You are on page 1of 25

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/226844920

Estimation of Earthquake Damage to Buried Pipelines


Caused by Ground Shaking

Article  in  Natural Hazards · September 2007


DOI: 10.1007/s11069-006-0002-1

CITATIONS READS
58 1,391

2 authors, including:

Selcuk Toprak
Gebze Technical University
84 PUBLICATIONS   952 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Component and System Based Robustness and Reliability Assessment of Pipelines View project

PM4WAT View project

All content following this page was uploaded by Selcuk Toprak on 20 February 2017.

The user has requested enhancement of the downloaded file.


Natural Hazards (2007) 40:1–24  Springer 2006
DOI 10.1007/s11069-006-0002-1

Estimation of Earthquake Damage to Buried


Pipelines Caused by Ground Shaking
SELCUK TOPRAKw and FILIZ TASKIN
Civil Engineering Department, Pamukkale University, Kinikli Campus, Denizli, 20070, Turkey

(Received: 9 August 2005; accepted: 11 January 2006)


Abstract. One of the most critical lessons of the recent earthquakes is the need for seismic
planning for lifelines, with appropriate supplies and back up systems for emergency repair
and restoration. Seismic planning, however requires physical loss estimations before the
earthquakes occur. Buried pipeline damage correlations are critical part of loss estimation
procedures applied to lifelines for future earthquakes. We review the existing pipeline
damage relationships only for ground shaking (transient ground deformations) in the light
of recent developments and evaluate them with Denizli City, Turkey water supply system.
Eight scenario earthquakes with four different earthquake magnitudes between M6 and M7
caused by two different fault ruptures (Pamukkale and Karakova-Akhan Faults) were used.
Analyses were performed by using Geographical Information Systems (GIS). This high
number of different scenario earthquakes made it possible to compare the pipeline damage
relationships at different ground shaking levels. Pipeline damage estimations for Denizli
City were calculated for each damage relationship and earthquake scenario. Relative effects
of damage relationships and scenario earthquakes on the results were compared and dis-
cussed. The results were presented separately for brittle, ductile, and all pipelines. It was
shown that the variation in ductile pipeline damage estimations by various relationships
was higher than the variation in brittle pipeline damage estimations for a particular
scenario earthquake.
Key words: seismic damage, Denizli, lifelines, loss estimation, pipelines, ground shaking,
transient ground deformation (TGD)

1. Introduction
During the devastating 1999 M7.4 Kocaeli and M7.2 Duzce, Turkey earth-
quakes, substantial water supply damage occurred in many cities. For
example, the entire water distribution system in Adapazari was damaged.
The water service could not be restored until many months after the earth-
quake. One of the most critical lessons of the earthquake is the need for
seismic planning, with appropriate supplies and back up systems for emer-
gency repair and restoration.

w
Author for correspondence: E-mail: stoprak@pamukkale.edu.tr
2 SELCUK TOPRAK AND FILIZ TASKIN

Physical loss estimation for future earthquakes is needed for a variety


of purposes, including emergency response, risk management, and hazard
mitigation. Methodologies for estimating potential pipelines damage use
relationships which are often called different names such as ‘‘fragility
curves’’, ‘‘damage functions’’, ‘‘vulnerability functions’’ or ‘‘damage rela-
tionships’’. These relationships are primarily empirical and obtained from
past earthquakes. In the last century several moderate to strong earth-
quakes caused substantial damage to buried pipeline systems. Especially,
the Northridge earthquake was an important event for a leap in the devel-
opment of pipeline damage relationships. The extensive earthquake dam-
age in the City of Los Angeles water supply system and availability of the
strong motion instruments throughout the area provided a unique oppor-
tunity to develop and improve damage correlations.
In the first part of the paper, a critical review of the existing pipeline
damage relationships for ground shaking were made in the light of recent
developments. Then, the relationships were evaluated with Denizli City,
Turkey water supply system. Denizli is an important industrial and tour-
ism centre and also one of the largest cities in the west part of Anatolia.
This paper focused on pipeline damage caused by ground shaking. Estima-
tions of damage caused by permanent ground deformations such as lique-
faction induced lateral spreading are covered in a subsequent paper. As
part of this study, the water pipelines of the city were digitized and ana-
lyzed using Geographical Information Systems (GIS). Eight scenario earth-
quakes with four different earthquake magnitudes between M6 and M7
caused by two different fault ruptures were used to determine the estima-
tions by various damage relationships. These faults, Pamukkale and
Karakova-Akhan Faults, are historically active faults and located at differ-
ent distances from Denizli city. By using high number of scenario earth-
quakes the damage relationships can be compared at different ground
shaking levels. Also it becomes possible to determine the relative effects of
damage relationships and scenario earthquakes on the results. In the
analyses herein, a distinction between pipelines regarding their behaviour
during the earthquakes were made. The damage to pipelines was estimated
for brittle, ductile, and all pipelines.

2. Pipeline Damage Correlations


2.1. CAUSES OF PIPELINE DAMAGE

Earthquake damage to buried pipelines can be attributed to transient


ground deformation (TGD) or to permanent ground deformation (PGD)
or both. TGD occurs as a result of seismic waves and often stated as wave
propagation or ground shaking effect. PGD occurs as a result of surface
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 3

faulting, liquefaction, landslides, and differential settlement from consolida-


tion of cohesionless soil. The relative magnitudes of TGD and PGD deter-
mine which one will have predominant influence on pipeline response.
TGD generally induces much smaller levels of pipeline strain and deforma-
tion than PGD. Nevertheless, because TGD covers a broader area than
PGD, its damage to pipelines is generally significant.
A good example where PGD was the responsible for high pipeline re-
pair rates is the 1971 San Fernando earthquake. Eguchi (1982) evaluated
the pipeline damage in North San Fernando Valley caused by the 1971
San Fernando earthquake. He compared the pipeline damage caused by
fault rupture, ground failure, and ground shaking. He concluded that al-
most half of the total distribution pipeline damage occurred in localized
areas where fault rupture and ground failure were observed.
A good example where TGD was responsible for high pipeline repair
rates is the 1985 Michoacan earthquake, which caused widespread damage
to the water supply system in Mexico City. Ayala and O’Rourke (1989) re-
ported that there was no soil liquefaction or PGD in the Mexico City area,
and they attributed the water system disruption primarily to seismic wave
propagation effects.
If PGD and TGD are similar in magnitude, they will contribute to
pipeline damage at comparable levels. Large TGD is often observed when
there are near-field large peak ground velocity pulses or where site
response characteristics result in amplified transient motion. For example,
Pease and O’Rourke (1997) studied pipeline damage in San Francisco
caused by the Loma Prieta earthquake and showed that transient lateral
shear strains in liquefied soils were the prime cause of horizontal displace-
ments and seismic damage to buried pipelines in the Marina.

