Issue 8, 2013 Journal of Civil Engineering and Architecture

Journal of Civil Engineering
and Architecture
Volume 7, Number 8, August 2013 (Serial Number 69)
David Publishing Company
www.davidpublishing.com
Publishing David

Publication Information:
Journal of Civil Engineering and Architecture (ISSN 1934-7359) is published monthly in hard copy and online by
David Publishing Company located at 9460 Telstar Ave Suite 5, EL Monte, CA 91731, USA.

Aims and Scope:
Journal of Civil Engineering and Architecture, a monthly professional academic journal, covers all sorts of researches
on structure engineering, geotechnical engineering, underground engineering, engineering management, etc. as well
as other issues.

Editorial Board Members:
Dr. Tamer A. El Maaddawy (Canada), Prof. San-Shyan Lin (China Taiwan), Dr. Songbai Cai (China), Dr. Xiaoyan Lei
(China), Prof. Vladimir Patrcevic (Croatia), Dr. Sherif Ahmed Ali Sheta (Egypt), Prof. Nasamat Abdel Kader (Egypt),
Prof. Mohamed Al-Gharieb Sakr (Egypt), Prof. Olga Popovic Larsen (Denmark), Prof. George C. Manos (Greece), Dr.
Konstantinos Giannakos (Greece), Pakwai Chan (Hong Kong), Dr. K. Muthukkumaran (India), Chiara Vernizzi (Italy),
Prof. Michele Maugeri (Italy), Dr. Giovanna Vessia (Italy), Prof. Valentina Zileska-Pancovska (Macedonia), Dr. J.
Jayaprakash (Malaysia), Mr. Fathollah Sajedi (Malaysia), Prof. Nathaniel Anny Aniekwu (Nigeria), Dr. Marta Sowik
(Poland), Dr. Rafael Aguilar (Portugal), Dr. Moataz A. S. Badawi (Saudi Arabia), Prof. David Chua Kim Huat
(Singapore), Dr. Vail Karakale (Waiel Mowrtage) (Turky), Dr. A.Senem Deviren (Turkey), Dr. Yasemin Afacan
(Turkey), Dr. Ming An (UK), Prof. Ahmed Elseragy (UK), Prof. Jamal Khatib (UK), Dr. John Kinuthia (UK), Dr.
Johnnie Ben-Edigbe (UK), Dr. Yail Jimmy Kim (USA), Dr. Muang Seniwongse (USA), Prof. Xiaoduan Sun (USA), Dr.
Zihan Yan (USA), Dr. Tadeh Zirakian (USA).

Manuscripts can be submitted via Web Submission, or E-mail to civil@davidpublishing.com or
civil@davidpublishing.org. Submission guidelines and Web Submission system are available at
http://www.davidpublishing.com, www.davidpublishing.org.

Editorial Office:
9460 Telstar Ave Suite 5, EL Monte, CA 91731, USA
Tel: 1-323-984-7526, 323-410-1082 Fax: 1-323-984-7374, 323-908-0457
E-mail: civil@davidpublishing.com; civil@davidpublishing.org; shelly@davidpublishing.com.

Copyright2013 by David Publishing Company and individual contributors. All rights reserved. David Publishing
Company holds the exclusive copyright of all the contents of this journal. In accordance with the international
convention, no part of this journal may be reproduced or transmitted by any media or publishing organs (including
various websites) without the written permission of the copyright holder. Otherwise, any conduct would be
considered as the violation of the copyright. The contents of this journal are available for any citation. However, all
the citations should be clearly indicated with the title of this journal, serial number and the name of the author.

Abstracted / Indexed in:
Database of EBSCO, Massachusetts, USA
Chinese Database of CEPS, Airiti Inc. & OCLC
Cambridge Science Abstracts (CSA)
Ulrichs Periodicals Directory
Summon Serials Solutions
ProQuest

Subscription Information:
$520/year (print) $360/year (online) $680/year (print and online)

9460 Telstar Ave Suite 5, EL Monte, CA 91731, USA
Tel: 1-323-984-7526, 323-410-1082 Fax: 1-323-984-7374, 323-908-0457
E-mail: civil@davidpublishing.com; civil@davidpublishing.org; shelly@davidpublishing.com.

www.davidpublishing.com
DAVID PUBLISHING
D

Jour nal of Ci vi l Engi neeri ng
and Archi t ect ure

Volume 7, Number 8, August 2013 (Serial Number 69)

Contents
Construction Materials Research
907 Freezing and Thawing Durability of Ultra High Strength Concrete
Jesus Muro-Villanueva, Craig M. Newtson, Brad D. Weldon, David V. Jauregui and Srinivas Allena
916 Enhancing the Mechanical Properties of Gap Graded Cold Asphalt Containing Cement Utilising
By-Product Material
Abbas Al-Hdabi, Hassan Al Nageim, Felecite Ruddock and Linda Seton
922 Deterioration of Concrete by the Oxidation of Sulphide Minerals in the Aggregate
Jose Duchesne and Benot Fournier
932 Behavior of Millstone Masonry Vaults Strengthened by Composite Materials
Maamoun Saade, Stephan Kesteloot, Chafika Djelal, Lotfi Hamitouche and Idriss Benslimane
940 Corrosion Study and Passive Film Characterization of 11% Cr F/M and 15% Cr ODS Steels
Masatoshi Sakairi, Sublime Ningshen, Keita Suzuki and Shigeharu Ukai
Traffic and Transportation
956 Effect of Traffic Speed on Stresses and Strains in Asphalt Perpetual Pavement
Daba S. Gedafa, Mustaque Hossain and Stefan A. Romanoschi
964 Stresses Analysis on a Rail Part
Cassio Eduardo Lima de Paiva, Jos Luz Antunes de Oliveira e Sousa, Luiz Carlos de Almeida, Paulo
Roberto Aguiar, Luiz Fernando de Melo Correia, Juliana Silva Watanabe, Rodrigo Moreira de Carvalho,
Creso de Franco Peixoto and Denis Palomo Paschoalin

973 Advanced Methods of Evaluating Benefits from Improved Flood Immunity in Queensland
Wayne Davies
Earthquake Engineering
992 Policy Analysis on Building Regulations and the Recovery of Earthquake and Tsunami Affected
Areas
Shoichi Ando
1006 Time Dependent Gaussian Equivalent Linearization of Duffing Oscillator Using Continuous
Wavelet Transform
Arunasis Chakraborty, Prateek Mittal and Sabarethinam Kameshwar
1018 An Earthquake Catalogue for El Salvador and Neighboring Central American Countries
(1528-2009) and Its Implication in the Seismic Hazard Assessment
Walter Salazar, Lyndon Brown, Walter Hernndez and Jos Guerra

Aug. 2013, Volume 7, No. 8 (Serial No. 69), pp. 907-915
J ournal of Civil Engineering and Architecture, ISSN 1934-7359, USA

Freezing and Thawing Durability of Ultra High Strength
Concrete
J esus Muro-Villanueva
1
, Craig M. Newtson
1
, Brad D. Weldon
1
, David V. J auregui
1
and Srinivas Allena
2

1. Civil Engineering Department, New Mexico State University, Las Cruces NM 88003, USA
2. Civil and Environmental Engineering Department, Washington State University, Richland WA 99354, USA

Abstract: Resistance to freezing and thawing of two UHSC (ultra high strength concrete) mixtures was evaluated in accordance with
ASTM C 666 Procedure A. The two mixtures (plain and fiber reinforced) were developed using materials local to southern New
Mexico, USA. Three different curing regimens were investigated for the mixture with fibers and one curing regimen was studied for the
mixture without fibers. All curing regimens included 24 h of ambient curing followed by four days of wet curing at 50
o
C, and then two
days dry curing at 200
o
C. At an age of seven days, one batch of fiber reinforced specimens was air cured at ambient conditions for the
following six days and then placed in a water bath at 4.4
o
C for 24 h prior to initiating freezing and thawing cycles. The second batch
was air cured from day seven to day 12, and then wet cured for one day at 23
o
C prior to being placed in the 4.4
o
C water bath. The final
batch was wet cured at 23
o
C from the seventh day to an age of 13 days and then placed in the 4.4
o
C water bath. The mixture with no
fibers was air cured from the seventh day to an age of 12 days and then wet cured for one day at 23
o
C prior to being placed in the 4.4
o
C
water bath. Higher moisture levels during curing produced greater initial dynamic elastic modulus values and durability factors at the
end of the freezing and thawing tests, with the greatest durability factor being 87.5. Steel fibers were observed to improve both
compressive strength and durability factor for UHSC.

Key words: Ultra high strength, freezing and thawing, durability, dynamic elastic modulus, quality factor.

1. Introduction

UHSC (ultra high strength concrete) is an advanced
fiber reinforced composite material, characterized by
compressive strengths greater than 140 MPa and
flexural strengths greater than 10 MPa at 28 days [1].
UHSC is produced with a high cementitious materials
content and a water-to-cementitious materials ratio
(w/cm) that is typically less than 0.25. Additionally,
the use of silica fume and HRWRA (high range water
reducing admixtures) along with pre-setting pressure
and post-setting heat treatments have been used to
produce high density concrete [1, 2]. A dense
microstructure results in very high compressive
strengths, including values greater than 200 MPa [3].
Along with the increase in strength, UHSC mixtures

Corresponding author: Craig M. Newtson, Ph.D., research
fields: concrete materials, reuse/recycling of materials, concrete
shrinkage, concrete durability, ultra-high performance concrete
and nondestructive evaluation of concrete. E-mail:
newtson@nmsu.edu.
may need to exhibit greater durability to justify the
high cost of the material. However, since UHSC is not
air-entrained, its ability to withstand cycles of freezing
and thawing is questionable and should be
investigated.
This paper presents results from freezing and
thawing tests conducted on two UHSC mixtures to
ensure that UHSC without air entrainment would
prove to be durable. The mixtures used in this study
were developed by Allena and Newtson [1] using
materials local to southern New Mexico, USA. Use of
local materials is an improvement in the sustainability
of UHSC over pre-packaged, commercially available
UHSC products because they offer substantial cost
savings. However, the chemical and physical
requirements for local materials are not as strict as for
commercial products. Additionally, no air-entraining
admixture is used in the UHSC mixtures.
Consequently, it is even more important to assess the
durability of UHSC produced with local materials to
DAVID PUBLISHING
D
Freezing and Thawing Durability of Ultra High Strength Concrete

908
determine whether or not it is adequate.
2. Background
2.1 ASTM C 666 Procedure A
ASTM C 666 assesses concrete resistance to
freezing and thawing by subjecting concrete
specimens to rapid repeated freezing and thawing
cycles under laboratory conditions. The specimens are
molded beams that are cured for 14 days prior to
testing. On the last day of curing, the specimens are
subjected to a 24 h conditioning period in a 4.4
o
C
water bath. After the conditioning period, the
specimens are subjected to freezing and thawing
cycles that are approximately 4 h in duration.
At intervals not to exceed 36 cycles, the specimens
are removed from the freezing chamber during the
thawed portion of the cycle for dynamic elastic
modulus testing. Fundamental transverse frequency
(used to compute dynamic elastic modulus), length
and mass are measured. Testing of the specimens
continues until each of the beams has been subjected
to at least 300 cycles of freezing and thawing, or until
the dynamic modulus of elasticity decreases to 60% of
its initial value or less.
2.2 Dynamic Elastic Modulus
Testing procedures for determining dynamic elastic
modulus are specified in ASTM C 215. To determine
the dynamic elastic modulus, concrete prisms are
excited over a wide range of frequencies using an
impact hammer or a transducer to identify the
frequency at which the maximum amplitude occurs.
The frequency at which the maximum amplitude
occurs is the resonant frequency. The concrete
specimen is generally assumed to be a
single-degree-of-freedom system, for which the
resonant frequency is referred to as the fundamental
frequency. The fundamental frequency is used to
compute dynamic elastic modulus, E
D
, with the
following equation:
2
r D
Cm E
(1)
where, C is a constant that accounts for Poissons ratio
and the geometry of the specimen, m is the mass of
the specimen and
r
is the measured fundamental
frequency.
When ASTM C 215 is used to monitor deteriorating
concrete, it is common to present results in terms of
the relative dynamic modulus, computed as follows:
) 100 (
o
n
E
E
RDM
(2)
where, RDM is the relative dynamic modulus after n
cycles of freezing and thawing, E
n
is the dynamic
elastic modulus after n cycles, and E
o
is the dynamic
elastic modulus at zero cycles of freezing and thawing.
After completion of freezing and thawing cycles, a
durability factor, DF, can be computed as:
M
N RDM
DF

(3)
where, N is the number of cycles imposed, and M is
the specified number of cycles (usually 300).
2.3 Quality Factor
When concrete is subjected to cycles of freezing
and thawing, degradation can be caused by distributed
microcracking in the concrete [4]. This microcracking
results in increased damping in addition to a decrease
in dynamic elastic modulus. Damping can be
quantified by the quality factor, Q, which can be
determined from a frequency response curve. An
example of a frequency response curve is plotted in
Fig. 1. Since a wide range of frequencies are excited
by the impact method, amplitudes at frequencies other
than the fundamental frequency are known. This
allows the quality factor, to be computed with the
following relationship:

1 2

r
Q
(4)
where,
1
and
2
are the frequencies on either side of
r
with amplitudes equal to 70.7% of the amplitude of
r
(Fig. 1). Quality factor is inversely proportional to
the damping ratio, , as shown in the following
equation:

2
1
Q
(5)

909

Fig. 1 Frequency response curve.

Relative quality factor, calculated with an
expression similar to Eq. (2), has been shown to be
useful for describing deterioration caused by freezing
and thawing [4, 5].
2.4 Ultra High Strength Concrete
Shaheen et al. [6] conducted a study on UHSC
mixtures with seven day compressive strengths up to
500 MPa. The mixtures included different carbon
fibers contents, presetting pressures, curing processes
and heat-treating temperatures. Results from ASTM C
666 Procedure A testing revealed that UHSC mixtures
with carbon fibers had durability factors of 100, which
represents no decrease in dynamic elastic modulus
after testing. Conversely, mixtures without fibers
tended to show a decrease in dynamic elastic modulus
[6].
Other freezing and thawing studies [7, 8] have used
various temperature ranges, specimen dimensions and
cycle durations. Results showed relative dynamic
modulus of 100 after up to 800 cycles of freezing and
thawing. The percentage of mass lost during testing on
both studies was extremely low or zero. It was
concluded that the UHSC mixture provided excellent
frost resistance.
Graybeal [9] presented results from ASTM C 666
Procedure A testing that showed that various steam
curing regimens as well as duration of exposure to
each treatment greatly influence UHSC durability
factors. Mixtures exposed to longer wet curing
regimens produced greater initial dynamic elastic
modulus, but tended to produce lower durability
factors.
Studies on curing of high performance concrete
mixtures have shown that the microstructure of the
surface concrete can be disturbed by the curing
method, resulting in a reduction in durability [10].
High performance concrete mixtures with silica fume
are more sensitive to curing methods than mixtures
without silica fume due to microcracking caused by
desiccation [10]. Hasni et al. [10] presented results
that show that high performance concrete mixtures
with high silica fume contents have less
microcracking when cured at a higher relative
humidity.
Non air-entrained concrete mixtures along with low
water to binder ratios may have satisfactory durability
against freezing and thawing if a compressive strength
of at least 24 MPa is achieved prior to the first
freezing cycle [10]. The use of a low water to binder
ratio ensures that all of the mixing water combines
with the cementitious material, resulting in a concrete
with low permeability that prevents critical saturation
of the paste. Therefore, the presence of an air void
system in the concrete is not essential [11].
3. Experimental Methods
The two UHSC mixtures used in this study were
developed by Allena and Newtson [1]. Compressive
strengths obtained from 51 mm cubes at 28 days were
170.3 MPa for the fiber reinforced UHSC mixture and
159.9 MPa for the plain UHSC mixture.
3.1 Materials
Type I-II portland cement, silica fume and local
alluvial sand from Las Cruces, New Mexico, USA
were used to produce the UHSC mixtures in this
research. Steel fibers with a length of 13 mm were
used to improve ductility. Workability was obtained
using a polycarboxylate-based HRWRA (Glenium
Frequency
A
m
p
l
i
t
u
d
e
1
1.0
0.707
r

2

910
3030 NS from BASF).
The local sand had a specific gravity of 2.53 and an
absorption of 1.2%. Table 1 presents the grain size
distribution of the sand. The sand was sieved to retain
particles between ASTM #30 and #200 (600 m and
75 m) sieves. The sand was first passed through a No.
30 (600 m) sieve and then washed over a No. 200
(75 m) sieve. The sand was then oven dried and
placed in an air tight container to ensure 0% moisture
content.
3.2 Concrete Mixtures
Two different concrete mixtures (with and without
fibers) were tested for freezing and thawing resistance
according to ASTM C 666 and for compressive
strength according to BS 1881. Saturated surface-dry
mixture proportions for both mixtures are presented in
Table 2.
3.3 Mixing and Curing
The dry constituents of each mixture were mixed
for 15 min using a 20 L pan mixer. Then, 75% of the
water was added and mixing continued for 5 min. The
HRWRA was then added, and mixing continued for
another 5 min. Next the remaining 25% of the water
was added and mixing continued for another 10 min.
The final step in the mixing process was to add the
steel fibers and mix for another 5 min.
Slump measurements were performed for both
UHSC mixtures in accordance with ASTM C 143.
The fiber reinforced UHSC mixture produced a slump
of 235 mm while the plain UHSC mixture had a
slump of 180 mm.
Freezing and thawing specimens were prepared
according to ASTM C 192. Each prism was prepared
in two layers. Consolidation was achieved by rodding
each layer every 1,400 mm
2
of the surface using a 16
mm rod. After specimens were rodded, they were
externally vibrated using a table vibrator to improve
consolidation. Theoretical densities for the UHSC
mixtures were 2,286 kg/m
3
for fiber reinforced UHSC
and 2,205 kg/m
3
for plain UHSC. Density of the
hardened UHSC was measured at 24 h and was found
to be 2,343 kg/m
3
for fiber reinforced UHSC and
2,261 kg/m
3
for plain UHSC. These values were
increases of 2.49% and 2.55% from the theoretical
density for the fiber reinforced and plain UHSC
mixtures, respectively.
A total of four 76 mm 102 mm 406 mm prisms
were produced from each batch for ASTM C 666
testing. The fiber reinforced UHSC mixture was used
to produce a total of 12 prisms (three batches) while
the plain UHSC mixture was used to produce four
prisms. In addition to the prisms, four 102 mm and
eight 51 mm cubes were prepared from each batch for
compressive strength testing at 14 days.
The specimens were cured for 14 days prior to
testing. During the first seven days, the curing
regimen was the same for all of the batches. The
specimens remained in their molds for 24 h at room
temperature 25
o
C. After demolding, the specimens
were heat cured in a waterbath at 50
o
C for the
following four days. Then, they were dry cured in an
oven at 200
o
C for two days. Curing at these elevated
temperatures to an age of seven days was necessary to
achieve the desired strength.

Table 1 Grain size distribution of sand.
Sieve No. Sieve size (m) Passing (%)
30 600 100.0
50 300 31.7
100 150 2.9
200 75 0.0
Table 2 Saturated surface-dry mixture proportions of UHSC mixtures.
Mixture
Cement
(kg/m
3
)
Silica fume
(kg/m
3
)
Fine sand
(kg/m
3
)
Steel fibers
(kg/m
3
)
Water
(kg/m
3
)
HRWRA
(l/m
3
)
w/c w/cm
Fiber reinforced 890 222 799 119 222 29.6 0.25 0.20
Plain 890 222 837 - 222 29.6 0.25 0.20


911

From an age of seven days to an age of 14 days,
three different curing regimens were investigated to
identify practices that might improve UHSC durability.
The three curing regimens that seemed most likely to
be used in practice were:
Regimen 1: Specimens were air cured for six days
at room temperature, 25
o
C, followed by 24 h of
conditioning in a water bath at 4.4
o
C;
Regimen 2: Specimens were air cured for five
days at 25
o
C, followed by one day of moist curing at
23
o
C, and then 24 h conditioning in a water bath at
4.4
o
C;
Regimen 3: Specimens were moist cured for six
days at 23
o
C, followed by 24 h of conditioning in a
water bath at 4.4
o
C.
Regimens 1 through 3 were used on the fiber
reinforced mixture, but only regimen 2 was used for
the plain UHSC mixture.
3.4 Testing Program
Cube specimens were tested for compressive
strength at 14 days according to BS 1881. Prism
specimens were tested for resistance to freezing and
thawing according to ASTM C 666 Procedure A. In
accordance with this testing method, a 3-h 31-min
cycle was selected. One full cycle of freezing and
thawing consisted of rapid temperature decrease from
4.4
o
C to -17.8
o
C in approximately 2 h 20 min. The
temperature was then held constant for 8 min before
raising the temperature back to 4.4
o
C in 53 min. The
temperature was then held constant at 4.4
o
C for 10
min. Testing was conducted during this 10 min
window.
The specimens were subjected to 300 cycles of
freezing and thawing (or less when the relative
dynamic elastic modulus of a specimen dropped
below 60). Length, mass and fundamental frequency
(to compute dynamic elastic modulus) measurements
were taken at intervals no greater than 36 cycles.
Distilled water was used in the conditioning water
bath and for immersion of the specimens during
freezing and thawing cycles.
4. Results and Discussion
4.1 Compressive Strength
Compressive strengths at 14 days for the fiber
reinforced and plain UHSC mixtures are presented in
Table 3. These values are less than the 28 day
strengths (170.3 MPa for fiber reinforced UHSC and
159.9 MPa for plain UHSC) reported by Allena and
Newtson [1] because testing was conducted at 14 days.
Fiber reinforced UHSC mixtures produced greater 14
day strengths than the plain UHSC mixture for
batches cured under the same regimen. The increase in
compressive strength caused by the fibers was
approximately 11% for 51 mm cubes and 17% for 102
mm cubes. It was observed that batches with wetter
curing regimens between ages of seven and 14 days
did not produce greater strength.
4.2 Freezing and Thawing
Initial dynamic elastic modulus results are also
presented in Table 3. Mixtures exposed to longer
periods of wet curing produced greater initial dynamic
elastic modulus values than mixtures with drier curing
regimens. Curing regimen 3 produced a 25% greater
initial dynamic elastic modulus than curing regimen 1

Table 3 Average compressive strength and initial dynamic elastic modulus values.
Mixture Curing regimen
Compressive strength (MPa)
Initial dynamic elastic
modulus (GPa)
102 mm cubes 51 mm cubes
Fiber reinforced 1 130.2 141.1 32.4
2 136.0 146.6 34.8
3 132.8 137.1 40.8
Plain 2 123.0 124.9 34.4


912

for the fiber reinforced UHSC. Fiber content did not
appear to influence initial values of dynamic elastic
modulus since the fiber reinforced and plain UHSC
mixtures had similar values when curing regimen 2
was used.
Figs. 2-4 show the relative dynamic modulus values
versus cycles of freezing and thawing for individual
specimens from each of the three curing regimens for
the fiber reinforced UHSC mixture. Results show a
consistent decline in relative dynamic modulus
throughout the duration of the testing. However,
specimens with drier curing regimens experienced a
more rapid decline in relative dynamic modulus than
specimens with wetter curing regimens. Wetter curing
regimens also seemed to produce less variation
between the relative dynamic modulus of specimens
within the same group.
Fig. 5 presents relative dynamic modulus values
versus cycles of freezing and thawing for the plain
UHSC mixture. This mixture was only tested under
curing regimen 2. The initial dynamic elastic modulus
values for this mixture were nearly the same as those
from the fiber reinforced UHSC mixture cured under
regimen 2. However, relative dynamic modulus of the
plain UHSC mixture declined more abruptly than the
relative dynamic modulus of the fiber reinforced
specimens cured according to regimen 2. Average
values for each of the four mixture and curing regimen
combinations investigated in this study are provided in
Fig. 6. As observed in Figs. 2-5, relative dynamic
modulus declined more rapidly for drier curing
regimens, as well as in the mixture without steel fibers.
This behavior is consistent with observations by
Graybeal [9] on specimens with wet versus dry curing
regimens.
The fiber reinforced mixture with curing regimen 1
reached an average relative dynamic modulus of 60 at
cycle 196. According to ASTM C 666 it is considered
to have failed. The average relative dynamic modulus
of the remaining mixtures did not drop below 60
before completing the testing period.
For the plain and fiber reinforced specimens cured
according to regimen 2, degradation of the plain
mixture occurred more rapidly than in the fiber
reinforced mixture. It was observed that the use of
steel fibers not only provided greater compressive

Fig. 2 Relative dynamic modulus for fiber reinforced
UHSC cured according to regimen 1.


0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
D
M
Specimen 1
Specimen 2
Specimen 3
Specimen 4
0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
D
M
Specimen 1
Specimen 2
Specimen 3
Specimen 4
0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
D
M
Specimen 1
Specimen 2
Specimen 3
Specimen 4

913

Fig. 5 Relative dynamic elastic modulus for plain UHSC
cured according to regimen 2.

Fig. 6 Average relative dynamic modulus for all batches.

strength, but also improved the resistance to freezing
and thawing by restraining the expansion of the
microcracks.
Fig. 7 presents the average relative quality factor
values versus cycles of freezing and thawing for the
four combinations of mixtures and curing regimens. It
was observed that the relative quality factor dropped
below 60 within 20 to 50 cycles for all of the
combinations.
As with decreases in relative dynamic modulus,
decreases in relative quality factor were more rapid for
specimens with drier curing regimens and for the
mixture without steel fibers. The most abrupt decline
in quality factor was observed for specimens subjected
to curing regimen 1. This abrupt decline
coincided with an early failure (based on relative
dynamic modulus) of the specimens cured using that
regimen.
Table 4 presents a summary of the average
durability factors for the four combinations of
mixtures and curing regimens. The durability factors
represent the relative dynamic modulus values at the
end of the freezing and thawing cycles (Fig. 6).
Durability factors were greater for wetter curing
regimens. Curing regimen 3 produced the greatest
durability factor (87.5). Although only four specimens
were tested for each batch, all of the average
durability factors were significantly different (95%
confidence) except the plain mixture and the fiber
reinforced mixture cured under regimen 1.
Minor scaling occurred at approximately 100 cycles
for all specimens, whereas spalling of corners only
occurred on the fiber reinforced UHSC mixture with
curing regimen 1 and on the plain UHSC mixture.
Elongation of prisms was relatively constant
throughout the duration of testing. Elongation
values at the end of testing ranged from 0.18% to
0.22%.
Mass of the freezing and thawing specimens
increased as water filled microcracks caused by
freezing and thawing. Fig. 8 shows the average
relative mass change for the specimens.

Fig. 7 Average relative quality factor for all batches.

Table 4 Average durability factors.
Mixture Curing regimen Durability factor
Fiber reinforced 1 39.7
2 72.6
3 87.5
Plain 2 55.1
0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
e
l
a
t
i
v
e

q
u
a
l
i
t
y

f
a
c
t
o
r
Fiber reinforced mixture, Regimen 1
Plain mixture, Regimen 2
0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
D
M
0
20
40
60
80
100
120
0 50 100 150 200 250 300
Cycles
R
D
M
Specimen 1
Specimen 2
Specimen 3
Specimen 4

914

Fig. 8 Relative mass for all batches.

The wettest curing regimen (regimen 3) resulted in
substantially less mass gain than the drier curing
regimens.
4.3 Modulus of Rupture
Modulus of rupture for the UHSC mixtures was
determined using ASTM C 78 testing procedures.
This test was performed using a simple beam with
third-point loading.
Allena and Newtson [1] reported the seven day
modulus of rupture strengths of undamaged specimens
(76 mm 102 mm 406 mm) to be 17.5 MPa for
fiber reinforced UHSC and 10.9 MPa for plain UHSC.
At the end of freezing and thawing cycles conducted
in the present work, the deteriorated test specimens
were tested for modulus of rupture. Results from this
testing are presented in Table 5. All of the specimens
lost at least 60% of the 7-day flexural strength. The
fiber reinforced mixture cured according to regimen 2
provided greater flexural strength after completion of
300 cycles of freezing and thawing than the plain
UHSC mixture. However, the percentage decrease in
modulus of rupture from the 7-day strengths given in
Allena and Newtson [1] were comparable for the plain
and fiber reinforced UHSC specimens cured according
to regimen 2. Specimens exposed to wetter curing
conditions also retained greater modulus of rupture
values.
Greater durability factors corresponded to greater
modulus of rupture values at the end of testing for results
Table 5 Modulus of rupture results after ASTM C 666
testing.
Mixture
Curing
regimen
Modulus of
rupture (MPa)
Percent of
7-day strength
Fiber 1 3.75 21.4
Reinforced 2 5.18 29.6
3 5.90 33.7
Plain 2 2.61 23.9

Fig. 9 Percentage of 7-day flexural strength versus
durability factor for fiber reinforced UHSC mixtures.

obtained within the fiber reinforced UHSC mixture.
Fig. 9 illustrates the correlation between durability
factor and retained modulus of rupture percentage,
showing that the relationship is nearly linear.
5. Conclusions
The effects of the different curing methods on
compressive strength were observed to be negligible.
Conversely, initial dynamic elastic modulus values for
concrete specimens with wetter curing conditions
were greater than those from specimens with drier
curing conditions. The presence of fibers did not
appear to influence initial dynamic elastic modulus.
Providing more moisture in the last seven days of
curing following the heat treatment increased the
durability of UHSC. It was observed that any increase
in curing moisture resulted in improved durability
factors. Wetter curing also resulted in less mass gain
during ASTM C 666 testing, greater modulus of
rupture at the end of the freezing and thawing cycles,
and less variation in relative dynamic modulus
between specimens from the same group. Durability
0
10
20
30
40
50
0 20 40 60 80 100
P
e
r
c
e
n
t

o
f

7
-
d
a
y

s
t
r
e
n
g
t
h
Durability factor
100
101
102
103
104
105
0 50 100 150 200 250 300
Cycles
R
e
l
a
t
i
v
e

m
a
s
s


915
of UHSC exposed to freezing and thawing was also
improved by the use of steel fibers.
Quality factors decreased sharply almost
immediately after freezing and thawing cycles were
initiated for all specimens. The quality factors
produced the same trends observed for relative
dynamic modulus but served as an earlier indicator of
deterioration. However, it is difficult to provide a
good correlation between quality factor and relative
dynamic modulus degradation because the decline in
quality factor was so rapid.
Modulus of rupture decreased more than relative
dynamic modulus. In fact, even when durability factor
was acceptable (greater than 60) modulus of rupture
still decreased by more than 60% from the seven day
strength values.
References
[1] S. Allena, C.M. Newtson, Ultra high strength concrete
mixtures using local materials, in: International Concrete
Sustainability Conference, National Ready Mixed
Concrete Association, Tempe, AZ, Apr. 2010,
http://www.concretetechnologyforum.org/2010CSCProcee
dings (accessed July 9, 2011).
[2] J.C. Scheydt, H.S. Mller, Microstructure of ultra high
performance concrete (UHPC) and its impact on durability,
in: The 3rd International Symposium on UHPC and
Nanotechnology for High Performance Construction
Materials, Kassel, Germany, 2012, pp. 349-356.
[3] M. Cheyrezy, P. Richard, Composition of reactive powder
concrete, Cement and Concrete Research 25 (7) (1995)
1501-1511.
[4] E. Vokes, S. Clarke, E. Janssen, Damping measurements
for nondestructive evaluation of concrete beams, in:
Proceedings of International RILEM Workshop on
Resistance of Concrete to Freezing and Thawing with or
Without Deicing Chemicals, UK, 1997, pp. 1-10.
[5] J. Muro-Villanueva, C.M. Newtson, B.D. Weldon, D.V.
Jauregui, S. Allena, Freezing and thawing durability of
ultra high strength concrete, in: International
Congress on Durability of Concrete, Trondheim, Norway,
2012.
[6] E. Shaheen, N.J. Shrive, Optimization of mechanical
properties and durability of reactive powder concrete, ACI
Materials Journal 103 (6) (2006) 444-451.
[7] W.Y. Ji, M.Z. An, G.P. Yan, J.M. Wang, Study on
Reactive Powder Concrete Used in Sidewalk System of
the Qinghai-Tibet Railway Bridge, 2007,
http://www.cptechcenter.org/publications/sustainable/jireac
tive.pdf (accessed July 11, 2011).
[8] J. Pirard, B. Dooms, N. Cauberg, Evaluation of durability
parameters of UHPC using accelerated lab tests, in: The
3rd International Symposium on UHPC and
Nanotechnology for High Performance Construction
Materials, Kassel, Germany, 2012, pp. 371-376.
[9] B. Graybeal, Material Property Characterization of
Ultra-High Performance Concrete, Federal Highway
Administration report No. FHWA-HRT-06-103, McLean,
VA, 2006.
[10] L. Hasni, J.L. Gallias, M. Salomon, Influence of the
curing method on the durability of high performance
concretes, durability of concrete, in: Third International
Conference, Nice, France, 1994, pp. 131-155.
[11] B. Mather, How to make concrete that will be immune to
the effects of freezing and thawing, International Concrete
Research & Information Portal 122 (1990) 1-18.

Enhancing the Mechanical Properties of Gap Graded
Cold Asphalt Containing Cement Utilising By-Product
Material
Abbas Al-Hdabi
1, 2
, Hassan Al Nageim
1
, Felecite Ruddock
1
and Linda Seton
3

1. School of Built Environment, Liverpool John Moores University, Liverpool L3 2ET, UK
2. Engineering Faculty, Kufa University, Alnajaf 54001, Iraq
3. School of Pharmacy and Bimolecular Science, Liverpool John Moores University, Liverpool L3 3AF, UK

Abstract: The little stiffness modulus, high voidage and long curing time has limited the use of CBEMs (cold bituminous emulsion
mixtures) in road and highways to pavement experiencing low traffic. The aim of this study is to improve the properties of gap
graded CRA (cold rolled asphalt) containing OPC (ordinary portland cement) as filler by addition of a by-product material as an
activator. OPC was added to the CRA as a replacement to the conventional mineral filler (0%-100%), while LJ MUA (Liverpool J ohn
Moores University Activator) was added as an additive in the range from 0%-3% by total mass of aggregate. Laboratory tests
included stiffness modulus and uniaxial creep test to assess the mechanical properties. The results have shown a considerable
improvement in the mechanical properties from the addition of LJ MUA to the CRA containing OPC especially for the early life
stiffness modulus that is the main disadvantage of the cold mixtures.

Key words: Stiffness modulus, creep stiffness, ordinary portland cement, cold rolled asphalt.

1. Introduction

HRA (hot rolled asphalt) surface course is a
continuous gap graded mixture of mineral filler, sand
and bitumen coarse aggregate is added to it. The
mechanical properties of the mixture are controlled by
the strength properties of the mortar, i.e., mineral filler,
sand and bitumen. The material is widely used for
surfacing major roads in UK because it provides a
dense, impervious layer resulting in a weather
resistant and durable surface able to withstand the
demands of todays traffic loads, providing good
resistance to fatigue cracking. Nevertheless, it might
experience some weakness to deformation
resistance [1].
Additionally, the use of CBEMs (cold bituminous
emulsion mixtures) for road construction,

Corresponding author: Abbas Al-Hdabi, Ph.D. student,
lecturer, research fields: roads materials and sustainable asphalt
technologies. E-mail: A.T.Al-Hdabi@2011.ljmu.ac.uk.
rehabilitation and maintenance is gaining interest day
by day, as these mixtures offer advantages over
traditional hot mixtures in different terms such as,
environmental impact, energy saving, cost
effectiveness, safety and cheap production processes.
In UK, todays use of cold mix takes less interest
compared with HMA (hot mix asphalt), as this
mixtures show low earlier strength to resist the
different traffic loads and low resistance to water
damage especially rainfall. Other countries such as
USA, European countries and Australia showed more
interest in the uses of the materials due to the above
advantages [2].
Chevron Research Company in California after
many research studies reported that full curing of cold
bituminous mixtures on site depending on the weather
conditions and curing times may extend from 2-24
months. Unfortunately, UK weather conditions are not
assistant to decrease the curing time, humid, cold and
rainy most time of the year [3].
DAVID PUBLISHING
D
Enhancing the Mechanical Properties of Gap Graded Cold Asphalt
Containing Cement Utilising By-Product Material

917
Several studies have been implemented on
improving the mechanical properties of the cold mixes
in terms of OPC (ordinary portland cement) addition.
Initial study conducted by Head [4] and focused on
the improvement on the Marshall Stability of the
modified cold asphalt mix. He stated that Marshall
Stability of modified cold asphalt mix increased by
about three times with the addition of 1% OPC
compared with un-treated mix.
Li et al. [5] conducted experiments to assess the
mechanical properties of a three-phase CAEC
(cement-asphalt emulsion composite). Through
experimental study, they showed that CAEC
possessed most of the characteristics of both cement
and asphalt, namely the longer fatigue life and lower
temperature susceptibility of cement concrete, and
higher toughness and flexibility of asphalt concrete.
Oruc et al. [6] implemented experiments to assess
the mechanical properties of emulsified asphalt
mixtures having 0-6% OPC. The test results revealed
considerable improvement with high OPC addition
percentage. Furthermore, they recommended that the
cement modified asphalt emulsion mixes might be
used as structural pavement layer.
Another study implemented by Wang and Sha [7]
indicated that the rise of cement and mineral filler
fineness has a positive impact on micro hardness of
the interface of aggregate and cement emulsion
mortar. Moreover, they concluded that the limestone
and limestone filler impact hardness value are
highly when compared with granite and granite
filler.
The main aim of this research is to improve the
mechanical properties of gap graded CRA (cold rolled
asphalt) containing OPC instead of conventional filler
by addition of a by-product material. To achieve this
aim, LJ MUA (Liverpool J ohn Moores University
Activator) was used as additive in the range from
0%-3% by total mass of the aggregate to the CRA
containing different amounts of OPC. Stiffness
modulus and uniaxial compressive cyclic creep tests
were used to assess the mechanical properties in this
investigation.
2. Experimental Program
2.1 Materials Properties
The coarse and fine aggregate used in this
investigation were crushed granite and their physical
properties are shown in Table 1. Two types of filler
were used in this study, traditional mineral filler
(limestone dust) and OPC, and a high silica (more
than 75% SiO
2
) by-product material with 10 microns
average particle size has been used as an additive, i.e.,
LJ MUA. The aggregate was dried, riffled and bagged
with sieve analysis achieved in accordance with BS
EN 933-1 [8].
The cationic slow setting bituminous emulsion
(K3-60) was used to produce the novel CRA to
develop high adhesion between aggregate particles. In
contrast, a 125-pen and 50-pen bitumen grades were
used to produce the traditional HRA. The properties
of the selected bituminous emulsion and bituminous
binder are shown in Table 2.
2.2 Selected Gradation
The aggregate gradation of wholly mixtures (CRA
and HRA mixtures) used in this study was based on
BS EN 13108-4 [9] for HRA, 55/14C gap graded
surface course mixture gradation has been used
in this work. The selected gradation is shown in
Table 3.
2.3 Preparation of CRA and HRA Mixtures
The design procedure for the new CRA mixtures
used in this investigation was based on the method
adopted by the Asphalt Institute (Marshall Method for
Emulsified Asphalt Aggregate Cold Mixture Design
(MS-14)) [10]. According to this procedure
pre-wetting water content, optimum total liquid
content at compaction and optimum residual bitumen
content were 5%, 15.16% and 7%, respectively.


918

Table 1 Physical properties of aggregate.
Material Bulk specific gravity (g/cm
3
) Apparent specific gravity (g/cm
3
) Water absorption (%)
Coarse aggregate 2.79 2.83 0.6
Fine aggregate 2.68. 2.72 1.6
Mineral filler 2.71 - -

Table 2 Bituminous binder and bitumen emulsion properties.
Bitumen emulsion (K3-60) Bituminous binder (40-60) Bituminous binder (100-150)
Property Value Property Value Property Value
Appearance
Black to dark brown
liquid
Appearance Black Appearance Black
Boiling point (C) 100 C Penetration at 25 C 43 Penetration at 25 C 122
Relative density at 15 C (g/ml) 1.05 Softening point (C) 54 Softening point (C) 43.6
Residue by distillation (%) 64 Density at 25 C 1.02 Density at 25 C 1.05

Table 3 Aggregate gradation for 55/14C gap graded surface course (BS EN 13108-4).
Sieve size, mm 20 14 10 2 0.5 0.25 0.063
Percent passing (specification range) 100 98-100 42-63 40 19-31 9-31 6
Percent by mass (passing mid) 100 99 52 40 25 20 6

Different percentages of OPC (0%, 1.5%, 3%, 4.5%
and 6% by total mass of aggregate) as a replacement
for the conventional mineral filler were used in
preparation of the specimens of CRA mixtures. On the
other hand, a by-product material (LJ MUA) was used
as an additive to the CRA mixtures containing OPC
with a range of 0%-3% by total mass of aggregate. In
contrast, HRA mixture samples were prepared with
the same aggregate type and gradation, 5.5% optimum
binder content was used according to the BS 594987
Annex H [11] for the 55/14C HRA surface course
design mixture. Both cold and hot mixes were
prepared to produce three specimens for each specific
mix. The cold mix specimens were mixed and
compacted at lab temperature, i.e., 20
o
C, while
125-pen and 50-pen hot mix specimens were mixed at
(150-160
o
C) and (160-170
o
C), respectively.
2.4 Curing of the CRA Samples
The conditioning of the CRA specimens is
depending on the procedure adopted by the Asphalt
Institute MS-14 [10]. The curing process involves two
stages: The first stage was achieved with 24 h @ 20
o
C as the sample needs to be left in the mould before
being extruded to prevent specimen disintegration;
while stage two was achieved with 24 h @ 40
o
C (the
samples have been left in the ventilated oven). After
these stages, the samples have been left in the lab (20
o
C) and tested at different ages, i.e., 2, 7, 14 and 28
days to indicate the indirect tensile stiffness modulus.
This curing process, i.e., 24 h @ 20C plus 24 h @
40C represents 7-14 days in the field as reported by
J erkins [12].
3. Testing
3.1 ITSM (Indirect Tensile Stiffness Modulus)
The test is conducted at 20
o
C in accordance with
BS EN 12697-26:2004 [13], Cooper Research
Technology HYD 25 apparatus was used (Fig. 1).
3.2 UCCT (Uniaxial Compressive Cyclic Test)
The UCCT is a destructive test used mainly to
evaluate the permanent deformation characteristics of
hot mixes. UCCT at 40
o
C was used to assess the
effect of addition of LJMUA to the CRA containing
OPC on creep performance. The test was conducted in
accordance with BS EN 12697-25 [14], also Cooper
Research Technology HYD 25 testing apparatus was
used (Fig. 2).

919

Fig. 1 HYD 25 indirect tensile apparatus.

Fig. 2 Creep test configuration.

4.1 ITSM Results
The specimens were tested according to BS EN
12697-26:2004 [13] at ages 2, 7, 14 and 28 days to
assess the effect of replacement of conventional
mineral filler with OPC as well as the addition of
LJ MUA to these mixtures, as shown in Figs. 3-5. It is
obviously shown that the addition of OPC instead of
filler in the CRA increased the stiffness modulus
considerably. This enhancement in ITSM results is
due to firstly, another binder was generated in
addition to the bitumen residue binder produced from
the hydration process of the hydraulic reaction of OPC
and secondly loss of the trapped free water which is
absorbed by OPC.
Also there is incredible enhancement to the
stiffness modulus from the addition of LJ MUA to the
CRA containing OPC within the early life of mixtures.
It is clearly shown that addition of 2% LJMUA to the
CRA containing 3% OPC increased the stiffness

Fig. 3 Influence of curing time and OPC percentage on
SM results.

Fig. 4 Influence of addition of 1% LJMUA to CRA
containing OPC after two days.

Fig. 5 Influence of addition of LJMUA to CRA containing
3% OPC after two days.

modulus more than three times and its value is more
than the target value for the 125-pen HRA, i.e., 1,941
MPa. This further improvement is due to the degree of
hydration of OPC is increased when the high silica
ash, i.e., LJ MUA were added in the CRA mixtures
and was working as an activating agent to OPC. This
finding corroborates the ideas of Dillshad [15], who
found similar effect by incorporating high silica
material in OPC system.
6,000
5,000
4,000
3,000
2,000
1,000
0
6,000
5,000
4,000
3,000
2,000
1,000
0
7,000
6,000
5,000
4,000
3,000
2,000
1,000
0
Curing time, days

920

Fig. 6 Creep strain versus number of pulse applications of specimens with different percent of OPC.

