Naturgefahren – Erdbebenrisiko

Nachdiplomkurs in angewandten Erdwissenschaft 15-19 Mai 2000

Seismische Gefährdungsanalyse
-------------------------------------------

Evaluation of earthquake hazard

Souad Sellami

SED, Institute of Geophysics, ETH-Hoenggerberg, CH-8093 Zurich,

[email protected]

Keyword: seismic hazard, probabilistic methods, sources zones, attenuation, frequency-

magnitude distribution, return period, probability of exceedance.

Introduction: Seismic Hazard Objective

The general theme of this course is earthquake risk. The concept of risk includes hazard
and vulnerability. The first part has dealt with earthquakes, where when they occur, how
big they are and why they happen. The second part is about the effects.

Hazard assessment is to evaluate, for a certain place, how frequent and how strong
earthquake will be felt, in order to take measure to reduce the possible damages. In other
terms, it is to qualify and quantify the level of ground motion in a site due to the earthquake.
Seismic hazard maps depict the levels of chosen ground motions that likely will, or will not,
be exceeded in specified exposure times. The ground motion can be the intensity of the
earthquake, displacement, velocity or acceleration of the seismic wave at the site.

Seismic hazard is determined by the following three factors:

• The distribution in time, space and size of the regional seismicity
• The attenuation of seismic waves at increasing distances from the location of the
earthquake
• The action of the shallow geology in the distortion of the seismic signal

The hazard can be estimated using deterministic or probabilistic methods. The probabilistic
method (Cornell 1968), broadly applied, will be first described. Alternative methods will be
presented in the discussion paragraph.

The three major elements of the probabilistic method are: 1) the characterisation of seismic
sources; 2) the characterisation of attenuation of ground motion; and 3) the actual
calculation of probabilities.
Probabilistic method basic principles

The objective of an earthquake-hazard analysis is to evaluate the probability of exceeding a

particular level of ground motion (such as a certain value of peak acceleration) at a site
during a specific time interval (such as 50 years).

The different steps of the probabilistic (Cornell) method are outlined on the sketch (figure
1.) below (Ruettener 95):

Figure 1.
1) the characterisation of seismic sources is usually achieved by
• compilation of an earthquake catalogue (a)
• delineation of the seismic sources (b)
• magnitudes-frequency distribution (c)
2) the characterisation of attenuation of ground motion is described by attenuation
functions (d)
3) computation of the probability analysis (c)

An earthquake-hazard analysis must incorporate the inherent uncertainty of the size,

location, and time of occurrence of future earthquakes, and the attenuation of seismic waves
as they propagate from all possible sources in the region to all possible sites.
Probability functions are required, they include the probability density function for the
magnitude probability FM(m), for the distance FR(r) to earthquake, and for the probability
that the ground motion exceed a certain value at a site given m and r.

Characterisation of seismic sources

The first element of seismic hazard assessment, the characterisation of seismic sources,
involves obtaining ‘robust’ answers to three questions, which have been addressed in the
first part of this course:
• Where do earthquakes occur?
• How often do earthquakes occur?
• How big can we expect these earthquakes to be?

In practice three physical parameters of a potential seismic source must be quantified in a

seismic-hazard analysis:
1. geometry of the source (or fault), (where)
2. rate of earthquake recurrence (how often) and
3. maximum magnitude, (how big).

Seismicity catalogues are the fundamental data base used to determine where, how often,
and how big earthquakes are likely to be. However and seismicity statistics are based on
geologically short catalogues. For magnitude above 6, the completeness is less than 1000
year in Switzerland (one of the best case) and less than 200 years in California). Therefore
other deformation data are examined. The results from seismic monitoring, the historic
record, geodetic monitoring, and the geologic record are combined to characterise seismic
sources. These data, when available, are used to interpret seismic source zones. Because
many interpretations of the input data are possible, large uncertainties are associated with
source characterisation.