2.2. DEVELOPMENT OF THE CORRELATIONS UP TO 1994 NORTHRIDGE


EARTHQUAKE, USA

Pipeline damage commonly expressed as repair rate, which is the number


of pipeline repairs in an area divided by the length of the pipelines in the
same area. Pipeline damage due to TGD effects is correlated with various
seismic parameters whereas damage due to PGD effects is correlated with
amount of ground movement or deformation. Toprak (1998) evaluated
pipeline damage correlations developed before the 1994 Northridge earth-
quake, USA, and used various seismic parameters to explore new relation-
ships between seismic intensity and pipeline damage. The seismic
parameters that he tried are Modified Mercalli Intensity (MMI), peak
ground acceleration (PGA), peak ground velocity (PGV), peak ground
displacement, spectral acceleration, spectral velocity, spectrum intensity,
4 SELCUK TOPRAK AND FILIZ TASKIN

and Arias intensity. A concise summary of his extensive review is provided


herein.
Katayama et al. (1975) performed one of the earliest correlations be-
tween seismic parameters and pipeline damage. He used PGA and didn’t
differentiate between TGD and PGD. This relationship was reanalyzed
later by Patelunas et al. (1977) and Erel et al. (1977) and published in a
logarithmic form.
Eguchi et al. (1981, 1983) were among the first researchers to correlate
pipeline repair rate with MMI, and they recommended that only wave
propagation damage be correlated with MMI. Furthermore, their relation-
ships were developed for different type of pipelines: cast iron (CI) and wel-
ded steel pipes with gas-welded joints (WSGWJ). Eguchi et al. (1983)
assumed that the asbestos cement (AC) and welded steel with caulked
joints (WSCJ) pipelines would have similar repair rates as CI pipelines, on
the assumption that joint failures would predominate. Later, Ballantyne
et al. (1990) and Eguchi (1991) introduced a bilinear pipeline damage rela-
tionship for different pipe types. Previously, Gori and Hays (1988) recom-
mended that pipe vulnerability models be revised downwards at intensities
IX and above. The main reason for this recommendation was to avoid the
problem of overestimating repairs as a result of linear extrapolation from
lower intensities. The bilinear plot has a break point at MMVIII. These
correlations with MMI were used widely because a substantial amount of
expert knowledge and existing damage data for earthquakes in the United
States existed in the form of MMI information (ATC 13, 1985). However,
developing correlations between pipeline repair rates and MMI involves
substantial uncertainties. Not only is the magnitude of MMI a subjective
measure of seismic intensity, but the areal extent of a given MMI is also
the product of subjective judgment regarding coverage. Pipeline repair rate
is influenced strongly by the total length of piping within a given zone.
Uncertainties associated with MMI coverage therefore affect the repair rate
estimates at the same time that subjective assessment of the MMI magni-
tude contribute to its uncertainty.
Barenberg (1989) recommended that pipeline damage be correlated with
a TGD parameter, such as velocity or velocity-induced ground strain in a
region of low intensity shaking, and with PGD in regions where relatively
high levels of permanent deformation are a result of surface faulting, lique-
faction, and other causes. In essence, he differentiated between wave prop-
agation damage and PGD damage. For areas affected principally by TGD,
he plotted repair rate with respect to maximum horizontal ground velocity.
Although Barenberg’s study used the best available database at the time of
his work, the information was limited primarily in terms of strong motion
measurements. For instance, four out of five data points used in the
regression were obtained by using velocities estimated from attenuation
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 5

relationships rather than directly from strong motion measurements


(O’Rourke, 1985). The relationships were obtained by using data primarily
from CI pipelines with relatively small diameters (less than 250 mm).
O’Rourke and Ayala (1993) expanded the database that Barenberg
(1989) used by including seven data points from two Mexican earthquakes
and one US earthquake. These additional data included CI, AC, concrete
and prestressed concrete cylinder pipelines primarily with large diameters
(larger than 500 mm). In essence, this relationship included pipelines with
various materials identified as brittle and with various diameters ranging
from about 80 to 1800 mm. HAZUS (FEMA, 1999) adopted O’Rourke
and Ayala (1993) relationship in its default analysis, hence the relationship
is referred as HAZUS relationship herein.

2.3. THE CORRELATIONS AFTER THE 1994 NORTHRIDGE EARTHQUAKE, USA

The Northridge earthquake provided a unique opportunity to develop and


improve correlations between pipeline damage and seismic parameters. The
research described in Toprak (1998), O’Rourke et al. (1998), O’Rourke
and Toprak (1997) represented the first time that comprehensive GIS anal-
yses were performed for a large US water supply with extensive earth-
quake damage and strong motion data. There was widespread pipeline
damage and deployment of strong motion instruments throughout the City
of Los Angeles at the time of the earthquake. The earthquake damage
included 15 transmission line, 74 trunk line (nominal pipe diameter
‡ 600 mm), and 1,013 distribution lines (diameter < 600 mm) repairs.
Furthermore, reliable information about pipeline characteristics, repair,
and strong motion measurements has been collected. As part of the re-
search, all of approximately 10,750 km of distribution lines and 1,000 km
trunk lines were digitized and incorporated in a GIS database that also
includes repair records and the corrected strong motion records of 165
seismograph stations.
Among the various seismic parameters, the most statistically significant
correlations were found for PGV (O’Rourke et al., 1998; Toprak, 1998).
PGV has a more direct physical interpretation in terms of its effects on
buried pipelines. PGV is correlated with axial strains experienced in the
soil due to seismic wave propagation as expressed in the following general
equation (Committee on Gas and Liquid Fuel Lifelines, 1984):
Vmax
eg ¼ ð1Þ
C
where eg is the maximum seismic ground strain, Vmax is the maximum
ground velocity, and C is the seismic wave-propagation velocity. Depend-
ing on the slippage developed between a pipeline and the surrounding soil,
6 SELCUK TOPRAK AND FILIZ TASKIN