Fig. 7 Influence of percent of LJMUA on ultimate creep
stiffness of CRA containing OPC.

4.2 UCCT Results
Figs. 6 and 7 show the results of the UCCT tests for
CRA containing cement with and without LJ MUA.
The creep strain versus number of pulse applications
for CRA containing OPC is shown in Fig. 6, while
Fig. 7 displays the effect of addition of 1% LJ MUA to
the CRA containing 1.5% and 3% OPC on creep
stiffness.
These figures reveal the positive effect of OPC
on the creep properties of CRA, specimens with
1.5% OPC decrease the creep strain incredibly
compared to the control specimens as well as HRA
mixtures. On the other hand, there is no ad the
addition of LJ MUA to the CRA mixtures containing
OPC on the creep performance of these ditional
outstanding from mixtures.
5. Conclusions
Adding the by-product material LJMUA to the
CRA containing OPC as mineral filler provides a
promising enhancement to the mechanical properties
of the new CRA mixtures (especially stiffness
modulus) to a level that they are comparable with
those of HRA mixtures.
The main conclusions drawn from this investigation
are as follows:
(1) The replacement of the conventional mineral
filler with OPC into the CRA mixtures improves
significantly the stiffness modulus results especially
with higher percentages of filler replacement (4.5%
and 6%);
(2) The addition of LJMUA increases considerably
the stiffness modulus of the CRA containing OPC.
Addition of 1% of LJ MUA to the CRA containing 3%
OPC improves the stiffness modules more than twice
and addition of 2% LJ MUA increases it to more than
three times (its value is more than the target point for
the conventional 125-pen HRA);
(3) The replacement of conventional mineral filler
with OPC can also improve the permanent
deformation resistance when compared to both control
CRA and traditional HRA. The creep stiffness
improves more than 6 times for the CRA containing
30,000
25,000
20,000
15,000
10,000
5,000
0
0 1,000 2,000 3,000 4,000

921
just 1.5% OPC compared with the control CRA;
(4) There is no further outstanding from the
addition of LJ MUA to the CRA containing OPC, i.e.,
the creep performance for the CRA containing OPC
with or without LJ MUA is better than the
performance of the control CRA as well as HRA
prepared with 50-pen and 125-pen bitumen.
References
[1] J .C. Nichollas, Asphalt Surfacings, Taylor and Francis
e-Library, London, 2004.
[2] I.N.A. Thanaya, Improving the Performance of Cold
Bituminous Emulsion Mixtures (CBEMs) Incorporating
Waste Materials, University of Leeds, UK, 2003.
[3] D. Leech, Cold Bituminous Materials for Use in the
Structural Layers of Roads in Traspotation Research
Laboratory, Project report 75, UK, 1994.
[4] R.W. Head, An informal report of cold mix research
using emulsified asphalt as a binder, in: Association of
Asphalt Paving Technologists Proceeding (AAPT), USA,
1974.
[5] G. Li, Y. Zhao, S. Pang, W. Huang, Experimental study
of cement-asphalt emulsion composite, Cement Concrete
Research 28 (5) (1998) 635-641.
[6] S. Oruc, F. Celik, V. Akpinar, Effect of cement on
emulsified asphalt mixtures, J ournal of Materials
Engineering and Performance 16 (5) (2007) 578-583.
[7] Z.S. Wang, A. Sha, Micro hardness of interface
between cement asphalt emulsion mastics and
aggregates, J ournal of Materials and Structures 43 (2010)
453-461.
[8] Determination of Partical Size Distribution-Sieving
Method-Test for Geometrical Properties of Aggregate,
British Standard Institution, BS EN 933-Part 1, London,
UK, 1997.
[9] Bituminous Mixtures Materials Specification-Hot Rolled
Asphalt, British Standard Institution, BS EN 13108: Part
4, London, UK, 2006.
[10] Asphalt Cold Mix Manual, Manual Series No.14 (MS-14),
Asphalt Institute, Maryland, USA, 1989.
[11] Asphalt for Roads and Other Paved Areas-Spacefication
for Transport, Laying, Compaction and Type Testing
Protocols, British Standard Institution, BS 594987,
London, UK, 2010.
[12] K. J erkins, Mix design consedrations for cold and
half-warm bitumious mixes with emphasis on foamed
asphalt, Doctoral Dissertation, University of Stellenbosch,
2000.
[13] Bitumenous Mixtures-Test Methods for Hot Mix
Asphalt-Stiffness, British Standard Institution, BS EN
12697: Part 26, London, UK, 2004.
[14] Bituminous Mixtures Test Methods for Hot Mix
AsphaltPart 25: Cyclic Compression Test, British
Standard Institution, London, UK, 2005.
[15] D.K.H. Amen, Degree of hydration and strength
development of low water-to-cement ratios in silica fume
cement system, International J ournal of Civil and
Environment Engineering 11 (5) (2011) 10-16.

Journal of Civil Engineering and Architecture, ISSN 1934-7359, USA

Deterioration of Concrete by the Oxidation of Sulphide
Minerals in the Aggregate
Jose Duchesne and Benot Fournier
Department of Geology and Geological Engineering, Universit Laval, Qubec G1Y 2S4, Canada

Abstract: Cases of degradation of concrete associated to iron sulphides in aggregates were recently recognized in the Trois-Rivires
area, Canada. The aggregate used to produce concrete was an anorthositic gabbro containing various proportions of pyrite, pyrrhotite,
chalcopyrite and pentlandite. Quantitative microanalysis on sulphide minerals show that pyrrhotite contains small amount of Ni, Co,
Cu and As substituting for Fe in the mineral structure. Considering element substitution, x value in the chemical formula (Fe
1-x
S) was
calculated to 0.099 in the pyrrhotite studied. Petrographic examination of damaged concretes showed the presence of oxidized
pyrrhotite. The observation of polished samples shows, in several cases, that the pyrite is intact while the pyrrhotite presents evident
signs of oxidation. In the presence of water and oxygen, pyrrhotite oxidizes to form iron oxyhydroxides and sulphuric acid. The acid
then reacts with the phases of the cement paste and provokes the formation of gypsum and ettringite. These minerals were observed
by SEM-EDS (scanning electron microscope/energy dispersive x-ray spectrometer) and their precipitation causes a volume increase
that creates expansion and cracking of the concrete.

Key words: Sulphide minerals, oxidation, expansion, petrographic examination, concrete durability.

1. Introduction

Over the past few years, rapid deterioration of
concrete foundations occurred in a few housing
developments in the Trois-Rivires area (Qubec,
Canada) two to four years after construction. More
than 400 residential owners have faced serious issues
related to the deterioration of their concrete
foundations and slabs. In some cases, the deterioration
was such that immediate remedial actions were
required.
The distressed concretes display map cracking on
the surface of the walls. A large number of concrete
samples were investigated. In all cases, the aggregate
material used for the concrete manufacturing was an
intrusive igneous rock, more precisely an anorthositic
gabbro, containing various proportions of sulphide
minerals mostly pyrite (FeS
2
) and pyrrhotite which is
a non-stoichiometric mineral of general formula
Fe
1-x
S, with x varying from 0 to 0.125 [1-3]. A

Corresponding author: Jose Duchesne, Ph.D., professor,
research field: concrete durability. E-mail:
Josee.Duchesne@ggl.ulaval.ca.
deleterious process involving the oxidation of
sulphide minerals is thought to have caused the
swelling and cracking of the affected concrete
elements.
This study reports the results of site inspections
along with concrete cores characterisation using
different petrographic tools including
stereomicroscope, polarizing microscope, SEM
(scanning electron microscope) and EPMA (electron
probe micro-analyser). The goal of that study is to
present a detailed characterization of the materials in
order to reach a better understanding of the
mechanisms involved. The next step will be the
establishment of a performance test enabling to
reliably screen aggregates that can cause deleterious
expansion/cracking due to oxidation in concrete.
2. Materials and Methods
2.1 General
The diagnosis of any concrete deterioration starts
with visual site inspection. Visual inspection of
DAVID PUBLISHING
D
Deterioration of Concrete by the Oxidation of Sulphide Minerals in the Aggregate

923
concrete foundations will include the identification of
any evidence of deformation, cracking (pattern and
intensity), exposure conditions including water
drainage, etc.. Visual inspection also helps selecting
sampling locations. Concrete samples (100-mm
diameter cores) were drilled through the foundation
walls for petrographic examination. Concrete cores
were examined for any sign of deterioration and some
cores were selected for further testing.
2.2 Materials
Crushed coarse aggregates were sampled directly
from selected stockpiles at the St-Boniface quarry
North of Trois-Rivieres, Qubec, Canada. Concrete
cores were drilled in different housing foundation
walls in the Trois-Rivires area. Crushed aggregates
were examined as they are and on polished sections.
Concrete cores were cut or broken for megascopic and
microscopic examinations.
2.3 Methods for Assessment and Analysis
(1) General
First, concrete cores were cut with a diamond blade.
Some surfaces were polished for stereomicroscope
observations. Selected sub-samples were impregnated
under vacuum with low viscosity resin (Epofix resin,
Struers) and polished for polarizing petrography (SiC
and loose alumina abrasive powders) [4]. To avoid
damage to the concrete during preparation, sections
were prepared with isopropyl alcohol as a lubricant
and excessive heating was avoided. Some concrete
pieces were broken in a random way for megascopic
examination and SEM observations. Polished samples
were carbon coated for EPMA and broken pieces were
coated with Au-Pd for SEM observations. Prior to
SEM observations, concrete samples were
heated in an oven kept at 40
o
C for a minimum of
24 h;
(2) EPMA and SEM
Polished sections were analysed in a CAMECA
SX-100 microprobe equipped with five WDS
(wavelength-dispersive spectroscopy) detectors (LIF
(lithium fluoride), TAP (thallium acid phthalate), PET
(pentaerythritol)) and one PGT (princeton gamma tech)
prism EDS detector. Operating conditions were set at
15 kV and 20 (observation) or 30 nA (analysis), at
high vacuum (< 10
-5
Torr). The instrument was
calibrated using a range of mineral standards, and
resulting detection limits are on the order of 0.02 wt%.
Standards were chosen based on their chemical
composition to be as close as phases analysed to avoid
matrix effect.
Concrete samples were observed under a JEOL
JSM-840A scanning electron microscope equipped
with an EDXA (energy dispersive x-ray analysis)
system. Operating conditions were set at 15 kV.
Images were taken in secondary electron mode.
3. Results
3.1 Visual Inspection
The deteriorated concrete displayed map cracking
on the walls with open cracks more pronounced at the
corners of the foundation walls (Fig. 1). Crack
openings often reach up to 10 mm and values as high
as 40 mm were reported. Major cracks are often
concentrated next to rain gutters, which highlight the
role of water or humidity in the reaction and
deterioration processes. Yellowish coloration was
often seen on the foundation walls. At some locations,
iron oxide was visible in the open cracks.
Deterioration causes an important breakdown of
concrete structures and many foundations had to be
replaced (Fig. 2). All the masonry and covering stones
were first removed to lighten the structure. Houses
were lifted up from their foundations and the later was
demolished and replaced. The remediation cost was
estimated to be close to the construction cost.
3.2 Petrographic Examination
Several samples of damaged concrete observed
under a stereomicroscope and a polarizing microscope
showed that the altered concretes were all made with

924
the same coarse aggregate containing a certain amount
of sulphide minerals. The aggregate used to produce
the deleterious concrete foundations was an
anorthositic gabbro containing various proportions of
sulphide minerals including pyrite, pyrrhotite,
pentlandite ((Fe,Ni)
9
S
8
) and chalcopyrite (CuFeS
2
).
Other constituting minerals consist mainly of calcic
plagioclase feldspars (CaAl
2
Si
2
O
8
), with lesser
amounts of biotite (K(Mg,Fe)
3
AlSi
3
O
10
(F,OH)
2
) and
pyroxene (XY(Si,Al)
2
O
6
Ca-rich, Na-poor, high-Mg
clinopyroxene) (Fig. 3).
The pyrite and pyrrhotite contents vary significantly
from one particle to another and can reach up to
5%-7% of the total coarse aggregate volume (Fig. 4).
The megascopic examination of concrete samples
showed alteration on pyrrhotite surfaces. Surfaces
were light brown and often covered by rust. Some
aggregates were completely disintegrated. The bond
between the aggregate particles and the cement paste
was often weak. The pyrite surface was unaltered even
if some particles were directly in contact with
particles of pyrrhotite.
Fig. 5 presents reflected light microscopy images
of the aggregate where iron sulphides, pyrite and

Fig. 1 Cracking in house concrete foundation.

Fig. 2 Replacement of the concrete foundation walls.

925

Fig. 3 Crushed coarse aggregate (anothositic gabbro).

Fig. 4 Stereomicroscopic view of the anorthositic gabbro aggregate.

Fig. 5 Reflected polarized light views of iron sulphide minerals (Py = pyrite; Po= pyrrhotite).

10 mm
Py

Po
(a) (b)
1 mm 1 mm

926

pyrrhotite, were in close contact and well
disseminated into silicate minerals. Very fine
inclusion of opaque minerals can be seen throughout
silicates. The observation of polished concrete
samples confirms, in several cases, that the pyrite was
intact while the pyrrhotite presented evident signs of
oxidation.
Fig. 6 presents BSE (back-scattered electron)
images of the sulphide minerals taken by EPMA.
Pyrite and pyrrhotite are seen in close contact with
each other. Pyrrhotite grains appear with a lighter
color than pyrite on the BSE image due to its higher
atomic density, with iron content higher and sulphur
content lower than for pyrite.
Fig. 7 presents a large grain of pyrrhotite with small
inclusions of pentlandite which appear whiter
compared to pyrrhotite in BSE image due to their high
atomic density. Both minerals showed extended
cracking. Black areas in the BSE images correspond
to resin or light silicates. Few flame-textured
pentlandite exsolved in pyrrhotite can be seen close to
grain boundary or along main cracks. Pentlandite is by

Fig. 6 Back-scattered electron images of sulphides minerals within the anorthositic gabbro (Py = pyrite; Po = pyrrhotite).

Fig. 7 Back-scattered electron images of a pyrrhotite grain with pentlandite inclusions gabbro (Po = pyrrhotite; Pe =
pentlandite).
100 m
Po
Pe
Pe
200 m

927
Table 1 Electron probe micro-analysis results measured on pyrrhotite (Po), pyrite (Py) and pentlandite (Pe) grains, results
are presented in weight percents.
S Cu Ni Fe Co As Total
Po-1 38.456 0.004 0.588 59.717 0.180 0.000 98.945
Po-2 38.693 0.000 0.503 59.020 0.255 0.011 98.482
Po-3 37.957 0.029 0.681 59.368 0.104 0.000 98.139
Po-4 38.052 0.021 0.674 59.225 0.303 0.000 98.275
Po-5 38.008 0.039 0.688 59.365 0.000 0.051 98.151
Po-6 38.644 0.008 0.363 59.631 0.000 0.015 98.661
MeanPo 38.302 0.017 0.583 59.388 0.140 0.013 98.442
Py-1 52.708 0.002 0.015 47.565 0.000 0.029 100.319
Py-2 52.821 0.000 0.000 47.720 0.000 0.021 100.562
Py-3 52.685 0.000 0.000 47.129 0.000 0.032 99.846
Py-4 52.554 0.023 0.000 47.993 0.000 0.013 100.583
Py-5 52.570 0.000 0.000 47.338 0.029 0.030 99.967
Py-6 52.580 0.000 0.000 47.082 0.134 0.051 99.847
MeanPy 52.653 0.004 0.003 47.471 0.027 0.029 100.187
Pe- 1 32.750 0.000 35.139 28.968 3.548 0.007 100.412
Pe -2 32.646 0.000 35.394 28.700 3.400 0.040 100.180
MeanPe 32.698 0.000 35.267 28.834 3.474 0.024 100.296
Pyrrhotite Fe
1-x
S where x =0.099

far more cracked than pyrrhotite.
Table 1 presents quantitative microanalysis
determined by EPMA punctual mode on sulphide
minerals. According to EPMA analysis, pyrrhotite
contains small amount of Ni, Co, Cu and As
substituting for Fe in the mineral structure. Analyses
are consistent and showed little variations between the
different grains analysed. Considering element
substitution, x value in the chemical formula of
pyrrhotite (Fe
1-x
S) was calculated to 0.099. Pyrite
contains only very small amount of Co and As while
content up to 3.47% Co was measured in the
pentlandite.
Fig. 8 presents stereomicroscopic views of the
deteriorated concrete. In general, the cement paste is
highly porous. This is not surprising considering the
minimum compressive strength of 15 MPa required
for plain concrete used in residential foundation
applications. High water-cement ratio in the order of
0.7 is often used for this application. Most of the
concrete samples were highly damaged, with
important cracking being observed around or through
the aggregate particles and the cement pastes. The
bond between the aggregate particles and the cement
paste is very weak and some concrete samples can be
easily broken in pieces by hand. They were often too
damaged to be polished.
Some aggregate particles were extensively cracked,
with some cracks extending from the aggregate
particles to the surrounding cement paste (Figs. 8a
and 8c). Most of the time, cracking occurred next to
sulphide rich particles. Sometime, aggregate particles
were mostly debonded, as in Figs. 8b and 8c and some
particles were completely disintegrated. Fig. 8d is an
enlargement showing rust (iron oxyhydroxide/iron
sulphate) and a whitish powdery secondary product
close to an air void.
Results obtained from stereomicroscopic
examinations confirm that the pyrrhotite grains were
mainly oxidized while pyrite grains were intact. The
nature/composition of the secondary reaction products
observed during stereomicroscopic examination was
determined by SEM observations.
Fig. 9 shows a pyrrhotite grain on the left end side


928

Fig. 8 Stereomicroscopic views of deteriorated concretes.

Fig. 9 SEM images of deteriorated concrete with a pyrrhotite grain (left)in contact with iron oxyhydroxide (right).

10 mm
10 mm
10 mm
10 mm
(b)
(c) (d)
(a)

929

Fig. 10 SEM images of deteriorated concrete with iron sulphate close to pyrrhotite grain.

Fig. 11 SEM images + EDXA analysis of gypsum present in deteriorated concrete.

of the micrograph, in close contact with an iron oxide,
hydroxide or oxyhydroxide. The precise nature of the
oxidation product can not be determined using EDXA
because hydrogen is not detected. The EDXA signal
(not presented) shows the x-ray lines of iron (Fe) and
oxygen (O). Fig. 10 shows a layer of an iron sulphate
next to a pyrrhotite grain. The EDXA signal shows the
x-ray lines of iron (Fe), sulphur (S) and oxygen (O)
(not presented). These secondary products are
associated to the reaction of pyrrhotite. Fig. 11
exhibits a grain of gypsum (calcium sulphate
dehydrate) with corresponding EDAX spectrum. SEM
observations have confirmed the identification of
secondary minerals/phases, mostly iron oxide,
hydroxide or oxyhydroxide, iron sulphate and gypsum
near altered pyrrhotite grains.
4. Discussion
Based on petrographic examination, pyrrhotite was
determined as the deleterious mineral phase in the
concrete samples. It is well-known in the mining
environment literature that sulphide minerals are
unstable in oxidizing conditions. Upon exposure to
water and oxygen, sulphide minerals oxidize to form
a

930
acidic, iron and sulphate-rich by-products according to
the following equations [5]:
Fe
1-x
S + (2 - x/2) O
2
+ xH
2
O
(1-x) Fe
2+
+ SO
4
2-

+ 2x H
+
(1)
The oxidation of ferrous iron (Fe
2+
) produces ferric
ions (Fe
3+
) that can precipitate out of solution to form
ferric hydroxide, if pH is not too low. Fe
2 +
is oxidized
and precipitated as ferric oxyhydroxides, principally
ferrihydrite and goethite.
Fe
2+
+ 1/4 O
2
+ 2H
+
Fe
3+
+ H
2
O (2)
Fe
3+
+ 3H
2
O Fe(OH)
3(s)
+ 3H
+
(3)
The oxidation reaction of iron sulphides occurs
only in the presence of oxygen and of humidity and
generates various mineralogical phases [5-7].
According to Grattan-Bellew and Eden [8] and
Shayan [9], the sulphuric acid generated through this
process reacts with the solids of the cement paste, and
particularly with the portlandite (Ca(OH)
2
), to form
gypsum according to the following equation:
H
2
SO
4
+ Ca(OH)
2
CaSO
4
2H
2
O (gypsum) (4)
The attack of concrete by sulphates resulting from
the oxidation of sulphide-bearing aggregates would
produce the crystallization of secondary ettringite
following the reaction with the alumina-bearing
phases of the hydrated portland cement paste (Eq. (5)).
According to Divet et al. [10], high pH conditions, as
those found in concrete, enhance iron sulphide
oxidation.
3(CaSO
4
2H
2
O) + C
3
A+ 26 H
2
O
C
3
A 3CaSO
4
H
32
(ettringite)

(5)
In a general way, secondary products most
frequently generated during the oxidation of iron
sulphides are the rust under all its forms (goethite
(FeOOH), limonite (FeO (OH) nH
2
O)), sulphates
including gypsum and ettringite. Degradation of
concrete is thus due to the combined mechanisms of
oxidation of iron sulphides followed by sulphatation
in the cement paste. Both reactions create secondary
minerals that cause expansion. According to Casanova
et al. 1996 [11], the later is by far more expansive. In
fact, during the formation of gypsum, the volume of
the resulting products represents a little more of
double of that of the starting solids.
According to the scientific literature, the cases of
deterioration of the concrete due to the oxidation of
sulphide minerals were reported for porous and
mechanically weak rocks such as black shale [11-14].
In the present case, the aggregate is an anorthosic
gabbro with low porosity and despite its good
mechanical performances, major damages were
noticed only two to four years after construction. As
mentioned before, the presence of oxygen and water is
necessary for the oxidation reaction. Conditions were
favourable for sulphide oxidation and it is possible
that presence of other sulphide minerals (pyrite,
pentlandite, chalcopyrite) can have played a role of
catalyst increasing the kinetic of pyrrhotite reaction.
The role or synergy of other sulphide minerals should
be investigated.
5. Conclusions
Cases of degradation of concrete associated to iron
sulphides-bearing aggregates were recently
recognized in the Trois-Rivires area, Canada. This
study reports result of site inspection along with
concrete cores characterisation using different
petrographic tools including stereomicroscope,
polarizing microscope, SEM and EPMA. Main results
show that:
Deteriorated concrete display map cracking,
yellowish coloration, open cracks more pronounced at
the corners of the foundation often next to drainage
elements;
The problematic aggregate is an anorthositic
gabbro containing iron sulphide minerals;
Iron sulphides are finely disseminated into
silicate minerals;
Pyrrhotite is oxidized while pyrite is not, the
former contains Ni, Co, Cu and As substituting for Fe;
Considering element substitution, x value in the
non-stoichiometric mineral chemical formula (Fe
1-x
S)
was calculated to 0.099 in the pyrrhotite studied;

931
Concrete samples are highly damaged, with
important cracking observed around and through the
aggregate particles and the cement paste, and some
particles completely disintegrated or debonded;
Secondary minerals identified are iron
oxide/hydroxide/oxyhydroxyde, iron sulphate and
gypsum;
The oxidation of pyrrhotite followed by
sulphatation of the cement paste seems to be the main
mechanism of concrete deterioration.
Further research is needed to develop a quality
control test to identify potentially deleterious
aggregates.
Acknowledgments
This study has been supported by the National
Science and Engineering Research Council of Canada
(NSERC) and by the Fonds de recherche sur la nature
et les technologies of the Province of Qubec
(FQRNT).
References
[1] J .P.R. de Villiers, D.C. Liles, The crystal-structure and
vacancy distribution in 6C pyrrhotite, American
Mineralogist 95 (2010) 148-152.
[2] D.C. Liles, J .P.R. de Villiers, Redetermination of the
structure of 5C pyrrhotite at low temperature and at room
temperature, American Mineralogist 97 (2012) 257-261.
[3] M. Becker, J . de Villiers, D. Bradshaw, The mineralogy
and crystallography of pyrrhotite from selected nickel and
PGE ore deposits, Economic Geology 105 (2010)
1025-1037.
[4] D. J ana, Sample preparation techniques in petrographic
examinations of construction materials: A state-of-the-art
review, in: ICMA Conference Proceedings of the 28th
Conference on Cement Microscopy, Denver, USA, 2006,
pp. 23-70.
[5] N. Belzile, Y.W. Chen, M.F. Cai, Y. Li, A review on
pyrrhotite oxidation, J ournal of Geochemical Exploration
84 (2004) 65-76.
[6] Y.L. Mikhlin, A.V. Kuklinskiy, N.I. Pavlenko, V.A.
Varnek, I.P. Asanov, A.V. Okotrub, et al., Spectroscopy
and XRD studies of the air degradation of acid-reacted
pyrrhotites, Geochimica and Cosmochimica Acta 66
(2002) 4057-4067.
[7] H.F. Steger, Oxidation of sulphide minerals: VII, Effect
of temperature and relative humidity on the oxidation of
pyrrhotite, Chemical Geology 35 (1982) 281-295.
[8] P.E. Grattan-Bellew, W.J . Eden, Concrete deterioration
and floor heave due to biogeochemical weathering of
underlying shale, Canadian Geotechnical J ournal 12
(1975) 372-378.
[9] A. Shayan, Deterioration of a concrete surface due to the
oxidation of pyrite contained in pyritic aggregates,
Cement and Concrete Research 18 (1988) 723-730.
[10] L. Divet, J.P. Davy, Study of pyrite oxidation risk in the
basic medium of concrete, LCPC Bulletin 204 (1996)
97-107. (in French)
[11] I. Casanova, L. Agullo, A. Aguado, Aggregate
expansivity due to sulphide oxidationI. Reaction
system and rate model, Cement and Concrete Research
26 (1996) 993-998.
[12] J . Berard, R. Roux, M. Durand, Performance of concrete
containing a variety of black shale, Canadian J ournal of
Civil Engineering 2 (1975) 58-65.
[13] J .S. Chinchon, C. Ayora, A. Aguado, F. Guirado,
Influence of weathering of iron sulfides contained in
aggregates on concrete durability, Cement and Concrete
Research 25 (1995) 1264-1272.
[14] C. Ayora, S. Chinchon, A. Aguado, F. Guirada,
Weathering of iron sulfides and concrete alteration,
Cement and Concrete Research 28 (4) (1998) 1223-1235.


Behavior of Millstone Masonry Vaults Strengthened by
Composite Materials
Maamoun Saade
1, 2
, Stephan Kesteloot
1
, Chafika Djelal
1
, Lotfi Hamitouche
2
and Idriss Benslimane
2

1. Laboratoire Gnie Civil et Go-Environnement (LGCgE)Lille Nord de France, IUT de Bthune, Bthune 62400, France
2. Structure & Rhabilitation, Bagnolet 93170, France

Abstract: Sewerage systems first appeared in Paris in the middle of 19th

century. Even if the majority of structures are still in
working order, their general state will deteriorate inexorably, and as reconstruction is not always possible for cost and social impact
reasons, rehabilitation is a solution adopted by many clients. It is necessary to resort to new rehabilitation techniques. Reinforcement
by bonding composite materials has many advantages compared to other techniques. The objective of the experimental campaign
presented in this paper is to study the addition of a lining by means of mortar reinforced by thin composite materials so as to restore
masonry structures. To that purpose, crushing tests on masonry vaults have been carried out. The application of a lining made of
mortar reinforced with composite materials has allowed increasing the breaking load and delaying the occurrence of the first cracks.
This article presents the characterization of the materials. Moreover, the results of the breaking tests applied to masonry vaults are
shown in this paper. A comparison with a traditional type of rehabilitation by a 6 cm-thick shotcrete lining will be performed.

Key words: Composite materials, masonry, fiber coating, sustainability, rehabilitation, sewerage systems.

1. Introduction

French man accessible main sewers have a total
length of several dozen thousand kilometers. Yet, some
of these sewerage systems have often exceeded their
useful life. The different forms of degradation present
on these structures (ageing, environment) require a total
or punctual rehabilitation. Most masonry sewerage
systems managers have to face the problem of
budgetary restrictions and have to raise substantial
volumes of finance so as to rehabilitate their structures
while seeking the best suited economic and technical
solutions. Main sewers, mainly made of millstone
masonry in big French cities show signs of damage, the
most often listed of which are surface deterioration,
structural degradations, leaks and cracking [1]. The
authors study mainly deals with egg-shaped, man
accessible wastewater facilities made of millstone and
known as main sewers. These structures consist of a
vault, abutments and a raft (Fig. 1).

Corresponding author: Maamoun Saade, Dr., Ph.D.
student, research field: reinforcement by composites materials.
E-mail: saade.ma@gmail.com.
The most widely used technique to rehabilitate the
majority of sewerage systems is to apply a 6 to 8
cm-thick reinforced shotcrete lining to the entire
section of the main sewer. The greatest disadvantages
of this technique lie in a significant reduction of the
hydraulic section, costs exceeding the budget,
problems of corroding reinforcement and buckling
steel. For these reasons, it is interesting to implement
a new reinforcement technique. The repair and

Fig. 1 Ovoid sewer.
Vault
Masonry
Coating
Abutment
wall
Invert
DAVID PUBLISHING
D
Behavior of Millstone Masonry Vaults Strengthened by Composite Materials

933
strengthening of the structures by bonding composite
materials have shown good results in the field of
construction and has recently been applied to concrete
sewerage systems [2].
The main goal of this research is to adapt this
rehabilitation process to masonry sewerage systems.
Besides, response times on site are cut down in order
to rehabilitate a greater length of network without
using corrodible materials. The project initiated by a
consulting firm and several French companies aims to
reduce costs, improve repair durability (without the
use of steel reinforcement) and reduce response times
on site.
2. Composite Materials in Structures
Over the past 30 years, researchers have studied the
possibility of using innovative materials like
composite materials to strengthen masonry structures.
Two types are commonly used in the construction
industry, extruded carbon plates and carbon fabrics.
Composite materials have considerable advantages
such as lightness, insensitivity to corrosion, ease of
application.
Many tests were carried out in Italy on masonry
vaults as part of the protection of the cultural heritage
against seismic hazards (CNR-DT, 2004).
(1) Aiello et al. [3] developed experimental studies
for the strengthening of masonry arches, indicating
that the application of CFRPs increases the strength of
the structure, and modifies the failure mechanism and
breaking load;
(2) Valluzzi et al. [4] presented the results of tests
performed on masonry arches made of brick and
reinforced with fabrics made from carbon or glass
fiber on the intrados or on the extrados. The breaking
loads are the same for both types of reinforcement but
the failure mechanism is different;
(3) Luciano et al. [5] showed the effectiveness of
masonry arches strengthened with carbon plates and
their numerical calculations reproduced the same
behavior for the tested arches.
However, bonding techniques for composite
materials are not directly applicable to masonry
sewerage systems due to the irregularity of the
substrate. It is necessary to apply a thin lining having
the appropriate characteristics to transmit the load of
the masonry onto the composite materials.
Tests on masonry vaults reinforced with a 3
cm-thick fiber coating have been conducted by Ref. [6],
in order to show the relevance of the putting in place
of a thin lining to rehabilitate main sewers.
Indeed in some cases, it is not necessary to
reinforce the structure but to simply get a facelift for
example during infiltration or exfiltration. A thin
lining is designed to homogenize the vault surfaces.
The study showed a significant increase in the load by
about 208% and a lower repair cost compared to the
method of the 6 cm-thick shotcrete.
Characterization tests of the materials used are
presented as well as tests on millstone masonry vaults,
coated and reinforced by composite materials (carbon
plates or fabrics) at the intrados. A comparison with a
traditional rehabilitation method by means of
reinforced shotcrete will be drawn.
3. Test Specimen Selection
Because of complexity of the material and the
geometry of the structure, it is necessary to adopt a
simplified geometry in order to perform laboratory
tests. Furthermore, a masonry ovoid sewer is not
stable under its own weight, hence the need to use
lateral stops. In most cases, reinforcement is necessary
with vertical loads that mainly put a strain on the vault.
The test specimen chosen for this study are vaults
built according to the numerical calculations made by
Ref. [7]. They must correspond to the upper part of
the ovoid, between the springs of the vault, close to
the two hinges, at abutment/vault level. The test
specimens are vaults representing the vault of a T210
ovoid sewer whose geometric characteristics are given
in Fig. 2. The vault has a width of 110 cm and a
length of 120 cm.


934

Fig. 2 Geometric characteristics of T210 ovoid section and test piece.

After these generalities, the mechanical properties
of the different materials (millstone, bonding mortar,
fiber coating and composite materials) are given in the
following paragraph. The results concerning the full
scale vaults are then discussed.
4. Experimental Tests
The mechanical properties of the millstone, the
bonding mortar, the fiber coating, the carbon plates
and sheets and their associated adhesives are as
follows:
(1) Millstone: The blocks used to build vaults;
(2) Bonding mortar: The bonding mortar between
the different blocks has been made according to
bibliographic research [8].
Table 1 presents the mechanical characteristics of
the millstone and the mortar.
All the results obtained during the characterization
tests show that millstone is a material with widely
disparate mechanical properties (different colors,
more or less important cavities, etc.). This mortar has
been chosen for the authors experimental campaign
in view of its mechanical properties, similar to those
used in the construction of sewerage systems in the
nineteenth century. The bonding mortar is similar to
the authors according to the three above given criteria
(modulus of elasticity, compressive strength and
flexural strength);
(3) Fiber coating: The coating should allow
transferring loads to the composite reinforcements.
The fiber coating selected for this study is one whose
mechanical characteristics are best suited for the
transmission of loads to composites. The fiber coating
is a fiber-reinforced mortar used for repairs. It is used
for resurfacing and restoration works. Table 2 presents
the mechanical characteristics of the fiber coating;
(4) Composite materials: For the study, the authors
strengthened the vaults by means of carbon plates and
sheets. These processes are intended to repair and
strengthen structures. The plates are composite
materials made of carbon fibers embedded in a polymer
matrix and the sheets are a carbon fiber matrix. Their
mechanical properties are summarized in Table 3.
To apply the composite reinforcements to the fiber
coating, the authors used a type of glue for each type
of epoxy-based reinforcement. The role of the
adhesive is very important in the case of
reinforcement by composite materials, due to the
transmission of forces into the structure of the
composite materials. The mechanical properties of the
different types of glue are summarized respectively in
Tables 4 and 5.

Table 1 Mechanical characteristics of millstone and
bonding mortar.
Designation Millstone Mortar
Modulus of elasticity (MPa) 2,500 5,931
Compressive strength (MPa) 12 to 26 0.9
Tensile stress (MPa) 1.47 -
Flexural strength (MPa) - 0.5

Table 2 Mechanical characteristics of fiber coating.
Designation Characteristics
Water content (%) 12
Flexural strength (MPa) 7.79
Compressive strength (MPa) 52.53
Tensile stress (MPa) 2.2
Tensile splitting (MPa) 4.31
Modulus of elasticity (MPa) 27,308
A = 210 cm
B = 105 cm
e = 25 cm

935
Table 3 Characteristics of the carbon plates and sheets.
Designation Carbon plate Carbon sheet
Modulus of elasticity > 165 GPa 230 GPa
Tensile strength > 2.8 MPa 3.5 MPa
Elongation at break > 1.7% 1.5%

Table 4 Characteristics of adhesives for the plates.
Modulus of elasticity 12,800 MPa
Flexural strength 30 MPa for two days at 20 C
Adhesion to concrete 4 MPa (break in the support)

Table 5 Characteristics of adhesives for the sheets.
Modulus of elasticity 3,800 MPa for seven days at 23 C
Compressive strength 55 MPa for two days at 20 C
Tensile strength 30 MPa for two days at 20 C
Adhesion to concrete
Greater than cohesive strength of
the support (at 24 h)

Unlike sheets that require a simple bonding (setting
up of the adhesive on the substrate prior to the
establishment of sheets and roller-pressing), the
implementation of a carbon plate is achieved by
double bonding. Therefore a thin layer of glue is
applied on the substrate and the thin strip. This double
bonding provides a uniform distribution of the glue.
Then the authors roll it in order to remove the excess
of glue.
Crushing tests will be performed on masonry vaults
strengthened through lining at the intrados. The
reinforcement will consist of 2 cm-thick
fiber-reinforced mortar strengthened with a layer of
composite materials (carbon plates or sheets).
Six vaults were built in accordance with the
construction techniques used at the time of Baron
Haussmann (mid-19th century) [9]. The authors used
the same composition of mortar for all vaults, and the
strength of the mortar was monitored during the
construction of the vaults. Millstone is a material with
very disparate mechanical properties. Each block is
unique (strength and sizes) to build the vaults.
5. Testing of Crushing Masonry Vaults
The experimental vaults are built through a stable
stacking-up of millstone stones jointed by mortar.
These vaults are semicircular with a diameter of 110
cm, a length of 120 cm and a thickness of 15 cm.
Crushing tests are conducted on the vaults by using a
compression machine. A forced displacement is
applied to the keystone. Using LVDTs (linear variable
differential transformers), placed at the intrados of the
vault, the authors were able to identify the movements.
The load is determined by the load cell of the machine
(Fig. 3).
Six vaults were subjected to a vertical displacement
to the key on the extrados. The tested vaults are:
(1) Two unreinforced vaults;
(2) Two coated vaults (2 cm thick) reinforced by
carbon plates;
(3) Two coated vaults (2 cm thick) reinforced by
carbon sheets.
6. Results of the Tests
The responses of the five crushed vaults are shown
in Fig. 4.
Table 6 shows first cracking loads and breaking
loads together with the first cracking load/breaking
load ratio.
The authors present the mode of failure for the
unreinforced vaults and reinforced by composite
materials.

Fig. 3 Crushing test.
Displacement sensor
Datalogger
Load cell

936

0
10
20
30
40
50
60
70
80
90
100
0 1 2 3 4 5 6 7
L
o
a
d

[
k
N
]
Deflection [mm]
Coated (2 cm thick) reinforced by carbon sheets
Coated (2 cm thick) reinforced by carbon plates
Unreinforced

Fig. 4 Response of tested vaults.

Table 6 Results of tests on vaults.
Tests N First cracking load (KN) Breaking load (KN) Deflection (mm) Ratio*
Unreinforced 8 9.32 0.8 -
Coated ( 2 cm thick) reinforced by plates 1 43 51.46 0.12 0.84
Coated ( 2 cm thick) reinforced by plates 2 70 85.85 1.76 0.82
Coated ( 2 cm thick) reinforced by sheets 3 45 57.24 5.35 0.79
Coated ( 2 cm thick) reinforced by sheets 4 63 67.76 1.94 0.93
Ratio* = first cracking load/breaking load.

Fig. 5 Cracking in mortar joints in the haunch at the extrados.

Fig. 6 Failure mode of the vault reinforced by carbon plates.
Deflection (mm)
L
o
a
d

(
K
N
)


937

Uncoated masonry vaults: In two tests on two
masonry vaults, the authors obtained a breaking load
of 10 KN and 9.32 KN for two vaults. Uncoated
masonry vaults had a low displacement of the
keystone. The appearance of cracks occurred in the
same way for both uncoated masonry vaults.
The two vaults presented the same structural
behavior mainly characterized by the formation of the
four hinges. The first crack appeared in the mortar
joints in the haunch at the extrados for a load of 16
KN, the haunches are located between the keystone
and the spring of the vault. Cracks then appeared in
the keystone at the intrados. These cracks propagated
into the joints of the masonry in Fig. 5.
Coated masonry vaults reinforced by carbon plates:
The cracks successively appeared under a load of 43
KN and 70 KN in Fig. 6. The authors can see a
separation, a detachment of the plates at the keystone
at the intrados in Fig. 7. The gains in load amount to
around 450% for vault n1 and 820% for vault n2 by
comparison with the unreinforced vaults. Cracks
appeared in the mortar joints in the haunch at the
extrados and in the middle of the abutments at the
extrados for the two tested vaults in Fig. 8. The ratio
between first cracking load and breaking load for
vault n1 is 0.84 whereas that of vault n2 is 0.82.
Coated masonry vaults reinforced by carbon sheets:
The cracks successively appeared under a load of 45
KN and 63 KN. It was possible to observe a
detachment of the fabrics at the keystone at the
intrados and in several other places in Fig. 9. The
gains in load are around 514% for vault n3 and 627%
for vault n4 by comparison with unreinforced vaults.
Cracks appeared in the mortar joints in the haunches
at the extrados for both vaults tested in Fig. 10. The
ratio between first cracking load and breaking load for
vault n3 is 0.79, it is 0.93 for vault n4.
This difference in loads can be explained by the
constructive aspect of the vaults (geometry) and some
heterogeneity in terms of strength (behavior) of the
masonry. From the first crack load, the vaults
gradually deteriorated and lost their rigidity.
The results depend on the geometry of each vault
(position and block size), one can consider that the

Fig. 7 Detachment of plates.

Fig. 8 Cracking in mortar joints in the haunch at the
extrados of vaults reinforced by carbon plates.

Fig. 9 Sheet detachment.
Cracking at the extrados
Detachment of plates
Detachment of sheets

938

Fig. 10 Cracking in mortar joints in the haunch at the
extrados of vaults reinforced by carbon sheets.
average ratio of the vaults reinforced by carbon plates
is 0.83 and that of the vaults reinforced by carbon
sheets is 0.86.
The cracks are continuous along the entire length of
the vault. The vaults reinforced by carbon plates
collapsed while those reinforced with carbon sheets
remained in place. However, a complete detachment
of the millstone blocks could be noted.
7. Comparison
Let people compare the breaking load of the
unreinforced vaults and reinforced ones by composite

8
43
70
63
45
10
51,46
67,76
57,24
85,85
0
10
20
30
40
50
60
70
80
90
100
unreinforced carbon plates carbon plates carbon sheets carbon sheets
L
o
a
d

[
K
N
]
First craking load
Breaking load

Fig. 11 Breaking load and first cracking load of unreinforced vaults and reinforced vaults.

materials. Fig. 11 shows the values of the first
cracking loads and breaking loads for each tested
vault (uncoated vault, coated vault reinforced by
carbon plates or sheets).
An increase in the breaking load was noted for each
type of reinforced vault compared to the unreinforced
vaults. The vaults reinforced by carbon plates support
greater breaking load because of their rigidity.
8. Conclusions
An experimental research conducted on millstone
masonry vaults, unreinforced and reinforced at the
intrados, has been presented.
The advantages of applying a composite material
are:
(1) The breaking load whether for reinforcement by
plates or by sheets has increased by six times;
(2) An increase in the durability of construction
works by avoiding the use of steel (corrodible
materials);
(3) Stiffness has an important role in the
reinforcement process: The more rigid the material is,
the more important the reinforcement is (comparison
between plates and sheets for example);
Cracking at the extrados
Unreinforced Carbon plates Carbon plates Carbon sheets Carbon sheets
51.46
85.85
67.76
57.24
L
o
a
d

(
K
N
)


939
(4) The hydraulic section (by comparison with the
shotcrete method) is maintained.
The reinforcement by composite materials is
determined by the quality of the composite/fiber
coating interface. To obtain a larger gain in terms of
first cracking load, it may be possible to contemplate
reinforcing the composite/fiber coating interface. This
will allow the composite reinforcement to work for a
long time.
The question of the sustainability remains posed.
Indeed, through the bibliography, there is risk of
premature detachment of complex fiber
coating/glue/composite. Through the authors research,
they used hydraulic mortars usually used in sewer
systems. Their durability is perfectly known and
appropriate. Regarding the sustainability of the
composite reinforcements with carbon fiber, Kesteloot
et al. [10] have made several tests in laboratory and in
site. The experimental site led during in his thesis
shows no deterioration until now (eight years bonding
in site).
Acknowledgments
The authors thank the Structure and Rehabilitation,
Freyssinet, Lanko, Sika and Weber and Broutin for
their financial and technical assistance in carrying out
this research project.
References
[1] RERAU-Restructuration des collecteurs
visitablesGuide technique, Tome 1 et 2, Projet National
RERAU, Edition Lavoisier, 2004.
[2] S. Kesteloot, C. Djelal, S. Baraka, I. Benslimane,
Modelling of sewerage systems strengthened with
composites plates, Construction and Building Materials
20 (2006) 158-168.
[3] M.A. Aiello, N. Galati, A. La Tegola, Collapse load of
masonry arches strengthened with FRP sheets, in:
Proceeding of the First National Conference on
Mechanics of Masonry Structures Strengthened with FRP
Materials: Modelling, Testing, Design, Control, Italy
2000, pp. 75-84.
[4] M.R. Valluzzi, M. Valdemarca, C. Modena, Behavior of
brick masonry vaults strengthened by FRP laminates,
J ournal of Composites for Constructions 5 (3) (2001)
163-169.
[5] R. Luciano, S. Marfia, E. Sacco, Reinforcement of
masonry arches by FRP materials, in: International
Conference on FRP Composites in Infrastructures, USA,
2002.
[6] M. Saade, S. Kesteloot, C. Djelal, L. Hamitouche, I.
Benslimane, Influence of a fiber coatings for the repair of
sewerage in masonry, in: Proceeding of the International
RILEM Conference on Advances in Construction
Materials through Science and Engineering, Hong-Kong,
Sep. 5-7, 2011, pp. 623-631.
[7] N. Khoufache, Contribution of the restructuring masonry
sewerage systems by composite materials, Ph.D. Thesis,
University of Artois, Dec. 2008. (in French)
[8] Annals of Bridges and Roads, Central Library Highways,
1834, 1836, 1854 and 1856. (in French)
[9] J . Hervieu, Pratical Treatise on the Construction of
Sewers, Paris BAUDRY & CIE Editeurs, 1897. (in
French)
[10] S. Kesteloot, Rehabilitation of sewerage works using
composite materials: Experimental study and
dimensioning using numerical calculations, Ph.D. Thesis,
University of Artois, Mar. 2005.