Geometry of the source

The identification of the seismogemic sources in the region is very important. In theory
earthquake sources are faults. In most of the places, the earthquake distribution does not
coincide with known fault visible at the earth surface. So in practice, the shape of a source
can be a fault, but they are surfaces (area zones) when active fault can not be recognised,
which is the most common case (for example in Switzerland), The Figure 2 (Rutenner 1995)
shows the historical seismicicity map (time span 1300-1994 and earthquake intensity V and
above) together with the geographic distribution of the seismic sources (Sägesser and
Mayer-Rosa, 1978). The shape of the source zone depends strongly on the earthquake
distribution ‘per extenso’ on the catalogue. Their design is subject to a part of subjectivity or
expert judgement (Schenk 1996). This is illustrated, for example, by the source zoning of
the Ibero-maghreb region (Jimenez et al 1999) shown on Figure 3.
Figure 2. Seismicicity (1300-1994) and sources zones (Ruettener 1995)

Figure 3. source delimitations in the Ibero-maghreb region (Jimenez et al 1999)

Rate of earthquake recurrence

The frequency-magnitude occurrence relationship help to characterise the activity of each

source. The rate of recurrence of earthquakes on a seismic source is assumed to follow the
Gutenberg-Richter relation

log10 n(M) = a-bM

where n(M) is the number of events per year having magnitudes greater than M. a and b are
constants defined by regression analysis. The slope of the magnitude-frequency Gutenberg-
Richter defines the "b value" parameter.

Maximum magnitudes

For a single source, the modified Gutenberg-Richter relation is

N(M)= an [1 - 1-e-b(M-Mlow) /1-e –b(Mupp-Mlow)]

Mupp and Mlow are the upper and lower bound magnitude on the source. Mlow, is the
magnitude below which no engineering-significant damage is expected and Mupp represents
the maximum expected magnitude. The maximum magnitude is related to the tectonic
setting, geometry, and type of the seismic source. Although no standard method exists for
assigning a maximum magnitude to a given fault, empirical correlation are used based on
the length of rupture of the fault, the total length of the fault trace or the area of the fault
rupture zone. In most cases, faults cannot be clearly recognised, maximum magnitude are
than deduced either from the earthquake catalogue, from the recurrence rate (extrapolating
the Gutemberg-Richter relationship) or from paleoseismicity studies.

To characterise each source zone, the following parameters are evaluated:

• Mupp and Mlow, the upper and lower bound magnitude on the source,
• the Gutenberg-Richter earthquake recurrence parameter (b-value),
• the activity rate an the number of event per year having magnitudes equal to or greater
than Mlow on the source, and
• additionally, the average hypocentral depth.

Characterisation of attenuation of ground motion

State-of-the-art estimates of expected ground motion at a given distance from an earthquake

of a given magnitude are the second element of earthquake hazard assessments. These
estimates are usually equations, called attenuation relationships, which express ground
motion as a function of magnitude and distance (and occasionally other variables, such as
type of faulting). Commonly assessed ground motions are maximum intensity, peak ground
acceleration (PGA), peak ground velocity (PGV), and several spectral accelerations (SA).
Each ground motion mapped corresponds to a portion of the bandwidth of energy radiated
from an earthquake. PGA and 0.2s SA correspond to short-period energy that will have the
greatest effect on short-period structures (one-to two story). PGA values are directly related
to the lateral forces that damage short period. Longer-period SA (1.0s, 2.0s, etc.) depict the
level of shaking that will have the greatest effect on longer-period structures (10+ story
buildings, bridges, etc.). Ground motion attenuation relationships may be determined in two
different ways: empirically, using previously recorded ground motions, or theoretically,
using seismological models to generate synthetic ground motions which account for the
source, site, and path effects. There is overlap in these approaches, however, since empirical
approaches fit the data to a functional form suggested by theory and theoretical approaches
often use empirical data to determine some parameters.

