Vous êtes sur la page 1sur 10

Natural Hazards 3: 49-58, 1990.

49
© 1990 Kluwer Academic Publishers. Printed in the Netherlands.

Attenuation Laws and Seismic Hazard Assessment*

J. E. P A P O U L I A
Earthquake Planning and Protection Organization of Greece, Athens, Greece

and

G. N. S T A V R A K A K I S
National Observatory of Athens, Seismological Institute, Athens, Greece

(Received: 12 May 1989)

Abstract. In the present study, the seismic hazard in the city of Patras, central Greece, is estimated. The
computations are based on a slightly modified version of the method proposed by Cornell, allowing the
introduction of individual attenuation laws for each seismic source.
The obtained results emphasize the dependence of hazard on attenuation and the importance of the use of
local attenuation laws in seismic hazard assessment.

Key words. Seismic hazard, local attenuation laws.

1. Introduction
Attenuation of strong ground motion with distance from the causative fault, is
a function of source characteristics, transmission path, geometrical spreading,
absorption coefficient, and local site conditions.
In practice, however, the estimation of the ground motion parameter is almost
always based on attenuation equations derived from regressions of observed motions
against earthquake magnitude, and distance from source to site. Because of their
importance, these equations have received much attention and are updated when new
data become available. Boore and Joyner (1982) reviewed the techniques used and
some of the recent results. Comprehensive regression analyses, including the new data,
were reported by Campbell (1981) and Joyner and Boore (1981).
The widely used seismic hazard models (Cornell, 1968; Algermissen and Perkins,
1976; McGuire, 1978) assume a general set of functions of the form
g~ =f(M, R), (1)
where gs is the ground motion at a site of interest, M is the earthquake magnitude, and
R is the site epicentral or hypocentral distance.
In the present study, an effort is made to incorporate the influence of individual
attenuation laws from each seismic source on seismic intensity hazard estimations.

*Paper presented at the 21st General Assembly of the European Seismological Commission, held in Sofia,
1988.
50 J.E. PAPOULIAAND G. N. STAVRAKAKIS

2. Seismic Hazard Model


To estimate the seismic hazard at a site of interest, the seismicity and other
characteristics of past regional earthquakes are reviewed to develop probabilistic
predictions about future events. This prediction requires (i) seismic source delineation,
(ii) determination of the recurrence model of earthquakes, (iii) consideration of
attenuation of ground motion, and (iv) contribution of all potential seismic sources to
future seismic load at the site of interest.
Seismic source delineation is attained by the correlation of spatial distribution of
epicenters with the tectonic features of the investigated region. This process requires
detailed knowledge of the tectonic regime, and is one of the most critical points in
seismic hazard assessment. Seismic sources are identified as 'point', 'linear', and 'areal'
sources, and seismicity over the whole area of each source is considered as
homogeneous (Cornell, 1968).
In the present model, the occurrence of earthquakes is thought to be a Poissonian
process. This assumption does not introduce any limitation in engineering practice and
implies that earthquakes follow the exponential distribution regarding their magnitude,
and are randomly distributed with regards to time. This process has been widely
adopted in all seismic hazard models.
A most critical point in seismic hazard estimation is the choice of the attenuation
model. Typically, the ground motion descriptor gs in Equation (1) is in terms of peak
acceleration, velocity or displacement. However, other measures have also been used
(McGuire, 1977). In the present model, intensity attenuation functions of the form
I s = I ( M , R), where I s is the site intensity, replaced the peak ground acceleration
functions. The advantage of this choice is that historical information on earthquake
damage can be interpreted in terms of modified Mercalli intensity. In this case, the data
base is greatly expanded as compared to the conventional ground motion parameters
(acceleration, velocity, displacement) requiring instrumental observations.
The variation in the tectonic regime, however, suggests a regionalization in intensity
attenuation. This is emphasized in the present study with the introduction of regional
attenuation laws based on local data from each seismic source of the proposed seismic
hazard model.
The mathematical formulation of the model has been described in detail in previous
studies (Cornell, 1968; Cornell and Merz, 1974), and will be omitted here.

3. Application of the Model


Cornell's methodology is applied to the city of Patras, central Greece, in order to obtain
a specification of the expected seismic intensity over a given period of time.
Figure 1 outlines the seismic epicenter distribution and the proposed seismic hazard
model for the city of Patras. The delineation of seismic sources has been attained with
regards to the model of the main tectonic features in central Greece (Mariolakos et al.,
1985) (Figure 2), and the main seismic fracture zones for shallow earthquakes in the
Aegean and surrounding region (Papazachos et al., 1985).
ATTENUATION LAWS AND SEISMIC HAZARD ASSESSMENT 51

21 22

[] []
° 0 ~t9
[]
o
ON
0 I

~ _

r 4 o" , '

I o° A

° " "<',._.., "4L<.