a certain percentage of the soil strain is transferred to the pipeline. Because


of this relationship, a good correlation between PGV and pipeline damage
would be expected.
PGV in previous correlations including Toprak (1998) is defined as the
larger of the two horizontal components recorded during the earthquake.
The parameter, however, can be defined in several different ways. In atten-
uation relationships, PGV is commonly defined as the geometric mean of
the two largest horizontal components (e.g., Campbell, 1997). PGV may
also be defined as the maximum vector magnitude of the two horizontal
components. O’Rourke and Jeon (2000) introduced pipeline damage rela-
tionships for the geometric mean PGV and maximum vector magnitude of
PGV. Their regressions indicate that the choice of PGV makes little differ-
ence in the statistical significance of the regressions. All are characterized
by coefficient of determination (r2) that are comparable, although the high-
est r2 is associated with maximum PGV. Coefficient of determination has
been used commonly in pipeline damage correlations as a measure of
goodness-of-fit (i.e. to show how much the variability in the RR values can
be explained by the observed value of PGV).
Toprak (1998) and O’Rourke et al. (1998) relationships were developed
primarily from CI pipeline damage although they made limited compari-
sons with damage for other pipe types. Also they considered effect of pipe
size on damage correlations by dividing the data into two groups as distri-
bution (nominal pipe diameter < 600 mm) and trunk lines (nominal pipe
diameter ‡ 600 mm). Table I gives their correlation equation for the CI
distribution pipelines whereas Figure 1 shows the relationship relative to
other relationships. Only HAZUS equation among the relationships devel-
oped before the 1994 Northridge earthquake is shown in Table and Figure,
because it is still being used widely. Other most recent damage relation-
ships shown in Table I and Figure 1 are explained below. The table also
shows the type of PGV, and pipe size and type that were used to develop
each relationship. The relationships in Figure 1 correspond to respective
CI, brittle mix, and mix type pipelines stated in Table I.
O’Rourke and Jeon (1999, 2000) went one step ahead and developed
separate relationships for CI, ductile iron (DI), AC, and steel pipelines.
They also developed relationships which uses pipe diameter (Dp) and PGV
together. They named the new PGV as scaled velocity which implies that
the PGV is scaled by pipe diameter. All these equations used maximum
PGV. Relationships for CI and DI distribution pipelines, as representatives
of brittle and ductile pipelines, respectively, are provided in Table I
whereas only CI relationship is shown in Figure 1. Also given in the table
is the CI pipeline damage relationship with respect to the geometric mean
PGV.
Table I. Pipeline damage correlations for ground shaking effects.

Pipeline damage correlations PGV Pipe type Pipe size Repair rate (number of repairs/km) equation

HAZUS Maximum Brittle mix Mix 0.0001*(PGV)2.25


Toprak (1998) Maximum CI Dp £ 600 mm 10(1.62*log(PGV))3.64)
O’Rourke and Jeon (2000) Geometric mean CI Dp £ 600 mm e(1.09*ln(PGV))6.12)
O’Rourke and Jeon Maximum CI Dp £ 600 mm e(1.21*ln(PGV))6.78)
(1999, 2000) DI Dp £ 600 mm e(1.84*ln(PGV))9.40)
O’Rourke and Jeon Maximum PGV scaled CI – 0.036*(PGV/Dp1.021)0.989
(1999, 2000) with pipeline diameter DI – 0.004*(PGV/Dp0.468)1.378
ALA (2001) Geometric mean Mix Mix 0.0024*PGV
0:1172 þ 0:7281
PGV
 Z 
1
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Pineda and Ordaz (2003) Maximum Brittle mix Dp ‡ 500 mm 2p 19:7811
1

ð1=2Þ½ðPGV51:8964Þ=19:78112
e dðPGVÞ
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES

O’Rourke and Deyoe (2004) Maximum Brittle mix Mix 0.0035*PGV0.92 for secondary wave
Brittle mix Mix 0.034*PGV0.92 for Rayleigh wave
7
8 SELCUK TOPRAK AND FILIZ TASKIN

1
Repair Rate (Number of Repairs/km)

0.1

HAZUS
Toprak 1998
O'Rourke & Jeon 2000, Geo. Mean PGV
O'Rourke & Jeon 1999, 2000, Max. PGV
ALA 2001
Pineda & Ordaz 2003
M.O'Rourke & Deyoe 2004-S Wave
M.O'Rourke & Deyoe 2004-R Wave
0.01
10 Peak Ground Velocity (cm/sec) 100

Figure 1. Buried pipeline damage relationships.

Trifunac and Todorovska (1997) developed pipeline damage relation-


ships using the 1994 Northridge earthquake data. Their relationships relate
the average number of water pipe breaks per km2 with the peak strain in
the soil or intensity of shaking at the site. However, direct comparison
between theirs and other damage relationships discussed herein couldn’t be
made because of differences in damage measures (e.g., repairs per kilometre
of pipeline versus repair per square kilometer of land area).
American Lifelines Alliance (ALA, 2001) project combined data from
12 US, Japan, and Mexico earthquakes and developed relationships for
wave propagation damage. The most common material in the database of
81 points is cast iron (38 data points) followed by steel (13), AC (10), DI (9),
and concrete (2). Another nine data points represent both CI and DI pipe
combined. In terms of pipe diameter, the database contains mostly those
sizes associated with distribution main systems; only eight data points were
identified as specifically for large-diameter pipe greater than 300 mm. The
ALA referred this equation as ‘‘backbone’’ pipe vulnerability function (see
Table I and Figure 1) and recommended that it be used when there is no
knowledge of the pipe materials, joinery, diameter, corrosion status, etc.
of the pipe inventory and when the evaluation is for reasonably large
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 9