The ground motion at a site, for example Peak Ground Acceleration depends on the
earthquake source, the seismic wave propagation and the site response. Earthquake source
signifies the earthquake magnitude, the depth and the focal mechanism, the propagation
depends mainly on the distance to the site. The site response deals with the local geology
(site classification); it is the subject of microzonation.

The basic functional (logarithmic) form for ground motion attenuation relationship is
defined as (Reiter 1990)

ln Y = lnb1 + lnf1(M) + lnf2(R) + lnf3(M,R) + lnf4(P) + ln ε

• Y is the strong motion parameter to be estimated (dependant variable), it is lognormal

distributed.
• f1(M) is a function of the independent variable M, earthquake source size generally
magnitude. In that case moment magnitude (M) is the preferred magnitude measure,
because it is directly related to the seismic moment of the earthquake and does not
saturate. But it can also be the epicentral intensity.
• f2(R) depends on the variable R, the seismogenic area source to site distance,
• f3(M,R) is a possible joint function between M and R. For example for an earthquake
with big magnitude the seismogenic area is large and the source to site distance may be
different.
• f4(Pi) are functions representing possible source and site effects. For example different
style of faulting in the near field may generate different ground motions values
Abrahamson and Shedlock (1997).
• ε is an error term representing the uncertainty in Y

These relationship are for a specific site classification (hard rock, soft rock, etc.). Hazard
values calculated for rock/stiff soil sites (the most common site classifications) are lower
than hazard values calculated for soil sites. A comprehensive review and application for
Switzerland can be found in Smit (1996).

Example of different attenuation functions used in the different countries of the Ibero-
maghreb region are shown on table 1 and figure 4 (Jimenez et al. 1999).

Table 1
Country Attenuation law

Algeria PGA = (190.67 e0.823 M) / (R+0.864 e0.463 M)1.561

Morocco log PGA = -1.02 + 0.25 M - 0.00255 (R2+7.32)1/2 - log(R2+7.32)1/2

Portugal I = 6.8 + 1.13 M - 1.68 ln (R+14)

Spain I = I0 + 12.55 - 3.53 ln (R+25)

Tunisia PGA = (5600 e0.8 M) / (R+40)2

Figure 4. Attenuation function used in the Ibero-Maghreb (Jimenez et al. 1999).

For the calculation of the seismic hazard map of the whole region, only one attenuation law
was considered Joyner and Boore (1981). This relationship is represented on the figure 4
with one standard deviation in log PGA. It could roughly represent an average of the
different laws used in the region.

To adjust the parameters of the attenuation function for intensities (ex. Portugal and Spain in
table 1), macroseismic data are needed. For example the isoseismal map of the 1356 Basel
earthquake (after Mayer-Rosa and Cadiot, 1979) shown on figure 5.

Figure 5. Isoseismal map of the 1356 Basel event (Mayer-Rosa and Cadiot, 1979)
The figure 6 depict the observed intensity distribution of an event (13 March 1964 in
Central Switzerland) together with the attenuation law, for subalpine region
(Ruettener1995). It shows a significant scattering of data around the attenuation function.
This scattering depends not only on physical effect like the directivity of the radiated energy
or the local geology but also on the quality of intensity data.

Figure 6. Intensity distribution (13/03/64) and attenuation low (Ruettener 1995)

Calculation of probabilities

The third element of hazard assessment, the actual calculation of expected ground motion
values, involves determining an annual frequency of exceedance of the ground motion
parameter of interest, then summing over the time period of interest.