IN 22 23

Fig. 1. Epicentral distribution, and proposed seismic hazard model for the city of Patras.

Because of the random earthquake epicentral distribution and the complicated


tectonic regime, sources A1, A 3 and A 6 are considered as areal.
The distribution of epicenters in source F 2 is approximately correlated with a fault
zone of this area, and the source is therefore considered linear.
The tectonic evolution in the Gulf of Corinth is mainly characterized by vertical
movements along normal faults of E - W direction, many of which are still active today.
The seismic epicenters are distributed along this major fault zone and source F 4 is
identified as being linear.
The distribution of epicenters in western Peloponnesus, however, can be considered
as random, and there is no good correlation between seismicity and well-defined
tectonic features. Consequently, line F 5 does not actually identify the exact position of
a fault, but represents the general orientation of the main neotectonic faults of the area,
which have been reactivated since the Upper Miocene.
52 J . E . P A P O U L I A A N D G. N. STAVRAKAKIS

str ko .0F. . . . . . . . . .

Kyparissia

Pylos

0
AIrlIIOL
0 50kin b

Fig, 2. Main tectonic features in central Greece (after Mariolakos et al. (1985).

All characteristic parameters of the seismic sources of the proposed model are
summarized in Table I, where b is the parameter of the exponential distribution of the
recurrence model of earthquakes estimated by regression analysis of the data of each
source, Rate is the mean annual number of events above a minimum magnitude for
each seismic source, Depth is the mean focal depth, M m i n is the minimum considered
earthquake magnitude, F i are the sources' geometrical characteristics, and Mmax is the
maximum probable magnitude for each source estimated on the basis of statistical
studies (Drakopoulos and Makropoulos, 1983; Makropoulos and Burton, 1984a).

Table I. Characteristic parameters of the seismic sources of the proposed seismic hazard model

Seismic b Rate Depth Mmin Fx F2 Fa F4 Mmax


source (km) (km) (km) (°) (°)

A1 0.694 2.329 23 4.0 68 110 252 289 7.0


F2 0.486 0.129 29 4.0 60 30 --74 - 7.2
Aa 0.963 0.618 36 4.0 52 100 31 75 7.2
F4 0.793 1.573 23 4.0 10 -- 30 110 - 7.2
F5 0.780 0.366 26 4.0 48 34 -- 56 - 7,2
A6 0.733 1,006 29 4.0 44 106 205 218 7.5
ATTENUATION LAWS AND SEISMIC HAZARD ASSESSMENT 53

T30
11
1

'45 ~ ~ ~ 65 70 "
M
i

T~

v
Zc~ IC

Iif
45
M 4~ 50 1~ 5~ 60 6 s 70

70

I
T
S~i.~.~c SoLmee F
1

60 65 5 0

45
M M ~

Fig. 3. Earthquake recurrence relationships for each seismic source of the proposed seismic hazard model.
54 J . E . P A P O U L I A A N D G. N. S T A V R A K A K I S

Finally, Figure 3 illustrates the earthquake recurrence relationships for each seismic
source of the proposed model.
By multiple regression analysis of the macroseismic data (Bulletins, 1900-1983) for
each seismic source, relationships of the following type are estimated
I s = a + bM + clogR, (2)
where I s is the site intensity, M is the earthquake magnitude, R is the hypocentral
distance, and a, b, c are regression coefficients. These are illustrated in Table II.
In this way, an effort is made to incorporate the different ways of attenuation of
intensities from different seismic sources of the proposed model.
The original computer program proposed by Cornell (1968) is so modified as to
consider these local attenuation laws in the assessment of seismic hazard of the city of
Patras.
For comparison reasons, seismic hazard is further estimated on the basis of an
average intensity attenuation relationship proposed by Papaioannou (1984).
On the basis of the recurrence model of earthquakes, seismicity, geometry,
geographical position of each seismic source, as well as the attenuation law, the return
period for different levels of the macroseismic intensity is estimated. The results of the
analysis are illustrated in Figure 4.
In both approximations, there is a good correlation between computed and observed
intensities. The critical point~ however, is the extreme value distribution of the
investigated parameter. Specifically, the return period of the minimum considered
seismic intensity seems to be underestimated, while the one of the maximum seismic
intensity is rather overestimated. This may be attributed to the data base in each
seismic source. Improvement of the estimated distribution presupposes increase of the
number of observations.
Furthermore, the results seem to be strongly dependent on the attenuation model.
The introduction of individual attenuation laws for each seismic source leads to more
realistic hazard values. This is particularly evident at the upper edge of the probability
distribution, namely the maximum intensity values. Specifically, the return period for
the intensity 7.0 MM is 48 and 70 years, after the first (use of local attenuation laws) and
second (average attenuation) approximation, respectively. These values increase to 111