inventory of pipelines comprising a water distribution system. When


certain information is available about the pipe material, joint type, soil
type, and pipe diameter, ALA equation is multiplied by a constant, K1.
For example, if the cast iron pipe is located in soils with uncertain soil
conditions, the constant, K1 is equal to 1. This reflects that the bulk of the
empirical data set is governed by cast iron pipe. ALA equations use the
geometric mean PGV.
Pineda and Ordaz (2003) developed a relationship to estimate damage
due to earthquakes in the primary water system in Mexico City. They
reanalyzed the damage data from the 1985 Michoacan earthquake by tak-
ing into consideration the recent developments in ground motion estima-
tion in the Valley of Mexico. Although the data was limited to Mexico
City and the 1985 Michoacan earthquake, the relationship is included here
as an example for pipeline damage by surface waves.
O’Rourke and Deyoe (2004) investigated why there is significant differ-
ence between HAZUS relationship and the other relationships developed
after the 1994 Northridge earthquake. They concluded that the most sig-
nificant difference between the data sets is seismic wave type. As shown in
Equation 1, seismic damage to buried pipelines is a function of ground
strain, which is a function of PGV and seismic wave-propagation velocity.
The latter is controlled by wave type, e.g., Rayleigh (R) wave which is a
type of surface wave or secondary (S) wave which is a type of body wave.
When plotted on repair rate versus ground strain, it appears that the scat-
ter of data points from Mexico and other earthquakes reduces substan-
tially. O’Rourke and Deyoe introduced a very simple criterion to establish
the wave type. It is assumed that R waves control if the ratio of epicentral
distance to focal depth is 5 or larger. In that case C is assumed equal to
500 m/s. On the other hand, it is assumed that S waves control if the
epicentral distance to focal depth ratio is less than 5. In that case C is
assumed equal to 3,000 m/s. According to these criteria, R waves presum-
ably control the 1985 Michoacan, Mexico data points while for all the
other data points S waves presumably control. In terms of PGV, they
introduce two different relationships, one to use in the case of R waves
and the other for S waves (see Table I and Figure 1).

3. Damage Estimations for Denizli Water Supply System


3.1. SEISMICITY AND GEOLOGY OF DENIZLI REGION

Denizli is located in the east of Aegean extensional province of Turkey.


This North–South extensional regime has formed primarily NW–SE and
E–W extending graben and horst systems bounded by normal faults. The
most active grabens in the province are NW–SE extending Gediz Graben
10 SELCUK TOPRAK AND FILIZ TASKIN

and E–W extending Menderes Graben (Figure 2a). The figure shows the
active faults in the province. Denizli basin is situated close to the conflu-
ence of the Gediz and Menderes Grabens. Figure 2b shows the general
geology of Denizli and its surroundings. The area consists of primarily
Neogene aged sedimentary rocks and Quaternary deposits.
Figure 2a also shows historic earthquakes that occurred in the region
before 1900 with their maximum intensity values and earthquakes between
1900–2004 with a magnitude greater than 5. The data were obtained from
Kandilli Observatory and Earthquake Research Institute KOERI (2004).
As shown in the figure, the Aegean extensional province had many historic
earthquakes with intensities of IX and X. Some of these earthquakes were
around Denizli and caused heavy damage and casualties. Altunel (2000)
summarized the historical earthquakes and their damage to Denizli and its
proximity. The historical development of the ancient cities in this area was
affected significantly by earthquakes. The ancient cities around Denizli,
e.g., the former Roman City of Hierapolis (modern Pamukkale), and faul-
ted archeological relics are the subject of many earthquake related research
activities Hancock and Altunel (1997). The most recent earthquake that
caused major damage around Denizli was in Menderes Graben in 1899
(Intensity IX). The recent earthquakes that occurred around Denizli was
smaller than M6 (Figure 2). However, many damaging earthquakes oc-
curred in the region. During the 1965 Denizli–Honaz earthquake (M5.7),
14 people were killed, 217 people were injured and 488 structures were
damaged. During the 1976 Denizli earthquake (M4.9), 4 people were kil-
led, 28 people were injured, and 3,200 structures were damaged (KOERI,
2004; MTA, 2004). The 1976 earthquake affected particularly the central
part of Denizli.
There have been some studies on the prediction of the size and location
of future earthquakes in the area. Aydan et al. (2001) evaluated the earth-
quake magnitude-occurrence relationships of the region and concluded that
earthquakes between magnitudes of 6 and 7.2 are highly probable at differ-
ent periods. In another study, Aydan et al. (2002) evaluated past earth-
quake data of Turkey and the crustal deformation measurements by
Global Positioning System (GPS). They calculated the mean stress distri-
bution in the western Turkey and concluded that the highest concentra-
tions occurred in Denizli region. They estimated the magnitude of a
possible earthquake in Denizli basin as M6.3. A recent liquefaction and
shaking hazard mapping work done for the Denizli municipality used this
magnitude earthquake for the scenario earthquakes (PAU, 2002) . In the
light of these studies, scenario earthquakes of M6.0, M6.3, M6.5,
and M7.0 were used herein for damage predictions of Denizli water supply
system.
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 11

Figure 2. Fault lines, past earthquakes, and geology around Denizli.


12 SELCUK TOPRAK AND FILIZ TASKIN

3.2. DENIZLI CITY WATER SUPPLY SYSTEM

Water distribution system in Denizli was commissioned in 1953. The water


supply system has been enlarged since then in proportion to the population
increase from 48,925 in 1960s to 275,480 in 2000s. In order to meet the
existing and future demand increase, multi million US dollars projects star-
ted in 1990s. In addition, Gokpinar Dam just next to Denizli was com-
pleted (Figure 3). Because of these investments, no water shortages are
expected in the near future.
Water is supplied to Denizli from four different sources (approximate
capacities are given in parenthesis): Derindere spring (250 l/s), Gokpinar
spring (1,100 l/s), Benlipinar spring (20 l/s), and many wells (400 l/s). The
springs are located outside the service area whereas the wells are located
within and outside of the service area. Water from these sources is col-
lected at water storage tanks, treated and released to the distribution sys-
tem. Some municipalities share the water brought from the same source.
Because of suitable topography of the area, the water distribution sys-
tem relies on gravity flow except for a few high elevation localities like
Camlik and Yenisehir. There are seven main pressure zones in the system
(Figure 3). These pressure zones are also divided into subzones, totaling
28. The subzones are shown with their identifying numbers in Figure 3.
The distribution system is designed so that the zones are independent of