The probability tool used is a poissonian model. The occurrence of a ground-motion

parameter at a site in excess of a specified level is poisson process if the occurrence of
earthquakes is poissonian. This means that any event is independent of the occurrence of all
other events. In theory, the catalogue has to be declustered, the foreschoks and afterschoks
should be removed

Probability of Exceedance

The annual mean number of events in which GM exceeds the specified ground motion level
Z is calculated by summing up the incremental contributions of the N sources, taking into
account:
• the annual mean rate of recurrence of earthquakes of magnitude Mi on each source
• given the occurrence of an earthquake of magnitude Mi, the distance from the rupture
surface to the site,
• given the occurrence of an earthquake of magnitude Mi at a distance of rj, the probability
that the value of ground motion at the site exceeds a specified level.
The probability (ex 10%) of exceeding a specified level of ground motion (such as a certain
value of peak acceleration) at a selected site within the time interval of interest (such as 50
years) is calculated by:
1/ combining the three probability functions,
2/ integrating over all possible earthquake, location, magnitude for a source
3/ integrating over all the sources

The probability of exceedance, in a specific time interval, that ground motion amplitude a*
is exceeded can be expressed as follows (McGuire, 1993)
P[A > a * in time t ] / t = ∑ v i ∫ ∫ G A m,r (a *)f m (m) f r (r m)dmdr
i

where ν i is the mean annual rate of occurence in source i, G is the probability that an
earthquake given m and r exceeds ground motion a* at a specific site. f m(m) is the
probability density function for magnitude, f r (r m) and is the probability distribution
function for distance.

Probability terms

Two equivalent results are typically calculated:

1/ The ground motion corresponding to a certain probability of exceedance in a specific
interval of time (exposure time) (see hazard for Brig, Ruettener, 1995 in Figure 7.a), or
2/ The ground motion having a specific average return period (id. Figure 7.b).

Figure 7.a and b. Hazard outputs of the historical method for Brig (Ruettener 1995)
 Return period = -T / ln(1-P(Zz))

p(Zz) desired probability of exceedance during the time T (exposure time)

Probability of non-
Probability of exccedance in approximate average return
exceedance in a given
a given exposure time period in years
exposure time

20 % in 10 years 80% in 10 years 50 years

10% in 10 years 90% in 10 years 100 years

10% in 50 years 90% in 50 years 500 years

10% in 250 years 90% in 250 years 2500 years

1% in 100 years 99% in 100 years 10 000 years

The probability of non-exceedance of 90% in 50 years corresponds to a return period of 475

years. It is frequently used to represent seismic hazard maps because 50 years is the average
life span of a building. Note that it is also the level required for European building codes
EC8. Longer return periods are chosen when dealing with critical lifeline systems like dams
etc..(see next contributions).

Incorporation of Uncertainties

The uncertainties of the basic input data must be taken into account. Uncertainties are
introduced either by lack of data or/and lack of knowledge. There are random uncertainties
which could be incorporated in the hazard curve calculation. There are also systematic or
modelling uncertainties for example for the choice of maximum magnitude, the correct
ground motion model. These uncertainties are taken into account by developing alternative
strategies and models in the interpretation of those input data for which significant
uncertainties are known to exist. For example, multiple source zone models may be defined.
Hazard calculations from each model are then combined using various schemes that produce
a weighted mean (or median) hazard value. It is the logic tree analysis (Frankel 1995).

Applications, alternative techniques and limits of earthquake hazard assessment

The probabilistc method, based on a poissonian model, does not depict a possible variation
of the seismicity in time, because of the hypothesis of stationarity of the model. Some
research is developing in this direction.

Seismic hazard assessment in low seismicity areas is much more subject to large errors than
in areas with high earthquake activity. This is specifically the case if the time span of the
available data catalogue is considerably smaller than the mean return interval of large
events, for which the hazard data has to be calculated.
Alternative techniques of the probabilistic method are mainly to avoid the delimitation of
source zone. The historical method for example, applied in Switzerland by Ruettener (1995)
estimates the ground motion at a site from each event of the historical catalogue. The
probability distribution of the historic occurrences of earthquake is calculated from the
earthquake rate.

The deterministic approach evaluates the maximum expected ground motion at a site,
resulting from the strongest potential earthquake at the nearer possible distance.
Deterministic approaches are often used to evaluate the hazard for a selected site. They are
in general more conservative. But they do not take into account uncertainties nor an estimate
of frequency of occurrence which is needed by decision makers for planning purposes.