Corrosion Study and Passive Film Characterization of
11% Cr F/M and 15% Cr ODS Steels
Masatoshi Sakairi
1
, Sublime Ningshen
1, 2
, Keita Suzuki
1
and Shigeharu Ukai
1

1. Faculty of Engineering, Hokkaido University, Hokkaido 060-8628, Japan
2. Corrosion Science and Technology Group, Indira Gandhi Centre for Atomic Research, Kalpakkam 603 102, India

Abstract: An ODS (oxide dispersion strengthened) steels are one of the most notable structural materials being developed for future
high-temperature energy production technologies, and several studies have been devoted to the development of ODS materials for
such applications. However, only little paper focuses on corrosion behavior of F/M (ferritic martensictic) and ODS steels. The
corrosion behavior of 11% Cr F/M steel and 15% Cr ODS steel were evaluated using electrochemical methods in borate buffer and 1
kmol m
-3
HNO
3
with or without NaCl and also in boiling 60% nitric acid. The corrosion resistance results clearly indicated the
influences of steel alloys composition and chloride ions. The XPS (X-ray photo-electron spectroscopy) results of the pre-passivated
surface revealed that the oxide formed were composed predominantly of Fe
2
O
3
along with Cr
2
O
3,
and Y
2
O
3
layers in ODS steel. The
corrosion rate measured in boiling nitric acid for 48 h for both the steels shows high corrosion rate in boiling condition. The SEM
(scanning electron microscopy) observation of the pit morphology after corrosion tests appears with shallow pit in both steel surfaces.
The corrosion degradation behavior in relation to the composition of the passive oxide film in different electrolytic solutions is
discussed in this paper.

Key words: F/M steel, ODS steel, corrosion, passive film, XPS analysis.

1. Introduction

Corrosion phenomena are defined as the
deterioration of a material, or its properties by
chemical or electrochemical interaction with its
environment, and this has become a problem for
construction/structural materials worldwide [1, 2].
Corrosion has an enormous economic and
environmental impact and causes unpredictable and
costly damage to everything from pipelines, bridges
and public buildings to oil and gas, chemical
processing, petrochemical, automobiles, power and
wastewater systems and including home appliances
[3, 4]. In addition to causing severe damage, corrosion
disrupts operations and requires extensive repair and
replacement of failed assets [1, 3, 4]. The annual cost
of corrosion worldwide is estimated to 3% to 4% of
the gross domestic product of industrialized countries

Corresponding author: Masatoshi Sakairi, Ph.D., research
fields: corrosion and protection, surface treatment and
electrochemistry. E-mail: msakairi@eng.hokudai.ac.jp.
[3, 4]. This figure is divided evenly between direct
costs (materials and structures) and indirect costs (loss
of productivity). About 90% of the corrosion is
associated with iron-based materials. Although
corrosion problems cannot be easily remedied, it is
estimated that corrosion related costs can be reduced
more than 30% by development and use of better
corrosion control technologies [2, 4].
The choices of suitable and appropriate materials
are one of the key objectives in Generation IV nuclear
reactor development. The F/M steels with 9%-12%
chromium and RAFM (reduced activation
ferritic/martensitic) steels are one of the most
promising materials being considered for higher
temperature applications for future advanced fast
reactor cladding and structural materials [5-7]. The
inherent and desirable properties of these steel alloys
include, high void swelling resistance, thermal
properties, helium embrittlement, micro-structural
stabilities and high recycling potential, which make
DAVID PUBLISHING
D
Corrosion Study and Passi ve Film Characterization of 11% Cr F/M and 15% Cr ODS Steels

941
them candidates for application in commercial fission
and fusion power plants [5, 6, 8, 9]. Interestingly, the
F/M steels do not suffer (or suffer less) from these
phenomena than the austenitic steels, and they can be
made in reduced activation compositions [6, 8, 9]. The
presence of fine and uniform precipitates obtained
during the fabrication process in these steel alloys
improved most of the above properties [6, 7].
However, one significant limitation of F/M steel is the
high temperature strength above 600
o
C dissolution
and precipitation coarsening leads to abrupt loss of
their tensile and creep property [6, 8, 9]. The expected
operating temperature of the first wall in future
Generation IV fusion reactor is above 700
o
C,
resulting in improved thermal efficiency up to
30%-40% [5-7]. The ODS based F/M or ferritic alloys
are the most promising materials with a potential to be
used at such higher temperatures by microstructure
modifications and also based on favourable
conventional properties and/or technical maturity
[6, 8].
The usefulness and application of any construction
and structural materials is based on its mechanical
performance. In addition, the corrosion resistance is
critical for the materials to be used in any
aggressive/corrosive environments. Corrosion control
is achieved by recognizing and understanding the
different forms of corrosion and its mechanisms. With
its many forms, causes and associated prevention
methods, corrosion obviously is highly complex and
requires extensive expertise and significant resources
to control, particularly in a nuclear environment [1, 2, 4].
Depending on specific environmental requirements,
two ODS families are presently developed one based
on 9% Cr martensitic ODS steels and other on
14%-18% Cr ferritic ODS alloys [5, 6, 8, 9]. Further,
one of the in-service degradation phenomena for these
steel alloys is uniform corrosion [10]. Hence, the
present work was undertaken to explore and
investigate the corrosion resistance behavior of 11%
Cr F/M steel and 15% Cr ODS ferritic steel in
different electrolytic environments, including in
boiling nitric acid solution.
2. Material and Experimental Methodology
2.1 Materials and Specimen Preparation for
Corrosion Test
The chemical composition of 11% Cr F/M steel (Cr
10.52, Ni 0.52, C 0.15, Mn 0.04, W 1.81, Mo 0.40, Fe
balance) and 15% Cr ODS steel (Cr 15.15, Ni 0.04, C
0.03, Mn 0.02, W 1.91, Al 3.90, Zr 0.57, Y
2
O
3
0.33,
Fe balance) all in mass % were used in the present
work. The specimens before corrosion experiment
were abraded mechanically with SiC paper up to 1000
grit SiC paper on all sides prior to mounting in an
epoxy resin, using a brass rod as an electrical
connection. The mounted specimens were then
abraded again up to 1000 grit SiC paper finish,
ultrasonically cleaned in acetone and subsequently
used for the experiment.
2.2 Corrosion Rate Measurements
The corrosion tests were conducted by immersion
of the 11% Cr F/M and 15% Cr ODS steels specimens
into 60% HNO
3
at boiling condition at different time
duration up to 48 h. After exposure, the weight
changes were measured and the corrosion rate was
calculated using an equation:
Corrosion rate (mm/y)
= (8.76 10
4
W) / D A T) (1)
where,
W = weight loss in milligrams;
D = metal density in g /cm
3
;
A = area of sample in cm
2
;
T = time of exposure of the metal sample in hours.
2.3 Electrochemical Measurements for Corrosion
Studies
All the corrosion samples were allowed to stabilize
under open circuit conditions for 30 min and
subsequently, the OCP (open circuit potential) was
measured as a function of time up to 1 h.

942
The potentiodynamic anodic polarization
experiments were carried out in different electrolytic
solutions: 0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O; 0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol
m
-3
Na
2
B
4
O
7
10H
2
O + 0.5 kmol m
-3
NaCl; 1 kmol m
-3
HNO
3
; and 1 kmol m
-3
HNO
3
+ 0.1 kmol m
-3
NaCl
solutions at room temperature (about 24
o
C) using
HABF 5001 potentiostat (Hokuto Denko, Japan). The
electrochemical cell consists of three electrodes,
reference electrode (Ag/AgCl in saturated KCl),
counter electrode (Pt.) and working electrode. The
electrode potential was anodically scanned at a scan
rate of 0.1667 mV s
-1
until the potential at which
breakdown occurred. The details of the experimental
methods have been already described elsewhere [9, 11].
2.4 Surface Analytical and Morphology
Characterization
The surface morphologies of the specimens were
observed with a SEM (Scanning Electron Microscopy,
Joel JSM-6510LA model) equipped with EDX
(energy dispersive X-ray) spectroscopy.
The XPS (X-ray photo-electron spectroscopy)
measurements were carried out by potential sweep to a
passive region. This was followed by holding
potentiostatically at the respective passive region for 1
h. The samples were immediately transferred into the
XPS analyzer chamber for measurement. XPS
measurements were performed using a JOEL
JPS-9200 model equipped with dual X-ray source.
MgK (1,253.6 eV) X-ray was used to generate the
photo electron. During these experiments, the base
pressure was maintained at 10
7
Pa. The X-ray gun
was operated at 100 W (10 kV, 10 mA) with a takeoff
angle = 0
o
(relative to the sample normal). A survey
spectrum was first recorded to identify elements
presence and then high-resolution spectra of the
following regions were recorded: oxygen (O 1s),
carbon (C 1s), chromium (Cr 2p),

iron (Fe 2p),
chlorine (Cl 2p) and yttrium (Y3d and Y 3p).
SpecSurf XPS operating software was employed to
collect and process all surveys and high-resolution
spectra.
3.1 Corrosion Rate Evaluation in Boiling Nitric Acid
Fig. 1 represents the corrosion rate obtained after
the immersion tests in boiling 60% HNO
3
for 48 h. It
was observed that all the specimens were covered with
a film of interference of pale dark color due to
oxidized oxide layers. This study was carried out to
evaluate corrosion resistance and its impact in boiling
nitric acid solution. Nitric acid is the main process
medium used in the reprocessing of the spent fuel by
PUREX process, and hence it is essential to
understand the performance of materials in such
environments [9, 12]. The obtained corrosion rates of
both steel alloys are high, and the corrosion rates are
in the range of about 2.0-3.5 mm/y for 11% Cr F/M
steel and 0.74-1 mm/y for 15% Cr ODS steel. This
thereby implies that the corrosion resistance of 15%
Cr ODS steel is lower than 11% Cr F/M steel. The
surface morphology of the samples after 48 h
immersion in nitric acid shown in Figs. 2a and 2b,
indicated different morphology of surface attack. The
surface appears mostly porous in 11% Cr F/M steel
(Fig. 2a), whereas the surface features are mostly
compact with a layer of oxide scale structure in 15%
Cr ODS steel (Fig. 2b). Thereby, such surface
morphology could have also effected on its corrosion
resistance.
3.2 OCP (Open Circuit Potential) Measurements
The measured OCP of 11% Cr F/M steel in
different electrolytic solutions are represented in
Fig. 3a. The OCP of 11% Cr F/M steel (Fig. 3a)
shows active potential in chloride containing media of
borate-buffer + 0.1 kmol m
-3
NaCl (-0.245 V) and 1
kmol m
-3
HNO
3
+ 0.1 kmol m
-3
NaCl (-0.215 V).
Similarly, the electrolytic solutions without chloride
also show active potential in borate-buffer (-0.190 V)
and 1 kmol m
-3
HNO
3
(-0.195 V). Moreover,
immediately upon immersion all the measured OCP

943

0.0
0.5
1.0
1.5
2.0
2.5
3.0
3.5
4.0
C
o
r
r
o
s
i
o
n

r
a
t
e

m
m
/
y
S teel alloys
60% HNO
3
- 48h
11% C r F /M steel
15% C r ODS steel

Fig. 1 The measured corrosion rate of 11% Cr F/M and 15% Cr ODS steels in boiling 60% HNO
3
of 48 h exposure.

Fig. 2 Typical SEM morphology of: (a) 11% Cr F/M steel; (b) 15% Cr ODS steel after corrosion test in boiling 60% HNO
3

(48 h).

shows steady state OCP value thereby indicating
stable film formation (Figs. 3a and b). The results of
the OCP measured for 15% Cr ODS steel specimen
obtained in different electrolytic solutions shown in
Fig. 3b is similar to the pattern observed in Fig. 3a.
The measured OCP shows active potential in 1
kmolm
-3
HNO
3
+ 0.1 kmol m
-3
NaCl (-0.215 V)
followed by measurement in 0.3 kmol m
-3
H
3
BO
3
+
0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O + 0.1 kmol m
-3
NaCl
(-0.165 V), 0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O (-0.150 V) and 1 kmol m
-3
HNO
3

(-0.140 V). The passive film stability of metals/alloys
in the corrosive environment depends upon both on
the chemical or electrochemical reactions occurring at
its surface. Less noble and active potential of both
steel alloys in chloride media is attributed to less
protective passive film due to Cl
-
[9, 11]. In both the
steel alloys, the OCP is the lower in chloride
containing solutions and higher in borate buffer
solution. According to the mixed potential theory [13],
the shift of OCP depends on the change of both anodic
reaction rate and cathodic reaction rate. Thus, both
(a) (b)
Still alloys
11% Cr F/M steel
15% Cr ODS steel
60% HNO
3
48 h
C
o
r
r
o
s
i
o
n

r
a
t
e

(
m
m
/
y
)


944
decreasing of anodic reaction and or increasing of the
cathodic reaction can lead to the shift of OCP.
3.3 Potentiodynamic Polarization Studies
The results of the potentiodynamic polarization
curves of 11% Cr F/M steel obtained in different
electrolytic solutions are shown in Fig. 4a. The
obtained electrochemical parameters are labeled in the
curve shown in Fig. 4a and values derive from the
polarization curves are listed in Table 1. In 11% Cr

0 10 20 30 40 50 60
-0.24
-0.22
-0.20
-0.18
-0.16
(a)
O
p
e
n

c
i
r
c
u
i
t

p
o
t
e
n
t
i
a
l
/
V

v
s

A
g
/
A
g
C
l
Time/min
11% Cr F/M steel
1 kmol m
-3
HNO
3
1 kmol m
-3
HNO
3
+0.1 k mol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
. 10H
2
O
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
. 10H
2
O
+0.5 kmol m
-3
NaCl

(a)

0 10 20 30 40 50 60
-0.24
-0.22
-0.20
-0.18
-0.16
-0.14
-0.12
(b)
O
p
e
n

c
i
r
c
u
i
t

p
o
t
e
n
t
i
a
l
/
V

v
s

A
g
/
A
g
C
l
Time/min
15% Cr ODS steel
1 kmol m
-3
HNO
3

1 kmol m
-3
HNO
3
+ 0.1 k mol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol m
-3
Na
2
B
4
O
7
. 10H
2
O
0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol m
-3
Na
2
B
4
O
7
. 10H
2
O
+ 0.5 kmol m
-3
NaCl

(b)
Fig. 3 The OCP measurement of: (a) 11 % Cr F/M steel; (b) 15 % Cr ODS obtained in different electrolytic solutions.

Time/min

15 % Cr ODS steel

1 kmol m
-3
HNO3
1 kmol m
-3
HNO3 + 0.1 kmol m
-3
NaCl
0.3 kmol m
-3
H3BO3 + 0. 075 kmol m
-3
Na2B4O7 10 H20
0.3 kmol m
-3
H3BO3 + 0. 075 kmol m
-3
Na2B4O7 10 H20 + 0.5 kmol m
-3
NaCl

Time/min

945

10
-9
10
-8
10
-7
10
-6
1x10
-5
1x10
-4
10
-3
10
-2
10
-1
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Borate +
0.1 k mol m
-3
NaCl
1 kmol m
-3
HNO
3
+
0.1 k mol m
-3
NaCl
Borate
1 kmol m
-3
HNO
3

(a)
E
corr
i
pass
E
BP

Potential/V vs Ag/AgCl
l
o
g

i
/
A

c
m
-
2
11% Cr F/M steel
1 kmol m
-3
HNO
3
1 kmol m
-3
HNO
3
+ 0.1 k mol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
.10 H
2
O Na
2
B
4
O
7
.10 H
2
O + 0.5 kmol m
-3
NaCl

10
-9
10
-8
10
-7
10
-6
1x10
-5
1x10
-4
10
-3
10
-2
10
-1
-1.0 -0.5 0.0 0.5 1.0 1.5 2.0
Borate +
0.1 kmol m
-3
NaCl
Borate
1 kmol m
-3
HNO
3
+ 0.1 kmol m
-3
NaCl
1 kmol m
-3
HNO
3
(b)
P otential/V vs Ag/AgCl
l
o
g

i
/
A

c
m
-
2
15% Cr ODS steel
1 kmol m
-3
HNO
3
1 kmol m
-3
HNO
3
+ 0.1 kmol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
0.3 kmol m
-3
H
3
BO
3
Na
2
B
4
O
7
. 10 H
2
O
+0.075 kmol m
-3
Na
2
B
4
O
7
. 10 H
2
O + 0.5 kmol m
-3
NaCl

(b)
Fig. 4 The potentiodynamic polarization curves behavior of: (a) 11% Cr F/M steel; (b) 15% Cr ODS steel in different
electrolytic solutions.

1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl

0.3 kmol m
-3
H
3
BO
3
+ 0. 075 kmol m
-3
Na
2
B
4
O
7
10 H
2
0 +0.5 kmol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+ 0. 075 kmol m
-3
Na
2
B
4
O
7
10 H
2
0

1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
0.3 kmol m
-3
H
3
BO
3
+0. 075 kmol m
-3
Na
2
B
4
O
7
10 H
2
0 +0.5 kmol m
-3
NaCl

0.3 kmol m
-3
H
3
BO
3
+0. 075 kmol m
-3
Na
2
B
4
O
7
10 H
2
0
(a)

946
Table 1 Polarization parameters obtains for 11% Cr F/M steel and 15% Cr ODS steel in different electrolytic solutions.
Electrolytic media E
corr
(V vs Ag/AgCl) i
pass
(A cm
-2
) E
BP
(V vs Ag/AgCl)
11% Cr F/M steel
1 kmol m
-3
HNO
3
-0.190
4.41 10
-6
(0.5 V)
7.06 10
-6
(0.5 V)
1.07 1st
1.62 2nd
1 kmol m
-3
HNO
3
+0.1 kmol m
3
NaCl -0.243 1.25 10
-3
(-0.1 V) 0.043
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10 H
2
O
-0.665 1.1 10
-6
(1.2 V) 1.23
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10 H
2
O +0.5 kmol m
-3

-0.517 3.11 10
-7
(0.1 V) 0.122
15% Cr ODS steel
1 kmol m
-3
HNO
3
-0.121
9.07 10
-7
(0.5 V)
8.96 10
-5
(1.4 V)
0.611 1st
1.64 2nd
1 kmol m
-3
HNO
3
+0.1 kmol m
3
NaCl -0.23
6.1 10
-7
(0.5 V)
1.84 10
-4
(1.4 V)
0.850 1st
1.50 2nd
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10 H
2
O
-0.66 1.5 10
-7
(0.5 V) 1.22
0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10 H
2
O +0.5 kmol m
-3

-0.420 3.05 10
-4
(0.5 V) 0.338

F/M steel (Fig. 4a), the change in the electrolytic
solution markedly changes the polarization behavior
and differences in i
pass
and E
BP
are evident. The
general shapes of most of the potentiodynamic curves
are similar, but differences in the E
corr,
i
pass
and E
BP

could be observed. In borate buffer and 1 kmol m
-3

HNO
3
media without chloride, the polarization curves
are characterized by a wider passive range. However,
drastic reduction in the apparent breakdown or pitting
potential and the sharp rise in the current density is
seen in 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl. In the
presence of chloride (Table 1), despite lower i
pass
of
3.11 10
-7
A cm
-2
(borate +0.1 kmol m
-3
NaCl),
higher i
pass
of 1.23 10
-3
A cm
-2
(1 kmol m
-3
HNO
3
+
0.1 kmol m
-3
NaCl) and low pitting potential (0.125
and 0.04 V) was observed. The breakdown
potential/pitting corrosion (E
BP
) is the potential above
which pits nucleated and grow. Hence, localized
corrosion such as pitting corrosion initiates above
certain critical potential above E
BP
and repassivate
below E
BP
. The higher the E
BP
, the more resistant are
the metal/alloy to pitting corrosion. In presence results,
lower pitting potential observed is attributed to low 11%
Cr content, surface inhomogeneity or possible
inclusions. Further, the presence of Cl
-
ions destroy
the passivating film, and the aggressiveness of
chloride is attributed to a number of reasons [14, 15]:
(1) Chloride is an anion of a strong acid, and many
metal cations are soluble in chloride solutions [9, 15];
(2) Chloride is a relatively small anion with a high
diffusivity and also acts as a contaminant, thereby it
interferes with passivation; and (3) the presence of
oxidizing agents in a chloride-containing environment
is extremely detrimental, and will additionally
enhance localized corrosion [14, 15].
In Fig. 4b, the typical potentiodynamic polarization
curves of 15% Cr ODS steel obtained in different
electrolytic solutions is represented, and the measured
polarization curves parameters are given in Table 1.
Higher E
BP
and lower i
pass
were observed in 1 kmol
m
-3
HNO
3
with or without chloride. However,
compared to 11% Cr F/M steel (Fig. 4a), the E
BP
in
the presence of chloride was not significant in 1 kmol
m
-3
HNO
3
+0.1 kmol m
-3
NaCl. Moreover, both the
steel alloys exhibited lower pitting corrosion
resistance in the presence of chloride media. Lower
pitting corrosion resistance of ODS alloy may also be
attributed to the presence of {Zr, Ti}C that has been
detected by EDS and XPS analysis. The presence of
such carbides can act as inclusion sites in ODS steels
[6, 16]. Once the pit initiates from such inclusion sites,
dissolution takes place accompanied by a local drop of
pH, and once a critical concentration in chloride is
attained, pitting corrosion is initiated [14, 15].

947
The typical SEM micrographs of 11% Cr F/M steel
and 15% Cr ODS steel specimens after corrosion tests
are represented in Fig. 5. In 11% F/M steel, the pit
morphology shows a deep shallow pit (Fig. 5a) and
along with uniform dissolution in nitric acid chloride
media (Fig. 5b). Similarly, in 15% Cr ODS, the SEM
morphology shown in Fig. 5c reveals that severe
pitting has occurred in chloride containing media with
lace-like structure. The features of such structure are
dependent on dissolution rate of pit solution chemistry
[14].
3.4 XPS Analysis of Passive Films in Alloy Steels
The changes in the properties of the passive film
affected the localized corrosion resistance leading to
pitting corrosion, crevice corrosion, intergranular
corrosion and SCC (stress corrosion cracking) [14, 17].
In the present work, the elemental compositions of the
surface film were determined by XPS. The wide
survey XPS spectra (not shown) of the passive layer
formed in different electrolytic solutions of 11% Cr
F/M steel reveals the presence of mostly Fe, Cr, O, C
and Cl depending on electrolyte used. Similarly, in the
wide survey spectra of 15% Cr ODS steel the presence
of Fe, Cr, O, Y, W, C and Cl core level peaks are
identifiable. In both steel alloys, it is evident that the
passive oxide film is consisted mainly of chromium
and iron and presence of W and Y was detected in

(a)

(b) (c)
Fig. 5 The SEM of surface and pit morphology after corrosion test: (a) 11% Cr F/M steel in 0.3 kmol m
-3
H
3
BO
3
+ 0.075
kmol m
-3
Na
2
B
4
O
7
10H
2
O + 0.5 kmol m
-3
NaCl; (b) 11% Cr F/M steel in 1 kmol m
-3
HNO
3
+ 0.1 kmol m
-3
NaCl; and (c) 15%
Cr ODS steel in 0.3 kmol m
-3
H
3
BO
3
+ 0.075 kmol m
-3
Na
2
B
4
O
7
.10H
2
O + 0.5 kmol m-3NaCl solutions.

948

15% Cr ODS steel.
In Fig. 6a, the appearance of broad metallic peak of
Fe
met
(706 eV) in a borate buffer (Fig. 6a(3)) and
borate +0.1 kmol m
-3
NaCl (Fig. 6a(4)) shows that the
oxide film layer formed is thin. This is followed by a
broad shoulder on the side of the higher binding
energies of Fe
2
O
3
(710 eV) and Fe(OH)
3
(712 eV).
Peak signal appearing in binding energy of 709 eV
may also be assigned to Fe
3
O
4
. However, it is not
possible to differentiate between Fe
2
O
3
(or their
hydrates) or Fe
3
O
4
on the basis of XPS [18]. In 11%
Cr F/M steel, the measured Fe atomic % concentration
was in the range 8%-18%. In the high-resolution
spectra of Fe 2p obtained for 15% Cr ODS steel
shown in Fig. 6b, the Fe 2p ionization reveals the
presence of metallic Fe
met
(706 eV), Fe (III) oxides
(710eV) and hydroxides (712 eV). The Fe metal peak
intensity (706 eV) was predominant is most of the
electrolytic solutions. However, relative concentration
of Fe in atomic % concentration varied in the range of
9-18 at % depending on the electrolytes used. Hence,
there appears to increase or decrease in the amount of
Fe, probably depending on whether the metal is in the
oxidized or reduced state. In Fig. 7a the Cr 2p spectra
of 11% Cr F/M steel can be fitted into two to three
components. The Cr 2p ionization consists of binding
energy at 574 eV that correspond to metallic Cr
met
, the
peak at 576 eV is attributed to Cr(III) oxides and
hydroxides (577 eV), respectively. The components of
the Cr 2p photoelectron peak located at 576 eV (Cr
2p
3/2
) and at 585 eV (Cr 2p
1/2
) are characteristics of
Cr
III
-O bonds in a Cr
2
O
3
compound. The Cr 2p peak
located at 577 eV (Cr 2p
3/2
) and at 587eV (Cr 2p
1/2
)
are characteristics of Cr-O bonds in a Cr
x
O
y

compound (> Cr
3
) [18]. The detected chromium
elements of oxide films are in the range of 8%-12%
and hence the Cr content in the oxide layer is low.
This thereby indicated that chromium is not
significantly enriched in the passive film. In Fig. 7b,
the spectra of Cr 2p 15% Cr ODS steel shows a broad
peak that appeared at 576 eV related to Cr
2
O
3
,
followed by a small shoulder related to Cr
met
(574 eV)
toward lower binding energy(Fig. 7b). In most media,
the Cr 2p can be fitted to three peaks corresponding to
the following components: the one at the lowest
binding energy (574 eV) corresponds to metallic Cr,
576 eV are attributed to the presence of Cr(III) oxides
(576 eV) and hydroxides (577 eV). The reported
corresponding binding energies are in good agreement
with the reported values [9, 11, 18].
In Fig. 8a, the overall O 1s spectra formed on 11%
Cr F/M steel can be resolved into three components:
oxide (530 eV), hydroxide (531 eV) and adsorbed
H
2
O (533 eV). Two prominent peaks observed in most
electrolytic media at 530 eV and 531 eV can be
attributable to O
2-
and OH
-
attached to metal ions.
However, binding energy appearing at 531 eV and at
530 eV may also be attributed to M-OH bonds and
M-O bonds [18]. In 15% Cr ODS steel shown in Fig. 8b,
the pattern is almost the same. Where, the O 1s
exhibited three components, oxide (530 eV),
hydroxide (531 eV) and adsorbed H
2
O (533 eV),
respectively.
In Figs. 9a and 9b, the spectrum of Cl 2p show
doublet peaks signals at 199 eV and 200 eV indicating
that Cl
-
ions are incorporated into the passive films in
both steel alloys. The measured Cl
-
ions
concentrations are in the range of 8%-9% and the
presence of such chloride ions are known to have
pronounced effects on the protective properties of the
passive film [14, 15]. Their presence enhances
hydrolysis and reduces the local pH, causing
dissolution of passive film. Hence, it is conceivable
that poor pitting corrosion observed in both the steel
alloys is attributable to the deleterious role of Cl
-
ions
in the passive oxide layer. The schematic
representations of passive film breakdown and
pit/corrosion initiation in these steel alloys are
represented schematically in Fig. 10. The event
leading to passive film breakdown has a very distinct
property of non-uniformity or occurs at a
limited number of sites in the whole surface. The most

949

( i ) 1 k m o l m
- 3
H N O
3

I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+ 0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O
7 2 8 7 2 4 7 2 0 7 1 6 7 1 2 7 0 8 7 0 4
( a )
F e
2
O
3
F e 2 p F e ( O H )
3
F e
2
O
3
F e
m e t
B i n d i n g e n e r g y ( e V )
( i v ) 0 . 3 k m o l m
- 3
H
3
B O
3
+ 0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O + 0 . 5 k m o l m
- 3

( i i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l

F e
2
O
3
F e
2
O
3

( i ) 1 k m o l m
- 3
H N O
3
F e 2 p
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l
( i i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O
7 2 8 7 2 4 7 2 0 7 1 6 7 1 2 7 0 8 7 0 4
( b )
F e ( O H )
3
F e
m e t
( i v ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O + 0 . 5 k m o l m
- 3

Fig. 6 The XPS spectra of Fe 2p detected for passive film layer on (a) 11% Cr F/M steel and (b) 15% Cr ODS steel in
different electrolytic solutions.
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(3) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O
(4) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3
728 724 720 716 712 708 704
Building energy (eV)
(a)
(b)
728 724 720 716 712 708 704
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(3) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O
(4) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3


950
C r
2
O
3
C r
met
C r(O H)
3
C r
2
O
3

C r 2p
(i) 1 kmol m
-3
HNO
3
592 588 584 580 576 572
(a)
I
n
t
e
n
s
i
t
y

(
a
.
u
)
B inding energy (eV)
(iii) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol
-3

Na
2
B
4
O
7
.10 H
2
O + 0.5 kmol m
-3

(ii) 1 k mol m
-3
HNO
3
+ 0.1 kmol m
-3
NaC l

(a)

C r
2
O
3
C r ( O H )
3
C r
2
O
3
C r
m e t
( i ) 1 k m o l m
- 3
H N O
3
( i i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O
5 9 2 5 8 8 5 8 4 5 8 0 5 7 6 5 7 2
( b ) C r 2 p
( i v ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O + 0 . 5 k m o l m
- 3

(b)
Fig. 7 The XPS spectra of Cr 2p detected for passive film layer on (a) 11% Cr F/M steel and (b) 15% Cr ODS steel in
different electrolytic solutions.
592 588 584 580 576 572
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(4) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3

(3) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O
592 588 584 580 576 572
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(3) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O

951
O 1 s
( i ) 1 k m o l m
- 3
H N O
3
H
2
O
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i i ) 1 k m o l m
- 3
H N O
3
+
0 . 1 k m o l m
- 3
N a C l
5 3 6 5 3 4 5 3 2 5 3 0 5 2 8
( a )
( i v ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O + 0 . 5 k m o l m
- 3
M e ( O H )
3
M e
2
O
3
( i i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+
0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O

(a)

O 1 s
( i ) 1 k m o l m
- 3
H N O
3
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l
( i i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+
0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O
5 3 6 5 3 4 5 3 2 5 3 0 5 2 8
( b )
M e
2
O
3
M e ( O H )
3
H
2
O
( i v ) 0 . 3 k m o l m
- 3
H
3
B O
3
+0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O
+ 0 . 5 k m o l m
- 3

(b)
Fig. 8 XPS spectra of O 1s detected for passive film layer on (a) 11% Cr F/M steel and (b) 15% Cr ODS steel in different
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol
m
-3
NaCl
(3) 0.3 kmol m
-3
H
3
BO
3
+
0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O
(4) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O
+0.5 kmol m
-3

536 534 532 530 528
536 534 532 530 528
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(3) 0.3 kmol m
-3
H
3
BO
3
+
0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O
(4) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3

952
204 202 200 198 196
C l
-
(a )
C l
-
C l 2p
I
n
t
e
n
s
i
t
y

(
a
.
u
)
B inding energy (eV )
1 kmol m
- 3
H N O
3
+ 0.1 kmol m
- 3
N aC l

(a)
C l
-
C l
-
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l
2 0 4 2 0 2 2 0 0 1 9 8 1 9 6
( b )
C l 2 p
( i i ) 0 . 3 k m o l m
- 3
H
3
B O
3
+ 0 . 0 7 5 k m o l
- 3

N a
2
B
4
O
7
. 1 0 H
2
O + 0 . 5 k m o l m
- 3

(b)
Fig. 9 The XPS spectra of Cl 1s detected for passive film layer on (a) 11% Cr F/M and (b) 15 % Cr ODS steel in chloride
containing electrolytic solutions.

Fig. 10 Schematic representation of passive film breakdown and pitting corrosion initiation in 11% Cr F/M and 15% Cr
ODS steels.
(1) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
204 202 200 198 196
(a) 11% Cr F/M steel
(b) 15% Cr ODS steel
204 202 200 198 196
(1) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
(2) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3
Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3

Cl
-
Cl
-

953

decisive factor to control initiation for the breakdown
are the imperfections sites on grain boundary,
non-metallic inclusions, or micropores in the film that
are responsible for the breakdown of the film.
Theories for passive film breakdown and pit initiation
have been categorized into three main mechanisms
[14, 15, 17] that focus on passive film penetration,
film breaking, or adsorption. Details of such
descriptions are beyond the scope of this work and are
discussed in detail elsewhere [14, 17].
The high resolution XPS spectra of Y 3p and Y 3d
of 15% Cr ODS steel are shown in Figs. 11a and 11 b.
In 1 kmol m
-3
HNO
3
and 1 kmol m
-3
HNO
3
+0.1 kmol
m
-3
NaClY 3d peak can be differentiated into two

Y 3 d
I
n
t
e
n
s
i
t
y

(
a
.
u
)
( i ) 1 k m o l m
- 3
H N O
3
1 6 4 1 6 2 1 6 0 1 5 8 1 5 6 1 5 4 1 5 2 1 5 0
( i i i ) 0 . 3 M H
3
B O
3
+ 0 . 0 7 5 M
N a
2
B
4
O
7
. 1 0 H
2
O
( a )

( i i ) 1 k m o l m
- 3
H N O
3
+ 0 . 1 k m o l m
- 3
N a C l

(a)
(i) 0.3M H
3
BO
3
+0.075M
Na
2
B
4
O
7
.10 H
2
O
306 304 302 300 298
(b)
I
n
t
e
n
s
i
t
y

(
a
.
u
)
Y 3p
Binding energy (eV)
(ii) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol
-3

Na
2
B
4
O
7
.10 H
2
O + 0.5 kmol m
-3

(b)
Fig. 11 The XPS spectra of (a) Y 3d and (b) Y 3p detected for passive film layer forms on 15% Cr ODS steel in different

(3) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O
(1) 1 kmol m
-3
HNO
3

(2) 1 kmol m
-3
HNO
3
+0.1 kmol m
-3
NaCl
164 162 160 158 156 154 152 150
(1) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol
m
-3
Na
2
B
4
O
7
10H
2
O
(2) 0.3 kmol m
-3
H
3
BO
3
+0.075 kmol m
-3

Na
2
B
4
O
7
10H
2
O +0.5 kmol m
-3

306 304 302 300 298

954

sub-peaks for Y
2
O
3
phase that appear at 158 eV (3d
5/2
)
and 160 eV (3d
3/2
). Whereas, in borate buffer solution
peaks of two different Y 3d and Y 3p are observed at
160 and 301 eV, respectively. The Y 3d doublets
binding energies of 158 eV and 160 eV show the
presence of Y
2
O
3
. The metallic Y 3d peak at 156 eV
are not observed. Similarly, the additional Y 3p
observed at 301 eV (Fig. 11b) not only show the
presence of Y
2
O
3
but may indicate its existent in more
than one chemical state [9, 11].
4. Conclusions
The corrosion resistance and passive film
characterization of 11% Cr F/M steel and 15% Cr
ODS steel in different electrolytic solutions has been
investigated. The results of the OCP plots of both steel
alloys are strongly dependent on the electrolytic
solutions. The OCP is less noble in chloride
containing solutions and more noble in borate buffer
solution in both the steel alloys. Active potential of
both alloys in chloride media is attributed to less
protective passive film due to Cl
-
.
The potentiodynamic polarization results
demonstrated that 15% Cr ODS steel offers higher
corrosion resistance than 11% Cr F/M steel in the
absence of chloride. In chloride containing solutions,
both steel alloys exhibited low pitting corrosion
resistance. However, pitting potential of 15% Cr ODS
steel is higher than 11% Cr F/M steel. The XPS
analysis of the oxide layers contains mostly Cr
2
O
3,

Fe
2
O
3
along with Y
2
O
3
in 15% Cr ODS steel. The
SEM micrograph of pit morphology mainly shows
shallow pit attack. The measured corrosion rate in
boiling nitric acid up to 48 h for both the steel alloys
shows high corrosion rate. In summary, the present
study provides detail insight into the effects of
different electrolytic media on the corrosion resistance
corrosion behavior of 11% Cr F/M and 15% Cr ODS
steels. In the presence of chloride, low pitting
corrosion was observed in both the steel alloys.
Acknowledgments
One author Dr. S. Ningshen gratefully acknowledged
Japan Society for the Promotion of Science, Tokyo,
Japan for the postdoctoral fellowship and financial
support.
References
[1] M.G. Fontana, N.D. Green, Corrosion, Engineering,
McGraw-Hill, New York, NY, 1967.
[2] L.L. Shreir, R.A. J arman, G.T. Burstein, Corrosion, 3rd
ed., Butterworth-Heinemann, Oxford, 1994.
[3] G.H. Koch, M.P.H. Brongers, N.G. Thompson, Y.P.
Virmani, J .H. Payer, Corrosion Costs by Industry Sector,
in: Corrosion Costs and Preventive Strategies in the
United States, Supplement to Materials Performance, J uly
2002, pp. 4-8.
[4] J .A. Richardson, J.L. Dawson, Economic aspects of
corrosion, Shreirs Corrosion 4 (2010) 3040-3051.
[5] S. Ukai, M. Fujiwara, Perspective of ODS alloys
application in nuclear environments, J . Nucl. Mater.
307-311 (2002) 749-757.
[6] S. Ukai, Oxide dispersion strengthened steels, Comp.
Nucl. Mater 4 (2012) 241-271.
[7] S. Ukai, T. Nishida, T. Okuda, T. Yoshitake, R & D of
oxide dispersion strengthened ferritic martensitic steels
for FBR, J . Nucl. Mater 258-263 (1998) 1745-1749.
[8] B. van der Schaaf, D.S. Gelles, S. J itsukawa, A. Kimura,
R.L. Klueh, A. Mslang, et al., Progress and critical
issues of reduced activation ferritic/martensitic steel
development, J . Nucl. Mat. 283-287 (2000) 52-59.
[9] S. Ningshen, M. Sakairi, K. Suzuki, S Ukai, The passive
film characterization and anodic polarization behavior of
11% Cr ferritic/martensitic and 15% Cr oxide dispersion
strengthened steels in different electrolytic solutions,
Applied Surface Science 274 (2013) 345-355.
[10] J . Bischoff, A.T. Motta, Oxidation behavior of
ferritic-martensitic and ODS steels in supercritical water,
J . Nucl. Mater 424 (2012) 261-276.
[11] S. Ningshen, M. Sakairi, K. Suzuki, S. Ukai, Corrosion
resistance and passive film characterization of 9 Cr oxide
dispersion strengthened steel in acidic and chloride
environment, in: Proc. of 6th J apan-China J oint Sem. on
Marine Corrosion, Tokyo, J apan, Mar. 2, 2012, pp. 7-16.
[12] B. Gwinner, M. Auroy, D. Mas, A. Saint-J evin, S.
Pasquier-Tilliette, Impact of the use of the
ferritic/martensitic ODS steels cladding on the fuel
reprocessing PUREX process, J . Nucl. Mat. 428 (2012)
110-116.
[13] C.W. Wagner, W. Traud, On the interpretation of

955
corrosion processes by superimposing of Electro
Chemical potential and the potential formation of mixed
electrodes, Z. Elektrochem 44 (1938) 391-402.
[14] G.S. Frankel, Pitting corrosion of metals: A review of the
critical factors, J. Electrochem. Soc. 145 (1998)
2186-2198.
[15] S. Ningshen, U.K. Mudali, R.K. Dayal, Electrolyte and
temperature effects on pitting corrosion of type 316LN
stainless steels, Brit. Corros. J. 36 (2001) 36-41.
[16] J. Isselin, R. Kasada, A. Kimura, T. Okuda, M. Inoue, S.
Ukai, et al., Effects of Zr addition on the microstructure
of 14% Cr 4% Al ODS ferritic steels, Mater. Trans. 51
(2010) 1011-1015.
[17] J. Kruger, Passivity of metals: A materials science
perspective, Inter. Mater. Rev. 33 (1988) 113-130.
[18] E. McCafferty, M.K. Bernettand, J.S. Murday, An XPS
study of passive film formation on iron in chromate
solutions, Corros. Sci. 28 (1988) 559-576.


Effect of Traffic Speed on Stresses and Strains in
Asphalt Perpetual Pavement
Daba S. Gedafa
1
, Mustaque Hossain
2
and Stefan A. Romanoschi
3
1. Department of Civil Engineering, University of North Dakota, Grand Forks ND 58202-8115, USA
2. Department of Civil Engineering, Kansas State University, Manhattan KS 66506, USA
3. Department of Civil Engineering, University of Texas-Arlington, Arlington TX 76019-0308, USA

Abstract: Increasing traffic volumes and loads as well as public expectation for a long-lasting transportation infrastructure have
necessitated designing perpetual pavements. The KDOT (Kansas Department of Transportation) conducted a field trial to investigate
the suitability of perpetual pavement concept for Kansas highway pavements. The experiment involved construction of four thick
pavement structures. To verify the approach of designing perpetual pavements on the basis of an endurance strain limit, the
pavements were instrumented with gauges for measuring tensile strains at the bottom of asphalt base layers at various speeds.
Pavements were also instrumented with pressure cells to measure stress on the top of subgrade. Pavement response measurements
under known vehicle load were performed in August 2006. FWD (Falling-weight deflectometer) was also used to collect deflection
data at 15 m intervals on the same date. FWD first-sensor (center) deflections were normalized and corrected to 20
o
C temperature
based on measured mid-depth pavement temperature. The result shows that strain and stress measurements show significant amount
of variations. Measurements in the thickest section are the most consistent. The higher the traffic speed, the lower the strains and
stresses. The difference between strains and stresses at 30 km/hr and 65 km/hr is higher than the difference between 65 km/hr and 95
km/hr. This shows the effect of speed on stresses and strains decreases as the speed increases. Softer binder in the asphalt base layer
results in lower strains, which confirms that softer binder results in higher fatigue life.

Key words: Traffic speed, stress, strain, perpetual pavement, deflection.