Pressure Zones %
U
Water Distribution System AKKALE
Transmission and Connection Lines
Main Lines
Distribution Lines
% Water Tanks
U %
U
6.4 %
U
%
Municipality Boundaries U
6.3 1.6 %
U

%
U
6.5 %
U

UCLER %
U

GUMUSLER 1.5
%
U
6.1 6.2 1.4
%
U %
U N
BEREKETLI
%
U
CLAR 5.4 1.4
%
U
5.2
%
U
%
U %
5.3
%
U 1.2 DENIZLI GOKPINAR
%U
5.1
%
U
%
U
%U
U %
U DAM
%
U
%
U
4.4 %
U

%
U %
U 1.3
%
U
4.3 %
U

4.1 U%4.2 4.5


%
%
U
U

%
U
SERVERGAZI
% UU
%
%
U1.2
%
U%
U KAYHAN

%U
U %
U
%
U%
1.1 KINIKLI
% U
%
U 1.3 2.4
1 0 1 2 Kilometers %
U
BAGBASI GOKPINAR
2.3
%
U
2.2
U%
21 U
%U
%

Figure 3. Map of Denizli City water supply system.


ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 13

each other. As a result, the damage can be contained in a particular zone.


Figure 3 shows the map of the water distribution pipelines of all munici-
palities in Denizli City. Also shown in the figure are transmission pipelines
which bring water from the sources to the city, connection lines which con-
nect one water storage tank to another, main lines which transfer water
from tanks to distribution lines and the locations of water storage tanks.
The water supply system maps of overall Denizli City were provided by
Denizli Municipality. These maps included the 1993 maps and some older
maps. The 1993 maps were prepared as part of the water supply system
development project for overall Denizli area. The older maps show the CI
pipelines. We also obtained many drawings and documents from Denizli
and other Municipality Water Works which showed the placement and
replacement of pipelines in Denizli City up to 2005. All these data were
used to prepare digital maps and database of Denizli City water supply
system. The GIS program Arcview (ESRI , 1996) was used primarily dur-
ing the data input and analysis phases.
Figure 4 presents charts that show the relative lengths of pipelines in the
water supply system with respect to pipe diameter and composition. It
should be noted that the vertical scale in Figure 4a is logarithmic scale. The
figures were developed using GIS database of above mentioned maps, docu-
ments, and construction bids up to 2005. The total length of pipelines about
1745 km. About 95% of the transmission and connection lines are made of
steel pipelines. The main and distribution lines are made of AC, CI, and
polyvinyl chloride (PVC). About 54 and 44% of them are made of AC and
PVC, respectively. The distribution lines consist of pipelines with diameters
between 65 and 200 mm whereas the main lines consist of pipelines with
diameters between 100 and 600 mm. The CI pipelines are the oldest pipe-
lines in the system. They primarily serve to the old parts of Denizli which
include important local business districts with high population density. In
recent years, water works of municipalities adopted a policy of switching
from AC pipes to PVC pipes in new placements as well as replacements.
Therefore, PVC pipelines are the newest pipelines in the system.

3.3. PIPELINE DAMAGE PREDICTION FOR DENIZLI CITY WATER SUPPLY SYSTEM

To determine the effect of different size earthquakes on damage estima-


tions and also to find out the sensitivity of the damage relationships to dif-
ferent ground shaking levels, scenario earthquakes of M6.0, M6.3, M6.5,
and M7.0 were used in the current work. Two different fault ruptures, one
on Pamukkale Fault and the other on Karakova-Akhan Fault were con-
sidered. The analyses were performed by using GIS. The mean PGV con-
tours for each scenario earthquake were developed by dividing Denizli
map into 11 km grid system and determining the PGV at corners of each
14 SELCUK TOPRAK AND FILIZ TASKIN

10000

1000
Length (km)

100

10

1
65
75
80
100
125
150
175
200
250
300
350
400
500
600
700
800
1000
1100
Pipe Diameter (mm)

(a) Length of Pipelines with Respect to Pipe Diameter

Transmission and
Connection Lines 5% Main Lines Steel 4% PVC 42 %
21 %

Asbestos
Distribution Lines 74% Cast Iron 2 %
Cement 52%

(b) Relative Lengths of Pipelines with (c) Relative Lengths of Pipelines


Respect to Pipe Composition with Respect to Type

Figure 4. Composition statistics of pipelines in Denizli City water supply system.

grid. Contours then were drawn from the spatial distribution of PGV val-
ues. The geometric mean PGV values were calculated by using Campbell
attenuation relationship. To determine the maximum PGV if the damage
correlation needs it, the mean PGV values were multiplied by 1.21 as pro-
posed by O’Rourke and Jeon (2000) after their evaluation of the North-
ridge earthquake database. This conclusion is close to that of Campbell
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 15

(1997). Campbell worked on a worldwide earthquake data set and con-


cluded that the geometric mean PGV values were 17% less than the maxi-
mum PGV values.
Figure 5 shows the water supply system superimposed on PGV zones.
Because the faults are located to the northeast of Denizli, there is a re-
peated pattern of peak ground velocity oriented from the northeast to the
southwest of Denizli. Length of pipelines in each zone was calculated using
GIS. The repair rates corresponding to each PGV zone was determined
using pipeline damage correlations as explained in the previous section.
The number of repairs were then calculated for each PGV zone by multi-
plying the corresponding repair rates and length of pipelines. The repairs
for each PGV zone were determined for brittle (AC and CI) and ductile
pipelines (steel and PVC) separately. By summing the repairs for each
PGV, the total number of repairs for a particular earthquake scenario was
obtained.
In the analyses of damage predictions using HAZUS (FEMA, 1999),
Toprak (1998), O’Rourke and Jeon (2000) for geometric mean PGV, Pineda
and Ordaz (2003) and O’Rourke and Deyoe (2004) correlations, ductile
pipelines were assumed to have 30% of the vulnerability of brittle materials
following the recommendations of FEMA (1999). That is to say repair rates