The probabilistic method allows to take uncertainties into account and is easily applied.
However, as the quality of the output depends strongly on the quality of the input
parameters (earthquake catalogues, strong motion relationships), and it might be that there is
little transparency on the quality and the integration of the data. On the other hand the
output reflects the state of the data and is subject to improvement.

Probabilistic approach can be applied to mapping the hazard for different probabilities and
exposure time and for different area sizes (local / regional (Figure 8) / global).

Global seismic hazard assessment maps: The variations in each element of the seismic
hazard assessment lead to differences in the estimated hazard along the national borders.
Some research programs (DACH GSHAP, SESAME) have been launched in order to
homogenise this hazard (Figure 9).

Figure 8. Earthquake hazard map of the Ibero-Maghreb region. PGA [m/s2] with
90% probability of non-exceedance in 50 years.
Figure 9. Horizontal peak ground acceleration seismic hazard map representing stiff
site conditions for an exceedance or occurrence rate of 10% within 50 years
(Gruenthal et al. 1999).
 Links

Non commercial computer programs for hazard analysis:

• SeisriskIII ( Bender and Perkins)
http://gldage.cr.usgs.gov/eq/html/swmain.html
• CRISIS (Ordaz) Bergen,Norway In SEISAN
http://www.ifjf.uib.no/Seismologi/software/seisan.html
• Wizmap (Musson) BSGS
http://www.gsrg.nmh.ac.uk/~phoh/wizmap.htm

Application projects:
• USGS national seismic hazard mapping
http://gldage.cr.usgs.gov/eq/
• Global Seismic Hazard Program
http://seismo.ethz.ch/GSHAP/

BSGS World seismic hazard service

http://www.gsrg.nmh.ac.uk/hazard/hazard2.htm
• A short guide to seismic hazard; a brief paper on the subject.
• Seismic hazard in the UK; a guide to UK seismicity, with examples.
 Lexicon (partly based on Hays 1980)

- Attenuation / Abminderung Daempfung / attenuation: decrease of the seismic

ground motion with the distance.

- b-value / b-Werte/ - : from the Gutenberg-Richter relationship, this parameter

indicates the relative frequency of earthquake of different sizes (magnitude) derived
from the historical seismicity.

- Catalogue / Katalog / catalogue: record of historical earthquake. It can be used for the
time span of completness for different magnitudes.

- Exceedance probability / Ueberschreitenswahrscheinlichkeit / probabilite de

dépasser ou d'exceder: the probability over some exposure time that an earthquake
will generate a value of ground shaking greater than a specific level.

- Exposure time / - / - : The period of time (period of interest ex 50 years) that a structure
or a facility is exposed to hazard. Linked to the design lifetime of the structure.

- Ground motion / Bodenbewegung / mouvement du sol: Parameter (acceleration,

velocity or displacement) which is evaluated by the hazard study can be also intensities
or acceleration at different frequencies (Spectral acceleration).

- Hazard / Gefaehrdung/ danger ou aléa: the phenomena accompagning the earthquake

which may cause the damage or loss in that case ground shaking

- Return period / Wiederkehrperiode / periode de retour: average period of time

between events causing ground motion that exceed a particular level at a site. It is the
inverse of annual probability of exceedance. Also mentioned as the reccurence interval

- Risk / Risiko/ risque: it is the possible probability of lost. It is a function of hazard *

vulnerability *(exposure)*(value)

- Seismic activity/ Aktivität/ activité sismique: (of a source) it is described with the
frequency/magnitude relationship ( Häufigkeitverteilung) characterise by the b-value.

- Source / Quelle / source: the source of energy release causing an earthquake. For
computing purpose, they can be areas, faults or points the source is characterise by its
activity and the maximal an minimal magnitudes

- Vulnerability/ Verwundbarkeit/ vulnerabilité: value of the structure exposed to the

hazard