Table II. Local attenuation laws for each seismic source of the proposed seismic
hazard model

Seismic source Attenuation law

a 1 Is = (4.08 _+ 0.73) + (0.60 _+ 0.11) M -- (1.62 _+ 0.63) log R


F2 I~ = (9.12 + 0.41) + (0.94 + 0.20) M - (5.22 _+ 1.51) log R
Aa Is = (2.10 _+ 0.55) + (0.87 _+ 0.25) M - (1.98 _+ 0.40) log R
F4 IS= (2.74 _+ 0.67) + (0.99 _+ 0.13) M - (1.95 _+ 0.35) log R
F5 I~ = (2.41 _+ 0.65) + (0.84 _+ 0.23) M - (1.35 _+ 0.41) log R
A6 I~ = (2.81 _+ 0.80) + (0.68 _+ 0.16) M -- (1.07 _+ 0.19) log R
ATTENUATION LAWS AND SEISMIC HAZARD ASSESSMENT 55

400 Seismic Hazard in Patras,central Greece ~ 400

Z
L
O

v
4 Maximum Seismic Intensity, I(mm)

-- Local Attenuation Laws


-- AveracJe A t t e n u a t i o n Law

Fig. 4. Return period for maximum expected seismic intensity for the city of Patras, using (i) local
attenuation laws, and (ii) an average law.

and 200 for 7.5 MM. It is evident that the application of an average attenuation law
leads to nonrealistic estimates, which represent only a mean return period, probably
enclosing a high degree of uncertainty, while the use of local attenuation laws
considerably improves hazard estimates.

4. Analysis of Uncertainties
One important advance in seismic hazard assessment is the recognition of the role of
uncertainties, such as uncertainties in source configuration, source seismicity, and
attenuation of ground motion.
To investigate the sensitivity of hazard to these uncertainties, a hypothetical seismic
source model is considered, and illustrated in Figure 5. The seismic sources
surrounding the site of interest, A, are of a certain area and position with regards to A.
The average return period, T~, of an intensity equal to or greater than i', is defined
(Cornell, 1968) as

T i = ( 1 / v C G ) e x p (~i/c2), i >~ i', (3)


where v is the average number of occurrences per unit length per year, and C, G are
constants related to the seismicity and geometry of the seismic source, respectively.
For each seismic source, the constant G is estimated for five different case models,
altering the b parameter, the source geometry, and the attenuation model.
As a first approximation, a mean b value equal to 0.75 - most common in the
investigated region - is considered. This is further altered to 0.9 and 0.6, values
56 J. E. P A P O U L I A AND G. N. STAVRAKAKIS

//

,,' ~ ' , , ~ ' Z

; ' 4,?,,

/t

Fig. 5. Hypothetical seismic source model.

characterizing seismic sources with a small number of high magnitude earthquakes and
a large number of low magnitude earthquakes, respectively. Next approximation is the
variation in source geometry, illustrated by a dotted line in the proposed model.
Finally, two different attenuation laws are used, the average one (Papaioannou, 1984)
and a local law derived by analysis of the macroseismic data of the area of interest.
The average return period, T~,for a certain intensity of 7.0 MM is then estimated for
all five cases. Both the parameters and the results are illustrated in Table IIIa-e.

5. Conclusions
in the present study, a slight modification of Cornell's original approach for seismic
hazard analysis is applied in the city of Patras, central Greece. Regional intensity
attenuation laws, based on local data from each seismic source of the proposed seismic
hazard model, are derived and incorporated in the analysis.
The results are compared with those derived from the application of an average
attenuation formula, leading to the conclusion that the use of local attenuation laws
results in significantly more realistic hazard estimates.
ATTENUATION LAWSAND SEISMIC HAZARDASSESSMENT 57

Table III. Resultsof the uncertainties' analysis: estimated return period for intensity 7.0 MM

Seismic b v c ct ro d G Return period


source x 10-4 x 1011 (°) (km) (km) (years)

(a) A1 0.75 8.1 1.7 36 172 57 1.6 x 10-s


A2 0.75 8.1 1.7 45 221 84 1.1 x 10-5 64
A3 0.75 8.1 1.7 43 99 44 0.1 x 10-5

(b) Ax 0.90 1.12 6.5 36 172 57 1.9 × | 0 - 7


A2 0.90 1.12 6.5 15 221 84 2.3 x 10-v 445
A3 0.90 1.12 6.5 13 99 44 4.2 × 1 0 - 7