Figure 5. Water supply system superimposed on the peak ground velocity zones from
M6.3 Pamukkale Fault rupture scenario earthquake.
16 SELCUK TOPRAK AND FILIZ TASKIN

for ductile pipelines for the same PGV is 30% of those for brittle pipelines.
For the analysis with other O’Rourke and Jeon models, CI and DI rela-
tionships were used, respectively, to determine repairs to brittle and ductile
pipelines. For ALA relationship, their suggested constants were used for
different pipelines.
Tables II and III present the pipeline damage estimations for Denizli
water supply system for different scenario earthquakes caused by Pamukkale
and Karakova-Akhan Fault ruptures, respectively. The total number of
repairs as well as the repairs for brittle and ductile pipelines were provided.
The results are shown for all damage relationships described in Table I.
HAZUS and Pineda and Ordaz (2003) relationships predict the highest
number of repairs for both fault rupture events. However, their result
should not be applicable to Denizli water supply system. As described
before, HAZUS relationship is significantly affected by Mexico earthquake
data and Pineda and Ordaz relationship was particularly derived from
Mexico City damage where surface wave effects dominate the damage. The
damage in Denizli is expected to occur as a result of secondary waves
rather than the surface waves.

3.4. DISCUSSION OF PIPELINE DAMAGE PREDICTION

Figure 6a, b show the variations in the predictions by the damage relation-
ships for each scenario earthquake. The results are presented for brittle,
ductile, and all pipelines. The percentile graphs use the data from Tables II
and III except the HAZUS and Pineda and Ordaz (2003) results. The low-
er boundary of the each box in the plots indicates the 25th percentile, a
line within the box marks the median, and the upper boundary of the box
indicates the 75th percentile. Whiskers above and below the boxes indicate
the 95th and 5th percentiles. The results show that the variation in brittle
pipeline damage predictions is less than that of ductile pipelines. The dif-
ferences among the brittle pipeline damage relationships significantly re-
duced after O’Rourke and Deyoe (2004) reevaluated HAZUS relationship
which is the most different one compared with other relationships. This
difference was pointed out previously in many publications, e.g., O’Rourke
and Jeon (2000). Considering the inherent variability in the pipeline sys-
tems such as pipe type, pipe diameter, soil conditions, and corrosion and
seismic uncertainties, Toprak (1998), O’Rourke and Jeon (2000), ALA
(2001), and O’Rourke and Deyoe (2004) for S-wave relationships are with-
in reasonable range of each other. For example at PGV of 20 cm/s, the
repair rates vary between 0.029 and 0.055 repairs/km and at PGV of
75 cm/s, the repair rate range is between 0.18 and 0.25 repairs/km.
They converge to a value of 0.1 repairs/km at PGV of about 45 cm/s. The
damage data for ductile pipelines in previous earthquakes were limited.
Table II. Pipeline damage prediction for Pamukkale Fault repture scenario.

Pipeline damage correlations Scenario earthquake

MW=6 MW=6.3 MW=6.5 MW=7

Brittle Ductile Total Brittle Ductile Total Brittle Ductile Total Brittle Ductile Total

HAZUS 109 28 137 189 48 237 268 66 334 510 123 634
Toprak (1998) 34 9 43 51 13 64 66 16 82 105 26 131
O’Rourke and Jeon (2000), 50 13 63 66 17 83 79 19 98 108 27 135
geometric mean PGV
O’Rourke and Jeon (1999, 47 21 68 63 33 96 77 42 119 109 71 180
2000), maximum PGV
O’Rourke and Jeon (1999, 65 48 113 84 67 151 98 83 181 133 122 255
2000), diameter scaled PGV
ALA (2001) 42 10 52 55 21 76 64 25 89 86 33 119
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES

Pineda and Ordaz (2003) 159 40 199 197 50 247 238 59 297 366 89 455
O’Rourke and Deyoe (2004) 58 14 72 73 18 91 85 21 106 111 27 138
17
18

Table III. Pipeline damage prediction for Karakova – Akhan Fault rupture scenario.

Pipeline damage correlations Scenario earthquake

MW=6 MW=6.3 MW=6.5 MW=7

Brittle Ductile Total Brittle Ductile Total Brittle Ductile Total Brittle Ductile Total

HAZUS 285 62 347 423 95 518 520 119 639 757 182 939
Toprak (1998) 69 16 85 92 21 113 107 25 132 140 34 174
O’Rourke and Jeon (2000), 81 19 100 99 23 122 109 26 135 131 32 164
geometric mean PGV
O’Rourke and Jeon (1999, 79 41 120 98 58 156 110 69 179 135 98 233
2000), maximum PGV
O’Rourke and Jeon (1999, 101 81 182 120 105 225 134 123 257 155 164 319
2000) diameter scaled PGV
ALA (2001) 66 24 90 79 29 108 86 33 119 102 40 142
Pineda and Ordaz (2003) 247 57 304 320 75 395 371 87 458 486 118 604
O’Rourke and Deyoe (2004) 87 21 108 102 25 127 107 25 132 131 32 163
SELCUK TOPRAK AND FILIZ TASKIN
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 19

350

Brittle Pipeline Ductile Pipeline All Pipeline


300

250
Number of Repairs

200

150

100

50

0
6 6.3 6.5 7 6 6.3 6.5 7 6 6.3 6.5 7
Earthquake Magnitude,Mw
(a) Pamukkale Fault Rupture
350
Brittle Pipeline Ductile Pipeline All Pipeline
300

250
Number of Repairs

200

150

100

50

0
6 6.3 6.5 7 6 6.3 6.5 7 6 6.3 6.5 7
Earthquake Magnitude, Mw

(b) Karakova – Akhan Fault Rupture

Figure 6. Pipeline damage estimations for scenario earthquakes.