1. Introduction

The increasing traffic volumes and loads as well as
the public expectation for a longer-lasting
transportation infrastructure have necessitated
designing flexible pavements with a life of up to 50
years. Perpetual pavements are expected to meet the
need for such a longer-lasting pavement. Thick and
stiff pavement layers in perpetual pavements reduce
the strains and stresses at the bottom of the asphalt
concrete layer, thereby reducing the potential for
cracking to be initiated at the bottom [1].
Fatigue cracks may develop in the surface layer and
propagate horizontally in the top lift. Because they

Corresponding author: Daba S. Gedafa, Ph.D., P.E.,
assistant professor, research fields: mechanistic-empirical
design, pavement management system, non-destructive testing,
materials engineering, and traffic engineering. E-mail:
daba.gedafa@engr.und.edu.
appear at the pavement surface, top-down cracks can
be observed, and action can be taken to eliminate
them. Asphalt overlays or inlays are the most common
solutions currently applied for top-down cracks. The
cracks that initiate at the surface are not eliminated
through the perpetual pavement concept. The concept
leads to pavement structures that crack only at the
surface and need repair only at the surface. This
failure mode leads to significant monetary savings for
the repair, rehabilitation, and reconstruction of these
pavements [2].
The KDOT (Kansas Department of Transportation)
conducted a field trial to investigate the suitability of
perpetual pavement concept for Kansas highway
pavements. The experiment involved construction of
four perpetual pavement sections. To verify the
approach of designing perpetual pavements on the
basis of an endurance strain limit (70 micro-strains at
DAVID PUBLISHING
D
Effect of Traffic Speed on Stresses and Strains in Asphalt Perpetual Pavement

957
the bottom of asphalt layer), the four pavements were
instrumented with gauges for measuring tensile strains
at the bottom of the asphalt base layers. Pavements
were also instrumented with pressure cells to measure
stresses on the top of subgrade. Pavement response
measurements under known vehicle load and FWD
(falling-weight deflectometer) were also performed [2].
2. Objectives
The main objective of the study was to investigate
the effect of traffic speed on stresses and strains in
perpetual pavement sections. Relationship between
the FWD first-sensor (center) deflection and measured
responses were also investigated.
3. Test Sections
Test sections in this study are the perpetual asphalt
pavement sections on US highway 75 (US-75). The
experiment involved construction of four thick
pavement structures on a new alignment of US-75
near Sabetha, Kansas. To verify the approach of
designing perpetual pavements based on an endurance
strain limit, four pavements were instrumented with
gauges for measuring strain at the bottom of the
asphalt base layers. A research team from Kansas
State University placed instrumentation systems in the
four pavement structures during their construction in
J une 2005. Table 1 shows the pavement structures.
The estimated design cumulative traffic for these
pavements is 2.6 million ESALS/lane for 10 years and
5.7 million ESALS/lane for 20 years [2].
The KAPA (Kansas Asphalt Pavement Association)
provided the design of Sections 1, 2, and 3. Section 4
was designed by KDOT. Sections 1 and 3 have the
same thickness, but a softer binder (performance
grade (PG 64-22)) SM (superpave mix) was used in
the construction of the base asphalt mix instead of PG
70-22, and a richer and more ductile asphalt mix was
used in the bottom lift of the base layer for Section 3.
It is expected this mix will have a longer fatigue life.
The test sections were constructed on a fill and each
was approximately 390 m long with approximately
150 m transition zones between them. The asphalt
paving work was done in June 2005. The project was
completed and the experimental sections were opened
to traffic at the beginning of November 2005 [3].
4. Data Collection
4.1 Stress and Strain Data
Stress and strain data have been collected on four
perpetual pavement sections. Each test section has
multiple gauges, with half of the sensors located in the
right wheel path (outer lane), and the other half
located 150 mm to the right. Pavement response
measurements were performed under 80 kN standard
truck load. On each pavement section, three sets of
five passes of the loading truck were performed. Five
passes were performed with the truck passing at about
30 km/hr, 65 km/hr, and 95 km/hr, in order to
determine the effect of vehicle speed on pavement
responses.
Horizontal strains and the vertical stress on top of
subgrade as well as the position of the loading truck

Table 1 Configuration of test sections.
Section 1 2 3 4
Acronym KAPA (Standard) High reliability KAPA 2 (Modified) KDOT
Surface course 40 mm, SM 9.5A
a
(PG 70-28)
Binder course 60 mm, SM 19 A (PG 70-28)
Base course 225 mm, SM 19A (PG 70-22)
175 mm, SM 19 A
(PG 64-22)
225, SM 19 A
(PG 64-22)
b

300 mm, SM 19A
(PG 64-22)
Stabilized subgrade 150 mm, 6% hydrated lime mixed with the natural soil
Natural subgrade High plasticity clay (A-7-6)
a
nominal maximum aggregate size is 9.5 mm and the gradation is above the maximum density line,
b
the bottom 3 was designed at 3%
air voids for a binder-rich layer (Pb =6.0%, design air voids =3% 2%, VFA =77%).

958

were recorded with a national instruments data
acquisition system at a rate of 300 records per second.
A sampling rate of 3,000 Hz was used and the average
value for ten samples were recorded. The data were
recorded in text format in separate files for each
passes of the vehicle and were then processed using
Microsoft excel [2].
4.2 FWD Deflection Data
FWD (falling-weight deflection) deflection data
used in this study were collected with a Dynatest 8000
FWD. The tests were done at 15 m intervals.
4.3 Pavement Mid-depth Temperature
Pavement mid-depth temperature was measured on
test sections during data collection session. The
thermocouple of a temperature gauge was lowered
into the holes drilled in the AC (asphalt concrete)
layers and filled with oil to measure the temperature at
the mid-depth of each AC layer at the time of
response measurements [2]. Measured mid-depth
temperature was used to correct normalized FWD
first-sensor (center) deflection data to a standard
temperature of 20
o
C.
5. Data Analysis
Structural capacities of flexible pavements are
determined from surface deflection measurements.
Currently the most widely used NDT (nondestructive
testing) system is the FWD (falling-weight
deflectometer) [4]. Measured deflections by FWD
must be corrected to a standard load, usually 40 kN [5]
and to a predefined environmental condition. The
most critical environmental factor affecting
deflections in flexible pavements is the temperature of
the AC layer [6-8]. Thus pavement deflection data
must be adjusted to a reference temperature [9]. A
two-step correction procedure typically needs to be
applied. The first step consists of measuring the
effective temperature of the AC layer, and the second
step consists of adjusting the surface deflection to a
reference temperature using a correction factor [10].
Normalized FWD center deflection data were
corrected to a standard temperature of 20
o
C based on
measured mid-depth temperature. It should be noted
that stress and strain measurements were not
made at exactly the same location as the FWD
deflection data.
6.1 Longitudinal Strain
Fig. 1 shows the effect of traffic speed on the
longitudinal strain. In general, the higher the FWD
center deflection, the higher the longitudinal strain at
the bottom of the asphalt layer. The higher the traffic
speed, the lower the longitudinal strain. Some of the
gauges were removed during construction and later
retrofitted in Section 3. That may be one of the
reasons why the trend in longitudinal strain in Section
3 is somewhat different.
Comparison of maximum and average longitudinal
strain at various speeds is shown in Fig. 2. The
minimum longitudinal strain was not included since it
is zero for all sections at all speeds. The higher the
traffic speed, the lower the longitudinal strain for all
sections. Section 4 shows the lowest maximum
longitudinal strain at all speeds, which may be due to
its highest thickness. The lowest average longitudinal
strain is observed in Section 3 at all speeds, which
shows that Section 3 has higher fatigue life because of
its softer base mix.
Table 2 shows the significant difference test results
for longitudinal strain at 5% significance level.
Longitudinal strain at 30 km/hr is not significantly
different from the longitudinal strain at 65 km/hr in all
sections except Section 4 whereas it is significantly
different from the longitudinal strain at 95 km/hr in all
sections except Section 3. Longitudinal strain at 65
km/hr mph is significantly different from longitudinal
strain at 95 km/hr in three out of the four sections.


959

(a) Section 1 (b) Section 2

(c) Section 3 (d) Section 4
Fig. 1 Effect of traffic speed on longitudinal strain.

Fig. 2 Comparison of longitudinal strain at various speed, L30, L65, and L95 is longitudinal strain at 30 km/hr, 65 km/hr,
and 95 km/hr speed, respectively.

0
5
10
15
20
25
30
35
40
45
0 50 100 150 200 250 300 350 400
L
o
n
g
i
t
u
d
i
n
a
l

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
Center Deflection (micrometer)
L30 L65 L95
0
5
10
15
20
25
30
35
40
45
50
0 100 200 300 400 500 600
L
o
n
g
i
t
u
d
i
n
a
l

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
L30 L65 L95
0
5
10
15
20
25
30
35
40
45
50
0 100 200 300 400 500 600
L
o
n
g
i
t
u
d
i
n
a
l

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
L30 L65 L95
0
5
10
15
20
25
30
35
0 50 100 150 200 250 300
L
o
n
g
i
t
u
d
i
n
a
l

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
L30 L65 L95
0
5
10
15
20
25
30
35
40
45
50
L30 L65 L95 L30 L65 L95 L30 L65 L95 L30 L65 L95
S1 S2 S3 S4
L
o
n
g
i
t
u
d
i
n
a
l

s
t
r
a
i
n

(
M
i
c
r
o
-
s
t
r
a
i
n
)
Sections at various speed
Max Avg
L30 L65
L95

960

Table 2 Significant difference test for longitudinal strain.
Sections
Comparison
group
Probability
(P-value)
Similar
Number of
data points
S1
L30
L65 0.0708 Yes
18 L95 0.0022 No
L65 L95 0.3655 Yes
S2
L30
L65 0.1761 Yes
15
L95 0.0274 No
L65 L95 0.0006 No
S3
L30
L65 0.2592 Yes
L95 0.1432 Yes
L65 L95 0.0127 No
S4
L30
L65 0.0002 No
L95 < 0.0001 No
L65 L95 0.0141 No
6.2 Transverse Strain
Fig. 3 shows the effect of traffic speed on
transverse strains at the bottom of the asphalt concrete
layer. It shows that the higher the center deflection,
the higher the transverse strain. In general, the
transverse strain also decreases with an increase in
speed. Transverse strains at 65 km/hr and 95 km/hr
are closer to each other than the transverse strains at
30 km/hr and 65 km/hr. The trend is somewhat
different in Section 3 as shown in Fig. 3c. This may
be due to lower number of gauges since some were
removed during construction and later retrofitted after
construction.


Fig. 3 Effect of traffic speed on transverse strain.
0
20
40
60
80
100
120
0 50 100 150 200 250 300 350 400
T
r
a
n
s
v
e
r
s
e

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
T30 T65 T95
0
50
100
150
200
250
0 100 200 300 400 500 600
T
r
a
n
s
v
e
r
s
e

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
T30 T65 T95
0
10
20
30
40
50
60
0 50 100 150 200 250 300 350 400 450
T
r
a
n
s
v
e
r
s
e

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
T30 T65 T95
0
10
20
30
40
50
60
70
80
90
100
0 50 100 150 200 250 300
T
r
a
n
s
v
e
r
s
e

S
t
r
a
i
n

(
m
i
c
r
o
-
s
t
r
a
i
n
)
T30 T65 T95
T30 T65
T95
T30 T65
T95

961

Fig. 4 Comparison of transverse strain at various speed, T30, T65, and T95 is transverse strain at 30 km/hr, 65 km/hr, and
95 km/hr speed, respectively.

Table 3 Significant difference test for transverse strain.
Sections
Comparison
group
Probability
(P-value)
Similar
Number of
data points
S1
T30
T65 0.0144 No
18 T95 0.034 No
T65 T95 0.5108 Yes
S2
T30
T65 0.3056 Yes
15
T95 0.2808 Yes
T65 T95 0.2896 Yes
S3
T30
T65 0.7875 Yes
T95 0.6202 Yes
T65 T95 0.0189 No
S4
T30
T65 0.001 No
T95 <0.0001 No
T65 T95 0.0092 No

Fig. 4 shows comparison of maximum and average
transverse strains at various speeds. Minimum
transverse strain was not included since it was zero in
all sections. The higher the speed, the lower the
transverse strain in all sections. The highest average
and maximum longitudinal strain is observed on
Section 2 since it is the thinnest section. The lowest
average and maximum transverse strain is observed in
Section 3. This confirms that softer binder in base
layer results in lower strain.
Table 3 shows the significant difference test results
for transverse strain. Transverse strain at 30 km/hr is
significantly different from the transverse strains at 65
km/hr and 95 km/hr in about half of the sections.
Transverse strain at 65 km/hr is significantly different
from the one at 95 km/hr in 50% of the sections.
6.3 Stresses on Subgrade
Fig. 5 shows the effect of traffic speed on the
stresses on subgrade. Stress on the subgrade is mostly
zero at lower center deflection in all sections except
Section 3. The higher the traffic speed, the lower the
stress (pressure) on subgrade. Stress on subgrade at 65
km/hr and 95 km/hr are closer to each other than
stresses on subgrade at 30 km/hr and 65 km/hr. There
is a negative stress on subgrade in Section 1, which
shows that it is tensile stress at that point.
Fig. 6 shows the comparison of minimum,
maximum, and average stresses on subgrade at
various speeds for all sections. Minimum stress on
subgrade is zero in all sections except in Section 1 at
65 km/hr and 95 km/hr where it is negative. Negative
minimum stress indicates a tensile stress. The highest
0
50
100
150
200
250
T30 T65 T95 T30 T65 T95 T30 T65 T95 T30 T65 T95
S1 S2 S3 S4
T
r
a
n
s
v
e
r
s
e

s
t
r
a
i
n

(
M
i
c
r
o
-
s
t
r
a
i
n
)
Max Avg

962
maximum and average stress on subgrade are
observed in Section 2 since it is the thinnest section.
The lowest maximum and average stress on subgrade
are observed in Section 3.


Fig. 5 Effect of traffic speed on stress on subgrade.

Fig. 6 Comparison of stress on subgrade at various speed, P30, P65, and P95 is pressure on the top of subgrade at 30 km/hr,
65 km/hr, and 95 km/hr speed, respectively.
-100
0
100
200
300
400
500
600
0 50 100 150 200 250 300 350 400
P
r
e
s
s
u
r
e

o
n

S
u
b
g
r
a
d
e

(
k
P
a
)
P30 P65 P95
0
200
400
600
800
1000
1200
0 100 200 300 400 500 600
P
r
e
s
s
u
r
e

o
n

S
u
b
g
r
a
d
e

(
k
P
a
)
P30 P65 P95
0
5
10
15
20
25
0 50 100 150 200 250 300 350 400 450
P
r
e
s
s
u
r
e

o
n

S
u
b
g
r
a
d
e

(
k
P
a
)
P30 P65 P95
0
50
100
150
200
250
300
350
400
450
500
0 50 100 150 200 250 300
P
r
e
s
s
u
r
e

o
n

S
u
b
g
r
a
d
e

(
k
P
a
)
P30 P65 P95
-200
0
200
400
600
800
1000
1200
P30 P65 P95 P30 P65 P95 P30 P65 P95 P30 P65 P95
S1 S2 S3 S4
S
t
r
e
s
s

o
n

s
u
b
g
r
a
d
e

(
k
P
a
)
Min Max Avg

1,200
1,000
800
600
400
200
0
-200
1,200
1,000
800
600
400
200
0
P
r
e
s
s
u
r
e

o
n

S
u
b
g
r
a
d
e

(
k
P
a
)

P30 P65
P95
P30 P65
P95

963

Table 4 Significant difference test for stress on subgrade.
Sections
Comparison
group
Probability
(P-value)
Similar
Number of
data points
S1
P20
P40 0.1334 Yes
18 P60 0.1175 Yes
P40 P60 0.0463 No
S2
P20
P40 0.3134 Yes
15
P60 0.1775 Yes
P40 P60 0.0887 Yes
S3
P20
P40 0.0834 Yes
P60 0.0006 No
P40 P60 0.0037 No
S4
P20
P40 0.2171 Yes
P60 0.1628 Yes
P40 P60 0.1698 Yes

Table 4 shows significant difference test results for
stresses on subgrade at various speeds. There is no
significant difference between the stresses on
subgrade at 30 km/hr and the other two speeds in all
sections except the stress on subgrade at 95 km/hr in
Section 3. Stresses on subgrade at 65 km/hr and 95
km/hr are significantly different from each other in
about half of the sections.
7. Conclusions
Based on this study, the following conclusions can
be made:
Strain and stress measurements show significant
amount of variation. Measurements in the thickest
section are the most consistent;
The higher the traffic speed, the lower the strains
and stresses in general. This confirms the viscoelastic
nature of asphalt;
The difference between strains and stresses at 30
km/hr and 65 km/hr is higher than the difference
between 65 km/hr and 95 km/hr in general. This
shows the effect of speed on strains and stress
decreases as the speed increases;
Softer binder in asphalt base layer results in
lower strains, which confirms that softer binder results
in higher fatigue life.
Acknowledgments
The authors would like to thank the Kansas
Department of Transportation for sponsoring this
study.
References
[1] Transportation Research Circular 503: Perpetual
Bituminous Pavements, TRB (Transportation Research
Board) of the National Academies, Washington, DC,
2001.
[2] S.A. Romanoschi, A.J . Gisi, M. Portillo, S. Dumitru, The
first findings from the Kansas perpetual pavements
experiment, Transportation Research Record: Journal of
the Transportation Research Board 2068 (2008) 41-48.
[3] M.M. Portillo, Measured and theoretical response of
perpetual pavement structures, Master Thesis, The
University of Texas at Arlington, 2008.
[4] A. Loizos, G. Zachos, Investigation of the applicability of
the L-FWD for flexible pavement layer materials,
IJ P-International J ournal of Pavements 5 (1-3) (2006)
62-74.
[5] K. Chatti, S.W. Haider, H.S. Lee, Y. J i, H. Salama,
Evaluation of nonlinear and dynamic effects on asphalt
pavement response under FWD loading, IJ P-International
J ournal of Pavements 2 (1-2) (2003) 88-99.
[6] Y.R. Kim, Y.C. Lee, Interrelationships among stiffnesses
of asphalt-aggregate mixtures, J ournal of the
Association of Asphalt Paving Technologies 64 (1995)
575-609.
[7] L. Shao, S.W. Park, Y.R. Kim, Simplified procedure for
prediction of asphalt pavement subsurface temperatures
based on heat transfer theories, Transportation Research
Record: J ournal of the Transportation Research Board
1568 (1997) 114-123.
[8] H.M. Park, Y.R. Kim, S. Park, Temperature correction of
multiload-level, falling-weight deflectometer deflections,
Transportation Research Record: J ournal of the
Transportation Research Board 1806 (2002) 3-8.
[9] D.H. Chen, J . Bilyeu, H.H. Lin, M. Murphy, Temperature
correction on falling-weight deflectometer measurements,
[10] D.Y. Park, N. Buch, K. Chatti, Effective layer
temperature prediction model and temperature correction
via falling-weight deflectometer deflections,


Stresses Analysis on a Rail Part
Cassio Eduardo Lima de Paiva
1
, Jos Luz Antunes de Oliveira e Sousa
1
, Luiz Carlos de Almeida
1
, Paulo Roberto
Aguiar
1
, Luiz Fernando de Melo Correia
1
, Juliana Silva Watanabe
1
, Rodrigo Moreira de Carvalho
1
, Creso de
Franco Peixoto
1, 2
and Denis Palomo Paschoalin
3

1. Geotecnical and Ttransportation Department & Department of Structure, School of Civil Engineering, University of Campinas,
Campinas 13083-852, Brazil
2. Ignatian Educational Foundation, So Bernardo do Campo 09850-901, Brazil
3. VALE S. A., Belo Horizonte 30140-001, Brazil

Abstract: The EFVM (Vitoria Minas Railroad) is one of the main railways in Brazil. It transports freight trains of ore, 220 wagons
each. These wagons have 2 boogies of 2 axles each and 32 metric tons on metre gauge. Elastic strains were measured on a special
part of this railway due to these trains. The main load to evaluate stresses and strains was a G 16 Locomotive, a C-C kind from Vale,
a Brazilian Company. The measurements were obtained by dynamic deflectometer installed on a main line of this railway, near
Ipatinga, a city from Minas Gerais, one of Brazil states. This track was equipped to obtain stresses under an equal repeated static load.
A simulation of the stresses was made under critical strain by Ferrovia 1.0 software. It was also made an evaluation of unequal results
from neighbor sleepers taking in comparison two equipped parts of this railway, one with compacted ballast and no compaction to
the other. The results were strain limited, avoiding breakage or damage to the studied rails. This work analyses these measurements
focusing on the improvement of track quality.

Key words: Railway track, field strain survey, rail stresses.

1. Introduction

Nowadays, the Brazilian railway loads have
increased constantly, even when the very high rate
goes to road transportation, practically carrying
almost all Brazilian passengers and the major part of
loads.
One of the main reasons that led to the reduction or
canceling of Brazilian long journey passenger trains is
the high traffic volume accumulated over the years on
lines sub-maintained to those practically abandoned
ones. As dealers revenue have not reached sufficient
amount for adequate track maintenance, it was
decided the cancellation of passenger services. In this
case, canceling passenger services, the operational
speed was reduced, minimizing rail failure risk by
means of the reduction of dynamic impact coefficient.
In this case, to balance the productivity, the reduction

Corresponding author: Cassio Eduardo Lima de Paiva,
professor, research field: transportation infrastructure. E-mail:
celpaiva@fec.unicamp.br.
of speed was counterbalanced with the increase of
longer trains.
This abandon state of passenger services was due to
some special features of Brazilian railway
privatization program, ran late 1990s. At that time
almost all Brazilian railways belonged to the
government. The absence of adequate level of
resources from federal budget to keep them running,
pushed the government to start a concession program.
Besides the concession granted 50 years to each
enterprise, that could be thought as a very large period
of concession, this length of time would not guarantee
money for a wide operation of passenger and load
trains, because the high level of worn rails and rotten
sleepers would require an excessive amount of
financial resources. If the dealers decided to continue
passenger services they needed to spend money that
could jeopardize the financial balance of its business.
This long period of concession, fifty years, could be
thought as a kind of sale, not an effective concession,
DAVID PUBLISHING
D

965
to reduce government costs. The adequate length of
time permits government and dealer to evaluate the
business, to know if the service quality reached the
expected level and if the profit would be fair to social
and economical approaches.
In order to optimize the growing transportation
volume, mainly concerning to ore, it was necessary to
settle new railway maintenance procedures. The
constant increase of traffic railway demands detailed
analyses on all Brazilian railways, from platform
bearing capacity to sleepers and rails lifetime. In this
maintenance scenario, rails stresses and strains are the
most important study, to guarantee transportation
continuous service and to keep enough capacity to
attend the growing demand. The new decision-making
was conceived to guide Brazilian railway maintenance
processes, which is a group of procedures and
techniques program as a director plan. To apply this
group of procedures, technicians selected parts of
EFVM, a metric gauge railway that crosses Minas
Gerais, one of the largest States of Brazil. In two lines
of EFVM, deflection measurements were taken and
the results were analyzed. In these railway parts, the
rails are fixed on steel sleepers. The main study was to
analyze rail stresses and strains from each studied
railway part. The principal difference between them
was ballast tamping levels.
Each studied railway part received a special name.
The line part named Case A presents a non tamped
ballast. In this line, repetition measurements were
very common. It was needed to remake the work
always that the high vibration and elevated deflections
generated non reliable group of values. The wide
deflection on its sleepers became difficult for
evaluations. The work on the other line, adequately
tamped, named Case B, occurred without any
difficulty. Probably the higher ballast stiffness helped
to avoid excessive vibration and wide deflection
amplitude to the sleeper.
The estimated stresses related to the Case A were
calculated considering sleepers under double spacing,
because it was significant the amount of sleepers that
did not show effective contact between bottom of rail
base and sleeper top. This criterion was adopted to
reduce the error between estimated and real stresses.
The priorities during the field measurements held
reliable data and, as much as possible, a significant
amount of values. Nevertheless, all measurements that
did not present good quality or adequate results were
discarded. For data acceptance or discharging decision,
it was used excessive vibration or wide sleeper
deflection as the main tools. Therefore, it could be
done the comparison of estimated and calculated
stresses, related to measured deflections and
calculated values by elastic models, respectively. The
deflections data were obtained by means of strain
gauges adhered to the rails.
2. Rail Deflections Caused by Loading
The measurements were taken at several parts of
EFVM near Ipatinga, a city between Belo Horizonte
and Vitoria, Capitals of Minas Gerais and Espirito
Santo States, respectively. Its traffic is composed by
heavy hall and passenger trains. This line is one of
two long journey passenger trains offered today in
Brazil. The main features of this railway at these
studied places are shown in Table 1.
According to Table 1, the profile of features of this
studied railway part, can be considered as a
representative one to the Brazilian railways. The TR
68, a Brazilian rail standard equivalent to 136 RE is
commonly applied to lines under high traffic, mainly
for those of ore trains. The Deenik rail fastening, an
elastic one, is also very applied for long extension of
rails all over the country. But, the steel sleeper of this
studied railway is not very common. The wooden
sleeper is the most common on Brazilian railways,
mainly from reforested eucalyptus wood, because it is
the cheapest Brazilian sleeper, despite its low capacity
to permit higher speeds. Nevertheless, it was decided
to evaluate this railway as a representative one,
because its traffic has also passenger trains, a profile


966

Table 1 Main features of the studied railway parts at EFVM.
Element or feature Kind Main characteristics
Rails
TR-68 (136 RE)
(Brazilian rail standard)
Brazilian rail standard is equivalent to the pattern 136 AREMA (America railway
engineering maintenance)
Rail fastening Deenik Elastic rail fastening
Sleepers UIC-865
Steel sleepers: 2.20 m length, 0.23 m width and 0.09 m height, 12 mm minimum
thickness on the surface, 600 mm spacing
Gauge Metric 1,000 mm
Ballast Crushed stone Thickness 0.39 m, grading curve correction

Fig. 1 Schematic views of G 16 locomotivebogies and axles of G 16 (C 30 M 3) [1].

Table 2 G 16 (C 30 M 3) locomotivemain features [1].
Traction kind Electro-diesel
Builder GE (general electric)
Wheel arrangement C-C kind
Power output 3,000 HP
Wheel diameter 914.4 mm
Total weight 1,253,670 kN (125,367 kgf)
Load per axle
209,000 to 220,000 kN (20,900 to
22,000 kgf)

of usage that is very important to be reestablished on
the Brazilian scenario, under strong social importance
approach. The metric gage is the most common all
over the country. The most common Brazilian rock
formations to get crushed stone for ballast are
composed by diabase, basalt, granite, gneiss or
limestone, which require particle size distribution
correction (grading curve correction) among other
procedures to meet high load capacity and good
drainage [2].
The rails of these two selected railway parts are
TR-68 (136 RE), the heaviest Brazilian standard.
They are fastened on steel sleepers. In each of these
studied parts were installed 12 gauges in different
parts of the rail, to measure deflections generated by
the traffic of the pattern locomotive, the GE C 30 M-3,
named G16 VALE [1]. The studied locomotive, a C-C
kind, has a gap of 1.680 mm between first and second
axles and 2.020 mm between the second and third
axles.
The general layout of this locomotive is shown in
Fig. 1 and its main features are shown in Table 2.
Dimensions Length (m)
A 4.55
B 2.76
C 3.53
D 1.59
E 10.42
F 18.65
G 1.91
H 0.08
I 3.54
J 0.75

967
The railway reaction to the locomotive movement
was evaluated by elastic deflection, which was
measured by two laser deflectometers installed in each
selected sleeper for measurements [3]. In each sleeper
selected, one is on the sleeper tip and the other, on the
ballast level, next to the sleeper. This procedure
allowed for strain basin design from the entire
locomotive shape, as shown in Fig. 2. The applied
field procedure approached a specific and repetitive
way to install the apparatus in each research site,
considering gauges, computer and connectors. This
adopted procedure was conceived to avoid
measurement differences generated by possible
divergences from received frequencies that could
generate inacceptable results.
For each pair of measurements, during rail
evaluation, the track and ballast elasticity modulus
was estimated, according to a retro-analysis model. In
Case A, the spatial twist on the sleepers, a common
movement generated by load application differences
on each rail of the studied sleeper, causes deflection
measurement losses. When this occurs, the laser
emission does not return to the signal receptor, losing
data. This problem was common when using laser
deflectometers for line strain measurements. In Case
B, the deflection values permitted to build the
deflection basin shown in Fig. 2.
The retro-analysis procedure considered the system
charged with the load amplitude and geometry of the
studied locomotive and also the track features. From
the deflection measurements, it was determined the
Winkler coefficient applied to the Case B, equals to
0.6 N/mm
3
(6 kgf/cm
3
).
The adopted Winkler coefficient was considered
adequate for railways working under good level of
maintenance. According to Eisenmamm, values
between 0.5 N/mm
3
(5 kgf/cm
3
) and 1.0 N/mm
3
(10
kgf/cm
3
) are considered as, respectively, good
condition of maintenance and very good condition
of maintenance [4].
In the two studied parts, Case A and Case B, six rail
gauges were installed in each rail [5]. Two of them
were fixed at the middle of sleepers spacing, i.e., in
the middle of the rail part length between two sleepers
support. In this situation, one gauge was stuck on the
head side, i.e., on the side surface of the head and the

Fig. 2 Deflection basinCase B.
Laser 1
Laser 2
-2.5
-2
-1.5
-1
-0.5
0
0.5
0 5 10 15 20 25 30 35 40
10-1-G16-BHV
Dormente de A; espamento 60 cm
D
E
F
O
R
M
A
O

(
m
m
)
DORMENTE
D
e
f
l
e
c
t
i
o
n

(
m
m
)

Distance (m)
10-1 G 16BHV Locomotive
Steel sleeper60 cm spacing deflection (mm)

968

other on the superior face of base rail. In this case, it
was expected to estimate the maximum force moment.
The other four gauges were positioned near sleeper
support, stuck on the rail the same way as the previous
ones, as presented at the B-B and C-C cross section in
Fig. 3. Therefore, the estimated stress was related to
the highest sheer force. Fig. 4 shows rails and
electrical devices, where it is easy to see these
considered dots before, places that were installed
gauges and its connectors. In this picture, the wires
are shown going to the right direction, from each
studied gauge. According to Fig. 3, the gauges were
named as Strain 0 and Strain 1 at central section,
Strains 2 to 5 near the sleeper support, Strain 2 and 3
to the left and Strains 4 and 5 to the right.
To the data field acquisition, it was used an
equipment model SCXI-1001 from national
instruments, with a computer caption card type
SCXI-1521B for deflections acquisition, receiving
data from electric gauges considering Wheatstone
Bridge Theory. The data acquisition program was
done in LabView and stored in a notebook [6].

Fig. 3 Gauges distribution in relation of rails and sleepers.

Fig. 4 Electrical devices on the rail side (Case A and Case B).

Sleepers spacing (cm)
Belo Horizonte
Vitria
Gauges 2 and 3
B B Cross Section
C C Cross Section
Gauges 0 and 1
0
1
4
5
Gauges 4 and 5
62.0 58.0 59.0 60.0

969
-700 10
-6
-467 10
-6
-233 10
-6
0
233 10
-6
467 10
-6
700 10
-6
157 157.5 158 158.5 159 159.5 160 160.5 161
Strain_1
Strain_0
t (s)
VIT BH

-700 10
-6
-467 10
-6
-233 10
-6
0
233 10
-6
467 10
-6
700 10
-6
157 157.5 158 158.5 159 159.5 160 160.5 161
Strain_3
Strain_2
VIT BH
t (s)

-700 10
-6
-466.7 10
-6
-233.3 10
-6
0
233.3 10
-6
466.7 10
-6
700 10
-6
157 157 158 158 159 159 160 160 161
Strain_5
Strain_4
VIT BH
t (s)

Case A Case B
Fig. 5 Longitudinal deflection graphs of six rail parts (Case A: left column; Case B: right column). Units: abscissastime
in seconds (s); ordinatesspecific deflection (10
-6
m/m).

700 10
-6
467 10
-6
233 10
-6
-233 10
-6
-467 10
-6
-700 10
-6
700 10
-6
467 10
-6
233 10
-6
-233 10
-6
-467 10
-6
-700 10
-6
t (s)

t (s)

700 10
-6
467 10
-6
233 10
-6
-233 10
-6
-467 10
-6
-700 10
-6
-233 10
-6
-467 10
-6
-700 10
-6
700 10
-6
467 10
-6
233 10
-6
-233 10
-6
-467 10
-6
-700 10
-6
-233 10
-6
-467 10
-6
-700 10
-6

700 10
-6
467 10
-6
233 10
-6
700 10
-6
467 10
-6
233 10
-6
t (s)
t (s) t (s)

t (s)

970

The Gauges 4 and 5 were installed in an equivalent
way of Gauges 2 and 3, but inside the other rail, as
shown in Fig. 3.
Fig. 5 shows the graphs of strain versus time [7].
These data were obtained from the six mentioned
gauges of Fig. 3, during G 16 locomotive passage,
loading and unloading the rails. The graphs on the left
column of Fig. 5 are applied to Case A and the graphs
of right column, to the Case B. Fig. 5 shows strains
which are influenced from the period of charge, which
means the time between each locomotive axle passing
at this studied point.
In Fig. 5, the high amplitude of deflections shown
on the left column, Case A related, compared to the
amplitude deflections shown on the right column,
Case B related, could be explained in relation of
tamping differences. The support rail deficiency of the
Case A line, observed during locomotive passage,
depends strongly to the adequate tamping absence and
to the material composition quality. The line of Case
B shown best line answers, which means, reduced
amplitude in comparison to the Case A.
On the deflection graphs of Case B are observed
straight proportionality between those measurements
from a point on the sleeper or from a point at the
middle of two sleepers. These results are also close to
the same value.
Concerned to Case B in Fig. 5, when the
locomotive axle groups are equidistant from the
studied section occur minimum strain on the middle
space of the sleeper, which could be reported as a
value of 20% of the maximum measured ones.
Concerned to Case A in Fig. 5, the strain
measurements at point 5 showed a difference of 30%
between the load of first and third axles of the same
locomotive group of axles because the support of the
sleeper under it was inoperative, i.e., this sleeper was
rotten and did not achieve to support the rail.
3. Stresses Evaluation
Other relevant feature that influences on the quality
line of both lines, in terms of bearing capacity, is the
use of crushed stone to ballast composition. Table 3
shows maximum strain for both two cases, permitting
to analyze focusing its limit values.
The calculated values of stresses were obtained
from the measured strains multiplied to an adopted
value of elasticity modulus (E), the known Youngs
Modulus, under elasticity theory model, shown at Eq.
(1) [8]. The values obtained are shown in Table 4.
E
(1)

where,
: rail stress as function of measured deflection;
E: rail elasticity modulus (Youngs Modulus), E =
205.000 N/mm
2
(2.050.000 kgf/cm
2
);
: elastic deflection (elastic strain) of rail or track
caused by the load pattern (according to the used
locomotive).
The maximum calculated stresses are 128.13
N/mm
2
(1,281 kgf/cm
2
) at Gauge 5 of Case A and
38.95 N/mm
2
(389 kgf/cm
2
) at Gauge 3 of Case B,
among values of Table 4. These values were calculated

Table 3 Maximum deflection values for each studied case (Case A and Case B).
Strain 0 Strain 1 Strain 2 Strain 3 Strain 4 Strain 5
Case A 215 10
-6
390 10
-6
290 10
-6
545 10
-6
385 10
-6
625 10
-6

Case B 117 10
-6
150 10
-6
117 10
-6
190 10
-6
153 10
-6
185 10
-6

Table 4 Maximum stresses to Case A and Case B.
Units
0

1

2

3

4

5

Case A
N/mm
2
44.08 79.95 59.45 111.73 78.93 128.13
kgf/cm
2
440.75 799.5 594.5 1,117.25 789.25 1,281.25
Case B
N/mm
2
23.99 30.75 23.99 38.95 31.37 37.93
kgf/cm
2
239.85 307.5 239.85 389.5 313.65 379.25


971

by Eq. (1), according to the adopted elastic deflection.
The graph of Fig. 2 shows how the deflections recover,
typical in an elastic medium.
The comparison of these two critical stresses permit
to analyze the risk of overburden any studied railway
part. The quotient of these two values, Case A
maximum stress by Case B maximum stress equals to
128.13/38.95 = 3.28, which means a surcharge of
more than three times of stress at rails of Case A than
rails of Case B. This relationship of stresses indicates
a critical state of use which will reduce dramatically
the rail lifetime.
The Case A and Case B data were loaded to
Ferrovia 1.0 software, a FEM (finite element model)
program developed for analysis in ballast on railways.
The main features of locomotive and each studied
railway part were considered than the sleeper spacing.
In this case, it was considered double spacing.
The railway ballast coefficient, its elasticity
modulus, was estimated under retro-analysis,
considering track-ballast measurement deflections
shown in Fig. 2. So, ballast elasticity, E
L
= 240
N/mm
2
(2,400 kgf/cm
2
) and platform elasticity, E
P
=
60 N/mm
2
(600 kgf/cm
2
). Table 5 presents results
from Ferrovia 1.0 and stresses related to Youngs
Modulus, permitting comparisons.
In Table 5, it is possible to calculate the difference
between stresses values from Ferrovia 1.0 and stress
values from measured deflections. For Case B,
stresses developed on the base rail and to the support
point (on the sleeper), the difference between stress by
software application and stress proportional to
deflection, is (34.3 31.40)/31.4 = 0.09, approximately
10%. Nevertheless, this studied point, is the main to
study shear efforts not flexural stresses. For Case A,
stresses developed on the base rail and to the mid-span
position, the difference between stress by software
application and stress proportional to deflection, is
(59.0 44.1)/59.0 = 0.25, approximately 25%. So,
these levels of errors can be admitted as acceptable
ones. Otherwise, for Case B, stresses developed on the
head of rail and to the support point (on the sleeper),

Table 5 Base rail and top head rail stresses (Case A and Case B).
Case Measurement position Units Force moment Units
Calculated stress:
Ferrovia 1.0
Measured stress
(Youngs
Modulus)
Measured stress
(Youngs
Modulus)
Head Base (foot) Head Base (foot)
A
On the sleeper

Nm 2507679 N/mm
2
75.0 55.3 128.1 78.9
kgfcm 255708 kgf/cm
2
750.2 553.3 1,281.25 789.25
Middle of sleeper
spacing
Nm 2671825 N/mm
2
79.9 59.0 80.0 44.1
kgfcm 272446 kgf/cm
2
799.3 589.6 799.5 440.75
B
On the sleeper
Nm 1927528 N/mm
2
46.6 34.3 39.0 31.4
kgfcm 196550 kgf/cm
2
466.4 342.8 389.5 313.65
Middle of sleeper
spacing
Nm 1874620 N/mm
2
56.1 41.4 30.8 24.0
kgfcm 191155 kgf/cm
2
560.8 413.6 307.5 239.85

Table 6 Differences of base rail and top head rail stresses (Case A and Case B).
Case
Measurement
position
Units
Calculated stress: (Ferrovia 1.0 software)
Measured stress
(Youngs
Modulus)
Measured stress
(Youngs
Modulus)
Head Base (foot) Head Base (foot)
A
On the sleeper N/mm
2
(75.0-128.1)/128.1 = 0.41 (55.3-78.9)/78.9 = -0.30 128.1 78.9
Middle of sleeper
spacing
N/mm
2
(79.9-80)/80 = -0.01 (59.0-44.1)/44.1 = 0.34 80.0 44.1
B
On the sleeper N/mm
2
(46.6-39)/39 = 0.19 (34.3-31.4)/31.4 = 0.09 39.0 31.4
Middle of sleeper
spacing
N/mm
2
(56.1-30.8)/30.8 = 0.82 (41.4-24)/24 = 0.73 30.8 24.0


972

the difference between stress by software application
and stress proportional to deflection, is (128.1
75)/128.1 = 0.42, approximately 42%. For Case B,
stresses developed on the head of rail and to the
mid-span, the difference between stress by software
application and stress proportional to deflection, is
(56.1 30.8)/30.8 = 0.82, approximately 82%. These
values are excessive high that were not envisaged in
the Ferrovia 1.0 software. Table 6 shows errors of all
calculated and measured stresses.
4. Conclusions
The behavior of ballast at the studied non tamped
railway parts was equivalent to a fluid (Case A). The
sleeper showed significant vertical displacement and
torsion effort. Unexpected stress peak was observed
on their rails, which could accelerate the rail lifetime
if this condition would be constantly repeated, as
successive cycles. The ballast densification is an
unknown effect, in relation of the layers and soil
features. The wide sleeper displacement practically
prohibited any laser measurement. As the laser was
stick on a reference beam and its bulkhead positioned
on the sleeper, any significant vibration could damage
the data quality.
The peak stresses occurrence probably is an effect
of the uneven line, caused by sleepers under
hypothesis of displacement. A new research is needed,
because, in this work, it was just adopted double
spacing sleepers to balance this problem.
The Case A measurements can be used as a model
to analyze other Brazilian railways that show high
level of rotten and missed sleepers among good ones.
In this paper, the amount of in situ measurements
under true greatness and the comparison of two
models of railway evaluation associated to a specific
FEM software permitted to conclude that the
traditional theoretical approach show good
consistency to the studied cases.
References
[1] Informaes Tcnicas das Loconotivas
C.V.R.D.Acervo Tcnico, Companhia Vale do Rio
Doce, 2002.
[2] C. Esweld, Modern Railway Track, 2nd ed., MRT
Productions, Zaltbommel, 2001.
[3] C.E.L. Paiva, C.F. Peixoto, L.F.M. Correia, P.R. Aguiar,
Evaluation between two Brazilian railway tracks, Journal
of Civil Engineering and Architecture 5 (46) (2011)
772-778.
[4] F.O. Rives, A.L. Pita, M.J.M. Puente, Treaty of
RailroadTrack Way, Rueda S.L., Madrid, 1977.
[5] K. Hoffmann, An Introduction to Measurements Using
Strain Gages, 1st ed., Hottinger Baldwin Messtechnik
GmbH, Darmastadt, 1989.
[6] National Instruments, LabView User Manual, National
Instruments Corporation, Austin, Texas, 2003.
[7] C.E.L. Paiva, L.C. Almeida, J.L.A.O. Sousa, Evaluation
of strains in railways on ballast from data collected in the
field, in: Proceedings 53rd Brazilian Congress of
Concrete, Florianpolis, 2011.
[8] S.P. Timoshenko, Strength of Materials, Trad. Jos
Rodrigues de Carvalho, Ed. USP, So Paulo,
1978.


Advanced Methods of Evaluating Benefits from
Improved Flood Immunity in Queensland
Wayne Davies
Department of Transport and Main Roads, Queensland 4000, Australia

Abstract: The regular occurrence and the high costs of flooding to both road agencies and communities is a strong impetus to
investigate the methodologies applied to evaluating flood immunity road projects. Very little literature exists on methods of
evaluating the benefits of improving flood immunity through better road infrastructure. This paper attempts to address some of the
numerous issues hindering the accurate evaluation of flood immunity road projects. The methodologies presented in this paper are
designed to evaluate benefits that are not normally included or not fully considered in evaluations; such benefits include reduced
flood related road accidents, costs of not travelling during the closure period and additional costs from unanticipated flood events. A
key focus of the paper is the consideration of the evaluation of improved flood immunity from a network perspective rather than the
typical approach of evaluating flood immunity projects in isolation to each other. The application of the proposed methodologies is
demonstrated with a hypothetical example of a typical rural network subjected to regular flooding. The results of the analysis,
conducted using the proposed methodologies, reveals that the currently applied methodologies account for less than half the likely
value of benefits of a package of projects that will prevent the complete isolation of communities during serious flood events.

Key words: Flood immunity, economic evaluation, cost benefit analysis, road closures.