(c) A1 0.60 5.01 7.4 36 172 57 4.3 × 10 -4


Az 0.60 5.01 7.4 15 221 84 3.2 x 10 -4
A3 0.60 5.01 7.4 43 99 44 5.5 x 10 -4

(d) A1 0.75 8.1 1.7 46 272 157 1.1 x 10 -6


A2 0.75 8.1 1.7 55 321 184 2.1 x 10 -6 490
Aa 0.75 8.1 1.7 53 199 144 3.5 x 10 -8

(e) Aa 0.75 0.8 2.1 36 172 57 5.0 x 10 -6


A2 0.75 0.8 2.1 45 221 84 5.4 x 10 -6 83
A3 0.75 0.8 2.1 43 99 44 4.8 x 10 -7

Furthermore, the importance of the incorporation of statistical parameter uncertain-


ties in seismic hazard assessment is investigated.
Among the parameters which effect seismic hazard, the b value seems to play a most
important role. The b parameter is estimated mainly by the analysis of observed data
from each seismic source. However, the data based on a given source, is often
incomplete and inhomogeneous in time, leading to large uncertainties in the estimation
of b.
Moreover, source configuration is a critical point of the analysis of hazard.
Uncertainties in tectonic interpretations lead to uncertainties in sources of future
earthquakes. Specifically, in regions with high seismicity and either complicated or
insufficient tectonic evidence, it is extremely difficult to delineate seismic sources in
a unique way. Such uncertainties strongly effect seismic hazard estimates, as pointed
out in the present study.
The examined hazard parameter is the macroseismic intensity. Previous studies
(Drakopoulos and Makropoulos, 1983; Makropoulos and Burton, 1984a) for the
investigated area are in terms of ground acceleration, therefore the results of the present
study cannot be compared directly with previous ones.
Finally, the analysis revealed that the attenuation law is a key parameter in seismic
hazard assessment, emphasizing the importance of use of local attenuation laws for
each seismic source instead of an average relationship.
It should be emphasized, however, that the primary problem with this technique is
that Modified Mercalli intensity describes ground motion on a very coarse scale which
58 J.E. PAPOULIA AND G. N. STAVRAKAKIS

is based on subjective evaluations of damage. Incorporation of local attenuation


relationships of the peak ground acceleration must wait for more strong motion data.

Acknowledgement
The authors express their thanks to Dr Vladimir Schenk, for his useful suggestions.

References
Algermissen, S. T. and Perkins, D. M.: 1976, A probabilistic estimate of maximum acceleration in rock in the
contiguous United States, U.S. Geol. Survey Open-File Rept. 76-416.
Boore, D. M. and Joyner W. B.: 1982, The empirical prediction of ground motion, Bull. Seism. Soc. Am. 72,
5269-5286.
Campbell, K. W.: 1981, Near source attenuation of peak horizontal acceleration, Bull. Seism. Soc. Am. 71,
2039-2070.
Cornell, C. A.: 1968, Engineering seismic risk analysis, Bull. Seism. Soc. Am. 58, 1583-1606.
Cornell, C. A. and Merz, H. A.: 1974, Seismic risk analysis of Boston, ASCE National Structural Engineering
Meeting, Cincinnati, Ohio, Preprint 2260.
Drakopoulos, J. and Makropoulos, K.: 1983, Seismicity and hazard analysis studies in the area of Greece,
University of Athens, Seismological Laboratory, Publ. No. 1.
Joyner, W. B. and Boore, D. M.: 1981, Peak horizontal acceleration and velocity from strong motion records
including records from the 1979 Imperial Valley, California, earthquake, Bull. Seism. Soc. Am. 71,
2011-2038.
Makropoulos, K. and Burton, P. W.: 1984a, Seismic hazard in Greece. I. Magnitude recurrence,
Tectonophysics 117, 205-257.
Mariolakos, I., Papanikolaou, D., and Lagios, E.: 1985, A neotectonic geodynamic model of Peloponnesus
based on morphotectonics, repeated gravity measurements and seismicity, Geol. Jb. 50, 3-17, Hanover.
McGuire, R.: 1977, Effects of uncertainty in seismicity on estimates of seismic hazard for the east coast of the
United States, Bull. Seism. Soc. Am. 67, 827-848.
McGuire, R.: 1978, FRISK-computer program for seismic risk analysis using faults as earthquake sources~
U.S. Geol. Survey Open-File Report, 78-1107.
Papaioannou, C. A.: 1984, Attenuation of seismic intensities and seismic hazard in Greece, PhD thesis:
University of Thessaloniki, (in Greek).
Papazachos, B. C., Hatzidimitriou, P. M., and Karacostas, B. G.: 1985, Seismic fracture zones in the Aegean
and surrounding area, Publication of the Geophysical Laboratory, University of Thessaloniki, No. 2, pp
1-10.

Vous aimerez peut-être aussi