Hence, the relationships have been generally obtained for brittle pipelines.
It is accepted from theoretical and observational point of view that ductile
pipelines are less prone to damage during earthquake shaking. Therefore,
20 SELCUK TOPRAK AND FILIZ TASKIN

brittle pipeline damage relationships were multiplied by a constant less


than one to determine the repair rates for ductile pipelines. For example,
HAZUS uses a factor of 30%. Although the Northridge earthquake dam-
age data set provided considerable data points for ductile pipeline relation-
ships, the data were still limited compared with the data of brittle
pipelines. The results shown in Figure 6a,b reflect this difference in the
damage relationships for brittle and ductile pipelines.
To show this point, two relationships can be compared using Denizli
water distribution pipeline damage predictions: O’Rourke and Jeon
(1999) geometric mean and maximum PGV relationships. As shown in
Table I, O’Rourke and Jeon maximum PGV relationships are available
for brittle and ductile pipelines whereas O’Rourke and Jeon geometric
mean PGV relationship is available only for brittle pipelines. To calculate
the ductile pipeline damage for the geometric mean PGV relationship,
HAZUS approach was used. Brittle pipeline damage estimations using
both models are very close (Tables II and III). However, significant dif-
ferences occur between the estimations for ductile pipeline damage chang-
ing by a factor of 2–3 depending on the size of the earthquake. This
illustrates that if the differences in various model’s approach for ductile
pipelines can be reconciled by additional data from the most recent
earthquakes, the total differences between the damage relationships could
be reduced substantially.
Figure 7 shows damage estimations for each relationship in terms of
number of repairs. The lower boundary of each bar corresponds to dam-
age estimates for M6.0 and the upper boundary corresponds to estimates
for M7.0. In essence, each bar shows the damage range predicted by each
relationship for scenario earthquake magnitudes between M6.0 and M7.0.
The predicted number of repairs increases by a factor of about 1.5–2.5 as
earthquake magnitude, in other words the shaking intensity increases. This
result indicates that the better the seismic intensity level is predicted for a
particular region the more realistic the damage estimates can be. Compari-
son of Figure 7 with Figure 6a,b for all pipelines shows that the variations
in damage estimations as a result of using different ground shaking levels
or different damage relationship are comparable.

4. Conclusions
Substantial progress have been made on the improvement of pipeline dam-
age relationships in the last decade. With the recent developments, the dif-
ferences between various damage relationship were reduced significantly.
The developments had important ramifications for officials who work to
reduce the risk from earthquakes and to prepare for emergency response
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 21

350
Pamukkale Fault Rupture
Mw7
300 Karakova-Akhan
Fault
250

200
Mw6
150

100

50

0
Toprak (1998)

O’Rourke and Jeon (2000)

O’Rourke and Jeon (1999,


2000) Diameter Scaled

ALA (2001)

O’Rourke and Deyoe


Geometric Mean PGV

Velocity

(2004)
O’Rourke and Jeon (1999,
2000) Maximum PGV

Figure 7. Pipeline damage estimation by the relationships for scenario earthquake


magnitudes between M6 and M7.

and recovery from an earthquake. This study shows that more need to be
done to reconcile the differences. To develop damage correlations, which
are representative of different earthquakes and localities, reliable informa-
tion should be available for pipe diameter and composition, repair loca-
tions, soil conditions and strong motion. In this regard, it is important to
speed up or encourage the efforts for development of databases for pipe-
lines and strong ground motion. Local soil conditions along with seismic
wave-propagation velocity should be part of the databases. Geographical
information systems are an important and useful tool to serve this purpose
whether it is at input or analyses stage. It is also important to get the best
possible damage data after the earthquakes. It should be ensured that
pipeline damage be collected with specific details during earthquake recon-
naissance studies.
22 SELCUK TOPRAK AND FILIZ TASKIN

Acknowledgements
The research reported in this paper was supported by Pamukkale Univer-
sity Research Center under Project No. 2003MHF006. The authors wish to
express their deep gratitude to Chief and employees of Water Works at
various municipalities, especially Denizli, Gumusler and Kinikli and Asst.
Prof. A. C. Koc of Civil Engineering Department for their interest and
assistance in assembling the water system data. Thanks are also extended
to Mehmet Genc of General Directorate of Disaster Affairs, Denizli
Branch, and Geology Department members of PAU, especially Prof. Halil
Kumsar for providing base maps for Denizli area and for their valuable
comments and discussions. S. Celik and S. Topal prepared the digital ver-
sion of the geology map. The writers also thank two anonymous reviewers
who suggested changes that clarified the presentation. The second writer
acknowledges the financial support provided by the Scientific and Techno-
logical Research Council of Turkey (TUBITAK).

References
Altunel, E.: 2000, Historical earthquake activity in and around Hierapolis, Ricerche
Archeologiche Turche Nella Valle Del Lykos, Lykos Vadisi Turk Arkeoloji Arastirmalari.
Congedo Editore (in Italian and Turkish). In: F. D’Andria and F. Silvestrelli (eds.),
pp. 299–325.
ALA (American Lifelines Alliance): 2001, Seismic fragility formulations for water systems,
Part 1-Guideline, http://www.americanlifelinesalliance.org.
ATC 13: 1985, Earthquake Damage Evaluation Data for California, Applied Technology
Council, Redwood City, CA.
Ayala, A. G. and O’Rourke, M. J.: 1989, Effect of the 1985 Michoacan Earthquake on water
system and other buried lifelines in Mexico, Buffalo, NY, NCEER 89-009.
Aydan, O., Kumsar, H. and Ulusay, R.: 2001, Denizli ve yakın cevresindeki jeotermal
kaynakların ve Denizli fayı zonundaki sıcaklık değisimlerinin yerkabuğundaki değisimlerle
_
ilgili olabilir mi? Jeoteknik III, 3. Izmir ve Çevresinin Deprem ve Jeoteknik Sempozyumu,
November 12–14, Izmir,_ Turkey (in Turkish).
Aydan, O., Kumsar, H. and Ulusay, R.: 2002, How to infer the possible mechanism and
characteristics of earthquakes from the striations and ground surface traces of existing
faults, Structural Eng./ Earthquake Eng. 19, 199–208.
Ballantyne, D. B., Berg, E., Kennedy, J., Reneau, R., Wu, D., Taylor, C. E., Crouse, C. B.,
Eguchi, R. and Tillman, C.: 1990, Earthquake loss estimation modeling of the Seattle
water system. Report No. 886005.00, Kennedy/Jenks/Chilton, Federal Way, Washington,
D.C.
Barenberg, M. E.: 1989, Correlations of pipeline damage with ground motions, J. Geotechnic.
Eng. 114, 706–711.
Campbell, K. W.: 1997, Empirical near-source attenuation relationships for horizontal and
vertical components of peak ground acceleration, peak ground velocity, and pseudo-
absolute acceleration response spectra, Seism. Res. Lett. 68, 154–179.
Committee on gas and liquid fuel lifelines: 1984, Guidelines for the Seismic Design of Oil and
Gas Pipeline Systems, ASCE, Newyork, NY.
ESTIMATION OF EARTHQUAKE DAMAGE TO BURIED PIPELINES 23

