firstly, it would provide an observational-based assessment of earthquake risk that
can be compared to estimates from computer modelling, which require a large
number of inputs and can be opaque to decision makers. If there are large differ-
ences in these estimates it may indicate that the computer models require calibra-
tion. Secondly, these estimates will enable differences in earthquake risk levels
between countries to be judged more easily and used to understand what the impact
of reducing risk could be in terms of the average number of buildings that could
collapse per year, for example.
Risk Targeting in Seismic Design Codes … 221

4 Conclusions

Crowley et al. (2013) propose that ‘Risk-targeted seismic design actions’ should be
considered for future versions of the Eurocodes. A call echoed by Formichi et al.
(2016) in a report on the background and application of Eurocodes, published on
behalf of the European Commission. This report also proposes that changes made in
other international seismic design codes should be considered when updating the
Eurocodes. It is, therefore, clear that application of the risk-targeting approach is
being seriously discussed in Europe. In consequence, additional research effort to
this end would provide valuable input for the development of risk-targeted design
maps for new buildings in Europe.
In this brief article, we have highlighted the critical issues that we believe need
to be solved before the risk-targeting approach for the development of seismic
design codes can be employed in practice. Some of these (e.g. development of
appropriate fragility functions) solely require engineering calculations but others
(e.g. choice of the acceptable level of risk) need input from other domains,
including decision makers. In the coming years we plan to tackle these issues,
particularly with respect to future Eurocodes.

Acknowledgements The second author of this article is undertaking a PhD funded by a

University of Strathclyde “Engineering The Future” studentship, for which we are grateful. We
thank Florin Pavel for sharing the submitted version of Vacareanu et al. (2017). We thank Roberto
Paolucci for discussions concerning empirical estimates of collapse risk.

References

Allen TI, Adams J, Halchuk S (2015) The seismic hazard model for Canada: past, present and
future. In: Proceedings of the tenth pacific conference on earthquake engineering building an
earthquake-resilient pacific, Sydney, Australia (Paper number 100)
ASCE (2005) Seismic design criteria for structures, systems and components in nuclear facilities.
American Society of Civil Engineers, ASCE Standard 43-05
ASCE (2010) Minimum design loads for buildings and other structures. ASCE Standard 7-10.
American Society of Civil Engineers, Reston, VA
Braverman JI, Xu J, Ellingwood BR, Costantino CJ, Morante RJ, Hofmayer CH (2007) Evaluation
of the seismic design criteria in ASCE/SEI standard 43-05 for application to nuclear power
plants. Agencywide Documents Access and Management System (ADAMS)—USNRC
Carydis PG, Tilford NR, Brandow GE, Jirsa JO (1982) The central Greece earthquakes of
February–March 1981. Report No. CETS-CND-018 of the Earthquake Engineering Research
Institute, Berkeley, CA, National Academy Press, Washington, DC
CEN (2004a) EN 1992-1-1:2004 Eurocode 2: design of concrete structures—Part 1-1: general
rules and rules for buildings. European Committee for Standardization, Brussels
CEN (2004b) EN 1998-1:2004 Eurocode 8: design of structures for earthquake resistance—Part 1:
general rules, seismic actions and rules for buildings. European Committee for Standardization,
Brussels
222 J. Douglas and A. Gkimprixis

Colombi M, Borzi B, Crowley H, Onida M, Meroni F, Pinho R (2008) Deriving vulnerability