1. Introduction

Roads in rural Queensland are frequently subject to
flooding. The accurate prediction of the occurrence of
these floods is not always possible as Queensland
frequently experiences unpredictable weather.
Flooding can potentially cause millions of dollars of
damage to road infrastructure [1] as well as millions
of dollars worth of damage to the economy and
society [2]. The severe flooding in early 2010 cost
TMR (transport and main roads) over $110 million in
damages across all regions [3]. Prior to floods in late
2010 and early 2011 more than 15% of Queensland
State projects were NDRRA (natural disaster relief
and recovery arrangements) and most of these projects
involve treating roads damaged by flooding [4].
Preliminary estimates of the total costs to road
infrastructure of the late 2010 and early 2011 floods
are expected to be close to $4.5 billion [5]. TMR [6]

Corresponding author: Wayne Davies, master, research
fields: transport economics, cost benefit analysis and road
project evaluation. E-mail: wmd03@yahoo.com.
have identified 36 sites along the Bruce Highway
1

between Brisbane and Cairns that flood during
major flooding events. Four of those sites resulted in
road closures of over five days in the recent 2011
floods.
The areas most impacted by floods are the South
West, Central West, Far North and Northern Regions
of Queensland. Table 1 contains the rivers that have
flooded most frequently and the number of times they
have flooded between 2000 and 2010.
The flooding of the Tully, Haughton, Herbert,
Murray and Burdekin rivers can cause road closures
along the Bruce Highway in the Northern Region,
while the flooding of the Mary River can cause road
closures along the Bruce Highway near Gympie
(Population: 42,820 [7]) and Maryborough
(Population: 21,501 [8]) possibly significantly
reducing access to highly populated areas. The

1
The Bruce Highway is a major highway linking Brisbane to
northern Queensland; the Bruce Highway has a length of over
1,600 km [9].
DAVID PUBLISHING
D
Advanced Methods of Evaluating Benefits from Improved Flood Immunity in Queensland

974
flooding of the Warrego River can cause reduced
access to Charleville (Population: 3,728 [10] from the
north. The flooding of the Thomson, Barcoo,
Diamantina and Georgina rivers as well as the Eyre
and Cooper creeks can cause road closures along the
Barkly and Landsborough highways, which form an
important freight route between Mount Isa (North
West Region) and Longreach (Central West Region).
The flooding of the Paroo and Bullo rivers can cause
road closures of the developmental roads connecting
the small towns of Quilpie, Thargomindah, Eromanga
and Eulo.
The projected NDRRA costs on state controlled and
national highways from 2010 to 2014 according to
TMR
2
[4] are given in Table 2.
Queenslands northern regions are expected to
experience the highest flood related road
infrastructure costs. Most of these costs can be
accredited to the severe flooding in these areas in
recent years, improving flood immunity will greatly
reduce these costs in the future.
Flood immunity projects are difficult to evaluate,
very little literature exists pertaining to the methods of
evaluating these projects. This paper proposes some
feasible methods of assessing the benefits and costs
associated with improved flood immunity. To
demonstrate how these methods can be applied, an
example of a bridge upgrade has been included in this
paper. The methods suggested in this paper can help
decision makers prioritize flood immunity
projects by identifying which road closures are more
costly.
2. Issues
The evaluation of a flood immunity project
involves the consideration of many factors. These
factors can vary in complexity depending on the
location of the project site and extent of the network.

2
These values are current as of November 17, 2010 according
to RIP Live databases, these figures may vary from those
published in the QTRIP (Queensland Transport and Roads
Investment Program).
The collection and organization of data often poses
the largest and trickiest part of evaluating a flood
immunity project. The analyst is confronted with a
number of issues, one such issue is the determining of
which data is relevant. This issue can not be easily
addressed as the data required depends on the detail
and the methodology applied to the evaluation.
Sources of data are not always available, thus, forcing
the analyst to either simplify the evaluation or
make assumptions. This section of the paper
indentifies relevant issues relating to networks, key
evaluation parameters, interrelated projects, flood
related accident costs and wider economic
benefits.
2.1 Networks
The network plays an essential role in the
evaluation of flood immunity projects. Good
knowledge of the network enables a more accurate
prediction of the benefits obtained from an
improvement in flood immunity. Important aspects of
the network that should be considered are the size of
the network, the existence of alternative routes, the
types of vehicles that can access these routes and the
types of vehicles utilizing the network.
When encountered with a flood, the road user
normally has three options, divert, wait or not travel
[11]. The decision of which option to take depends on
which option the road user deems to cost
3
the least. If
floods can be anticipated, the road user is more likely
to select an optimal strategy during the flood, thus
making the prediction of road user behavior easier to
make. If the flood is unanticipated, road user behavior
is less likely to be optimal and road user costs will be
higher during a flood [12].
In the case of an anticipated flood, the analyst
should obtain the optimal road user response to
flooding [13]. The factors, which influence this response,

3
Costs in this case refer to the road users perceived costs,
which includes travel time costs.

975

Table 1 Flooding frequency of rivers in Queensland.
Ranking
River Region 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010 Total
1 Paroo South West 6 0 2 1 1 1 0 5 3 4 4 27
2 Bulloo South West 7 1 2 0 1 1 0 4 3 2 4 25
3 Cooper Creek Central West 5 1 1 1 2 0 1 3 3 3 4 24
4 Tully Far North 3 2 0 1 3 1 2 3 2 4 0 21
5 Georgina Central West 4 1 0 0 3 2 0 2 1 4 2 19
6 Diamantina Central West 4 1 0 0 2 0 0 3 1 5 3 19
7 Thomson Central West 5 2 1 0 1 0 0 1 2 3 4 19
8 Barcoo Central West 5 2 1 0 1 0 0 1 2 1 3 16
9 Haughton Northern 3 0 1 0 0 1 1 1 3 2 3 15
10 Condamine South West 2 0 0 2 1 0 0 0 4 0 5 14
11 Eyre Creek Central West 3 0 0 0 2 2 0 2 0 3 1 13
12 Don River Whitsunday 2 0 1 0 0 1 1 2 3 1 2 13
13 Herbert Northern 2 1 1 0 1 0 1 2 1 2 1 12
14 Warrego South West 2 0 1 1 1 0 0 1 1 0 5 12
15 Murray Northern 0 0 0 0 3 0 2 2 1 3 0 11
21 Burdekin Northern 1 1 1 0 0 1 0 1 0 1 2 8
23 Mary Wide Bay 0 1 0 0 1 0 0 1 1 1 2 7
Source: Adapted from Bureau of Meteorology [14].

Table 2 Projected NDRRA costs for the next five years prior to the 2011 floods.
NDRRA OSCR (other stated controlled roads) National
Regions Cost ($) No. Projects Cost ($) No. Projects
Darling Downs 6,159,000 43 476,000 6
Metropolitan 1,129,000 30 111,000 3
Fitzroy 14,540,000 33 450,000 2
Northern 88,886,000 4 33,159,000 7
Far North 278,684,000 174 45,703,000 15
North West 55,697,000 27 33,682,000 16
South West 20,501,000 15 - 0
Central West 29,233,000 41 2,340,000 9
Mackay/Whitsunday 27,495,000 138 17,747,000 19
Wide Bay/Burnett 481,000 12 25,000 2
North Coast 2,039,000 56 111,000 3
South Coast 5,231,000 13 550,000 1
Total 530,075,000 586 134,354,000 83
Average cost per project 904,565 1,618,723
Source: TMR [15] RIP Live accessed November 17, 2010.

Fig. 1 Simple mud map of a road network. Legend: Red linesQ2 Bridges/Culverts; Brown linesQ10 Bridges (mass
limit 20 tonnes); Blue linesQ20 Bridges; Green linesQ50 Bridges.

976

are the road users value of time, the flexibility of
travel options
4
, the duration of the road closure, the
length of the diversion route if a diversion route exists,
the type of vehicle the road user is driving and any
other factors particular to the road or network. Data
may or may not be available for all these factors and
in these cases, the analyst may wish to exclude the
factor or make some assumptions to its value.
In the case of an unanticipated flood, the analyst
could still proceed in determining the optimal road
user response but consider this response as a best-case
scenario. Additional factors need to be considered
such as the impact of imperfect information about
durations of road closures and the most appropriate
diversion routes considering other roads may have
also been impacted by the flood. In the case of
unanticipated flooding, sensitivity analysis around
possible road user behavior could improve the
robustness of the evaluation.
Some routes will attract road users with a higher
value of time. These routes are likely to have a high
percentage of commercial vehicles. Depending on the
type of business located in the region, commercial
vehicle road users could value time very differently
depending on the urgency of the journey. If the flood
can be anticipated, the road users with a high value of
time can either travel before or after the floods arrive.
If the flood is unanticipated, the timing of journeys
can not be easily changed. Considering this logic, the
road user costs of unanticipated flooding are likely to
be higher than the costs of anticipated flooding.
The length of the diversion route plays a significant
role in determining the behavior of the road user.
Long diversion routes that add to the cost of the
journey are unattractive and may cause more road
users to wait or not travel. In some cases, diversion
routes do not exist at all, forcing road users to wait or
not travel. Some networks allow for multiple

4
Flexibility of travel options refers to the road users
flexibility in travelling at different times or to different
locations.
diversion routes, which change depending on the
extent of the flooding and mass limits of the roads.
Fig. 1 is a mud map of a simple rural network with
multiple possible diversion routes.
Route ACDB is the fastest route between Towns A
and B but unfortunately approximately every two
years the sections CD and DB are closed because of
flooding. The route ACFB is the next fastest route but
a bridge along section CF has a mass limit of 20
tonnes. Therefore, heavy vehicles will divert using the
route AEFB, which is considerably longer than ACDB.
The route ACFB may also be flooded on occasions as
section CF is closed 1/5 the time CD and DB are
closed. Route AEFB will rarely flood, thus, providing
road users with a reliable route from A to B.
Anticipated floods will not create a great
disturbance to this network as viable diversion routes
exist for road users who wish to travel or if trips are
more flexible, road users can reschedule trips to
periods before or after the flooding. Unanticipated
floods could cause a greater disturbance to this
network. Some road users may need to backtrack to
Town B if the bridge along DB is impassable. Road
users may not know if route ACFB or AEFB should
be taken if information is not available to whether the
bridge along CF is passable, thus, causing further
delays. Town D could be isolated from the rest of the
network without important resources. The costs to
isolated towns or communities are not fully recognized
in TMR cost benefit analysis methodology.
2.2 Key Parameters and Their Impact on Project
Benefits in Currently Practiced TMR Cost Benefit
Analysis
Many flood immunity projects do not have enough
quantifiable benefits when analyzed with currently
practiced cost benefits analysis approaches. The four
key determinants of quantifiable benefits for flood
immunity projects are AADT (annual average daily
traffic), AATOC (annual average time of closure),
ADC (average duration of closure) and the length of

977
the diverting route.
AADT is a key determinant of benefits in all road
project evaluations. A road with an AADT of 10,000
vehicles will have fifty times more benefits than a
road with 200 vehicles assuming ceteris paribus.
Flooding is predominantly an issue in rural areas
where AADT is normally low. The benefits per
vehicle maybe large but if the AADT is low, the
overall benefits will not cover the costs of building a
bridge or culvert to reduce road closures.
The AATOC is a parameter unique to projects
involving road closures. The number of hours a road
is closed has a large impact on the total benefits of the
project. The benefits normally captured in a cost
benefits analysis are the immediate benefits of
reduced road closure times. These benefits come in
the form of reduced waiting time, reduced VOC
(vehicle operating costs) and reduced TTC (travel
time cost). Costs of not travelling during a road
closure and the long-term growth of communities are
excluded from analysis. A standard AATOC is
normally applied to all vehicle types. In reality, road
closure times can vary depending on the vehicle type.
A road may have a load limitation for a period of time
(dry-back period) after the road has been reopened
to cars [11]. This load restriction imposes a
cost on heavy vehicle road users that is often not
considered.
The ADC is another parameter unique to projects
involving road closures. The ADC can have a large
impact on the overall benefits of a project as the ADC
influences road user behavior. If road closures are
long, a higher percentage of road users will either
divert or not travel. If the road closures are short,
more road users will wait for the roads to re-open.
This behavior will vary depending on the extent of
information available regarding the duration of the
road closures. If flooding is anticipated, the durations
of the road closures are likely to be known and road
users will act more efficiently. If the flooding is
unanticipated, road users may choose less efficient
options.
The costs per road user during an anticipated road
closure are likely to decrease as road users shift from
not travelling to diverting as diversion routes become
more viable and then as the duration shortens road
users will switch from diverting to waiting.
Shown in Fig. 2 is an example of a possible cost
structure to a road user depending on when the road
user encounters a road closure. Initially, the road user
would not travel, as the cost of diverting is higher than
the value of the journey to the road user. After
duration of x, an alternative diversion route becomes
available, the cost of taking this diversion route is less
than the cost of not making the journey
5
. The road
user will choose to travel along the diversion route
until the cost of waiting and the cost of the journey
along the original route is less than the cost of
travelling along the diversion route. The red line in
Fig. 2 illustrates the road users optimal behavior, as
the road becomes closer to re-opening.
In the case of an unanticipated flood, the duration
of closure is uncertain and alternative routes are
uncertain, too. The parameters in Fig. 2 can therefore
be roughly estimated, and are likely to be inaccurate.
This inaccuracy will cause road users to make
inefficient choices, thus resulting in higher costs per a
road closure. Currently practiced cost benefit analysis
will not accurately detect these inefficiencies.
Another key determinant of overall benefits for a
flood immunity project is the length of the diversion
route. Long diversion routes could greatly increase the
costs of road closures, if the ADC is long. A further
complicating issue is that diversion routes could
change depending on the extent of flooding. If the
flood is small and localized, the diversion route could
be short as other roads are not flooded. If the flood is
large and covers a large area, the diversion route could
become longer or no longer exist if the entire local
network is flooded. Determining the diversion routes

5
The cost of not making the journey is the difference between
the value of the journey to the road user and the cost of the
journey to the road user when the road is not closed.

978

Fig. 2 Possible road user behavior during a road closure.

Fig. 3 Mud map of an example of project dependency.

to use in the evaluation can be complicated and often
the shorter diversion routes are selected, thus the
additional costs of diverting during larger floods are
negated
6
.
2.3 Interrelated Projects
If one part of a network is inundated, it is likely that
other parts of the network are also inundated. It is
possible that other flood immunity projects will be
proposed to resolve flooding in other parts of the
network. When a project is first suggested, the
diversion route may be large but this diversion route
could be significantly reduced when another route is
upgraded.
Another important aspect that should be considered
is the flood immunity along the entire route.
Improving the flood immunity on one part of the route
may only improve the flood immunity of the entire
route to a certain extent. Other sections of the route
could still be prone to occasional flooding. This other

6
This conclusion was derived from reading a sample of BCA
reports conducted in recent years.
section of the route may be upgraded later. The
uncertainty surrounding this could make evaluating
the current flood immunity project difficult, thus
invoking the use of even more assumptions. The
optimal choice of diversion route may or may not be
changed by the upgrade of the other part of the route.
If the author assumes the same legend applied to
Fig. 1 is applied to Fig. 3, the diversion route in the
base case is AEFB for road users travelling between
Town A and B. It can be seen that on average CD is
flooded every 10 years. This is inconsequential in the
base case for vehicles travelling between Town A and
B, as whenever CD is flooded DB is also flooded. In
the project case, if the bridge along DB is upgraded to
a Q50 bridge
7
, the flooding along CD becomes
relevant, as ACDB, the shortest route, is not always
viable, therefore, the route AEDB is taken instead.
2.4 Flood Related Accidents
From 1788 to 1996, at least 2,213 people have died

7
A Q50 bridge is built to a standard where it will only flood
during a Queensland 50-year or worse flood event.

979
as a result of over 926 recorded floods [16].
According to Ashley and Ashley [17], 63% of flood
fatalities occur in vehicles. Jonkman and Keiman [18],
identified that 33% of all flood related activities
involved drowning in vehicles, most of these occurred
because road users attempted to cross flooded roads. If
the authors apply these rates to the fatalities that occur
in Australia, approximately an average of 1.5 people
die in their vehicles and approximately an average of
0.75 people die crossing flooded roads per flood.
Improving flood immunity would decrease the
occurrence of these fatalities.
2.5 Wider Economic Benefits
Flooding reduces road access and therefore will
likely reduce the economic growth of regions isolated
by floods. Improved flood immunity may generate
traffic that will directly benefit from improved access.
This generated traffic will likely produce wider
economic benefits to the community. Department for
Transport [19] has conducted extensive research into
wider economic benefits in general. Wider economic
benefits of improved flood immunity could be
substantial and should be considered for further
research.
3. Approach
To overcome some of the issues discussed in the
previous section, this paper offers some approaches
and methodologies. The extent of these approaches
can be applied depend on the network and the
necessity of a detailed analysis. This section also
explains how these same methodologies can be
applied to determine which roads are more feasible for
flood immunity upgrades.
A simple mud map of the network can be used to
ascertain the extent of the network to be analysed. The
mud map only needs to contain the network of roads
that are likely to be flooded and the diversion routes
that road users are likely to use during a flood. Figs. 1
and 3 are examples of such mud maps. All sections of
the mud map, which are prone to flooding, should be
identified. Data should be collected on the extent of
the flooding and types of flood immunity treatment
that these sections have received or are proposed to
receive.
With the use of relevant data and mud maps,
diversion routes can be determined based on the
extent of the flooding. Road and traffic parameters
need to be established for the diversion routes. Once
all the available information is gathered, an informed
decision to the exact approach can be made. Complete
information will enable a detailed analysis, while
incomplete information may require assumptions to be
made or the analysis could be simplified to cater to the
information available.
3.1 The Model
A flooding model has been constructed to estimate
the costs of flooding to road agencies and road users.
This model expands upon the current methodology
applied to flood immunity projects. This model
attempts to address most of the deficiencies in the
current model. This model is intended to work in
synchronisation with current CBA models by applying
the values per VKT (vehicle kilometre travelled)
derived from these models. The issues addressed in
the flooding model are those of network impacts,
unanticipated flooding, underestimated flooding
benefits, accounting for different diversion routes by
different vehicle types, flood related accidents and
assessing the impact of interrelated projects.
The road agency costs applied to this model are the
costs to the road agency to restore the road to an
acceptable standard. These costs are averaged and
allocated annually. Road user costs are the standard
costs for road closures (diverting, not travelling and
waiting costs). These costs are calculated across the
network for each route for various severity of flooding.
Severity of flooding has been categorized as minor,
moderate and major according to the Bureau of
Meteorology cited by Attorneys-Generals

980
Department [20]. Each category has been divided into
anticipated and unanticipated flooding.
The model requires TTC and VOC outputs per
VKT per route from conventional cost benefit analysis
models. Considering the potential number of routes
impacted by flooding, the model does not
endogenously calculate TTC and VOC, these values
can be entered directly into the flooding model. The
length of the original route and the diversion route are
required to determine the additional costs of diverting.
The ADC is used to determine the costs of waiting for
the road to re-open during a road closure. In the case
of an unanticipated flood, road users are assumed to
have imperfect information about the duration of
closure. To account for the imperfect information, an
upper and lower bound of the expected duration of
road closure have been incorporated into the model.
The maximum cost of a road user is willing to pay to
make a journey
8
is used in the model to estimate the
elasticity of demand the road users have for travel
along the route. The additional costs of diverting, the
duration of closure and maximum cost of a road user
is willing to pay to make a journey are used to
determine the road user behavior during a flood.
3.2 Calculation of Road User Behavior and Flooding
Costs
Fig. 2 presented a relationship between the duration
of the road closure and road user behavior. The
flooding model is used to calculate the optimal road
user behavior given perfect (anticipated flood) and
imperfect (unanticipated flood) information. Road
user behavior given perfect information can be
assumed to be optimal. Road users will select the
option that minimizes their perceived costs. To derive
this behavior, perceived costs of diverting, waiting
and not travelling need to be estimated. Costs of
diverting are TTC, VOC, accident and externality.
TTC are normally accepted as perceived [21], VOC

8
This is the value of the destination to the road user, this value
can vary quite considerably depending on the nature of the
journey.
are accepted by some as perceived and not by others,
accidents and externalities are not normally
considered perceived. For the purpose of this model,
TTC and VOC are assumed to be perceived. TTC and
VOC per vehicle are calculated by Eqs. (1) and (2),
respectively:
) ((
) ( ) ( ) ( DR DR DR km Div
Dist AdjF TTC TTC

) ( ) ( ) ( OR OR OR km
Dist AdjF TTC
9
(1)

) ( ) ( DR DR km Div
Dist TTC VOC

) ( ) ( OR OR km
Dist TTC
(2)
TTC
Div
=Additional TTC per vehicle for diverting;
TTC
km(DR)
=TTC per km of the diversion route;
TTC
km(OR)
=TTC per km of the original route;
AdjF
(DR)
=TTC adjustment factor of the diversion
route;
AdjF
(OR)
=TTC adjustment factor of the original
route;
Dist
(DR)
=Section length
10
(km) of the diversion
route;
Dist
(OR)
=Section length (km) of the original route;
VOC
Div
=Additional VOC per vehicle for diverting.
TTC is multiplied by an additional factor to account
for a potential difference in the value of travel time for
different types of floods. This parameter should have
a higher value for floods that are unanticipated. The
perceived costs of diverting is the sum of the
additional TTC per vehicle and the additional VOC
per vehicle (Pcost
Div
= TTC
Div
+ VOC
Div
). The
perceived costs of diverting are used to determine
when the option of waiting becomes more favourable
than diverting. The point in time when waiting is more
favourable to diverting has been denoted as the
equilibrium waiting time, waiting before the
equilibrium is reached is considered not feasible if the
authors assume road users select the least cost option.

9
Costs of using the diversion route are assumed to exceed the
costs of using the original route, as the original route is the
preferred route during a non-flood event.
10
The length of road is between the start and finish of the
defined diversion route and original route.

981
The equilibrium waiting time is given in Eq. (3):
|
|
.
|
\
|
= DC
Cost
MaxCostNT
Cost
PCost
WT
perhr perhr
Div
E
, , min
(3)
where,
WT
E
=Equilibrium waiting time (hours);
PCost
Div
=Perceived costs of diverting;
Cost
per hr
=Cost of one hour of waiting;
MaxCostNT = Maximum perceived cost of not
travelling;
DC =Duration of closure.
The equilibrium waiting cost is assumed to equal
the perceived cost of diverting or if a diversion does
not exist, the equilibrium waiting time is the
minimum cost of not travelling or the duration of the
road closure.
The maximum cost of not travelling
11
has been
used to determine a linear demand function for travel.
The maximum cost of not travelling is the maximum
value of the journey to the road user minus the cost of
travelling along the original route. The percentage of
road users not travelling when waiting is not feasible
or when a diversion route does not exist, NT is given
in Eq. (4):
|
|
.
|
\
|
= 1 , , min
MaxCostNT
Cost DC
MaxCostNT
PCost
NT
perhr
Div
12
(4)
The percentage of road users diverting when
waiting is not feasible (D) or the percentage of road
users waiting when a diversion route does not exist
(W) equals to one minus (NT). The formulae for NT
and D are expressed graphically in Fig. 4.
The percentage of road users opting to not travel
when waiting is not feasible is XY/0Y and the
percentage of road users opting to divert when waiting
is not feasible is 0X/0Y.
The costs of diverting are calculated as the sum of
perceived costs, accident costs and externalities, the
formula is given in Eq. (5):

11
Maximum cost of not travelling can be adjusted according to
the duration of the road closure depending on whether the
opportunity cost of the alternative activity is sensitive to time.
12
This equation holds true, as the demand function is assumed
linear.
+ + = ) (
Div Div Div
y Externalit Accident PCost TDC
24
) ( D AADT WT DT
E

(5)
Accident costs are derived based on crash costs per
model road state defined in the Transport and Main
Roads cost benefit analysis model, which are derived
from Austroads paper AP-R184 [22]. Externality costs
are calculated based on vehicle kilometres travelled
with 2007 unit values provided by Austroads paper
IR-156/08, which are updated using the CPI
(consumer price index) provide by the ABS
(Australian Bureau of Statistics) [23].
The average costs of waiting per road closure are
half the perceived costs of diverting if the author
assumes the traffic volume is evenly distributed across
the duration of the road closure. For road users who
either not travel or wait when diverting is not feasible,
the costs of waiting are calculated as an eighth the
perceived costs of diverting. An eighth is derived by
halving the costs of not travelling and the costs of
waiting when diverting is not feasible, which is half
the costs of waiting when diverting is feasible, which
is half of the perceived costs of diverting. The costs of
waiting are given in Eq. (6):
+

=
24
) 1 (

2
1
NT AADT WT
PCost WC
E
Div
24
8
1
NT AADT WT
PCost
E
Div

13
(6)
The costs of not travelling are derived from the
perceived costs of diverting and the percentage of
vehicles not travelling, the formula is given in Eq. (7):
+

=
24
) (

2
1
NT AADT WT Dc
PCost NTC
E
Div
24
8
1
NT AADT WT
PCost
E
Div

(7)
The average costs of not travelling per road closure
are half the perceived costs of diverting if the authors
assume the demand for travel to locations that require
the use of the closed route is linear. For road users
who either not travel or wait when diverting is not
feasible, the costs of not travelling are calculated as an

13
If a diversion route does not exist, PCost
Div
should be
replaced with Min(MaxCostNT, DC Cost
perhr
), this same
assumption is applied to Eq. (6).

982

Fig. 4 Linear demand function used to determine number of road users not travelling.

eighth of the perceived costs of diverting
14
. In
Appendix A, the formulae are expressed and
explained in three-dimensional graphs. The road
closure costs are calculated annually by multiplying
the diverting, waiting and not travelling costs by the
AEP (annual exceedance probability)
15
.
Costs of repairing roads damaged by floods
(NDRRA) are allocated to the relevant sections of
roads damaged by flooding. Both a base case and a
project case estimate should be established. These
costs should be lower in the project case than the base
case and should be considered as maintenance benefits
of improved flood immunity. Flood related accident
costs have been calculated for major floods by
multiplying the average number of flood related
fatalities described in Section 2.4 by the Austroads
unit cost per fatality [24]
16
. The removal of mass
restrictions on upgraded roads could provide
additional benefits during the flood period as well as

14
The cost of waiting and the costs of not travelling are
assumed equal for the period when diverting is not feasible.
15
AEP is the probability that flooding will occur, this
probability is calculated with the formula |
.
|
\
|
=
ARI
AEP
1
exp 1 ,
ARI is the average recurrence interval of flooding [25].
16
Formula applied to the model is given Appendix A.
improved access all year round. The benefits of the
removal of mass limits during the flood period can be
assessed in the flood model but benefits for the rest of
year should be calculated independently to this model.
3.3 An Example of the Application of Methodologies
Discussed
To demonstrate the capturing of benefits using the
currently practiced cost benefit analysis approach and
the approach suggested in this paper, two hypothetical
bridge projects have been assessed. They have been
assessed using CBA6, partial endogenous road user
behavior CBA model and the proposed flood
immunity model.
CBA6 is the Transport and Main Roads CBA tool,
the partial endogenous road user behavior (PERUB)
CBA model is an excel based spreadsheet derived
from the NIMPAC algorithms harmonized in
Austroads paper AP-R264/05 [26] with partially
endogenously derived road user behavior during road
closures.
The approaches suggested in this paper have not
been applied to an actual evaluation, therefore, a
hypothetical case study of a plausible scenario has

983
been derived. The flood immunity projects evaluated
in this case study are two bridge upgrades in an area
of low AADT with limited access during severe
floods.
The collection of data of the types of floods that the
region experiences are important as the nature of the
floods could greatly influence the results of the
evaluation. In the instance of this case study, data of
the types of floods are assumed available. The full set
of data applied to the evaluation is given in
Appendix B.
In this example, TTC and VOC costs per vehicle
have been extracted from TMR cost benefit analysis
software (CBA6). The weighted average costs per
vehicle for each section have been entered into the
model using the formula given in Eq. (8):
=
=
=
=
|
|
.
|
\
|
+
|
|
.
|
\
|
+

=
n i
i
i
n i
i
i
r
Vehicles No
r
Vehicles No icle CostPerVeh
WAC
1
1
) 1 (
.
) 1 (
.

(8)
where,
i =Year of evaluation;
r =Discount rate.
This average is applied to the model and
extrapolated over the life of the project based on
projected growth rates and discount rates. Data is not
always available for all the relevant factors, hence,
assumptions should be made. To obtain a more
accurate result, the number of assumptions should be
kept to a minimum.
3.4 Results of Models
The results obtained from CBA6 and PERUB are
quite similar. Both models consider an original route
and a diversion route between C and B. In this model,
all vehicles are able to divert to avoid the road
closures. The mud map applied to these models is
given in Fig. 5. The proposed model is applied to a
larger part of the network as shown in Fig. 6.
Final and intermediate destinations of vehicles are
required as inputs to this model. CBA6 and PERUB
only require the AADT of each section of road. The
results of the three models are presented in Table 3
and input data is provided in Appendix B.
The benefits derived from the proposed model are
considerably higher than both those obtained from
CBA6 and PERUB. The main source of this
difference is that the costs of not travelling are not
fully identified by the other two models.
Town D loses access to all five other neighboring
towns when any form of flooding occurs. The impacts
of lack of access to Town D are ignored by CBA6 and
PERUB, which are focused on road users travelling

Fig. 5 Mud map of network considered in CBA6 and PERUB methodologies.
Diversion route
Original route

984

Fig. 6 Mud map applied to the example. Legend: Red linesQ2 Bridges/Culverts; Brown linesQ10 Bridges (mass limit
20 ton); Blue linesQ20 Bridges; Green linesQ50 Bridges.

Table 3 Results generated by models.
Results PERUB CBA6 Proposed
Total costs $3,692,585 $3,692,585 $3,692,585
Capital $5,769,231 $5,769,231 $5,769,231
Maintenance costs -$1,454,079 -$1,454,079 -$1,454,079
Residual value -$622,567 -$622,567 -$622,567
Total benefits $343,423 $760,222 $6,250,695
VOC savings $134,870 $51,991 $103,138
TTC savings $128,570 $59,967 $53,197
Accident cost savings -$15,477 -$2,145 $17,976
Flood related accident cost savings NA NA $365,255
Emission cost savings $7,902 NA NA
Environment cost savings $4,845 NA NA
Secondary cost savings $251 NA NA
Other cost savings $80,191 NA $33,035
Waiting time cost savings (diversion) $71 $650,409 $565,772
Not travelling cost savings (diversion) $2,201 NA $5,112,322
Indicators
NPV (net present value) -$3,349,162 -$2,932,363 $2,558,110
NPV (excluding NDRRA savings) -$4,803,241 -$4,386,442 $1,104,030
NBIR (net benefit investment ratio) 0.42 0.49 1.44
BCR (benefit cost ratio) 0.09 0.21 1.69
BCR (excluding NDRRA savings) 0.07 0.15 1.21

the entire route between CB. CDB also acts as a heavy
vehicle diversion route for heavy vehicle road users
travelling between CF during periods of no flooding.
When CDB is flooded, CAEF is the only viable
diversion route for the heavy vehicles; this route is an
additional 55 km. If the two bridges along CDB are
sufficiently upgraded, CDB can act as a diversion
route for road users utilizing other parts of the
network, which are only affected during major floods.
During minor and moderate floods, five trips are no
longer possible forcing road users not to travel.
During a major flood, nine trips are no longer possible.

985
After the two bridges are upgraded, all trips become
available to all vehicle types as CDB becomes a
possible diversion route for other flood-affected
areas.
The projected reduced NDRRA costs, which can be
included in either of the above three models, adds an
additional $1.45 million worth of benefits to the
projects. This value substantially improves the NPV
of all three models. In this example, a clear
discernable value of projected NDRRA costs has been
allocated to the project site. In reality, collection and
allocation of such costs can be difficult, and estimates
of these costs need to be made using historical data
from the region.
The proposed flooding model can be used to assess
the impact of upgrading different combinations of
bridges to achieve the optimal bridge upgrade
program for particular areas. Upgrading of the bridges
along section CD and DB have been evaluated
separately and compared with the option of upgrading
both bridges. Table 4 shows for the comparisons.
Upgrading of only one bridge greatly improves
access during minor and moderate floods but does not
provide many benefits during major floods, this is
especially the case for the upgrading of the bridge
along section CD. During heavy floods, Towns C and
D become isolated from the other towns. If the bridge
along Section DB is the only bridge upgraded, Town
C becomes completely isolated from the network.
Both bridges need to be upgraded for complete access
to the regional network. The NPV of upgrading both
bridges is greater than the sum of NPVs from both the
CD and DB bridge upgrades. Upgrading of only one
bridge also produces negative TTC, VOC, accident
and externality benefits, these negative benefits occur
because the upgrading of only one bridge provides
improved access to towns C and D through long
diversion routes rather than the original routes.
The proposed model has a number of limitations,
and the largest of these is the availability of data to
feed into the model. Currently, traffic volumes are
available per section of road but not per route. This
data can be made available through transport studies
but these studies are often costly to conduct. The

Table 4 Options analysis.
Results Both bridges DB bridge CD bridge
Total costs $3,692,585 $1,846,292 $1,846,292
Capital $5,769,231 $2,884,615 $2,884,615
Maintenance costs -$1,454,079 -$727,040 -$727,040
Residual value -$622,567 -$311,283 -$311,283
Total benefits $6,250,695 $2,800,296 $2,746,188
VOC savings $103,138 -$188,936 -$90,690
TTC savings $53,197 -$99,765 -$59,385
Accident cost savings $17,976 -$84,712 -$37,841
Flood related accident cost savings $365,255 $182,628 $182,628
Externality cost savings $33,035 -$68,092 -$25,659
Waiting time cost savings(diversion) $565,772 $304,284 $289,259
Not travelling cost savings (diversion) $5,112,322 $2,754,889 $2,487,877
Indicators
NPV $2,558,110 $954,003 $ 899,895
NBIR 1.44 1.33 1.31
BCR 1.69 1.52 1.49
Benefits according to flood type
Minor floods $1,780,819 $1,269,691 $1,424,025
Moderate floods $1,235,212 $621,674 $652,502
Major floods $3,234,664 $908,930 $669,661
Total $6,250,695 $2,800,296 $2,746,188

986

quantity of data required for the proposed model is
quite substantial compared to current models such as
CBA6. The collection of this data could be time
consuming and the results of the analysis will take
longer to obtain. The proposed model does not
generate values for travel time or vehicle operating
costs per km of travel, these values need to be
calculated prior to using the proposed model.
4. Conclusions
The costs to the road network and the costs to
society from flooding can be enormous. The solution
of improving flood immunity can greatly reduce these
costs if the right projects are introduced at the right
places. The current practiced methodology does not
sufficiently account for the possible benefits
obtainable from a large number of flood immunity
projects. The model proposed in this paper addresses
some of the issues surrounding the evaluation of
benefits of flood immunity projects. These issues
include identifying the impact of a project on the
immediate network, identifying the impact of flood
severity, valuing flood related accident costs,
incorporating NDRRA flooding costs, valuing
unanticipated flood costs, valuing costs of not
travelling and the determination of endogenous road
user behavior. As seen from the results presented in
Section 3.4, addressing the issues mentioned greatly
increases the quantifiable benefits of flood immunity
projects that relieve isolation to communities during
flood events. Incorporating generated traffic and wider
economic benefits would further enhance the
assessment of benefits of some flood immunity
projects and would be an appropriate next step
forward from this paper.
References
[1] Earth Science Australia, Flood Damage Assessment Case
StudyJ ohnstone River Flooding, Earth Science
Australia, http://earthsci.org/flood/j_flood04/flood/
(accessed Apr. 21, 2010).
[2] Geoscience Australia, Flood Information, Australian
Government, 2010, http://www.ga.gov.au/hazards/flood/
[3] TMR Infrastructure Program Performance Report,
Department of Transport and Main Roads, Queensland
Government, 2010.
[4] Roads Implementation Program Development 2009-10 to
2013-14, Department of Transport and Main Roads,
Queensland Government, 2010.
[5] ABC Far North Queensland, Flood Threat Eases on Qlds
Tablelands, 2011, http://www.abc.net.au/news/stories
[6] Flooding Locations, Flood Closures and Flood Immunity
Standards for the Bruce Highway (Brisbane to Cairns)
(Presentation), Department of Transport and Main Roads,
[7] National Regional Profile: Gympie (R) (Local
Government Area), Australian Bureau of Statistics, 2010,
http://www.abs.gov.au/AUSSTATS/ (accessed Apr. 4,
2011).
[8] Maryborough (Urban Centre/Locality), 2006 Census
QuickStats, Australian Bureau of Statistics, 2007,
http://www.censusdata.abs.gov.au (accessed Apr. 4,
2011).
[9] Bruce Highway, http://en.wikipedia.org/wiki/
Bruce_Highway (accessed Mar. 4, 2011).
[10] Charleville (Urban Centre/Locality), 2006 Census
QuickStats, Australian Bureau of Statistics, 2007,
http://www.censusdata.abs.gov.au (accessed Apr. 4,
2011).
[11] Flooding on Roads in Queensland, Department of
Transport and Main Roads, Queensland Government,
2010.
[12] N. Cernaianu, Risk management in road transport,
munich personal RePEc archive, Technical Paper,
University of Bucharest, Romania, 2010.
[13] Cost Benefit Analysis Manual, Department of Transport
and Main Roads, Queensland Government, Brisbane,
2011.
[14] Queensland Flood Summary 2000 onwards, Bureau of
Meteorology, Australian Government, 2011,
http://www.bom.gov.au/hydro/flood/qld/fld_history/flood
sum_2000 (accessed Mar. 30, 2011).
[15] Roads Implementation Program Development 2009-10 to
2013-14 Live, Department of Transport and Main Roads,
[16] L. Coates, Flood fatalities in Australia, 1788-1996, Aust.
Geogr. 30 (1999) 391-408.
[17] S.T. Ashley, W.S. Ashley, Flood fatalities in the United
States, J ournal of Applied Meteorology and Climatology
47 (2008) 805-818.
[18] N.J . J onkman, I. Kelman, An Analysis of the Causes and
Circumstances of Flood Disaster Deaths, Blackwell

987
Publishing, UK, 2005, pp. 75-97.
[19] Transport, Wider Economic Benefits and Impacts on
GDP, Department for Transport, 2006, London.
[20] Australian Emergency Management Terms Thesaurus,
Attorney-Generals Department, Australian Government,
2009, http://library.ema.gov.au (accessed May 30, 2010).
[21] National Guidelines for Transport Systems Management
in Australia, 4 Appraisal of initiatives, Commonwealth of
Australia, Australian Transport Council, Canberra, 2006.
[22] Economic Evaluation of Road Investment Proposals:
Improved Predictions Model for Road Crash Savings,
Austroads, AP-R184, 2001.
[23] 6401.0Consumer Price Index, Australian Bureau of
Statistics, Australia, J une 2010.
[24] Austroads Internal Report: Update of RUC Unit Values
to J une 2007, Austroads, IR-156/08, Austroads, 2008.
[25] ARI and AEP, Bureau of Meteorology, Australian
Government, 2010, http://www.bom.gov.au/hydro/has/
ari_aep.shtml (accessed Apr. 21, 2010).
[26] Economic Evaluation of Road Investment Proposals:
Harmonisation of Non-Urban Road User Cost Models,
Austroads, AP-R264/05, 2005.

Appendix A: Mathematical and diagrammatical derivation of flooding costs.
Mathematical derivations of formulae using Figs. A1-A4:
Formula: TDC (total diverting costs)
) (
Div Div Div
y Externalit Accident PCost TDC
24
) ( D AADT WT DC
E

Calculated using Fig. A2:
DE CD AB BC TDC ) (
where,
BC =Perceived costs;
AB =Accident and externality costs;
CD =Duration of closureequilibrium waiting time;
DE =Number of road users diverting per hour.

Fig. A1 Diversion costs.

988

Fig. A2 Waiting costs for road users who choose to travel.

Fig. A3 Not travelling costs when waiting is not a feasible option.

Fig. A4 Waiting and not travelling costs when diverting is not a feasible option.


989
Formula: total waiting costs:
+

=
24
) 1 (

2
1
NT AADT WT
PCost WC
E
Div
24
8
1
NT AADT WT
PCost
E
Div

Calculated using Figs. A2 and A4:
)
4
1
(
2
1
)
2
1
( OP NO MN FI FH GF WC + =

where,
GF =Perceived costs;
FH =Equilibrium waiting time;
FI =Number of road users diverting per hour;
MN =Perceived costs;
NO =Equilibrium waiting time;
OP =Number of road users not travelling per hour.
Formula: total NTC (not travelling costs):
+

=
24
) (

2
1
NT AADT WT Dc
PCost NTC
E
Div
24
8
1
NT AADT WT
PCost
E
Div

Calculated using Figs. A3 and A4:
)
4
1
(
2
1
)
2
1
( OP NO MN KX XL JK NTC + =

where,
JK =Perceived costs;
XL =Duration of closure - equilibrium waiting time;
KX =Number of road users not travelling per hour;
MN =Perceived costs;
NO =Equilibrium waiting time;
OP =Number of road users not travelling per hour.
Formula: FRAC (flood related accident costs):
Fatality per Cost AccRate AEP
Flooded Area Total
Considered Sections Flooded
FRAC =

Appendix B: Evaluation input data.
Table B1 Input data according to section.
Sections Pte cars Com cars Non-artic
17
Articulated B-doubles Pte cars Length MRS
18
AC 100 77 9 5 4 5 32 10
CD 105 79 8 7 10 6 40 10
DB 121 91 15 7 9 4 38 10
AE 84 65 7 4 6 2 92 8
EF 184 148 22 9 3 2 34 12
FB 171 140 20 5 4 2 25 12
CF 65 49 8 8 0 0 62 10
830 649 89 45 36 21 323 72

17
Non-articulated vehicles
18
MRS (Model Road State)

990
Table B2 Input data according to route.
Routes AADT Pte cars Com cars Non-artic Articulated B-doubles Length
19
Route MRS
20
ACDB 60 50 4 2 2 2 110 10
AEFB 0 0 0 0 0 0 151 10
CDB 10 5 1 2 2 0 78 10
CAEFB 0 0 0 0 0 0 183 10
DB 15 12 2 0 1 0 38 10
ACD 15 10 2 1 1 1 72 10
CD 20 14 1 2 2 1 40 10
AC 15 8 2 2 1 2 32 10
ACF 10 9 1 0 0 0 94 10
AEF 4 0 0 0 2 2 126 9
CF 30 22 4 4 0 0 62 10
CDBF 5 0 0 0 3 2 103 10
AE 80 65 7 4 4 0 92 8
EFB 120 108 10 2 0 0 59 12
EF 15 10 3 1 1 0 34 12
FB 10 8 2 0 0 0 25 12
CFE 25 18 3 4 0 0 96 11
CAE 3 0 0 0 1 2 124 9
DBFE 20 12 6 2 0 0 97 11
DBF 16 12 2 1 1 0 63 11
Total 473 363 50 27 21 12

Table B3 Project data.
Project data Base case Project case
Safe operating speed 100 km/h 100 km/h
Roughness 100 NRM 100 NRM
Curvature Straight Straight
Gradient Flat Flat
Discount rate 4% 4%
Capital cost per bridge - $1,500,000
Number of proposed bridges 0 2
Bridge standard Q2 Q50
AATOC (annual average time of closure) 108 0
ADC (average duration of closure) 170 0

Table B4 Flood data.
General flood data Base/project case
Minor flood costs (anticipated) $10,000
Minor flood costs (unanticipated) $20,000
Moderate flood costs (anticipated) $100,000
Moderate flood costs (unanticipated) $120,000
Major flood costs (anticipated) $5,000,000
Major flood costs (unanticipated) $6,000,000
Average duration of closure Light vehicles Heavy vehicles
Minor 120 180
Moderate 180 270

19
Route length is the sum of section lengths comprised in the route.
20
Route MRS is the weighted average of MRS of the sections comprised in the route.

991
(Table B4 continued)
General flood data Base/project case
Major 360 540
ARI (average recurrence interval)
Minor (anticipated) 2 years
Minor (unanticipated) 10 years
Moderate (anticipated) 20 years
Moderate (unanticipated) 20 years
Major (anticipated) 40 years
Major (unanticipated) 40 years
Maximum willingness to pay per km (anticipated flood) $3/km
Maximum willingness to pay per km (unanticipated flood) $4/km
Life of new bridge 100 years
Exogenous road user behavior applied to CBA6
Waiting 5%
Diverting 90%
Not travelling 5%


Policy Analysis on Building Regulations and the
Recovery of Earthquake and Tsunami Affected Areas
Shoichi Ando
Disaster Management Policy Program, National Graduate Institute for Policy Studies (GRIPS), Tokyo 106-8677, Japan

Abstract: The paper examines the situation and causes of recent big earthquake and tsunami disasters especially in the case of the
Great East J apan Earthquake in 2011, and analyzes the damage to extract lessons on safety of buildings and recovery of cities from
the view point of building regulations such as the Article 39 of the Building Standard Law. In addition, the Article 8 of the Ordinance
of City Planning Law resulted in not so effective against tsunami in March 2011. Control mechanisms of building construction
should be integrated into socio-economic, institutional, technical and other policy tools. In order to mitigate earthquake risk, all
stages of building construction, from location, planning and construction to maintenance are important. Awareness creation is
instrumental for building culture of safety and demands for intervention in disaster mitigation. The demands ultimately help in
creating conducive environment for policy intervention, in realizing institutional mechanism of building code enforcement and land
use control for the municipal authorities and in creating demand for competent professionals.