Eguchi, R. T.: 1982, Earthquake performance of water supply components during the 1971
San Fernando Earthquake. Technical Report 1396-2a, Wiggins JH Company, Redondo
Beach, CA.
Eguchi, R. T.: 1991, Seismic hazard input for lifeline systems. In: E. H. Vanmarcke (ed.),
Structural Safety, Vo1.10. Elsevier Science Publishers, pp. 193–198.
Eguchi, R. T., Philipson, L. L., Legg, M. R., Wiggins, J. H. and Slosson, J. E.: 1981,
Earthquake vulnerability of water supply systems. Technical Report No. 80-1396-3,
J. H. Wiggins Company, Redondo Beach, CA.
Eguchi, R. T., Taylor, C. and Hasselman, T. K.: 1983, Seismic component vulnerability
models for lifeline risk analysis, Technical Report No. 82-1396-2c, J.H. Wiggins Company,
Redondo Beach, CA.
Environmental Systems Research Institute (ESRI).: 1996, Using ArcView GIS, Environmental
Systems Research Institute Inc., Redlands.
Erel, B., Patelunas, G. M., Niece, J. E. and Oppenheim, I. J.: 1977, Measuring the earthquake
performance of urban water systems. In: Proceedings, Technical Council on Lifeline
Earthquake Engineering Specialty Conference, Los Angeles, CA, pp.183–198.
FEMA (Federal Emergency Management Agency): 1999, Earthquake loss estimation
methodology, HAZUS 99: technical manual, prepared by the National Institute of
Building Sciences.
General Directorate of Mineral Research and Exploration (MTA).: 2004,
http://www.mta. gov.tr.
Gori, P. L. and Hays, W.W.: 1988, Assessment of regional earthquake hazards and risks along
the Wasatch, Front, Utah. Open-File Report 88-680. U.S. Coast and Geodetic Survey,
Washington, D.C.
Hancock, P. L. and Altunel, E.: 1997, Faulted archaeological relics at Hierapolis (Pamukkale),
Turkey. J. Geodynamics. 24, 1–4: 21–36.
Kandilli Observatory and Earthquake Research Institute (KOERI).: 2004, http://www.
koeri.boun.edu.tr.
Katayama, T., Kubo, K. and Sato, N.: 1975, Earthquake damage to water and gas
distribution systems. In: Proceedings, U.S, National Conference on Earthquake Engineer-
ing, Ann Arbor, MI, pp. 396–405.
O’Rourke, T. D.: 1985, Factors affecting the performance of cast iron pipelines, A review of
U.S. observations and research investigations. Contractor Report 18, Transport and Road
Research Laboratory, Crowthorne, U.K.
O’Rourke, M. J. and Ayala, G.: 1993, Pipeline damage due to wave propagation,
J. Geotechnic. Eng. 119, 1490–1498.
O’Rourke, M. and Deyoe, E.: 2004, Seismic damage to segmented buried pipe, Earthquake
Spectra 20, 1167–1183.
O’Rourke, T. D. and Jeon, S. S.: 1999, Factors affecting the earthquake damage of
water distribution systems, Optimizing post-earthquake lifeline system reliability. In:
W. M. Elliott and P. McDonough (eds.), Proceedings Fifth U.S. Conference on Lifeline
Earthquake Engineering, ASCE, Seattle, WA, pp. 379–388.
O’Rourke, T. D. and Jeon, S.S.: 2000, Seismic zonation for lifelines and utilities, Invited
Keynote Paper on Lifelines. In: Proceedings Sixth International Conference on Seismic
Zonation, Palm Springs, EERI CD ROM, CA.
O’Rourke, T. D. and Toprak, S.: 1997, GIS assessment of water supply damage from the
Northridge earthquake, In: J. D. Frost (ed.), Geotechnical Special Publication, ASCE,
New York, NY, pp. 117–131.
24 SELCUK TOPRAK AND FILIZ TASKIN

O’Rourke, T. D., Toprak, S. and Sano, Y.: 1998, Factors affecting water supply damage
caused by the Northridge earthquake, In: Proceedings of the 6th US National Conference
on Earthquake Engineering, Seattle, WA, USA, pp. 1–12.
PAU (Pamukkale University): 2002, Geological, geotechnical, and hydrogeological properties
of Denizli Municipality lands, Geology Department, Engineering Faculty, Pamukkale
University (in Turkish).
Patelunas, G. M., Erel, B. and Thiers, G. R.: 1977, Vulnerability of Urban Water Systems to
Seismic Hazard, GAI Consultants, Inc., Monroeville, PA, Project 74-536-4.
Pease, J. W. and O’Rourke, T. D.: 1997, Seismic response of liquefaction sites, J. Geotechn.
Geoenviron. Eng. 123, 37–45.
Pineda, O. and Ordaz, M.: 2003, Seismic vulnerability function for high-diameter buried
pipelines: Mexico City’s primary water system case, ASCE Int. Conf. Pipeline Eng.
Construction 2, 1145–1154.
Sun, S.: 1990, Geology and coal resources between Denizli and Usak, MTA Report No 9985.
Toprak, S.: 1998, Earthquake Effects on Buried Lifeline Systems, PhD Thesis. Cornell
University, Ithaca, NY.
Trifunac, M. D. and Todorovska, M. I.: 1997, Northridge, California, earthquake of 1994:
Density of pipe breaks and surface strains, Soil Dynamics Earthquake Eng. 16, 193–202.

View publication stats

You might also like