curves using Italian earthquake damage data. Bull Earthq Eng 6(3):485–504. https://doi.org/10.
1007/s10518-008-9073-6
Crowley H, Weatherill G, Pinho R (2013) Suggestions for updates to the European seismic design
regulations. SHARE Deliverable 2.6. Universita degli Studi di Pavia
Douglas J, Ulrich T, Negulescu C (2013) Risk-targeted seismic design maps for mainland France.
Nat Hazards 65(3):1999–2013
EPISCOPE (2014) Inclusion of new buildings in residential building typologies: steps towards
NZEBs exemplified for different European countries. EPISCOPE Synthesis Report no.
1 (Deliverable D2.4). Contract N°: IEE/12/695/SI2.644739. ISBN 978-3-941140-42-4
Fajfar P, Dolšek M (2012) A practice-oriented estimation of the failure probability of building
structures. Earthq Eng Struct Dynam 41:531–547
FEMA (2009) NEHRP recommended seismic provisions for new buildings and other structures
(FEMA P750). Federal Emergency Management Agency
Formichi P, Danciu L, Akkar S, Kale O, Malakatas N, Croce P, Nikolov D, Gocheva A,
Luechinger P, Fardis M, Yakut A, Apostolska R, Sousa ML, Dimova S, Pinto A (2016)
Eurocodes: background and applications: elaboration of maps for climatic and seismic actions
for structural design with the Eurocodes. EUR, https://doi.org/10.2788/534912
Goulet C, Haselton C, Mitrani-Reiser J, Beck J, Deierlein G, Porter K, Stewart J (2007) Evaluation
of the seismic performance of a code-conforming reinforced-concrete frame building—from
seismic hazard to collapse safety and economic losses. Earthq Eng Struct Dynam 36:
1973–1997
ICC IBC (2003) International building code (IBC), international code council (ICC), IL, USA
ICC IBC (2012) International building code (IBC), international code council (ICC), IL, USA
Iervolino I, Spillatura A, Bazzurro P (2017) RINTC project: Assessing the (implicit) seismic risk
of code-conforming structures in Italy. In: Papadrakakis M, Fragiadakis M (eds) COMPDYN
2017, 6th ECCOMAS thematic conference on computational methods in structural dynamics
and earthquake engineering, Rhodes, Greece
Judd JP, Charney FA (2014) Earthquake risk analysis of structures. In: Proceedings of the 9th
international conference on structural dynamics, EURODYN 2014, pp 2929–2938
Karababa FS, Pomonis A (2011) Damage data analysis and vulnerability estimation following the
August 14, 2003 Lefkada Island, Greece, earthquake. Bull Earthq Eng 9:1015–1046. https://
doi.org/10.1007/s10518-010-9231-5
Kennedy RP (2011) Performance-goal based (risk informed) approach for establishing the SSE site
specific response spectrum for future nuclear power plants. Nucl Eng Des 241:648–656
Labbé PB (2010) PSHA outputs versus historical seismicity: example of France. In: Proceedings
of fourteenth European conference on earthquake engineering
Liel AB, Luco N, Raghunandan M, Champion CP (2015) Modifications to risk-targeted seismic
design maps for subduction and near-fault hazards. In: 12th international conference on
applications of statistics and probability in civil engineering, ICASP12, Vancouver, Canada
Luco N, Ellingwood BR, Hamburger RO, Hooper JD, Kimball JK, Kircher CA (2007)
Risk-targeted versus current seismic design maps for the conterminous United States. In:
SEAOC 2007 convention proceedings
Martins L, Silva V, Crowley H, Bazzurro P, Marques M (2015) Investigation of structural fragility
for risk-targeted hazard assessment. In: 12th international conference on applications of
statistics and probability in civil engineering, ICASP12, Vancouver, Canada
Ramirez CM, Liel AB, Mitrani-Reiser J, Haselton CB, Spear AD, Steiner J, Deierlein GG,
Miranda E (2012) Expected earthquake damage and repair costs in reinforced concrete frame
buildings. Earthq Eng Struct Dynam 41:1455–1475
RELUIS and EUCENTRE (2014–2018) Research project DPC—RELUIS/EUCENTRE, technical
report and deliverables
Sengara IW, Sidi ID, Mulia A, Muhammad A, Daniel H (2016) Development of risk coefficient for
input to new Indonesian seismic building codes. J Eng Technol Sci 48(1):49–65
Risk Targeting in Seismic Design Codes … 223

Silva V, Crowley H, Bazzurro P (2016) Exploring risk-targeted hazard maps for Europe. Earthq
Spectra 32(2):1165–1186
SNI (2012) Tata cara perencanaan ketahanan gempa untuk struktur bangunan gedung dan non
gedung. Badan Standardisasi Nasional, ICS 91.120.25; 91.080.01. Report 1726:2012
(In Indonesian)
Tsang HH, Wenzel F (2016) Setting structural safety requirement for controlling earthquake
mortality risk. Saf Sci 86:174–183
Tsang HH, Lumantarna E, Lam NTK, Wilson JL, Gad E (2017) Annualised collapse risk of
soft-storey building with precast RC columns in Australia. Advancements and Challenges,
Mechanics of Structures and Materials, pp 1681–1686
Ulrich T, Negulescu C, Douglas J (2014a) Fragility curves for risk-targeted seismic design maps.
Bull Earthq Eng 12(4):1479–1491
Ulrich T, Douglas J, Negulescu C (2014b) Seismic risk maps for Eurocode-8 designed buildings.
In: Proceedings of the second European conference on earthquake engineering and seismology
United States Nuclear Regulatory Commission (2007) A performance-based approach to define the
site specific earthquake ground motion. Tech report 1:208
Vacareanu R, Pavel F, Craciun I, Coliba V, Arion C, Aldea A, Neagu C (2017) Risk-targeted maps
for Romania. J Seismolog. https://doi.org/10.1007/s10950-017-9713-x
Žižmond J, Dolšek M (2017) The formulation of risk-targeted behaviour factor and its application
to reinforced concrete buildings. In: Proceedings of the 16th world conference on earthquake
engineering, Santiago, Chile, 9–13 January (Paper no. 1659)