Key words: Great East J apan Earthquake, Building Standard Law, City Planning Law.

1. Introduction

The impact of earthquake can be reduced by policy
measures such as earthquake resistant building design
and construction standards, proper planning,
education and trainings. The risk increasing as
urbanization is adding extra pressures on building
construction. In the case of tsunami disaster, the
BSL (Building Standard Law) of J apan allows local
governments to create their own regulation based on
its Article 39. However the regulation has not been
utilized so often against tsunami. Moreover, the
Article 8 of ordinance of the CPL (City Planning Law)
resulted in not so effective in tsunami affected areas of
the Great East J apan Earthquake on March 11, 2011
except some areas in Sendai Plain.
The paper examines the outline situation and causes
of recent large scale earthquakes and tsunamis and
analyzes the responses and reconstruction processes.
Then, it shows how responded to the disaster, and
how can people avoid the same disasters in the future

Corresponding author: Shoichi Ando, Dr., professor,
research fields: disaster management and sustainable building
policy. E-mail: ando@grips.ac.jp.
based on the valuable lessons from the Great
Hanshin-Awaji (Kobe) earthquake in 1995, Sumatra
Tsunami in 2004, Pakistan earthquake in 2005, J ava
earthquake in 2006, Peru earthquake in 2007, Sichuan
earthquake in 2008, Haiti earthquake in 2010 and the
Great East J apan Earthquake. Since many large scale
disasters have happened during these ten years, this
paper tries to comprehensively summarize the reasons
of damages and what are the lessons in terms of
earthquake and tsunami safety of buildings and cities.
The demands ultimately help in creating conducive
environment for policy intervention, in realizing
institutional mechanism to mitigate disasters for the
municipal authorities and creating demand for
competent professionals. In J apan, demands for
seismic assessment and retrofit of houses that follow
old seismic codes before 1981, could be realized
mainly by raising awareness.
2. Damage Analysis of the Great East Japan
Earthquake
Firstly, the damage of the Great East J apan
Earthquake on March 11, 2011 is analyzed in this
DAVID PUBLISHING
D
Policy Analysis on Building Regulations and the Recovery of Earthquake and Tsunami Affected Areas

993

Table 1 Earthquake and Tsunami Disasters with over 15,000 casualties (1960-2011).
No. Country: Earthquake Year Mg Death (A) Collapse (B)
Casualty ratio
(A/B)%
1 China: Hebei (Tangshan Earthquake) 1976 7.8 242,800
2 Indian Ocean Tsunami 2004 9.0 226,408 470,000 48.1
3 Haiti Earthquake 2010 7.0 222,576 (189273) (117.6)
4 China: Sichuan (Wenchuan Earthquake) 2008 8.1 87,576 5,461,900 1.6
5 Pakistan Kashmir Earthquake 2005 7.6 73,328 272,000 26.9
6 Peru: Chimbote, Huaras 1970 7.8 66,794 (>15,000)
7 Iran: Manjil Earthquake, Rudbar 1990 7.7 35,000
8 Iran: Kerman, Bam Earthquake 2003 6.7 31,830 55,000 57.8
9 Armenia: Spitak Earthquake 1988 6.8 25,000
10 Guatemala Earthquake 1976 7.5 22,870
11 India: Bhuj Earthquake (Gujarat) 2001 8.0 20,023 (339,000) (5.9)
12 J apan: Great East J apan Earthquake 2011 9.0 19,295 127,185 15.2
13 Iran: Tabas Earthquake 1978 7.4 18,220
14 Turkey: Kocaeli Earthquake 1999 7.8 17,118 (60,000) (28.5)
15 China: Yunnan Earthquake 1970 7.8 15,621
16 J apan: Great Hanshin-Awaji Earthquake 1995 7.3 6,434 111,123 5.7
Source: the BRIs World Seismic Disaster Catalog, and United Nations; Events in No. 2-5, 8, 11 and 12 represent the earthquake
disasters that occurred in 2001-2011; Part of collapse data [1-3]; (): not confirmed data.

section to verify the current policy measures of
building regulation against tsunami disasters. Table 1
shows ratio of casualties per collapsed houses (A/B)
during recent huge earthquakes. Fig. 1 shows ratio of
casualties per population in the inundated area by
affected municipality including dead and unknown
persons as of Dec. 30, 2011 reported by the National
Police Agency. Blue columns in Fig. 1 indicate
Sanriku area where main geographical condition is
rias coast that was suffered from severe damages,
while green columns imply plain areas:
(1) The maximum ratio of human damage
(casualties including death and unknown) per
population of the inundated areas in a municipality is
recorded as 12% in Onagawa town. Otsuchi town and
Rikuzen Tkata city claimed the following large ratio;
(2) Since no damage by tsunami can be observed
outside of inundated areas, Figs. 1 and 2 represent
human damage and physical damage per inundated
area, respectively. The density of human and physical
damages in Otsuchi town (Fig. 3) and Onagawa town
(Fig. 4) are the severest. Kesennuma city and
Kamaishi city follow them as the areas of massive
damages both in human and physical aspects;

0
2
4
6
8
10
12
14
O
n
a
g
a
w
a
t
o
w
n
R
i
k
u
z
e
n
T
a
k
a
t
a
c
i
t
y
O
t
s
u
c
h
i
t
o
w
n
K
a
m
a
i
s
h
i
c
i
t
y
N
a
t
o
r
i
c
i
t
y
Y
a
m
a
m
o
t
o
t
o
w
n
Y
a
m
a
d
a
t
o
w
n
M
i
n
a
m
i
S
a
n
r
i
k
u
N
a
m
i
e
t
o
w
n
M
i
n
a
m
i
S
o
m
a
c
i
t
y
S
o
m
a
c
i
t
y
I
s
h
i
n
o
m
a
k
i
c
i
t
y
K
e
s
e
n
n
u
m
a
c
i
t
y
H
i
g
a
s
h
i
M
a
t
s
u
s
h
i
m
a
M
i
y
a
k
o
c
i
t
y
S
e
n
d
a
i
c
i
t
y
S
h
i
n
c
h
i
t
o
w
n
O
f
u
n
a
t
o
c
i
t
y
I
w
a
n
u
m
a
c
i
t
y
CasualtiesperPopulationofinundated
area(municipalitiesmorethan100casualties,2%)
Sanriku area(Rias)
South ofSendai(Plain)

Fig. 1 Ratio of casualties by municipality.

0
200
400
600
800
1000
1200
O
n
a
g
a
w
a
t
o
w
n
O
t
s
u
c
h
i
t
o
w
n
Y
a
m
a
d
a
t
o
w
n
S
e
n
d
a
i
c
i
t
y
I
w
a
k
i
c
i
t
y
K
e
s
e
n
n
u
m
a
c
i
t
y
O
f
u
n
a
t
o
c
i
t
y
K
a
m
a
i
s
h
i
c
i
t
y
M
i
y
a
k
o
c
i
t
y
M
i
n
a
m
i
S
a
n
r
i
k
u
I
s
h
i
n
o
m
a
k
i
c
i
t
y
T
a
g
a
j
o
c
i
t
y
R
i
k
u
z
e
n
T
a
k
a
t
a
T
a
n
o
h
a
t
a
v
i
l
l
a
g
e
T
a
k
a
h
a
g
i
c
i
t
y
TotallyCollapsedHousesper
InundatedArea(unit/km2)morethan200
Sanriku area(Rias)

Fig. 2 Totally collapsed houses per inundation area.

1,200
1,000
800
600
400
200
0

Totally collapsed house per inundated area
(unit/km
2
) more than 200

Casualties per population of inundated area
(municipalities more than 100 casualties, 2%)

994

Fig. 3 Otsuchi town, Iwate prefecture (Apr. 2011).

Fig. 4 Onagawa town, Miyagi prefecture (Apr. 2011).

0
100
200
300
400
500
600
0 200 400 600 800 1000
U
n
k
n
o
w
n
R
a
t
i
o
(
U
n
k
n
o
w
n
p
e
r
1
0
0
0
c
a
s
u
a
l
t
i
e
s
)
TotallyCollapsedRatio(Totallycollapsedper1000damagedunits)
CharacteristicsoftheGreatEastJapanEq.(1)
Sanriku
Sendaiplain
Fukushima
Ibaraki
Tanohata
TOtsuchi
Onagawa
MinamiSanriku
Ofunato
RikuzenTakata
Namie
Sendai
Ishinomaki
Kesennuma
Kamaishi

Fig. 5 Missing ratio and totally collapsed ratio.

0
200
400
600
800
1000
1200
1400
1600
1800
2000
0 2000 4000 6000 8000 10000
N
u
m
b
e
r
o
f
C
a
s
u
a
l
t
i
e
s
(
p
e
r
s
o
n
)
NumberofTotallyCollapsedHouses(unit)exceptSendai,Ishinomaki
CharacteristicsoftheGreatEastJapanEq.(2)
Sanriku
Sendaiplain
Fukushima
Ibaraki
Onagawa
Otsuchi
MinamiSanriku
Kesennuma
Ofunato
RikuzenTakata
Kamaishi
HigashiMatsushima
Natori
Yamamoto
MinamiSoma
Miyako
Soma
Iwaki
Watari
Tagajo

Fig. 6 Totally collapsed houses and casualties.

(3) The gravity of physical damage can be measured by the totally collapsed ratio. The density of
Characteristics of the great east J apan earthquake (2)
0 2,000 4,000 6,000 8,000 10,000
2,000
1,800
1,600
1,400
1,200
1,000
800
600
400
200
0
Number of totally collapsed houses (unit) except Sendai, Ishinomaki
0 200 400 600 800 1,000
Totally collapsed ratio (totally collapsed per 1,000 damaged units)
Characteristics of the great east J apan earthquake (1)
M
i
s
s
i
n
g

r
a
t
i
o

(
u
n
k
n
o
w
n

p
e
r

1
,
0
0
0

c
a
s
u
a
l
t
i
e
s
)


995
houses and population in the inundation area of these
cities was lower than in Otsuchi and Onagawa town.
If the ratio of unknown per human damage will
represent severity of human damage, Onagawa town
reached 39% as the highest ratio and Otsuchi town,
Minami Sanriku town follow high ratio. They are the
municipalities that are ranked in Fig. 2 as the heavily
damaged areas;
(4) Figs. 5 and 6 show the severity of damage by
municipality, classifying the characteristics of regions.
The indicator of Fig. 6 may represent characteristics
of damage and will help to compare with other
disasters. The proposed indicator is calculated as
number of human damage (as a sum of death and
unknown) per 100 totally collapsed houses in each
affected municipality. Coburn, Spence and Pomonis
defined similar ratio as Lethality Ratio in 1992. The
following data is formed after extracting
municipalities that have larger fluctuation because of
its smaller numbers;
(5) The ratio varies almost 100 times under this
indicator. Rikuzen-Takata city recorded more than 50
persons human damage per 100 totally collapsed
houses, while Sendai citys indicator shows around 2
persons. Fig. 6 tries to classify the damages however
there was not so clear characteristics from above
mentioned data. Analysis on other big disasters in the
world is conducted utilizing the same indicator
(casualty ratio: number of casualties per 100 totally
collapsed houses) in the latter section.
3. Damages and Recovery of Recent Big
Earthquake Disasters
(Basic data of each disaster are shown in the Table 1.)
3.1 Indian Ocean Tsunami (2004)
Numerous people including tourists from Europe
and North America were drowned and died by Indian
Ocean Tsunami. The tsunami caused by the Sumatra
Earthquake on December 26, 2004, killed in total
approx. 230,000 people in 12 countries. One of the
reasons why the tsunami brought so many victims lies
in the tourists from the western countries at Christmas
vacation who did not notice the danger of tsunami.
This tsunami provided an opportunity for the western
countries to acknowledge natural disasters. WCDR
(World Conference on Disaster Reduction) that was
held in Kobe in J anuary of 2005, only less than a
month after the Indian Ocean Tsunami, proved such
global awareness. In total, 158 countries have
participated in the WCDR, and the HFA (Hyogo
Framework for Action) 2005-2015 was adopted.
From experiences in Aceh (Fig. 7) after the 2004
tsunami, many lessons can be learned. There are huge
amount of construction sites of new houses in the
areas. However, because of so rapid reconstruction of
public facilities and houses, some houses were
constructed with lower earthquake resistance and
lower quality than before. Some completed recovery
houses were still vacant in some residential zones as
no infrastructure was provided as of Dec. 2007.

Fig. 7 Aceh, Indonesia in 2005 (by UNCRD).

Fig. 8 Barakot, Pakistan (Mar. 2006).

996
3.2 Pakistan Earthquake (2005)
Kashmir (Pakistan) earthquake occurred on October
8, 2005 at 9 a.m.. Most of victims were killed by
collapse of houses, schools and other buildings.
Barakot is the most heavily damaged city by the
Pakistan Earthquake that occurred on Oct. 8, 2005.
More than 70% of buildings in the city were collapsed.
This Kashmir area has been suffered from boundary
conflict with India, and it snows in winter as located
at the foot of Himalayan Mountains with 7,000 m
above sea level.
Almost 20,000 children have lost their lives
because of the collapsed school building during
school time. The seismic damage has concentrated on
masonry houses. Buildings that have been constructed
with basic rules such as using high quality materials
and structural components are firmly connected and
maintained properly, no damages were observed even
in the heavily affected area (Fig. 8).
3.3 Java Earthquake (2006)
Java earthquake claimed more than 5,000 casualties
on May 27, 2006 early in the morning. More than
200,000 houses were destroyed, and crushed the
residents. It is said that most of houses in Indonesia
were constructed with timber a half century ago. After
the timber became an international commodity,
Indonesian people could not use timber as the price
has increased at an international level, and people
became to use low-cost and low quality bricks that are
burnt with rice hulls. This issue resulted in a cause of
heavier seismic damages in Indonesia. Therefore, Mr.
Teddy Boen who graduated from IISEE in 1963 as the
third courses participant and other structural experts
in Indonesia are disseminating localized earthquake
resistant construction methods (Fig. 9:
semi-permanent recovery house).
3.4 Wenchuan Earthquake (2008)
Wenchuan Earthquake occurred on May 12, 2008
in the afternoon. Around 90,000 people were killed
mainly in Sichuan Province of China. Many of them
claimed their lives by being pressed with collapsed
buildings particularly their own houses (Fig. 10).
The affected area extended from cities that have
hundreds of thousand populations to villages in the
mountain regions more than 500 km long range of
areas. It was just before the Beijing Olympic Game,
the Chinese government quickly responded, such as
the rapid provision of temporary houses. It was ten
times larger in number of units within the same period
compared with the case of the 1995 Great
Hanshin-Awaji Earthquake. The reconstruction of
houses finished within two years although the original
recovery plan needed three years.
One of the factors why the recovery has been
completed so fast is Partner Assistance system that
has contributed to recovery of a heavily affected city

Fig. 9 Java, Indonesia (Sept. 2006) reconstruction.

Fig. 10 A village in Sichuan, China (June 2008).

997
through the support by a designated province in China.
This system was applied from the initial stage to
whole reconstruction and proved effective, partly
because of competitive atmosphere of the system. The
traumatic care of victims and recovery of cultural
heritages are included. Retrofitting of vulnerable
buildings remains a major issue as a lesson.
4. Domestic and International Comparison
of the Damages
Figs. 11 and 12 show the comparison of disasters in
different areas using a proposed indicator. From the
indicator that sets forth number of casualties per 100
totally collapsed houses, the range of numbers varies
from 60 to 2 or 1 in both figures that Fig. 11 shows
difference among municipalities in the Great East
J apan Earthquake affected areas and Fig. 12 indicates
difference of recent huge disasters (index of disaster
severity by number of casualties per collapsed houses
[4]).
The following observation can be pointed out from
above-mentioned figures and field visits:
(1) Number of Fig. 11 represents the inclination
of Fig. 6. Although there may exist slight difference
of judgment and definition of a totally collapsed
house among municipalities, large difference of
value can not be explained. From the data in Fig. 13, it
is not so clear the difference of each municipality. The
difference of prefectures may be analyzed (although
the highest casualties in Iwate prefecture in the Fig. 12
can not be explained from the Fig. 13 that indicates
the awareness of risk in Iwate was the highest);
(2) Fig. 11 shows lower casualty (mortality) ratio
in Ofunato city even though it is located in Sanriku.
The reasons why the ratio in Soma city, Natori city
and some other plain areas resulted in higher ratio
compared with average of Iwate or Sanriku region
may also provide social or historical reasons of the
damages. Since this paper focuses on institutional
aspect, further research is executed;
(3) Fig. 12 indicates also the wide range of
difference of casualty ratio in the recent huge disasters.
In general, tsunami disasters claimed rather higher
ratio than the other cases. One of the reasons may be
the frequency of a disaster (Huge tsunamis occur
every thousand years while huge earthquakes can be
experienced once per several hundred years, as it is
said disaster comes when it is forgotten);
(4) Fig. 12 also suggests that the large difference of
casualty ratio between the higher cases such as Iran
and Pakistan (and in the case of Haiti, the casualty
ratio became over 100 persons by official statistics)

0
10
20
30
40
50
60
70
R
i
k
u
z
e
n
T
a
k
a
t
a
c
i
t
y
O
t
s
u
c
h
i
t
o
w
n
S
o
m
a
c
i
t
y
K
a
m
a
i
s
h
i
c
i
t
y
N
a
t
o
r
i
c
i
t
y
O
n
a
g
a
w
a
t
o
w
n
Y
a
m
a
m
o
t
o
t
o
w
n
N
a
m
i
e
t
o
w
n
Y
a
m
a
d
a
t
o
w
n
M
i
n
a
m
i
S
a
n
r
i
k
u
t
o
w
n
I
w
a
n
u
m
a
c
i
t
y
H
i
g
a
s
h
i
M
a
t
s
u
s
h
i
m
a
S
h
i
n
c
h
i
t
o
w
n
I
s
h
i
n
o
m
a
k
i
c
i
t
y
K
e
s
e
n
n
u
m
a
c
i
t
y
M
i
y
a
k
o
c
i
t
y
M
i
n
a
m
i
S
o
m
a
c
i
t
y
O
f
u
n
a
t
o
c
i
t
y
T
a
g
a
j
o
c
i
t
y
W
a
t
a
r
i
t
o
w
n
I
w
a
k
i
c
i
t
y
S
e
n
d
a
i
c
i
t
y
Casualtiesper100TotallyCollapsedHouses
(Allmunicipalitiesmorethan100casualties)
Sanriku area(Rias)

Fig. 11 Number of casualties per 100 totally collapsed houses in 2011 in Japan.
Casualties per 100 totally collapsed houses
(All municipalities more than 100 casualties)

998

Fig. 12 Number of casualties per 100 totally collapsed houses of the world recently.

Fig. 13 Reasons (chance) of evacuation on March 11, 2011.

Fig. 14 Suburb of Pisco, Peru (Aug. 2007).

Fig. 15 Bam, Iran after 2003, reconstruction.
Number of casualties per
100 totally collapsed houses

999
and lower cases in China 2008, Peru 2007 and Java
2006. The case of Kobe in 1995 set forth the middle
level of casualty ratio. The following (5) and (6) are
the tentative hypotheses based on the current data;
(5) Figs. 8-10 and Fig. 14 that were taken one week
or a few months after each disaster show the typical
damage of houses. Fig. 9 of Java, Indonesian case
shows a reconstructed house. The collapsed wall was
seen at the back right side. That was the former house
with very thin brick wall. Not only in Indonesia but
also in China (Fig. 10) and Peru (Fig. 14) people
constructs one story house with light roof materials
and thin wall, especially in rural regions. The local
residents use traditional construction systems;
(6) Fig. 15 in Iran and Fig. 8 in Pakistan indicate
modern construction methods. In case of Iran, people
uses steel frame brick infill structures. As the damage
of pure (dried) brick houses was severer than the steel
frame type, many reconstructed buildings adopted the
steel frame structure. In the case of Pakistan in 2005,
reinforced concrete buildings like schools, hospitals
and apartment houses were collapsed and caused
many casualties. It means that even modern structure,
it causes severe human damage if the structure was
not properly designed and constructed.
Based on the data on damages of the Great East
Japan Earthquake from the view point of building
control and urban planning, i.e., Figs. 16-19, further
observations can be pointed out as follows [5]:
(1) Fig. 16 shows two exceptionally large damaged
cities in terms of physical damage. Both Sendai city
and Ishinomaki city are classified as the area of Urban
planning with Area Division. That means the
pressure of development and increase of population
are expected in these cities. Therefore, in order to
effectively invest resources into designated UPA
(urbanization promotion area) without investing into
UCA (urbanization control area) (almost ten times
difference of casualty ratio between two cities) [6];
(2) Fig. 17 clearly indicates the characteristics of
urban planning with area division (red columns).
Accumulated houses existed in the area were damaged.
The damaged houses include collapsed houses, half
collapsed houses and partially damaged houses.
However, top six municipalities in the Fig. 17 can not
be found in the Fig. 18. That means in urbanization
promotion areas especially in the Sendai plain,
housing damage in tsunami inundation areas turned
out large number, while human damage was not so
severe if compared with Sanriku where there is no
UPA except Onagawa town;
(3) Fig. 18 shows the same data as Fig. 12 with
classification by urban planning type. As same as the
Fig. 16, all municipalities in Fig. 18 established urban
planning. That means heavily damaged areas on
houses (Fig. 16) and human (Fig. 18) were basically

0
5000
10000
15000
20000
25000
30000
S
e
n
d
a
i
c
i
t
y
I
s
h
i
n
o
m
a
k
i
c
i
t
y
K
e
s
e
n
n
u
m
a
c
i
t
y
I
w
a
k
i
c
i
t
y
H
i
g
a
s
h
i
M
a
t
s
u
s
h
i
m
a
M
i
n
a
m
i
S
o
m
a
M
i
y
a
k
o
c
i
t
y
O
f
u
n
a
t
o
c
i
t
y
R
i
k
u
z
e
n
T
a
k
a
t
a
c
i
t
y
M
i
n
a
m
i
S
a
n
r
i
k
u
O
t
s
u
c
h
i
t
o
w
n
K
a
m
a
i
s
h
i
c
i
t
y
O
n
a
g
a
w
a
t
o
w
n
N
a
t
o
r
i
c
i
t
y
Y
a
m
a
d
a
t
o
w
n
W
a
t
a
r
i
t
o
w
n
Y
a
m
a
m
o
t
o
t
o
w
n
T
a
g
a
j
o
c
i
t
y
S
o
m
a
c
i
t
y
TotallyCollapsedHouses
(municipalitiesmorethan1000units)
UrbanPlanwithAreaDivision
UPA withoutAreaDivision

Fig. 16 Totally collapsed houses by municipality.
30,000
25,000
20,000
15,000
10,000
5,000
0
Totally collapsed houses
(Municipalities more than 1,000 units)
30,000
25,000
20,000
15,000
10,000
5,000
0

1000
0
1000
2000
3000
4000
5000
6000
I
w
a
k
i
c
i
t
y
R
i
f
u
t
o
w
n
S
e
n
d
a
i
c
i
t
y
S
h
i
o
g
a
m
a
c
i
t
y
T
a
g
a
j
o
c
i
t
y
M
a
t
s
u
s
h
i
m
a
O
n
a
g
a
w
a
t
o
w
n
O
o
t
s
u
c
h
i
t
o
w
n
K
e
s
e
n
n
u
m
a
c
i
t
y
I
s
h
i
n
o
m
a
k
i
c
i
t
y
Y
a
m
a
d
a
t
o
w
n
K
a
m
a
i
s
h
i
c
i
t
y
N
a
t
o
r
i
c
i
t
y
M
i
y
a
k
o
c
i
t
y
O
f
u
n
a
t
o
c
i
t
y
M
i
n
a
m
i
S
a
n
r
i
k
u
S
h
i
c
h
i
g
a
h
a
m
a
H
i
g
a
s
h
i
M
a
t
s
u
s
h
i
m
a
T
a
n
o
h
a
t
a
v
i
l
l
a
g
e
R
i
k
u
z
e
n
T
a
k
a
t
a
N
o
d
a
v
i
l
l
a
g
e
M
i
n
a
m
i
S
o
u
m
a
I
w
a
i
z
u
m
i
t
o
w
n
Y
a
m
a
m
o
t
o
t
o
w
n
I
w
a
n
u
m
a
c
i
t
y
W
a
t
a
r
i
t
o
w
n
S
o
u
m
a
c
i
t
y
S
h
i
n
c
h
i
t
o
w
n
N
a
m
i
e
t
o
w
n
K
u
j
i
c
i
t
y
F
u
t
a
b
a
t
o
w
n
O
o
k
u
m
a
t
o
w
n
N
a
r
a
h
a
t
o
w
n
DamagedHousesperInundationArea(unit/km2)
UrbanPlanningwith Area Division
Land UseDistrictswithoutAreaDivision
UrbanPlanningwithout LandUseDistrict
OutsideofUrbanPlanningArea

Fig. 17 Damaged houses per Inundation area.

0
10
20
30
40
50
60
70
R
i
k
u
z
e
n
T
a
k
a
t
a
c
i
t
y
O
t
s
u
c
h
i
t
o
w
n
S
o
m
a
c
i
t
y
K
a
m
a
i
s
h
i
c
i
t
y
N
a
t
o
r
i
c
i
t
y
O
n
a
g
a
w
a
t
o
w
n
Y
a
m
a
m
o
t
o
t
o
w
n
N
a
m
i
e
t
o
w
n
Y
a
m
a
d
a
t
o
w
n
M
i
n
a
m
i
S
a
n
r
i
k
u
t
o
w
n
I
w
a
n
u
m
a
c
i
t
y
H
i
g
a
s
h
i
M
a
t
s
u
s
h
i
m
a
S
h
i
n
c
h
i
t
o
w
n
I
s
h
i
n
o
m
a
k
i
c
i
t
y
K
e
s
e
n
n
u
m
a
c
i
t
y
M
i
y
a
k
o
c
i
t
y
M
i
n
a
m
i
S
o
m
a
c
i
t
y
O
f
u
n
a
t
o
c
i
t
y
T
a
g
a
j
o
c
i
t
y
W
a
t
a
r
i
t
o
w
n
I
w
a
k
i
c
i
t
y
S
e
n
d
a
i
c
i
t
y
Casualtiesper100TotallyCollapsedHouses
(Allmunicipalitiesmorethan100casualties)
UrbanPlan(AreaDivision)
UPAwithoutAreaDivision

Fig. 18 Casualties per 100 totally collasepd houses.

Fig. 19 Totally collpased houses/casualties.
0
200
400
600
800
1,000
1,200
1,400
1,600
1,800
2,000
0 2,000 4,000 6,000 8,000 10,000
Number of totally collapsed houses (unit) except Sendai and Ishinomaki
Characteristics of the Great East Japan Earthquake (2-2)
UPADivision
noDivision
nolanduse
outofUPA
Rikuzen Rikuzen
Kesennuma
Higashi Matsushima
Otsuchi
Kamaishi
Natori
Onagawa
Minami
Sanriku
Minami
Soma
Iwaki
Ofunato
Miyako
Yamamoto
Soma
Tagajo
Watari
Damaged houses per inundation area (unit/km
2
)
6,000
5,000
4,000
3,000
2,000
1,000
0
Casualties per 100 totally collapsed houses
(All municipalities more than 100 casualties)

1001

controlled under urban planning system with rather
strict building control systems;
(4) In Fig. 19, not so specific difference between
urban planning with area division and without area
division can be observed.
5. Japanese Building Regulations against
Disasters
Area division system and building control
systems that are prescribed in the CPL (City Planning
Law) and the BSL (Building Standard Law) are
expected to play significant roles to prevent tsunami
and earthquake disasters. However, before the Great
East Japan Earthquake, a few cases of DHA (Disaster
Hazard Area) under the Article 39 of BSL were
applied to prevent tsunami disaster. The reason why
DHA has not been so popular in the case of tsunami
can be explained as follows:
(1) DHA aims to prevent disasters utilizing locally
applicable control codes through designation of the
area. There exist approximately 17,800 DHA in Japan
(2007 MLIT), however most of them were designated
against land slide to restrict housing construction in
the steep slope areas. DHA against tsunami risk was
not established except a few cases as the frequency of
occurrence is quite rare and residents do not agree to
prohibit from building their houses. There is no
national financial support;
(2) As shown in the Table 2, DHA provides
permanent restriction while other building control
system in the disaster affected area like the building
control based on the Article 84 of BSL, sets normally
two months limitation or in the case of the Great East
Japan Earthquake maximum eight months control.
DHA controls will not be necessary for the area
without any development pressure;
(3) DHA was sometimes used in the recovery
projects after damaged disasters. In the case of Aonae
area of Okushiri town after a big tsunami of the off
coast of south-west of Hokkaido earthquake in 1993,
DHA was introduced to the high risk area in the old
residential zone after the new hilly safe area was
developed utilizing the Collective Removal project
against Disasters with subsidies from national
government (by MOC (Ministry of Construction),
current Ministry of Land, Infrastructure and
Transport). This was the unique case after tsunami
under DHA;
(4) As shown in the Appendix, Iwate prefecture
requested all affected municipalities to set DHA to the
heavily tsunami affected area in April 2011. However
Kamaishi city decided not to use DHA in July 2011
and other municipalities are also reluctant to apply
DHA. On the contrary, Miyagi prefecture set building
control in large areas using the Article 84 of BSL.
Sendai city and Yamamoto town utilize DHA to
control building construction in tsunami hazardous
areas.
As shown in the Table 2, the basic direction toward
reconstruction of Miyagi prefecture and Iwate
prefecture seems to select different way as the case of
building restriction in early stage. It seems that

Table 2 Comparison of building control based on the BSL Article 39 and Article 84.
Building Standard Law Article 39 (DHA: Disaster Hazard Area) Article 84 (Control in Affected Area)
Designation of area Based on bylaw of local governments By Specialized Admin. Authority *1
Duration of control Permanent measures Max. two months *2
Construction control
Prohibit housing, limit other building (no national
intervention)
Prohibit/limit building construction in the project
planned area
Application to great east
Japan earthquake
Iwate: Urge municipality to set bylaw
Miyagi: Pref. started to plan to apply
Iwate pref.: No application
Miyagi: Applied to 5 municipalities
Response of
municipalities
Iwate: Mayors are prudent (negative)
Miyagi: Part of Minami-sanriku town
Miyagi: Enterprises were embarrassed then try to
permit some construction
Applied cases Hokkaido, Okushiri town, Aonae area Great Hanshin-Awaji earthquake (Kobe etc.)
*1 Specialized Administrative Authority: Mayors that put building officer, Governor of prefecture in other cases;
*2 Based on a special law, max. six months additional extension was possible this time, Miyagi extended till Nov. 11.

1002

Miyagi prefecture aims to improve urban structure using
this opportunity especially in the coastal zones, while
Iwate prefecture seems to be struggling to maintain
population in the tsunami affected areas and then
restriction of building construction in Iwate prefecture
is not so strict compared to Miyagi prefecture because
the population decrease trend is expected severer in
the remote regions from big cities. However, it may
be caused simply because of the difference of urban
planning settings of both prefectures, i.e., Miyagi
prefecture sets Area Division and most of coastal
areas are prohibited to construct buildings while there
is no Area Division (UPA and UPC) and construction
of buildings is not so strictly controlled in the coastal
cities and towns in Iwate prefecture.
6. Recovery Policy Measures for Tsunami
Affected Areas in Japan
Total 40 types of project systems are designated as
the principal project in the Law on the Special Area
for Reconstruction of the Great East Japan Earthquake
that was established in December 2011. Major project
systems are Collective Removal Project against
Disasters, Recovery Base Project against Tsunami
(a new system based on the new Law on Tsunami
Disaster Management Regional Development),
Land Readjustment Project for Urban Renewal,
Public Operated Housing for Disaster Victims and
Improvement Project of Small-Scale Residential
Area. Characteristics are summarized in Table 3.
The application of 40 project types will be selected
by the local government within this fiscal year (by
March 2012) and those reconstruction projects will be
executed within three-five years utilizing mainly
national supplementary budget that will be provided
to the local governments of affected areas soon.
The author tries to analyze execution process of
reconstruction projects from urban planning view point.
As for December 30, 2011 national government has
announced these systems and budgetary framework to
the local governments. Now local governments are
discussing with community on the concrete projects
for reconstruction based on the basic plans and future
prospect in each community that have been basically
created in 2011.
The basic issue of these areas is the trend of
decreasing population that will be heavily influenced
to decide future urban plan of reconstruction areas.
Table 4 shows estimated population based on a
research study by Hayashi and Saito in 2011. While

Table 3 Characteristics of major basic project systems to be applied to reconstruction (in 2011).
Basic
projects
Collective Removal
project against
Disasters
Recovery Base project
against Tsunami (new)
Project on Land
Readjustment for Urban
Recovery
Public Operated
Housing after
Disasters
Improvement of
Small-Scale
Residential Area
Subsidies
Cost for public
works incl. land
development except
sell land
Total mounding cost,
development of evacuation
building and public works
etc.
Cost for public works
incl. land, totally
mounding (40
persons/ha)
Cost for public
houses for low
income people, aid
for the aged
Public works, clear
vulnerable units,
community facilities
etc.
Area
No relation to urban
planning
Principally within UPA
(urban planning area)
Within UPA (urban
planning area)
No relation to urban planning
Scale
More than five
(usual 10) houses
Principally 2 projects per
urban, and approx. 20 ha per
project
No condition No condition
More than 15 (5)
units, with 50%
vulnerable units
Condition
Designation of
disaster hazard area
is requisite
Define area for land
purchase, step by step
extension will be possible
Consolidated area to
develop road system,
division of project area
No condition
Consolidated area is
preferable
Process
Agreement of MLIT
minister on removal
plan
Planning decision as urban
facility, project approval of
prefecture (or MLIT)
Urban planning
procedures are needed
(from planning decision
to liquidation)
No need for urban
planning permit,
need financial MLIT
agreement
No need for urban
planning permit, need
financial MLIT
agreement
Aid ratio All costs will be covered (special case by national grant + special tax)
Depend on the tax condition (local debt will
be reimbursed by rent)

1003
Table 4 Comparison of population and urban reconstruction of Miyagi and Iwate prefectures.
Miyagi prefecture Iwate prefecture
Total population 2,360,218 persons (2005) 1,385,041 persons (2005)
Estimated pop. 1,894,000 persons (2040) [7] 962,000 persons (2040) [7]
Ratio (2040/2005) -19.8% (affected area46.8%) -30.5% (affected area48.8%)
Aged ratio (05-40) 20.0% (2005) 34.3% (2040) 24.6% (2005) 38.0% (2040)
Basic concept for
Reconstruction (part
of land use and
development)
Miyagi Pref. Recovery Plan [8]:
Recovery focusing on tsunami disaster management
of coastal areas applying removal to high land,
separation of work and home, multiple protection
against tsunami from the lessons
Iwate Pref. Recovery Basic Plan [9]:
Based on agreements with residents, improvement of
residential area for safety and development connected
with land use plan considering tsunami disaster
management
Current situation
(building control)
Pref. set building control based on City Planning Law
etc. after BSL Article 84.
Pref. has recommended municipalities to use BSL
Article 39, but no execution

population of Miyagi prefecture will decrease less
than 20%, total population of Iwate prefecture will
decrease more than 30% by 2040 from 2005 within
these 35 years. In addition, decrease of regional
population is estimated almost 50% in the tsunami
affected areas both in Miyagi prefecture and Iwate
prefecture. The reconstruction plan will be influenced
by such regional population prospect as shown in the
Table 4.
7. Limits of Japanese Building Regulations
against Tsunami
Both of the BSL (Building Standard Law) and the
CPL (City Planning Law) in Japan lay down several
Articles related to recovery processes after disasters.
However, the term of tsunami appears only once in
the BSL and in the case of CPL, it does not contain
the term of tsunami in the law before 2011, because
most of recovery processes were prepared against
urban fire in these laws. The Japanese history of urban
disaster focuses on spread of fire in the city since Edo
era and large-scale urban fire also occurred recently in
Kobe city in 1995 at the Great Hanshin-Awaji
Earthquake. Therefore, current urban planning system
in Japan seems to deal with tsunami disaster
management not so clearly. The devastated tsunami
that occurred on March 11, 2011 resulted in the first
huge tsunami disaster during almost one century of
history of both legal systems.
There is no UPA in the 2011 tsunami affected areas
in Iwate prefecture, while most of Sendai plain in
Miyagi prefecture is clearly divided into UPA and
UCA. Most of Sanriku areas are not applied Area
Division (to classify into UPA and UCA) even in
Miyagi prefecture like Minami-sanriku town and
Kesennuma city, because there was not so much
pressure on increase of population. Area Division
system should be applied to the area where urban
settings are already formulated or where urban
development is expected within the next 10 years.
Especially in the latter UPA case, the standard of
setting area prescribes not to principally designate
tsunami hazardous areas based on the Article 8 of the
Ordinance of CPL. However, in 1964 when the
current CPL was established there existed few
scientific and engineering knowledge on tsunami to
establish legal system, and with other reasons such as
indicated below, the regulations do not function well
in reality.
The reasons why CPL did not function well against
tsunami in spite of Article 8 of its ordinance are as
follows:
(1) Social reasons:
When the UPA was firstly designated, the areas
where consensus could be easily obtained were
selected. Already urbanized city areas had the first
priority;
UPA was basically set in the regions that rapid
increase of population occurred;
Many new public facilities and care houses for
the elderly were constructed in the UPC areas with
tsunami risks, because of consolidation of

1004
municipalities and progress of aged society.
(2) Physical reasons:
People thought that it will be easy to evacuate
from tsunami, since the most advanced information
technology and early warning systems are established;
People who lived in the coastal area thought that
seawalls, river gates and tsunami evacuation buildings
would protect residents from tsunami disasters.
(3) Economic reasons:
UPA tends to be designated in and around
already urbanized area, because of efficient use and
maintenance of infrastructure, such as roads, sewerage
and water supply;
To move to safer area, in particular change of
urban setting is almost impossible in normal time even
though all people recognized their high risk (partly
because of financial reason of municipality);
The movement to protect agricultural land
becomes weaken as demand of agricultural land will
not increase and because of higher exchange ratio
(strong yen) and excessive domestic rice production
etc..
(4) Institutional reasons:
The Article 8 of the ordinance of CPL does not
include tsunami hazard application to already
urbanized areas. It regulates only newly urbanized
(planned) areas;
In the regulation of Article 8, the terms of
principally and hazardous area against tsunami
are used without clear definitions. There is no detail
explanation in the MLIT guideline of urban planning.
8. Conclusions
Concentration of population into urbanization area
and rapid improvement of social infrastructures due to
the economic growth mostly in emerging countries
would be common now in the world economy. As
frequent earthquake and tsunami disasters during the
21st century (e.g., Sumatra Tsunami in 2004, Kashmir
Earthquake in 2005, Sichuan Earthquake in 2008, and
the Great East J apan Earthquake and Tsunami in 2011)
proved such demands for disaster management. In
each on-site observation, the collapse of buildings
caused major damages in the earthquake related
disasters. Therefore, quality of new houses and in the
case of tsunami location of new urban area is the key
to successful recovery. Moreover seismic retrofitting
became popular in J apan after the Great
Hanshin-Awaji Earthquake, while there are few
cases of retrofitting of existed buildings in many
countries.
Further field investigation will be needed to follow
the reconstruction processes in the affected areas of
the Great East J apan Earthquake. Comparison of
prefectures and cities provide lessons on the
reconstruction process under the different urban
planning settings and conditions, and will be helpful
to improve urban planning systems. One sided control
mechanism for implementation of building can not
solve the problems. Building control has to be
integrated with socio-economic, institutional,
technical and other tools to achieve safety of buildings
and built-environment. In order to mitigate earthquake
risk, all stages of building construction, from location,
planning, construction to maintenance, are important.
Awareness creation is instrumental for building
culture of safety and demands for intervention in
disaster mitigation. The demands ultimately help in
creating conducive environment for policy
intervention, in realizing institutional mechanism of
code enforcement and land use control for the
municipal authorities and in creating demand for
competent professionals.
Acknowledgments
The authors acknowledge support from the BRI
(Building Research Institute), the IISEE (International
Institute of Seismology and Earthquake Engineering)
for providing opportunities to visit tsunami affected
areas in April (Iwaki, Miyako, Kamaishi, Sendai,
Ishinomaki, Onagawa, Natori and Iwanuma) and in
September (Sendai, Ishinomaki, Onagawa and

1005
Minami-sanriku). The author also appreciates the
University of Tokyo for providing supports to visit
Kessennuma, Rikuzentakata, Ofunato and Kamaishi
in November 2011 and January 2012. These supports
make possible to summarize the paper.
References
[1] T. Utsu, World Earthquake Catalogue, The University of
Tokyo, Japan, 1990, p. 243.
[2] T. Utsu, A List of Deadly Earthquakes in the World:
1500-2000, International Handbook of Earthquake and
Engineering, Seismology Part A, Academic Press, San
Diego, 2002, pp. 691-717.
[3] Earthquake Hazards Data Web Site,
http://earthquake.usgs.gov/research/external/reports/08H
QGR0102.pdf etc. (accessed Jan. 1, 2013).
[4] A.W. Coburn, R.J.S. Spence, A. Pomonis, Factors
determining human casualty levels in earthquakes:
Mortality prediction in building collapse, in: The 10th
World Conference on Earthquake Engineering (WCEE),
University of Cambridge, UK, 1992.
[5] Reference Data, Report of Special Investigation
Committee on Earthquake and Tsunami Measures Based
on the Lessons from the Tohoku Earthquake, Cabinet
Office, Sept. 28, 2011, pp. 35-43.
[6] Guideline of Urban Planning, MLIT (Ministry of Land,
Infrastructure, Transport and Tourism), Government of
Japan, 2011.
[7] N. Hayashi, S. Saito, Prospect of Population in Iwate,
Miyagi and Fukushima prefectures, SERC Discussion
Paper, SERC11023, 2011. (in Japanese)
[8] Miyagi Pref. Recovery Plan Web Site,
http://www.pref.miyagi.jp/juutaku/saigaijouhou/2011122
1plan_honbun.pdf (accessed Jan. 1, 2013).
[9] Miyagi Pref. Recovery Plan Web Site,
http://www.pref.iwate.jp/view.rbz?nd=2974&of=1&ik=1
&pnp=2974&cd=35093 (accessed Jan. 1, 2013).


Time Dependent Gaussian Equivalent Linearization of
Duffing Oscillator Using Continuous Wavelet Transform
Arunasis Chakraborty
1
, Prateek Mittal
2
and Sabarethinam Kameshwar
3

1. Department of Civil Engineering, Indian Institute of Technology Guwahati, Guwahati 781039, India
2. Department of Urban and Regional Planning, University of Illinois at Urbana-Champaign, Urbana IL 61801, USA
3. Department of Civil Engineering, Rice University, Houston TX 77251, USA

Abstract: Evolutionary response analysis of Duffing oscillator using Gaussian equivalent linearization in wavelet based
time-frequency frame work is presented here. Cubic (i.e., odd type) non-linearity associated with stiffness and damping is modeled.
The goal of this research is to develop the mathematical model of an equivalent linear system which is applicable for different
non-stationary input processes (i.e., either summation of amplitude modulated stationary orthogonal processes or digitally simulated
non-stationary processes). The instantaneous parameters of the ELTVS (equivalent linear time varying system) are evaluated by
minimizing the error between the displacements of non-linear and equivalent linear systems in wavelet domain. For this purpose, three
different basis functions (i.e., Mexican Hat, Morlet and a modified form of Littlewood-Paley) are used. The unknown parameters (i.e.,
natural frequency and damping) of the ELTVS are optimized in stochastic least square sense. Numerical results are presented for
different types of input to show the applicability and accuracy of the proposed wavelet based linearization technique.

Key words: Statistical linearization, wavelet analysis, Duffing oscillator, evolutionary process.

1. Introduction

Stochastic response characterization of both linear
and nonlinear dynamical systems has remained an
active area of research in the last few decades. While
theories of stationary random vibration analysis of
linear system are well developed [1-3], nonlinear
random vibration analysis offers several challenges and
hence remains on open area of research. Caughey [4, 5]
developed mathematical models for linear and
nonlinear systems based on the theory of Markov
processes. In these studies, the systems were subjected
to white noise excitation and the response was
evaluated by solving the FPK
(Fokker-Planck-Kolmogrov) equation. Hammond [6]
used evolutionary spectral representation of a
non-stationary process to study the response of a linear
oscillator. Orabi [7] developed nonstationary response

Corresponding author: Arunasis Chakraborty, Ph.D.,
assistant professor, research fields: earthquake engineering,
random vibrations and wavelet analysis. E-mail:
arunasis@iitg.ernet.in.
analysis of nonlinear systems using the functional
series expansion method. Recently, Cross et al. [8]
developed approximate solution for frequency
response function of Duffing oscillator using FPK
equations to model higher order resonance. Here, it
may be mentioned that close-form solutions based on
FPK equations are limited to certain class of input. For
nonlinear systems, equivalent linearization is the most
popular tool for response characterization.
In equivalent linearization, the nonlinear system is
replaced by an equivalent linear system whose
parameters are evaluated in stochastic least square
sense [9, 10]. Caughey [11] developed approximate
solution for weakly nonlinear systems using equivalent
linearization. He considered nonlinearity associated
with velocity and displacement. Iwan et al. [12] used
equivalent linearization to solve the response of
nonlinear system under nonstationary excitations.
Iyengar [13] developed analytical model for the
stochastic response of Duffing oscillator excited by
white noise. A comprehensive review of statistical
DAVID PUBLISHING
D
Time Dependent Gaussian Equivalent Linearization of Duffing
Oscillator Using Continuous Wavelet Transform

1007
linearization may be found in the book by Robert et al.
[14]. Elishakoff et al. [15] and Elishakoff [16]
suggested the use of both potential energy and
complementary energy as the criteria for linearization.
Sobiechowski et al. [17] derived expression response
of Duffing oscillator when the input is non Gaussian.
Pradlwater [18] used local linearization coupled with
Gaussian superposition to estimate the probability
density function of a nonlinear system. Mickens [19]
used combined equivalent linearization and
perturbation technique to solve the response of
nonlinear systems. Crandall [9] used non Gaussian
input for statistical linearization of Duffing oscillator
and other oscillators. Roy et al. [20] generalized locally
transversal linearization and developed multi step
transversal linearization method for solving nonlinear
oscillators. Lacquaniti et al. [21] proposed probabilistic
linearization based on FPK equations for nonlinear
systems subjected to additive and multiplicative
random excitations. Li et al. [22] used equivalent
linearization based on equivalence of energy dissipated
by two systems. In view of these models for random
vibration analysis, it may be mentioned that
conventional techniques often face difficulty to model
both amplitude and frequency nonstationarity. For this
purpose, wavelet based time-frequency analysis has
gained popularity in the recent past. Basu et al. [23-26]
and Chakraborty et al. [27] used wavelet based
time-frequency analysis to evaluate the response of
randomly excited linear and nonlinear systems.
Wavelet transform has time localization property
which makes it more suitable when the input has time
varying amplitude and frequency content. Spanos et al.
[28] used different wavelet basis functions to model the
nonstationary process. From these studies, one may
conclude that stochastic response analysis of nonlinear
systems especially in the light of both amplitude and
frequency nonstationary offers several challenges and
needs further attention.
With this in view, present work aims to develop
wavelet based equivalent linearization technique for a
Duffing oscillator subjected to nonstationary random
process. The time localization characteristics of the
wavelet transform is proposed to be utilized to capture
the fluctuation of amplitude and frequency content of
the input signal to evaluate the instantaneous
parameters of the time varying equivalent linear
system.
2. Problem Formulations
In this paper, the Duffing oscillator shown in Fig. 1
is considered for evolutionary response analysis. The
governing equation of motion is given by:
( )
g
x x x g x = + , (1)
The time function ( ) t x
g
in the above equation
represents the nonstationary acceleration at the support.
Function ( ) x x g , in Eq. (1) is given by:
( )
2
2
2 3
1
2 2 ,
n n n n
x x x x x g e e qe qe + + + = (2)
Parameters q and e
n
are the damping ratio and
un-damped natural frequency of the linear system.
Parameter
1
and
2
in Eq. (2) control the degree of
nonlinearity. The nonlinear part of the stiffness and
damping varies with the cubic power of the
displacement and velocity, respectively. Using
statistical linearization, the Duffing oscillator as
described in Eq. (1) is replaced by an equivalent linear
system whose governing equation is given by
g eq eq eq
x x x x = + +
2
2 e e q (3)
In the above equation, q
eq
and e
eq
are the damping
ratio and un-damped natural frequency of the equivalent

Fig. 1 Duffing oscillator.
( ) t x
g

( ) x g
( ) x g
( ) t x
m

1008
linear system. For simplifying mathematical operations,
a new parameter, p
cq
= p
cq
cq
is introduced. Hence
the equation of motion for equivalent linear system
takes the form:
x + 2p
cq
x +
cq
2
x = -x
g
(4)
The unknown parameters in the above equation are
evaluated in least square sense by minimizing the
expected value of the square of the error between Eq. (1)
and Eq. (4). However, as the input process is assumed
to be non-stationary (whose exact mathematical model
may or may not be known), the unknown parameters in
Eq. (4) are bound to vary with time. In view of this
non-stationary nature, the time-frequency localization
characteristics of wavelet based signal processing is
proposed to be utilized here to evaluate the unknown
time varying parameters in Eq. (4).
2.1 Brief Overview of Continuous Wavelet Transform
Wavelet transform has been widely used for
different applications of signal processing in the recent
past. Plenty of books and literatures [1, 29, 30] are
available that describe wavelet transform and its
applications in details. Here, a brief overview of the
wavelet transform is presented for continuity. Let f(t)
be a square integrable function in L
2
(R) (i.e., f
2
(t)dt <
), then using continuous wavelet transform f(t) can be
converted into two dimension with parameters a and
b, which is given by:
w
(o, b) = ] (t)
u,b
-
-
(t)Jt (5)
where, * denotes the complex conjugate and
( ) t
b a
*
,
is the dilated and time localized version of the
mother wavelet +(t):
u,b
-
=
1
|u|

-
[
t-b
u
(6)
Parameter a dilates the wavelet basis and provides
frequency resolution while parameter b localizes it in
time (i.e., around t = b). The original time signal f(t)
can be traced back from the wavelet coefficients using
the following relationship:
(t) =
1
2nC
] ]
1
u
2
-
w
(o, b) [
t-b
u
du
|u|
Jb(7)
Parameter C
+
in the above equation is the
admissibility criteria for +(t) to be a wavelet basis and
is given by:
C
= ]
|+(o)|
2
|o|
J
-
< (8)
In the above equation, +(e) is the Fourier transform
of wavelet basis+(t). It may be noticed that parameters
a and b are continuous. However, following
discretization scheme is adopted for parameters a
and b in the present study for the purpose of
numerical analysis:
_
o
]
= o
]
b
= (i -1)b
_ (9)
It can be shown that the above discretization scheme
reduces Eq. (7) to the following form:
(t) =
K b
u
]
]
w
(o
]
, b
)_
t-b
i
u
]
] (10)
The constant K in Eq. (10) is given by:
K = (o
2
- 1) 4nC
o , (11)
In this context, selection of wavelet basis is an
important step as it affects the frequency and time
resolution of the wavelet transform. In this paper, three
different basis functions are used to evaluate the
wavelet coefficients in Eq. (5).
(1) Modified form of little-wood Paley basis
function (MLP)
This basis function is constructed using the real part
of the harmonic wavelet basis [1]. The time and
frequency signature of this basis are given by:
( )
( ) ( )
( )
( )
s s
=
ot e t
t o
e
t ot
o t
1 2
1
sin sin
1
1
t
t t
t
(12)
From the above expression, one can notice that +(e)
has compact support in frequency domain;
(2) Mexican Hat basis function (Mex)
The time and frequency description of the Mexican
Hat basis function is given by:
( ) ( )
( ) ( )

=
=
2
2
4
2
2
4
2
2
3
2
1
3
2
e
e
t
e
t
e
e t t
t
(13)
(3) Morlet basis function (Mor)

1009
The time and frequency description of the Morlet
basis function is given by:
( ) ( )
( )
( )
( ) ( )
|
|
.
|
\
|
+ =
=
+

2 2
2
2 2
2 2
2
2
cos
c c
c
e e
e
t e t
c
t
e e e e
e e
e
e
(14)
Parameter e
c
in the above equation represents the
central frequency of this basis and is assumed to be 5
rad/s throughout this paper.
2.2 Non-stationary Input in Wavelet Domain
In this paper, the support motion is modeled as
amplitude modulated stationary process which is
mathematically expressed as Refs. [31, 32]:
x
g
= ] A
(, t) c
ot
-
J0
()
n
=1
(15)
where, A
r
(e, t) represents the r
th
amplitude modulation
function dependent on time and frequency. In above
equation, dG
r
(e) presents the r
th
orthogonal stationary
Gaussian process which is given by:
E|J0
(). J0
-
(
i
)] =
=

S
]]
()J
u
; =
i
; =
i
(16)
Taking wavelet transform of both sides of Eq. (15), it
can be shown that wavelet coefficients of ( ) t x
g
can
be expressed as Ref. [28]:
w
x
g
(o
]
, b) = ] A
(, b) c
ob
J0
-
()
n
=1
(17)
In Eq. (17), the frequency parameter has been
discretized and hence a
j
represents the j
th
frequency
scale while the time parameter b is kept continuous.
It can be seen that the wavelet coefficients
( ) b a x W
j g
,
at a scale a
j
has a similar mathematical
expression as that of ( ) t x
g
. The orthogonal process in
wavelet domain can be shown as:
E|J0
(). J0
-
(
i
)] = S
()J
= u
; =
i
; =
i
(18)
where, S
() = 2no
]
|(o
]
)|
2
S
]]
(). In the above

equation, +(ea
j
) represents the Fourier transform of
the wavelet basis function at the j
th
frequency scale a
j
.
In this study, two different examples of the input
processes are considered. In the first example, the
system is excited by an amplitude modulated stationary
process. The amplitude modulation A(t) and the
stationary processes S
ff
(e) are given by:
A(t) = y(c
-ut
- c
-[t
) (19)
S
]]
() = S
o
_
4q
g
2
[
o
o
g
,
2
_1-[
o
o
g
,
2
_
2
+4q
g
2
[
o
o
g
,
2
_ (20)
In the second case, EL-Centro ground motion which
is modeled as the summation of different amplitude
modulated stationary orthogonal process is used. The
envelop function and the power spectral density for the
recorded El-Centro ground motions as suggested by
Conte et al. [33] is used in the analysis and is expressed
as:
A
(t) = o
(t -
)
[
r
c
-y
r
(t-
r
)
E(t -
) (21)
S
]]
() =
v
r
2n
j
1
v
r
2
+(o+q
r
)
2
+
1
v
r
2
+(o-q
r
)
2
[ (22)
For brevity, the detailed description of the terms in
Eqs. (21) and (22) are omitted here. The reader may
refer to Conte et al. [33] for further details.
2.3 Wavelet Based Linearization
In this section, a LTVS (linear time varying system)
is designed in wavelet domain in such a way that the
error between the Duffing oscillator and the LTVS is
minimized with respect to its parameters in stochastic
sense. The first step in this process is to express the
governing equations in wavelet domain. On
transforming both sides of Eq. (1) in wavelet domain,
one can show that:
2
b
2
w
x(o
]
, b) +w
g(o
]
, b) = w
x
g
(o
]
, b) (23)
Similarly, wavelet transforming both sides of Eq. (3)
one gets:
2
b
2
w
x(o
]
, b) +2p
cq
b
w
x(o
]
, b) +

cq
2
w
x(o
]
, b) = w
x
g
(o
]
, b) (24)
It can be seen that
cq
and p
cq
are the unknowns
in the above equation. The difference between Eqs. (23)
and (24) represents the error. On taking expectation of
the square of this error at i
th
time instant and j
th
scale,
one can show that:
E|e
]
2
] = E|{2p
cq
w
x (o
]
, b
) +
cq
2
w
x(o
]
, b
) -
w
g(o
]
, b
)]
2
] (25)
The gross error at i
th
time instant (i.e., for all j) in Eq.
(25) is differentiated with respect to the unknown

1010
parameters
cq
2
and p
cq
to form two simultaneous
equations which is given by:
o
cq
2
1
u
]
]
E|e
]
2
] = u (26)
q
cq
1
u
]
]
E|e
]
2
] = u (27)
On simplifying above expression and applying
properties of wavelet transform and Gaussian process,
it can be proved that:
cq
i
2
=
n
2
_1 + Sz
2
K
u
]
E|w
2
x(o
]
, b
)]
]
_ (28)
p
cq
i
= p
n
_1 + Sz
1
K
u
]
E|w
2
x (o
]
, b
)]
]
_ (29)
From Eqs. (28) and (29), it can be noticed that the
solution for natural frequency and damping of the
LTVS need second moment of the displacement and
velocity in wavelet domain. Therefore, to solve for the
unknown parameters, the evolutionary response of the
LTVS system is first modeled in wavelet domain. With
this in view, the response W
+
x(a
j
, b) from Eq. (24) in
wavelet domain at scale a
j
may be written as:
w
x(o
]
, b) = ] b(b - )
b
0
w
x
g
(o
]
, ) J (30)
In the above equation, h() represents the impulse
response function. Substituting Eq. (17) in Eq. (30) and
simplifying, one can show that
w
x(o
]
, b) =
] b(b -)
b
0
] A
(, ) c
o:
-
J0
() J
n
=1

= ] H
(, b) c
ob
-
J0
()
n
=1
(31)
where, M
r
(e, b) is given by:
H
(, b) = ] b(b - )
b
0
A
(, ) c
o(:-b)
J (32)
It is assumed that amplitude modulations in Eq. (15)
are independent of frequency i.e., ( ) ( ) t A t A
r r
= , e .
Moreover, it may be noticed that in Eq. (30), ( ) t A
r
, e
is a slowly varying function of time as compared to
( ) h
. Therefore, as the wavelet transform is localized
around t = b , A
(, ) may be approximated as
A
(b) and may be considered constant. Therefore,

( ) b M
r
, e may be approximated as
H
(, b) = - A
(b) E() (33)

where, ( ) e H is the frequency response function of
the LTVS and is given by:
E() =
1
(o
cqi
2
o
2
) +(2q
cqi
o
cqi
o)
(34)
Substituting Eq. (33) in Eq. (31) and multiplying
both sides of Eq. (31) by its complex conjugate, one
can show that:
E|w
2
x(o
]
, b) ] =
4no
]
|A
(b)|
2
] |E()|
2
|(o
]
)|
2
S
]]
() J
0
n
=1
(35)
On differentiating both sides of Eq. (31), it can be
shown that:
w
x (o
]
, b) =
b
] H
(, b) c
ob
-
J0
()
n
=1
(36)
Substituting the expression for H
(, b) and
performing differentiation using chain rule, one gets:
w
x (o
]
, b) =
] - A
(b) E()i c
ob
-
J0
()
n
=1
+
] - A
(b) E() c
ob
-
J0
()
n
=1
(37)
Again it may be noticed that as A
(b) is
approximated as slowly varying function at around
t = b, its time derivative i.e., A
(b) can be neglected.

Therefore, considering the first term and multiplying
Eq. (37) by its complex conjugate and taking the
expectation, one can show that:
E|w
2
x (o
]
, b) ] =
4no
]
|A
(b)|
2
] |E()|
2

2
|(o
]
)|
2
S
]]
() J
0
n
=1
(38)
Now all the expressions required to evaluate
cq
i

and p
cq
i
, from Eqs. (28) and (29), have been defined.
Hence two error functions can be formulated as:
e
1,
2
= _
cq
2
-
n
2
_1 +Sz
2
K
u
]
E|w
2
x(o
]
, b
)]
]
__
2
(39)
e
2,
2
=
_
p
cqi
-p
n j
1 +S
K
u
]
E|w
2
x (o
]
, b
)]
]

[_
2
(40)
To find the parameters of the equivalent linear
system, the two error functions will have to be
minimized simultaneously. This can be done by
defining a new error function which is the sum of
squares of the original error functions:
e = e
1,
2
+ e
2,
2
(41)
To minimize the error function expressed in Eq. (41),
in-built functions of optimization toolbox in MATLAB
are used. When the natural frequency and damping of
LTVS is calculated, then PSDF (power spectral density
function) of the displacement of the nonlinear system
may be evaluated using the following relation [28]:


1011
S
xx
(, b) = o
]
(b)|(o
]
)|
2
n
]=1
(42)
where, a
j
(b) is evaluated from the following matrix
equation at each time instant:
_
K
1,1
K
1,2
K
1,n
K
2,1
K
2,2
K
2,n

K
n,1
K
n,2
K
n,n
_ _
o
1
(b)
o
2
(b)
o
n
(b)
_ =

l
l
l
l
l
E|w
(o
1
, b)
2
]
E|w
(o
2
, b)
2
]
E|w
(o
n
, b)
2
]
1
1
1
1
1
(43)
The coefficients of the above matrix can be
evaluated from the wavelet basis function and are given
by:
K
,s
= 4no
] |(o
)|
2
|(o
s
)|
2
J
0
(44)
Finally, the higher order moments of the response of
the Duffing oscillator may be evaluated once the
displacement PSDF is calculated using Eq. (42).
3. Numerical Results and Discussion
In this section, numerical analysis of the wavelet
base equivalent linearization of Duffing oscillator is
carried out. Two different example cases are solved to
study the accuracy of effectiveness of the proposed
method. The parameters
1
and
2
in Eq. (2) are
assumed to be 0.1 throughout this numerical study. In
the first example, a single amplitude modulated
stationary processes as discussed in Eqs. (19) and (20)
is considered. The natural frequency and the damping,
i.e., e
g
and q
g
of the input process as defined in Eq. (20)
are assumed to be 12 rad/s and 0.4, respectively. The
parameters of the envelop function in Eq. (19) (i.e., o,
| and ) are 0.25, 0.5 and 4, respectively. For these set
of values, the envelope function attains a peak value of
1 at 2.5 s. The natural frequency and critical damping
ratio, i.e., e
n
and q in Eq. (1) are assumed to be 40
rad/s and 5%, respectively. Using these values, wavelet
based statistical linearization is performed to evaluate
the time varying parameters of the equivalent linear
system. Three different wavelet bases are used namely
MLP, Mexhat and Morlet. The details of the
parameters for these bases are listed in Table 1. Fig. 2
shows the instantaneous natural frequency and
damping ratio of the equivalent LTVS. It may be
noticed that the values of the equivalent parameters
obtained from Eqs. (28) and (29) using three different
bases match closely. Finally, the instantaneous
normalized mean square values, i.e., | | | |
2 2
/
i i
x E x E
of the equivalent LTVS are compared with simulation
results.
A close match is observed between them which
indicate that the accuracy of the proposed wavelet
based model. In this context, 100 samples of the
earthquake accelarogram are used for the simulation.
The wavelet based linearization model is then further
used to study the stochastic response of the Duffing
oscillator. Fig. 3 shows the variation of peak mean
square value for different e
n
. A close match between
the wavelet results using different bases and the
simulation is also observed in this case. Fig. 4 shows
the instantaneous PSDF at 3.5 s and 7 s. It may be
noticed that as the MLP basis has compact support in
frequency domain, the PSDF obtained using this basis
function shows band localization. Nevertheless, it
matches with the results obtained using other basis
functions, i.e., Mexhat and Morlet in mean sense.
In the second example, El-Centro ground motion is
used, which is modeled by Eqs. (21) and (22). For the
details of this model, one may refer to Ref. [33]. The
El-Centro ground motion is modeled as summation of
amplitude modulated stationary orthogonal processes.
The details of the parameter of the twenty one different
amplitude modulations and stationary processes are

Table 1 Model parameters for wavelet transform.
Basis function J
MLP 2
1/4
-15 to 6
Mex. Hat 2 -8 to 6
Morlet 2
1/2
-9 to 5


1012

Fig. 2 Parameters of equivalent linear system for
n
= 40 rad/s and = 5% under amplitude modulated stationary excitation:
(a)
eqi
; (b)
eqi
; (c) normalized mean square value.

Fig. 3 Peak mean square value vs.
n
for = 5% for amplitude modulated stationary excitation.
(a)
(b)
(c)
40.0002
40.0001
40
0

1013

Fig. 4 PSDF of the LTVS for
n
= 40 rad/s and = 5% for amplitude modulated stationary excitation: (a) at t = 3.5 s; (b) at t
= 7 s.

given in Ref. [33]. The natural frequency and damping,
i.e., e
n
and q are assumed to be 40 rad/s and 10%,
respectively.
Fig. 5 shows the instantaneous values of the natural
frequency and damping of the equivalent LTVS for
three different basis functions. The values of the
parameters in both the cases match close for different
bases. Fig. 5c shows the instantaneous normalized
mean square values of the equivalent LTVS using three
bases and simulation and the wavelet results closely
match with the simulation which, in turn, proves the
accuracy of the proposed linearization scheme. Fig. 6
shows the peak mean square value of the response for
different natural frequencies. It can be observed that at
high frequency, the response evaluated by three
different bases closely matches with simulation results.
While at low frequency range, MLP basis function
provides better results.
Fig. 7 shows instantaneous PSDF at 3.5 s and 7 s for
different bases. Here also the MLP basis shows step
like frequency localization as it has compact support in
frequency. Nevertheless, the results match with each
other in average sense. Finally, Figs. 8-10 show
response PSDF evaluated using three different bases.
From these figures, one may easily conclude that the
proposed wavelet based linearization scheme
successfully models the amplitude and frequency
non-stationary nature of the response of the non-linear
system. In this context, it may be noticed that results
obtained using MLP and Morlet bases are similar as their
peak values and frequency localization match closely.
Although the peak value of the PSDF obtained by
this basis function matches with other two, the
frequency localization is not distinct as compared to
MLP and Morlet bases. Nevertheless, the area under
the PSDF at different time (i.e., instantaneous mean
square values) obtained by three different bases are
same, which is also reflected in Figs. 5 and 6. All these
results establish that the proposed model of the wavelet
based statistical linearization can be adopted for the
evolutionary response analysis of the Duffing
oscillator.

(a)
(b)

1014

Fig. 5 Parameters of equivalent linear system for
n
= 40 rad/s and = 5% under El Centro excitations: (a)
eqi
;

(b)
eqi
; (c)
normalized mean square value.

Fig. 6 Peak mean square value vs.
n
for = 10% for El Centro excitations.

Fig. 7 PSDF of the LTVS for
n
= 40 rad/s and = 5% for El Centro excitations: (a) t = 3.5 s; (b) at t = 7s.
(a)
(b)
(a)
(b)
(c)

1015

Fig. 8 Displacement PSDF using LP basis function for
n
= 40 rad/s and = 10% for El Centro excitations.

Fig. 9 Displacement PSDF using Morlet basis function for
n

Fig. 10 Displacement PSDF using Mexican Hat basis function for
n


1016

4. Conclusions
The non-stationary response of a Duffing oscillator
is evaluated using combined statistical linearization
and continuous wavelet transform. As amplitude
modulated stationary process is used at input, the
parameters of the equivalent linear system vary with
time. These parameters are evaluated by minimizing
the mean square error between the displacements of
the nonlinear system and the LTVS in wavelet domain.
In this context, the input-output relation of a LTVS
under non-stationary input is developed using time
localization characteristics of wavelet transform,
which is the uniqueness of the present work. The
time-frequency analysis does not impose any
restriction on the choice of the basis function. Three
different basis functions are used to evaluate the
displacement PSDF of the nonlinear system which
matches closely with the simulation results. Moreover,
as the linearization scheme is developed using the
wavelet coefficients of the input process, the proposed
linearization technique is applicable for either
summation of amplitude modulated stationary
orthogonal process or digitally simulated signals,
which makes it more general. All these prove the
efficiency and accuracy of the proposed wavelet based
linearization to evaluate the stochastic response of the
Dufffing type oscillator.
References
[1] D.E. Newland, An Introduction to Random Vibration,
Spectral and Wavelet Analysis, Longman Scientific and
Technical, Essex, 1983.
[2] N. Nigam, S. Narayanan, Application of Random
Vibrations, Narosa Publishing House, New Delhi, 1994.
[3] L. Lutes, S. Sarkani, Random Vibrations,
Butterworth-Heinemann Burlington, UK, 2004.
[4] T. Caughey, Derivation and application of Fokker-Planck
equation to discrete dynamics system subjected to white
random excitation, Journal of Acoustic Society of
America 35 (1963) 1683-1692.
[5] T. Caughey, Non-linear theory of random vibrations,
Advanced Applied Mechanics 11 (1971) 209-253.
[6] J. Hammond, Evolutionary spectra in random vibrations,
Journal of the Royal Statistical Society 35 (1973)
167-188.
[7] A.G. Orabi, A functional series expansion method for
response analysis of nonlinear systems subjected to
random excitations, Journal of Non-linear Mechanics 22
(1987) 451-465.
[8] E. Cross, K. Worden, Approximation of the Duffing
oscillator frequency response function using the FPK
equation, Journal of Sound and Vibration 330 (2011)
743-756.
[9] S. Crandall, On using non-Gaussian distributions to
perform statistical linearization, International Journal of
Non-linear Mechanics 39 (2004) 1395-1406.
[10] J.R. Booton, The analysis of nonlinear control system
with random inputs, in: Proceedings of the Symposium
on Nonlinear Circuit Analysis, Polytechnic Institute of
Brooklyn, 1953, pp. 369-397.
[11] T. Caughey, Equivalent linearization techniques, Journal
of the Acoustical Society of America 35 (1963)
1706-1711.
[12] W. Iwan, A. Mason, Equivalent linearization for systems
subjected to non-stationary random excitation,
International Journal of Non-linear Mechanics 35 (1980)
71-82.
[13] R.N. Iyengar, Stochastic response and stability of the
Dufffing oscillator under narrow band excitation, Journal
of Sound and Vibration 126 (2) (1988) 255-263.
[14] J. Roberts, P. Spanos, Random Vibration and Statistical
Linearization, John Wiley and Sons, New York, 1990.
[15] I. Elishakoff, X.T. Zhang, An appraisal of different
stochastic linearization criteria, Journal of Sound and
Vibration 153 (1992) 370-375.
[16] I. Elishakoff, Multiple combinations of the stochastic
linearization criteria by the moment approach, Journal of
Sound and Vibration 237 (3) (2000) 550-559.
[17] C. Sobiechowski, L. Socha, Statistical linearization of the
Duffing oscillator under non-Gaussian external excitation,
Journal of Sound and Vibration 231 (2000) 19-35.
[18] H.J. Pradlwarter, Non-linear stochastic response
distributions by local statistical linearization,
International Journal of Non-linear Mechanics 36 (2001)
1135-1151.
[19] R. Mickens, A combined equivalent linearization and
averaging perturbation method for non-linear oscillator
equations, Journal of Sound and Vibration 264 (2003)
1195-1200.
[20] D. Roy, R. Kumar, A multi-step transversal linearization
(MTL) method in non-linear structural dynamics, Journal
of Sound and Vibration 287 (2005) 203-226.
[21] S. Lacquaniti, G. Ricciardi, Aprobabilistic linearization
method for non-linear systems subjected to additive and

1017
multiplicative excitations, International Journal of
Non-linear Mechanics 41 (2006) 1191-1205.
[22] W. Li, C. Chang, S. Tseng, The linearization method
based on the equivalence of dissipated energies for
nonlinearly damped structural systems, Journal of Sound
and Vibration 295 (2006) 797-809.
[23] B. Basu, V.K. Gupta, Non-stationary seismic response of
MDOF systems by wavelet transforms, Earthquake
Engineering and Structural Dynamics 26 (1997)
1243-1258.
[24] B. Basu, V.K. Gupta, Seismic response of SDOF system
by wavelet modeling of non-stationary processes,
Journal of Engineering Mechanics 124 (10) (1998)
1142-1150.
[25] B. Basu, V.K. Gupta, Stochastic seismic response of
SDOF systems through wavelets, Engineering Structures
22 (2000) 1714-1722.
[26] B. Basu, V.K. Gupta, Wavelet based stochastic seismic
response of a Duffing oscillator, Journal of Sound and
Vibration 245 (2001) 251-260.
[27] A. Chakraborty, B. Basu, Non-stationary response
analysis of long span bridges under spatially varying
differential support motions using continuous wavelet
transform, ASCE Journal of Engineering Mechanics 134
(2) (2008) 152-162.
[28] P. Spanos, G. Failla, Evolutionary spectral estimation
using wavelets, Journal of Engineering Mechanics 130
(2004) 952-960.
[29] R. Rao, A. Bopardikar, Wavelet Transforms, Introduction
to Theory and Applications, Addison Wesley Longman,
Massachusetts, 1995.
[30] Y.T. Chan, Wavelet Basics, Kluwer Academic Publishers,
Boston, 1995.
[31] M.B. Priestley, Power spectral analysis of non-stationary
random processes, Journal of Sound and Vibration 6
(1967) 86-97.
[32] M.B. Priestley, Evolutionary spectra and non-stationary
processes, Journal of the Royal Statistical Society, Series
B 27 (1965) 204-237.
[33] J. Conte, B. Peng, Fully nonstationary analytical
earthquake ground-motion model, Journal of Engineering
Mechanics 123 (1997) 15-24.

An Earthquake Catalogue for El Salvador and
Neighboring Central American Countries (1528-2009)
and Its Implication in the Seismic Hazard Assessment
Walter Salazar
1
, Lyndon Brown
2
, Walter Hernndez
3
and Jos Guerra
4
1. Seismic Research Centre, The University of the West Indies (UWI), St. Augustine, Trinidad and Tobago
2. Earthquake Unit, The University of the West Indies (UWI), Mona, Kingston, Jamaica
3. Geological Survey/General Directorate of Environmental Observatory (DGOA), San Salvador, El Salvador
4. Oficina de Planificacin del rea Metropolitana de San Salvador (OPAMSS), San Salvador, El Salvador

Abstract: Merging all available databases, a comprehensive and updated earthquake catalogue for El Salvador and surrounding areas
has been compiled, containing a total of 2,584 events for the period 1528-2009, covering the geographic window delimited by the
coordinates 11.0
-16.5
N and 85.5
-92.0
W, focal depths of 0.0 to 304 km, and the moment magnitudes in the interval 5.0 M
W

8.1. Events in the catalogue are distributed into six seismogenic sources taking into consideration the tectonic regime affecting El
Salvador, the interplay and complexities between shallow crustal, intraplate and interface subduction seismicity has been thoroughly
investigated, primarily with the aim of developing detail criteria to delimit the seismogenic sources in order to perform a consistent
seismic hazard assessment. A uniform hazard spectrum for San Salvador and seismic hazard maps and their uncertainty have been
calculated for the horizontal component of ground motion for rock site conditions using zone and zone free methods. The references
cited in the article constitute a comprehensive list of sources of information on the tectonics and seismicity of El Salvador and
neighboring Central American countries.

Key words: Earthquake catalogue, tectonics, seismogenic sources, Gutenberg-Richter relationships, seismic hazard.

1. Introduction

In countries affected by earthquakes, the
compilation of an accurate seismic catalogue turns into
an inevitable part of the seismological investigations.
Once compiled, the seismological catalogue serves a
whole variety of purposes, including the evaluation of
the seismic hazard inherent to the region under
consideration. Cataloguing of El Salvador seismicity
has been attempted before, Martnez [1] published the
first comprehensive catalogue for El Salvador
including historical seismicity since 1520 sizing the
earthquakes and then assigning a modified Mercalli
intensity, this work was the main reference for future

Corresponding author: Walter Salazar, doctor of
engineering, research fellow, research fields: earthquake
engineering and engineering seismology. E-mail:
walter.salazar@uwiseismic.com.
special studies in the region which later re-calculate the
origin parameters of the destructive events.
Algermissen et al. [2] made the first probabilistic
seismic hazard assessment for the country, however,
their seismological data base is not accessible from the
article. Alfaro et al. [3] compiled an earthquake
catalogue including historical events covering the
period 1538-1986 based on the work of White [4] for
the Guatemalan seismicity in the boundary of the North
American plate and the Caribbean Plate and some
publications in process by the same author at that time
with updated works listed in Table 1 for both, the
subduction zone and the volcanic chain in Central
America (Section 4). To homogenize the catalogue,
they employed the surface wave magnitude M
S

(magnitude scale) incorporating mainly the ISC
(International Seismological Centre) and the PDE

An Earthquake Catalogue for El Salvador and Neighboring Central American Countries

1019

Table 1 List of references employed to compile the earthquake catalogue for El Salvador and surrounding areas.
Code Reference
Historical seismicity (1500-1900)
U Muoz and Udias (2006) [7]
G Guzmn-Speziale et al. (2005) [8]
W White et al. (2004) [9]
W White (1984) [4]
E Leeds (1974) [10]
N NOAA (National Oceanic and Atmospheric Administration) significant earthquakes of the world [11]
Special studies (1900-1994)
A Ambraseys and Adams(1996) [12]
M Ambraseys (1995) [13]
W White and Harlow (1993) [14]
S Molnar and Sykes (1969) [15]
World standard catalogues (1900-2009)
I ISC (International Seismological Centre), 2010, England [16]
P Program of PDE (preliminary determinations of epicenters), from NEIC (National Earthquake Information Center) [17]
D DNA (Decade of North America) GeologyNGDC [18]
H IPGH (Instituto Panamericano de la Geografa y la Historia) [19]
C Global CMT (centroid moment tensor) catalogue [20]
Local and regional catalogues (1984-2009)
L
Local earthquake catalogue for El Salvador: Geological Survey/General Directorate of Environmental Observatory (DGOA),
El Salvador (1984-2009)
R Regional earthquake catalogue for Central America: CASC (Central American Seismic Center) (1994-2009)

(Preliminary Determination of Epicenters) catalogues.
Singh et al. [5] compiled an earthquake catalogue
based also on world standard catalogues and retrieved
the historical information presented in Ref. [3]. This
catalogue was used to perform the seismic hazard
assessment for the building code regulations of El
Salvador [6]. However, to evaluate the seismicity and
the seismic hazard in the region, Singh et al. [5] did not
homogenize the magnitude to a single scale, they
adopted the maximum reported value of magnitude
amongst different available determinations. A common
feature of the works of Alfaro et al. [3] and Singh et al.
[5] are historical events prior to 1900 with magnitudes
up to 7.9 in the volcanic chain of Central America (its
location shown in Fig. 1). According to White et al. [14]
and the notorious macro-seismic compilation of
Feldman [21], earthquakes in this source are moderate
in size with maximum magnitudes reaching up to 6.9
due to the high degree of fragility of the crust in Central
America, which is reflected with the presence of many
faults of short lengths along the volcanic belt.
Lindholm et al. [22] performed a seismic hazard
assessment based on the catalogue of Rojas et al. [23]
for the Central American Isthmus using the M
S
for the
period 1500-1990. More recently Benito et al. [24]
updated the catalogue of Rojas et al. [23] to perform a
new seismic hazard assessment for the Central
American region in terms of M
W
magnitude, but the
whole catalogue itself is not accessible yet for the
general public (personal communication with Eng.
Alvaro Climent, Trinidad 2011). Grases [25] presented
useful macro-seismic information and instrumental
determinations for destructive earthquakes in the
Caribbean and the Central American region for the
period 1502-1990 listing magnitudes m
b
and M
S
.
Salazar et al. [26] compiled an instrumental seismic
catalogue for the period 1898-1994 compiling new
seismological information that were not included in the
previous works, this catalogue and the correspondent
geometrical delimitations for the seismic sources were
employed in the seismic hazard assessment presented
by Bommer et al. [27] in terms of the magnitude M
S
.

1020
They performed the seismic hazard assessment
employing normalized values of peak ground
acceleration discussing the spatial distribution of the
hazard in the country using both zone and zone-free
methods. Later Salazar [28] updated the work of
Salazar et al. [26] extending the period of observation
covering the years 1898-2001 including the burst of
seismicity related with the earthquake disasters of
January 13 and February 13, 2001 [37], employing Ms
as the homogenized magnitude as well and
incorporating for the first time in the region as local
and regional network information.
In keeping with modern requirements of cataloguing
seismicity, this article presents an updated catalogue
based on Salazar [28] including new historical,
instrumental, geological and tectonic information that
has been analyzed and published in the last decades.
Attempts have been made to ensure that the final
catalogue is as homogeneous as possible in terms of
magnitudes originally reported on five different

Fig. 1 Tectonic map of El Salvador and neighboring Central American countries, solid and open triangles indicate the DVF
(dominant volcanic front) and the BVF (behind volcanic front), respectively [29, 30], MF: Mixco Fault; GG: Guatemalan
Graben; IG: Ipala Graben; HD: Honduran Depression; JCF: Jocotn-Chamelecn Fault; JF: Jalpatagua Fault; GF: Gulf of
Fonseca; ESFZ: El Salvador Fault Zone; TF: Tiscapa Fault [30-36]; position of the Middle American trench [12]. The inner
rectangle delimitates the geographical window employed for the seismic catalogue (11.0
-16.5
N and 85.5
-92.0
W), the lakes

are represented by the gray color areas.

1021

magnitude scales, namely M
W
, M
S
, m
b
, M
D
and M
L
. A
new empirical relationship amongst magnitudes
M
W
-M
S
, M
W
-m
b
and M
W
-M
D
have been calibrated
specifically for El Salvador.
The catalogue includes the solutions of the local and
regional seismological networks in El Salvador and
Central America. Events were also separated as
subduction events in interface and intraplate shocks in
order to keep with actual requirements in seismic
hazard estimations. An ASCII (American Standard
Code for Information Interchange) format version of
the catalogue is available that also lists whether or not
other natural phenomena such as volcanic eruptions,
tsunamis or eclipses coincided with the occurrence of
the earthquakes (as shown in Appendix).
2. Data Sources
The source of the information used to compile the
catalogue is presented in Table 1. Events within this
geographic window are delimited by the coordinates
11.0
-16.5
N and 85.5
-92.0
W which cover inland

events of El Salvador, Guatemala, Honduras and
Nicaragua and offshore earthquakes in the Central
American Isthmus (Fig. 1).
To extend the catalogue back in time (1500-1900),
the six studies were used, listed in Table 1 (from Code
U to N in this order of priority) which based the
determinations on macro-seismic data, public reports
and early historical records. The authors choose the
origin parameters listed in Ref. [9] for the Santa Marta
earthquakes in Guatemala of July 29, 1773 (M 7.6) and
its aftershock on December 12, macro-seismic
information [25, 38] confirm that the main shock were
felt in western El Salvador and Mexico suggesting that
this earthquake is related with the subduction zone,
while Guzmn-Speziale et al. [8] located the events in
the Guatemalan Grabens with moderate magnitudes of
6.0 and 5.7, respectively (as shown in Fig. 1 and
Section 4). Leeds [10] also assigned magnitude values
for volcanic eruptions in Nicaragua, however, in this
study, only the events for which there is an evidence of
strong shaking and damage of infrastructure are
included. Similarly, only subduction events listed in
Ref. [9] which has a seismogenic source classification,
namely, interface or intraplate subduction origin are
included in this catalogue. For the historical seismicity
related with the boundary of North American and the
Caribbean Plate in the Guatemalan Faults (Fig. 1) the
earthquake epicenter was adopted as the centroid of the
isoseismal maps as presented by White [4]. Leeds [10]
listed historical subduction earthquakes with
magnitudes above 7.0, only the earthquakes were
included for which depth range estimation is available
classifying them as either inter and intraplate shocks
(as shown in the earthquakes classification scheme in
the Section 4).
For the instrumental period, in most cases several
hypocentral determinations for one earthquake were
available constituting a initial merge catalogue of about
50,000 solutions, each set of apparent duplicates was
examined visually and a decision was taken as to which
event should be included as primary in the depurated
catalogue: as a first, priority the special studies were
chosen which recalculate the origin parameters for
destructive earthquakes (Code A to S) and as a second
priority, the standard world catalogues (Code I to C)
following the hierarchy listed in the Table 1 from top to
bottom. Local and regional catalogues determination
(code L and R) were used to include the shallow-crustal
earthquakes in the volcanic chain in Central America
that are not listed in the special studies covering the
years 1984-2009, using these catalogues aftershocks
sequences were included with the destructive
earthquakes for events located inside the volcanic belt
in Central America.
It is noted that the CASC solutions are also included
in the ISC catalogue.
3. Estimating Moment Magnitude
To homogenize the catalogue, a unique and
consistent magnitude is used for each event, the
moment magnitude M
W
was taken directly when it was

1022
listed in the CMT catalogue.
In the cases in which M
W
was not available, the value
of M
W
was obtained from the surface wave magnitude
M
S
making use of the empirical relationships suggested
by Ambraseys and Adams [12] between M
S
and the
seismic moment M
o
calibrated for Central America,
later the value of M
W
was obtained using the classical
relation of Hanks et al. [39] between M
W
and M
o

(Table 2). For the case of historical earthquakes, only
the minimum reported values (usually M
S
) were

Table 2 Empirical magnitude relations employed to homogenize the earthquake catalogue to the moment magnitude M
W
.
Empirical relation Reference
M
W
= 4.363 0.155 M
S
+ 0.077 M
S
2
This work
M
W
= 0.050 + 1.089 m
b
This work
M
W
= 2.189 + 0.660 M
D
This work
M
W
= 3.518 + 0.371 M
L
This work
log M
o
= 24.578 0.903 M
S
+ 0.17 M
S
2
+ 0.0043 (h 40) p, p = 0, h < 40 km, and p = 1, h > 40
km
Ambraseys and Adams [12]
M
W
= 2/3 log M
o
10.7 Hanks and Kanamori [39]
Log M
o
= 17.53 + 0.63 M
L
+ 0.10 M
L
2
Hanks and Boore [40]

Fig. 2 Orthogonal magnitude relationship between: (a) magnitudes M
W
and M
S
; (b) M
W
and m
b
; (c) M
W
and M
D
; (d) M
W
and
M
L
employing 233, 393, 198 and 154 events respectively, sigma () denotes the standard deviation, a comparison with other
empirical relations in the world is presented [40-44].
(a) (b)
(c)
(d)

1023

selected because the assessment of higher intensities
later converted to magnitudes could be have been the
influence from a diet of exaggeration of the local press
reports, conquerors and clerical authorities letters
asking for financial recovery support to political
governors in Spain. Since a specific depth value is
required in the empirical relationships of Ambraseys et
al. [12] between M
o
and M
S
, earthquakes listed in the
standard world catalogues with a fixed depth value of
33 km can not be incorporated directly in this scheme,
in this case, a empirical magnitude relationship was
employed developing in this work by applying a
orthogonal regression between 233 magnitude pairs of
M
S
and M
W
for all depths of the catalogue (Fig. 2 and
Table 2). Some historical destructive earthquakes prior
to 1900 lack of a depth value determination, if the
earthquake is located in the volcanic chain or in the
interface subduction zone according to the seismogenic
source delimitations developed in this work, a p = 0
was assigned based on the Ref. [12] relations. There is
one historical earthquake dated on December 19, 1862
with Ms = 8.10 cited by White et al. [9] that occurred in
the intraplate subduction zone based on the analyzed
macro-seismic information. The reported M
S
value was
adopted as a good approximation of M
W
for these
events.
When M
S
was not given, the body wave magnitude
m
b
was selected in order to calculate M
W
employing the
empirical relation developed in this work for 393 pairs
of magnitudes m
b
-M
W
via application of an orthogonal
regression scheme

(Fig. 2 and Table 2). When M
W
, M
S
,
and m
b
were not available, the duration M
D
was used as
listed in the solutions of the local DGOA and regional
CASC catalogues (Table 1) employing the empirical
relation developed in this work using 198 pairs of M
D

and M
W
(Table 2). Similarly, when the local magnitude
M
L
was only available, the empirical relation between
M
L
and M
o
proposed by Hanks and Boore [40] was
employed and later obtain the M
W
by using the cited
relation by Hanks and Kanamori [39]. The authors did
not employ the relation between M
L
and M
W
developed
in this work employing 154 pairs of magnitude due to
the high dispersion observed in the M
L
data and the
high deviation from relations of other parts of the world
[40-42].
The final depurated catalogue contains 2,584 events
for the period 1528-2009 covering focal depths of
0-304 km and in a magnitude range 5.0 M
W
8.1 (Fig.
1). In Tables 3 and 4, the contributions are presented
for each sub-catalogue as listed in Table 1 and basic
statistics of magnitudes conversions.
4. Tectonic Setting and Definition of
Seismogenic Sources Delimitation
Once the catalogue has been compiled, each event
must be associated to a seismogenic source in the
region and make the geometrical delimitations of each
of them. The six seismogenic sources established that
affect the study area based on tectonics and geological
features that have been already studied by different
authors [14, 15, 45-52]. From their studies, the
following sources are detected in El Salvador and
surrounding areas (Fig. 1):
Zone 1: volcanic chain: The seismicity of this source
is concentrated mainly in the upper part of the crust in
the Caribbean Plate (depth 25 km), and within a
nearly continuous belt of 20-30 km width along the
axis of the principal quaternary volcanoes. The
earthquakes in the volcanic chain frequently appear in
clusters with no discernible main shock (swarms).
Their location lays parallel to the subduction Middle
American Trench (Figs. 1, 3 and 4). Destructive
earthquakes in Central America are located in the DVF
(dominant volcanic front) composed by major
stratovolcanoes and calderas, located not more than
200 km from the subduction trench. More distant to the
north, an irregular zone of volcanic edifices formed in
the BVF (behind volcanic front) characterized by the
presence of small, basaltic volcanic, mostly
monogenic cinder cones and associated lava fields [30]
where the lowest level of seismicity is observed in
the area.

1024

Table 3 Number of events for each source of information regarding with epicenter location (see the Code in Table 1).
Code U G W E N A M S I P D H C L R Total
Number 1 44 44 12 2 21 86 104 1,773 407 5 26 2 24 33 2,584

Table 4 Magnitude contribution, the star (*) denotes magnitude conversion to moment magnitude M
W
from the original
values of M
S
, m
b
, M
D
and M
L
using the empirical relations listed in Table 2.
Magnitude M
W
*M
S
*m
b
*M
D
*M
L
Total
Number 388 1,020 1,156 8 12 2,584

White et al. [14] and more recently Corti et al. [34]
suggested that the seismicity in the volcanic chain
could be interpreted as the result of a right-lateral shear
zone, driving an oblique component of convergence
between the Caribbean and the Cocos Plates. However,
recent research suggests that the strain field along the
Central America Arc Volcanic Front is the result of a
slight compression along the convergent
Cocos-Caribbean margin, and the extension in the
back-arc region, along the Guatemalan-Honduras
Grabens of northern Central America (Zone 3 shown in
this section). This field is resolved a strike-slip faulting
along the volcanic belt [13, 53].
Faults with left lateral and right lateral mechanisms
are located perpendicular and parallel to the volcanic
chain axis, respectively. The last destructive
earthquake in El Salvador of February 13, 2001 (M
S
=
6.5) had a right lateral fault mechanism in the E-W
fault segment of San Vicente with evidence of surface
rupture [54]. The destructive earthquake in San
Salvador City on October 10, 1986 (M
S
= 5.4) had a
left lateral mechanism with N-E striking plane. The
increase of coulomb failure stress also enhances the
possibility of infrequent fault slip on arc-faults that are

Fig. 3 Epicenters (solid circles) and geometrical delimitations of the seismogenic Zone 1: volcanic chain, Zone 2: Guatemalan
faults and Zone 3: Honduran-Guatemalan grabens, the epicenters in the background seismicity are denoted by the open
circles.

1025

located between the volcanic centers [55] as the case of
the destructive earthquake in Nicaragua on December
23, 1972 (M 6.3) along the Tiscapa Fault [49, 56].
Strike slip earthquakes also take place in this area of
the Volcanic Chain in Guatemala (e.g, September 18,
1991 M 6.2). In El Salvador, Corti et al. [34] suggest a
late Pleistocene-Holocene slip rate of about 11 mm/yr
along Berlin City, while Lyon-Caen et al. [57]
proposed a similar value of 10 mm/yr of dextral slip
across the Mid-American Volcanic Arc. More recently
[58] suggested a slip rate of 4 mm/yr for the modern
volcanic zone in El Salvador. Fig. 3 shows the
geometrical delimitation and related epicenters for the
volcanic zone in the study area. A tsunami was reported
for the upper crustal earthquake of August 3, 1951 with
a M 6.15 located in the Golf of Fonseca.
Salazar et al. [59] employing strong motion data for
El Salvador confirmed that the source spectra match
those obtained from the point source model after
setting the stress drop of 75 bars for upper-crustal
inland earthquakes in the volcanic chain, and that the
seismic waves attenuate rapidly in the volcanic zone
with a faster rate of 10 times in comparison of the
subduction zone earthquakes. From this, the authors
confirm the source and path characteristics via
inspection of the isoseismal maps for destructive
inland-earthquakes in the country;
Zone 2: Guatemalan faults: These constitute the
boundary between North America Plate and the
Caribbean plate. Most of the events that have their

Fig. 4 Geometrical delimitation and focal mechanisms for shallow earthquakes, GF: Guatemalan faults (Zone 2), HGG:
Honduras and Guatemalan grabens (Zone 3), VC: volcanic chain (Zone 1), IFS: interface subduction (Zone 4), OR: outerise
earthquakes (Zone 6). The blue focal mechanism inside IFS: earthquakes with an inverse focal mechanism (T1) with slip vector
between 45
and 135
within 20
azimuth of the Central American Trench, purple focal mechanism inside OR: earthquakes
with an normal focal mechanism (N1) with slip vector between -45
and -135
within 20
azimuth of the Central American

Trench. Other focal mechanism belongs to strike slip faults (in VC and GF), and normal faults (HGG), the depth contours of
the upper part of the Cocos plate are presented based on SLAB 1 [60].


1026

Fig. 5 Epicenters (solid circles) and geometrical delimitations for the Zone 4: interface subduction and Zone 6: outer rise
earthquakes, epicenters that lay outside the geometrical delimitation correspond to the center of the maximum isoseismal
contours for historical earthquakes.

origin in the faults are shallow (less that 35 km). They
are characterized by left lateral strike slip mechanism.
The Motagua and Polochic faults cross Central
Guatemala in an arc concave to the north (Figs. 1, 3 and 4)
and it extends in the Caribbean Sea to the north east in
the area known as the Swan Fracture [30, 61]. The
maximum magnitudes listed in catalogue dated on July
22, 1816, August 4, 1856 and April 19, 1902 all with
magnitude 7.8 and 7.6, respectively. The largest
reported destructive earthquake in the last hundred
years in this boundary had a magnitude 7.5 on February
4, 1976, producing a surface rupture along 230 km of
the Motagua Fault. Franco et al. and De Mets et al.
[62, 63] estimated the North American-Caribbean
plate relative motion to 18-20 mm/yr. Plate boundary
deformation may have been localized further south
along the inactive Jocotn-Chamelecn fault zone in
northwestern Honduras, which has been fragmented by
the east-west extension of the Honduran-Guatemalan
north trending Grabens. Two tsunamis are reported in
the catalogue corresponding to this source dated on
August 4, 1856 (M 7.8) and another one on February 4,
1976 both hit the Gulf of Honduras [12, 64]. Fig. 3
shows the geometrical delimitation and related
epicenters for the Guatemalan earthquakes;
Zone 3: The Honduran-Guatemalan grabens: This
zone extends approximately from the Guatemalan
Graben to the Honduras Depression [31] where
infrequent shallow earthquakes (less than 35 km)
occurred in the Chortis Block (Figs. 1, 3 and 4), this
zone is characterized by extensional deformation
reporting normal focal mechanism. This seismogenic
source is delimited by the Motagua-Polochic fault
system to the west, the Swan Fracture zone to the north,
the Guayape Fault to the east and the volcanic chain to
the south, conforming the western rifted highlands
according to physiographic provinces suggested by
Marshall [35]. The largest reported earthquake in this

1027
zone occurred in October 15, 1743 and had an
estimated magnitude of 6.7. According to Ref. [36], the
Graben is bounded by en-echelon oblique-normal
faults which form part of the northwest-trending zone
of early Tertiary faults.
The most prominent extensional structure is the
Ipala Graben, situated in southeastern Guatemala and
northwestern El Salvador [30, 51, 52]. Three main
volcanoes are located in this Graben, namely, Ipala,
Suchitn and Chingo, there are also almost hundredth
of monogenetic volcanoes cinder cones with associated
basaltic and andesite basaltic lavas. The age of this
cones mostly are Holocenic to Pleistocene cluster
chiefly in domains of extensional tectonic regime
perpendicular to the DVF (dominant volcanic front)
[30]. Other important feature is the existing of a
rhyolitic domes in the center of the Ipala Graben, which
converts it in a bimodal volcanism [51, 52]. The Ipala
Graben is by far the largest of several areas and with
more volcanic cluster behind the volcanic front of
Central America, maybe because it is located in the
higher inflection of the Jocotn-Chamelecn fault
curvature. A shallow earthquake (depth 5 km) with M
L

4.6 ( M
w
4.3) occurred on December 20, 2006 in this
source producing severe damage in adobe houses and
engineering structures. This event occurred during a
swarm from Dec. 17, 2006 to January 8, 2007 with a
total of 157 felt earthquakes and a total of 1,119 events.
Casualties were not reported [65, 66]. Preliminary
analysis [66] indicates a normal focal mechamism for
the main event.
This seismogenic source also includes a zone of
infrequent earthquakes in an extensional tectonics in an
area bounded by the Guatemala faults to the north and
north-west (with its bo undary in the Jocotn
Chamelecn Fault [67]), the volcanic chain to the
southeast and the Honduras Depression on the east,
near the Trifinio forest in the junction of Guatemala,
Honduras and El Salvador in the western part of the
Ipala Graben.
Guzmn-Speziale et al. [8] has calculated a rate of
approximately of 8 mm/year E-W opening while
Rodrguez et al. [68] employed GPS measurements at
37 stations in Honduras and El Salvador suggesting 3.3
mm/year in Central Honduras, 4.1 in western Honduras
and 11-12 mm/year in southern Guatemala,
lvarez-Gmez et al. [53] suggested that the state of
the extensional stress of this wide Graben area is a
result of: (1) the drift direction of the eastern part of the
Chortis Block toward the east-northeast; (2) the
curvature of the Guatemalan Faults and a low degree of
coupling in the subduction Middle American Trench.
Fig. 3 shows the geometrical delimitation and related
epicenters for this seismogenic source.
The earthquakes in the Chortis Block also appear in
clusters with no discernible mainshocks. Normal-faults
earthquakes are reported in the Guatemala Graben and
the Mixco Fault to the West of Guatemala City in the
aftershock sequence of the destructive earthquake on
February 4, 1976 in the Motagua Fault [67];
Zone 4-5: Subduction: The Middle American trench
that runs about 150 km and almost parallel to the coast
(Figs. 1 and 5), where the Cocos plate submerges itself
underneath the Caribbean plate, reaching depths of 300
km below the Central America Isthmus [15, 45, 47, 48].
Convergence of these two plates takes place at a rate of
7-8 cm/year reaching maximum magnitudes of about
8.1. The shallow focus earthquakes along the inclined
seismic zone yield underthrust focal mechanisms
(Fig. 4) with depth less than 60 km [12], Zone 4 is
declared as an interface subduction. The focal
mechanisms of intermediate events (60 km < depth
120 km) with its origin beneath the trench axis indicate
that there is occurrence of normal faulting resulting
from initial flexure of the down going Cocos plate with
a steeply NE-dipping of an average of 45
(Fig. 6).
However, mixed-motions earthquakes with thrust and
strike slip indicate bending of the subducting slab at
deeper depths as such Zone 5 is declared as the
intraplate subduction zone (Fig. 7).

1028
Large intraplate shocks occurred in El Salvador
offshore on December 12, 1862 and on January 13,
2001 with magnitudes 8.10 and 7.70, respectively.
Panagiotopouls [69] suggested the occurrence of a
large earthquake of M 7.5 in El Salvador with the
highest probability magnitude value among all the
seismogenic sources in the Central American and
Caribbean Sea based on the time- and
magnitude-predictable model during the decade
1993-2002, the January 13, 2001 confirmed his
analysis. The largest interface event in the catalogue
offshore El Salvador is dated on September 7, 1915
with a magnitude 7.8 [25]. Sometimes the thrust
earthquakes are accompanied by destructive tsunamis,
as the case of the recent event in the coast of Nicaragua
on September 2, 1992 with M 7.60 [64] with an
abundant foreshock activity. The subduction
earthquake on March 5, 1719 with M 7.2 occurred
during a lunar eclipse with an abundant foreshock
activity producing severe damage in the towns of San
Vicente, San Salvador and Cojutepeque [38]. Contrary
to the subduction zone, the magnitudes of the
earthquakes in the volcanic chain are moderate. The
maximum magnitude observed in the last century in the
volcanic region was 6.9 on July 14, 1930 in Guatemala,
geographically associated with the E-W Jalpatagua
right lateral strike-slip fault. However, Canora et al. [58]
suggested a maximum magnitude of 7.6 in El Salvador
Front Zone (ESFZ in Fig. 1) with a recurrence interval
of about 750-800 years based on paleoseismological
investigations.
To delimit geometrically the interface and intraplate
zones, the depth determinations listed in this catalogue,
the focal mechanism listed in the CMT catalogue and

Fig. 6 Geometrical delimitation and focal mechanisms for ITS (intraplate subduction) Zone 5: normal focal mechanism (N1)
with slip vector between -45
and -135
within 20
azimuth of the Central American Trench, the depth contours of upper part
of the Cocos plate are presented based on SLAB 1 [60], the intraplate subduction zone is divided in three sub sources according
to the depth range of the subducted slab, namely (1) Zone 5a: 61-120 km (vertical stripes); (2) Zone 5b: 121-180 km (inclined
stripes); (3) Zone 5c: 181-304 km (vertical stripes).


1029

Fig. 7 Epicenters (solid circles) and geometrical delimitation of seismogenic Zone 5: intra-plate subduction (see cross sections
AB-AB, CD-CD and EF-EF in Fig. 8).

the geometrical features of the subduction trench and
the depth to the subducted slab (SLAB 1.0 model [60])
was employed simultaneously. The interface
subduction source is declared as an area geographically
delimited by the Middle American Trench to the south
[12] and the Central American Coast to the north which
incorporates the epicenters of events with a component
of thrust faulting with depth less than 60 km.
Employing the definition of Hauksson [70], a
reverse-slip vectors is identified as having rakes of 45

to 135
and within 20
of the strike of the Middle

America Trench [71]. The northern limit of the
interface subduction zone coincides with the southern
limit of the Volcanic Chain (Figs. 3 and 5). Some
historical earthquakes epicenter are located outside the
geometrical delimitation that occurred on 1648, 1609,
1765/10/24, 1773/07/29 and 1773/12/13 (Santa Marta
earthquakes in Guatemala), 1870/06/12, 1874/09/03
and 1915/09/07 and its aftershocks. The epicenter
location is assumed to be the center of the isoseismal
maps presented by the authors or where the maximum
intensity was reported amongst different observations
[9, 10].
On the other hand, the intra-plate source is delimited
geographically as an area that incorporates the
epicenters of events with a component of normal
faulting with depth greater than 60 km to the south and
the bottom of the subducted slab to the north [45]
(Fig. 7). Following the definition of Hauksson [70], a
normal-slip vectors is established as having rakes of
-45
to -135
and within 20
of the strike of the Middle

America Trench. When the earthquakes are located in
the overlapped inter-intraplate zone and have a lack of
focal mechanisms, these are classified as the interface
and intraplate earthquakes based on the depth value.
Interface earthquakes were considered with depth less
or equal than 60 km and intraplate events with depth
greater than 60 km. The overlapped zone is the result of
the inclination of the subducted slab. The rest of the
events are separated into the interface and intraplate
sources using the correspondent geometrical
delimitations based on epicentral determinations. A

1030
depth of 33 km is fixed by the world standard
catalogues when the depth is indeterminate based on
only on the first-arriving P-waves, the events are
assigned to the intra-plate subduction zone when the
epicenter lies in the overlapped inter-intraplate zone. In
order to have a realistic tectonic model for the
intraplate earthquake source, three sub sources have
been assigned according to the depth range of the
subducted inclined slab (Fig. 6), namely, (1) Zone 5a:
61-120 km; (2) Zone 5b: 121-180 km; (3) Zone 5c:
181-304 km.
Salazar et al. [59] employing strong motion data for
El Salvador confirmed that the source spectra match
those obtained from the point source model after
setting the stress drop of 300 bars for intra-plate
earthquakes and that the geometrical spreading factor
accounts for most of the attenuation. The source and
path characteristics can be defined via inspection of the
isoseismal maps for destructive subudction
earthquakes in the country;
Zone 6: Outer-rise: These earthquakes occurred
south of the subduction trench in front of the Central
America coast at shallow depth less than 50 km.
Outer-rise earthquakes are caused by stresses in the
subducting oceanic plate induced by bending as the
plate enters the trench, flexure of the plate elevates the
sea floor, creating an oceanic feature that dips parallels
to the oceanic trench. The tensional stresses in the
oceanic crust can create large normal faults (Figs. 4
and 5). Crustal stresses caused by earthquakes on the
interplate thrust fault in subduction zones can also be
transferred to the outer rise, triggering earthquakes on
normal faults that are already close to failure.
Besides the interface subduction zone, this
seismogenic zone is identified as a potential source of
tsunamis, these earthquakes are infrequent but their
damage can be devastated as the cases of the
earthquakes of 1933 in Sanriku with a M 8.6 (3,000
victims) and significance damage in the Hawai Island.
An earthquake in 1977 in Sumba with a M 8.3
produced 189 deaths in Indonesia. On September 29,
2009, an earthquake occurred in front of Samoa Island
next to the Tonga Trench with M 8.1 with a
considerable amount of victims.
The background seismicity is located at the north of
Guatemala with a very low level of disperse earthquake
locations and also located in Central west Nicaragua,
characterized by earthquakes with relative low
magnitude and shallow depth (Fig. 3).
The 2,584 events in the earthquake catalogue were
separated into the six seismogenic sources and
background seismicity as explained before, yielding
2,071 for the subduction zone (974 for the interface and
1,097 for the intraplate zones), 155 events for the
Volcanic Chain, 78 for the Guatemalan faults, 79 for
the Honduran-Guatemalan Grabens, 187 for the
outer-rise and 14 for the background seismicity. Figs. 3-7
illustrate the geometrical delimitation for both shallow
and deep zones covering the subduction, the
upper-crustal volcanic zone, transform and intra-plate
faulting and extensional graben zones. This
configuration comprises the subduction trench to the
south and the deepest part of the Cocos Plate to the
north. The geometrical delimitation for upper-crustal
seismicity comprises related epicenters and main
geological structures such as volcanoes and seismic
faults. The background seismicity is introduced in
order to keep shallow events that could not be
associated with a specific seismic source. Typical cross
section showing the subduction and volcanic chain
events are presented in Fig. 8 and the focal mechanisms
for each seismogenic sources are presented in Figs. 4
and 6.
5. Seismicity Evaluation
5.1 The Catalogue Completeness and Recurrence
Relationships
In this section, one specific aspect of quality control
for the catalogue is addressed, namely the assessment
of the magnitude of completeness M
c
which is
defined as the lowest magnitude at which 100% of the


1031

Fig. 8 Hypocenter cross sections AB-AB, CD-CD and EF-EF (Fig. 7) in the Central American Isthmus (1898-2009).

earthquakes in a space-time volume are detected [72].
A number of methods have been applied to test the
completeness of the catalogues, based on the
assumption that earthquakes of magnitude greater than
a stated value follow a Poisson probability distribution
suggesting de-clustering of the catalogue before
computing the completeness method, in this regard, the
foreshocks and aftershocks were removed from the
catalogue. The method of Gardner et al. [73] was
applied to decluster the catalogue. After applying the
algorithm, the final catalogue contains 894
independent events with magnitude M
w
between 5.0
and 8.1 covering the years 1528-2009. To compute the
completeness period, it was suitable to apply the visual
method of Tinti et al. [74] for each seismogenic source,
suggesting that the earthquakes within a magnitude
interval (i.e., 5.0-5.5, 5.5-6.0, 6.0+, etc.) are complete
back to a specific year if the cumulative annual
numbers of events for this magnitude range are
approximately linear back to that date (Fig. 9). A
magnitude bin size of 0.5 units has been used for all
zones except for the volcanic chain, for which a
magnitude bin size of 0.25 was used to revise the
non-linear trend in the G-R (Gutenberg-Richter)
recurrence relationship [3, 27, 28, 75]. In Table 5, the
estimated period of completeness is presented with the
correspondent magnitude M
c
for each seismogenic
source of this catalogue.
In the attainment of a G-R recurrence relationship,
the regional seismicity is characterized by adopting the
expression:
BM A N = ) ( log
10
(1)
where, N represents the accumulative number of
earthquakes per year above magnitude M, and A and B
set two constants derived from regression analysis. It is
worth mentioning that the computation of the N values
must be consistent with the completeness period for
each magnitude bin. Constant B is the measure of the
relative abundance of large to small shocks and
characterizes the degree of fracture of the crust and it
could be related with the thermal gradient constituting
the seismogenic source or the degree of coupling
between two plates. Figs. 10 and 11 illustrate the
Gutenberg-Richter recurrence relationships after
applying the completeness analysis and Table 6 shows
the A and B recurrence values. It is noted that the two
options are presented for the G-R recurrence
relationships based on the clustered and declustered
catalogue (Fig. 10).
When using the declustered catalogue, a B value
yields 0.84 and 0.94 for the intraplate subduction zone
for the range depth of 60-120 km and 121-180 km,
respectively, confirming the results of the work [76]
for Central America employing data for the years
1964-1994. In the case of the interface zone the high B
value of 1.01 coincides with the results of Ghosh [77]
for the subduction zone, in Costa Rica and Benito et al.
[24] in Central America suggesting a moderate seismic

1032
coupling between the Cocos-Plate and the Caribbean
plate confirming also the previous analysis of
lvarez-Gmez et al. [53] who affirm a slight
compression between the two plates. What is not clear
at this time, if such slight compression yields
maximum magnitudes below 8.0 in the interface and
what would be the possibility of having a megathrust
event in the Central American trench of M 9.0 as the
recent cases in Japan in 2011 and Sumatra earthquakes
in 2004. During the preparation of this article, two
earthquakes occurred offshore El Salvador and
Guatemala; both inside of our geographical window of
study in the interface subduction zone with M 7.3 and
M 7.4 on August 27 and November 7, 2012,
respectively. Employing the recurrence relationship
with the correspondent standard deviation, the
recurrence interval for a magnitude M 7.3 and above
yields 21 years. Taking the last century as the most
reliable set of data, three earthquakes with M 7.3 and
above are registered in our catalogue for the period
1915-1992. By extending the period from 1915 to 2012
(98 years) to incorporate the last earthquakes on Aug.
2012 and Nov. 2012, it will obtain a recurrence interval
of almost 20 years (98 yr over 5 earthquakes), nearly
the recurrence value employing the G-R relationships
for this zone.
The tri-linear G-R relationships in the volcanic chain
reflects for the first part (B
1
= 1.02) a large number of
small earthquakes related to destructive ones, the
middle part with very low B
2
value reflects the effects
of clusters or the abundance of intermediate size
earthquakes (for the declustered catalogue B
2
= 0.42)
and the third part suggests small number of moderate to
big shocks related to the clusters and the small shocks
(B
3
= 1.46). It is worth mentioning that same tri-linear
trend is obtained when using a 0.5 magnitude bin size
(Fig. 10). These results for crustal seismicity differ
greatly with the linear trends in the G-R relationships
obtained by Benito et al. [24] for the volcanic chain in
Guatemala, El Salvador and Nicaragua with B values

Fig. 9 Illustrative example of the method of Tinti et al. [74]
to compute the completeness period for the whole intraplate
subduction zone, the method suggests that the earthquakes
within a magnitude interval (i.e., 5.0-5.5) are complete back
to a specific year if the cumulative annual numbers of events
for this magnitude range are approximately linear back to
that date, the number of earthquakes in this magnitude
range is utilized to obtain the Gutenberg-Richter relations
presented in Fig. 10 and Table 6.

Table 5 Estimation of completeness period employing the visual cumulative method for all the seismogenic sources employing
0.25 and 0.5 of magnitude bin size for the seismogenic sources (1) and (2-6), respectively.
Zone
M
W

5.13 5.38 5.63 5.88 6.13 6.38 6.63 6.88 5.25 5.75 6.25 6.75 7.25 7.75 8.25
1 1984 1991 1838 1856 1839 1847 1717 1930 - - - - - - -
2 - - - - - - - - 1978 1928 1750 1907 1785 1816 -
3 - - - - - - - - 1961 1918 1689 1743 - - -
4 - - - - - - - - 1980 1975 1964 1867 1776 1765 -
5 - - - - - - - - 1970 1960 1930 1898 1916 1907 1862
6 - - - - - - - - 1975 1978 1918 1911 1907 - -
All sources - - - - - - - - 1961 1838 1689 1717 1776 1765 1862


1033

Fig. 10 Gutenberg-Richter (G-R) recurrence relationships (Table 6), in the case of the volcanic chain (Zone 1), the triangles
correspond to magnitudes bin size of 0.5 and the circles to magnitudes bin size of 0.25 for which the G-R relationships was
obtained, denotes the standard deviation in the regression. N denotes the number of earthquakes per year ad M
w
the moment
magnitude. The G-R relationships are presented for both the clustered (with foreshocks and aftershocks) and the declustered
catalogue (removing the foreshocks and aftershocks), see the correspondent legend in Zone 4 graph.
5 5.5 6 6.5 7 7.5 8
0.01
0.1
1
10
100
Clustered catalogue
Clustered Regression
Declustered catalogue
Declustered Regression
5 6 7 8 9
M
W
0.01
0.1
1
10
100
N
5.2 5.6 6 6.4 6.8 7.2
0.01
0.1
1
10
N
5 6 7 8
0.01
0.1
1
5.2 5.6 6 6.4 6.8
0.001
0.01
0.1
1
N
Zone 4: Interface Subduction
Zone 5: Intraplate Subduction
Zone 1: Volcanic Chain
Zone 2: Guatemalan Faults
Zone 3: Honduran Guatemalan
Grabens
B=0.68
B=1.62
B=1.25
B=1.05
B
1
=1.76
o=0.14
o=0.23
o=0.15
o=0.15
o=0.085
B
2
=0.47
B
3
=1.58
5 5.5 6 6.5 7 7.5
M
W
0.01
0.1
1
10
Zone 6:Outer rise
B=1.14
o=0.18
B
1
=1.02
B
2
=0.42
B
3
=1.46
o=0.10
B=1.01
B=0.89
o=0.06
o=0.06
B=1.38
o=0.12
B=0.59
o=0.19
B=0.98
o=0.23

1034

Fig. 11 Gutenberg-Richter (G-R) recurrence relationships for subsources in the intraplate subduction (Zone 5) zone for the
declustered catalogue (Table 6).

Table 6 Constants A and B for the Gutenberg and Richter (G-R) relationships (log N = A BM), M
min
: minimum magnitude,
M
max1
: maximum magnitude listed in the catalogue; M
max2
: maximum magnitude based on paleoseismological investigations
[56] for Zone 1 (volcanic chain*) or M
max1
+ an increment of 0.5 units for zones 2-6;
min
: activity rate (number of earthquakes
per year) above M
min
. Noted that for the intraplate subduction Zone 5, the G-R relationships is presented for the whole
intraplate zone (5) and the intraplate subsources, namely, Zone 5a: 61-120 km; Zone 5b: 121-180 km and Zone 5c: 181-304 km
depth (Fig. 6).
Zone
Variable
A
1
B
1
A
2
B
2
A
3
B
3
M
min
M
max1
M
max2

min
A B
1 5.10 1.02 1.72 0.42 8.05 1.46 5.0 6.9 7.6* 1.0 - -
2 - - - - - - 5.0 7.6 8.1 0.66 2.77 0.59
3 - - - - - - 5.0 6.7 7.2 1.20 6.98 1.38
4 - - - - - - 5.0 7.8 8.3 10.00 6.05 1.01
5 - - - - - - 5.0 8.1 8.6 12.02 5.53 0.89
5a - - - - - - 5.0 8.1 8.6 6.61 5.02 0.84
5b - - - - - - 5.0 7.5 8.0 2.95 5.17 0.94
5c - - - - - - 5.0 7.8 8.3 1.12 3.59 0.71
6 - - - - - - 5.0 7.1 7.6 2.57 5.31 0.98

between 0.67-0.97. It is noted that non-linear trends in
the G-R for the volcanic chain have been obtained
before [3, 27, 28, 75]. In order to use the catalogue in
the calculation of the seismic hazard, the catalogue
must be declustered (Section 5.1) by employing a
Poisson probability function. However, applying the
de-cluster algorithm [73] to the whole catalogue
eliminates destructive earthquakes in the volcanic
chain as an interpretation of aftershock of offshore
subduction earthquakes (i.e., September 1915-June
1917, January 2001-February 2001). However, a low B
value for intermediate earthquakes is obtained even
applying the descluster scheme in the volcanic chain
(Fig. 10), the G-R trilinear trend is observed for both
the clustered and declustered catalogue suggesting the
same seismicity pattern but with a decrease in the
activity rate of earthquakes after the declustering
process for the whole range of magnitudes. Note that
the decluster scheme affects the activity rate for only
small and moderate magnitudes for the rest of
seismogenic sources.
Employing the G-R relationship for the volcanic
chain using the clustered catalogue, results yield log N
= 9.12 1.58 M for M > 6.2, and for a M 7.6 the
recurrence interval results in 773 years, nearly the
750-800 years suggested by Canora et al. [54] for this
maximum magnitude based on paleoseismological
investigations. When employing the G-R relationships
for the declustered catalogue the recurrence interval of
this earthquake yields 1,000 yr. A similar high B value
5 6 7 8
M
W
0.01
0.1
1
10
N
5 6 7 8
M
W
5 6 7 8
M
W
B=0.84 B=0.94
B=0.71
Zone 5a:
Intraplate subduction 60-120 km
Zone 5b:
Zone 5c:
o=0.09 o=0.11
o=0.07

1035
of 1.38 is obtained for the Honduran Guatemalan
Grabens and coincides with the short length of the
geological faults in the broad area of extensional
tectonics.
The lowest B value of 0.59 yields in the
Guatemalan Faults suggests a low degree of fragility
represented by faults with hundreds of kilometers in
length in northern Central America, however, there is
a lack of events in the 6.0-6.5 magnitude bin in this
source including just two earthquakes for the whole
period of the catalogue suggesting a similar tri-linear
trend for strike slips earthquake regimes in Central
America. While the tri-linear trend in the volcanic
chain is caused by the abundance of moderate
earthquakes in clusters, the tri-linear trend in the
Guatemalan faults is caused by the lack of seismicity
of medium size events. The same suspicious tri-linear
trend is observed for the outer-rise Zone 6 and yields
the largest standard deviation in the G-R relationships
amongst all the seismogenic sources (Fig. 10).
5.2 Earthquake Interactions
The catalogue can be used to identify possible
earthquake interactions among the seismogenic
sources that affected the region. For example, in 1982
the Honduran Grabens, the Subduction Zone and the
Volcanic Chain produced earthquakes of M 5.3-5.6
(April 27 and September 29), M 7.3 (June 19) and M
6.1 (December 1), respectively. Besides, after all large
subduction earthquakes in the twentieth and the current
century, destructive volcanic chain events have
occurred within less than four years [37] as the cases of
the subduction earthquake in 1915 and the subsequent
earthquakes in 1917 in San Salvador and the
Guatemala Graben, the sequence of volcanic chain
earthquakes in Jucuapa-Chinameca and Gulf of
Fonseca in 1951 were preceded by the interface
earthquake in Guatemala offshore on October 23, 1950.
Martnez-Daz et al. [33] investigate the existence of a
mechanism of static stress triggering driven by the
interaction of normal faults in the Cocos Plate and
strike-slips faults inside the volcanic arc of El Salvador.
The results suggested causative relationships between
both kinds of earthquakes as the destructive ones
occurred in January-February 2001. Earthquake
interactions are reported also for the Guatemalan
earthquake of February 4, 1976 and the aftershock
sequences located in the Guatemalan Grabens [67].
Also in Managua Nicaragua, the January 8, 1968
earthquake of M 5.0 precedes the destructive
earthquake on December 23, 1972. It is not clear how
to incorporate the earthquake interactions in future
seismic hazard assessment for El Salvador and
neighboring countries.
6. Seismic Hazard Implications
In order to test the reliability of this catalogue, the
authors performed a preliminary seismic hazard in
terms of hazard maps and uniform hazard spectral
accelerations for the horizontal component of
motion.
Two different computation methodologies have
been adopted: the standard and classical
Cornell-McGuire approach based on the definition of
appropriate seismogenic sources and the Poisson
probability distribution [78, 79] and the zone-free
Kernel approach developed by Woo [80], which
overcomes the ambiguities related with the definition
of the seismic sources employing directly the location
and size of the earthquakes listed in our catalogue.
Since the zone-free method is not based on a
Poissonian model of earthquake recurrence, the entire
catalogue which includes foreshocks, aftershocks and
clusters may be used in the hazard computation. In this
sense, the use of the free-zone probabilistic approaches
is needed as an alternative method. As a first instance,
the inclusion of aftershocks in the hazard assessment
can be justified due to catastrophic consequences after
the main shock. Just for citing some examples, in the
case of the destructive earthquake of October 10, 1986
in San Salvador City, an aftershock on October 13
produced the collapse of a children hospital and other

1036
buildings in the city. The aftershock sequence of
January 13, 2001 produced the collapse of many
residential houses along El Salvador [37], such
situation has been observed in several parts of the
world as in the recent earthquakes in Haiti (2010),
Chile (2010) and New Zealand (2010, 2011).
Seismic hazard calculations for the Cornell method
have been performed using EZ-FRISK 7.62 code. The
Kernel method has been applied using a modified
version of the original numerical FORTRAN code
KERFRACT developed by Woo [80].
The seismic loads are calculated for: (1) stiff ground
conditions (NEHRP site classification B) and flat
topographic surface representing outcropping ground
conditions; (2) 5% structural damping of the critical; (3)
two return periods RP = 475 and 2,475; (4) three
spectral periods (PGA, 0.2 s and 1.0 s); (5) 1,947
computational points corresponding to a resolution of a
spatial grid of 0.050 degrees with an average spacing of
the nodes of about 5.5 km; and (6) a truncation value of
3 sigma (3) in the GMPEs (ground motion prediction
equations). The attenuation relationships of Kanno et al.
[81] was used to compute the hazard and made use of
the seismogenic zone delimitations and the G-R
relationships developed in the previous sections. The
number of earthquakes within different magnitude
ranges expected in each seismogenic zone has been
obtained using the bounded G-R relationships [82]. For
the case of the free zone method, the catalogue is
divided in two sub catalogues, namely: (1) crustal
events; (2) subduction zone events that are separated in
interface and intra plate events. This subdivision was
necessary because the depths of the shallow and deep
zones can not be averaged because this would result in
an intermediate unrealistic depth. The uncertainties in
epicenter location and magnitude are taken from Ref.
[44] and the completeness periods were adopted for all
the sources listed in Table 5. The results of the two
probabilistic methods are averaged with equal weights
employing the maximum magnitudes M
max1
and M
max2

listed in Table 6.
Figs. 12-14 show the seismic hazard contour maps
developed for El Salvador in terms of PGA with 10%
probability of exceedance in 50 years (RP = 475 years),
SA at 0.2 s and 1.0 s with 2% probability of exceedance
in 50 years (RP = 2,475). PGA with RP = 475 years
has been considered since it is the widely used
parameter to express the results of the PSHA
(probabilistic seismic hazard assessment), while the
computing of the spectral acceleration at periods of 0.2
s and 1.0 s and RP = 2,475 years is consistent with
International Building Codes [83, 84]. The covariance
COV is the standard deviation of the estimated PGA
(or spectral ordinate) at each point divided by the mean
value at that point (COV = /) and tells you the
relative breath of the Gaussian distribution used to
represent the variability for the two probabilistic
methods and correspondent maximum magnitudes
used in the hazard assessment [85].
The covariance for the PGA and the period of 0.2 s
varies from 0.04 to 0.14 inland El Salvador, yielding a
high level of reliabiltiy in our hazard calculations
and also for our catalogue. For the period of 1.0 s, a
level of covariances 0.14 to 0.24 (Table 7) was
observed.
A comparison between the UHS (uniform hazard
spectrum) computed for San Salvador Metropolitan
Area for RP = 475 and 2,475 years and the design
response spectrum recommended by the North
American Building Codes [83, 84] for rock site
conditions, site class B, is provided in Fig. 15. It is
noticed that both probabilistic methods yield almost the
same spectral values (Fig. 15) as expected according to
the calculated covariance in the maps, it is worth
nothing that the three linear trends in the G-R
relationship for the volcanic chain were weighted
according to the probability of magnitude occurrence
yielding a total of 1.0 when summing all magnitudes
bin contributions employed in the Cornell hazard
analysis computation (e.g., for the B
1
, B
2
and B
3
parts
the weights were calculated as 0.75, 0.09 and 0.16,
respectively). The United States spectral shape has


1037

(a)

(b)
Fig. 12 The seismic hazard contour map and its uncertainty: (a) seismic hazard map for the peak ground acceleration (g) and
475 years return period; (b) acceleration uncertainties in terms of covariance (COV).

(a)
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.22
0.26
0.3
0.34
0.38
0.42
0.46
0.5
Pacific Ocean
Honduras
Guatemala
EL SALVADOR
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.04
0.065
0.09
0.115
0.14
0.165
0.19
0.215
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.95
1.1
1.25
1.4
1.55
1.7
1.85
2
EL SALVADOR
Pacific Ocean
Honduras
Guatemala

1038

(b)
Fig. 13 The seismic hazard contour map and its uncertainty: (a) seismic hazard map for spectral acceleration (g) at the period
of 0.2 s and 2,475 years return period; (b) acceleration uncertainties in terms of covariance (COV).

(a)

(b)
Fig. 14 The seismic hazard contour map and its uncertainty: (a) seismic hazard map for spectral acceleration (g) at the period
of 1.0 s and 2,475 years return period; (b) acceleration uncertainties in terms of covariance (COV).
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.04
0.07
0.1
0.13
0.16
0.19
0.22
0.25
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.36
0.41
0.46
0.51
0.56
0.61
0.66
0.71
EL SALVADOR
Pacific Ocean
Honduras
Guatemala
-90 -89.5 -89 -88.5 -88 -87.5
13
13.5
14
14.5
0.14
0.172
0.204
0.236
0.268
0.3
0.332
0.364

1039
0 1 2 3 4 5
0.5 1.5 2.5 3.5 4.5
Period (s)
0
0.4
0.8
1.2
0.1
0.2
0.3
0.5
0.6
0.7
0.9
1
1.1
1.3
1.4
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)
AVERAGE
Woo Mmax
1
Woo Mmax
2
Cornell Mmax
1
Cornell Mmax
2
IBC 2012
0 1 2 3 4 5
0.5 1.5 2.5 3.5 4.5
Period (s)
0
0.5
1
1.5
2
2.5
0.25
0.75
1.25
1.75
2.25
A
c
c
e
l
e
r
a
t
i
o
n

(
g
)
AVERAGE
Woo Mmax
1
Woo Mmax
2
Cornell Mmax
1
Cornell Mmax
2
IBC 2012

(a) (b)
Fig. 15 Design elastic response spectra for 5% of critical damping for San Salvador Metropolitan Area (13.70
N, 89.21
W)
for (a) 475 years return period; (b) 2,475 years return period.

Table 7 Degree of reliability for different values of
covariance COV.
Covariance Degree of reliability
0 COV < 10% Very high
10% COV< 20% High
20% COV< 30% Medium
30% COV < 40% Low
COV 40% Very low

been scaled to the spectral acceleration at 0.2 s and 1.0
s computed in this study for San Salvador employing
IBC formulas.
It is of practical interest to see that the spectrum
calculated in this study is not totally in good agreement
with the elastic response spectrum [83, 84], especially
for short period components less than 0.3 s. The
differences are attributed due to the abundance of
earthquakes in the volcanic chain for moderate shallow
magnitudes yielding an underestimation of seismic
loads when applying IBC spectral shapes factors. It is
suggested that such differences are the consequences of
employing spectral shapes calibrated for California
seismicity in areas of subduction and volcanic
environments. The level of uncertainty increases at
longer periods and the correspondent maximum
magnitudes in the analysis.
7. Conclusions
El Salvador and surrounding areas are characterized
with a high seismic activity yielding about 40
earthquakes per year with magnitude 5.0 or above that
generates in six seismogenic sources in the region with
a whole variety of focal depth, fault plane solutions and
expected maximum magnitudes. In this regard, the
most critical aspect in the upper-crustal seismicity
generated in the volcanic chain is the possibility of a
M
max
of 7.6 every 750-800 years according to the new
paleoseismological information and its coincidence
with the recurrence time for these events employing
our G-R recurrence relationships. Long return periods
of earthquake ground motion (e.g., 4,975 years that
corresponds to 1% of probability of exceedance in 50
years life time of the structure) might be considered the
maximum possible earthquake and future critical
facilities in the country.
Despite large interface thrust earthquakes have not
been reported in El Salvador and neighboring Central
American countries since historic times, the worldwide
seismic activity trend suggests that large megathrust
events of magnitude 8.0-9.0 are possible in subduction
trenches worldwide (e.g., recent Sumatra and Japan

1040
Earthquakes in 2004 and 2011, respectively). In this
regard, it is recommended to introduce a characteristic
models for this seismogenic source based on slip rates
and the inclined geometry of the slab beneath Central
America in the future hazard calculations.
This research suggests that either the interface
subduction Trench, the outer-rise zone offshore or the
Guatemalan faults constitutes clear sources for
triggering tsunamis. However, tsunamis have been also
reported due to upper crustal earthquakes located in the
Gulf of Fonseca in El Salvador.
The earthquake catalogue could be used to calibrate
time-space empirical formulas for main-aftershock
sequences for El Salvador and the surrounding areas.
The local magnitude calculations for Central America
deserve further revision at the monitoring centers in the
Isthmus (Fig. 2) observing a high dispersion in the
actual data and a very high deviation when comparing
with other parts of world.
The horizontal peak ground acceleration expected on
stiff ground, with a 10% probability of exceedance in
50 years, which corresponds to 475 years return period,
ranges between 0.26 g and 0.55 g. The highest hazard is
observed at the western coast of the country
presumably by the effects of the close megathrust
interface zone, the effects of the 1915 earthquake
interface subduction zone with M 7.8 in the free zone
method yields a high influence in the hazard results.
The low values of covariance obtained in the seismic
hazard maps employing zone and free-zone methods
for 475 and 2,475 years return period suggest an
acceptable degree of reliability for both the
seismogenic source delimitations and the earthquake
catalogue itself employing appropriate weights for the
tri linear G-R relationships in the volcanic chain. At the
same time the low error observed suggests that both
methods can be used for the hazard calculation in the
country respecting their assumption and limitations for
the earthquake probability of occurrence.
When using the free-zone method, special caution
must be taken into account regarding with the
epicenters errors of historical earthquakes for the
subduction zone, since the epicenters sometimes
correspond to the center of the isoseismal curves inland
instead being offshore Central America (e.g.,
epicentral error of 30 km were used for interface
earthquakes prior 1900), instead, a recalculation of the
epicentral locations for subduction historical
earthquakes offshore is needed.
Employing the IBC formulas scaling the spectral
values for 0.2 s and 1.0 s in the maps must be doing
with caution for El Salvador and Central American
countries. The authors propose that the seismic
coefficients must be calibrated for the volcanic and
subduction conditions inherent to the region if the
future codes would anchor the spectral shapes using
control periods and its correspondent spectral
acceleration level.
The complete catalogue is available in ASCII format
upon request to the correspondent author and its format
is listed in the Appendix. An advantage of the
catalogue is that for each earthquake it is possible to
identify the original source of information where the
data was taken according to the code listed in the left
column in Table 1, making this easy to use, revise and
update.
The authors expect that the catalogue can be used,
updated and corrected according to the suggestions
presented in this report as new information become
available, for both historical and instrumental
seismicity. Salazar et al. [86] also compiled
macro-seismic information from the newspapers of the
National Library of El Salvador and the Indian Sevilla
Archives in Spain that could help to complement
previous historical earthquake research in the region.
As a parallel research work to the seismicity evaluation,
new GMPEs (ground motion prediction equations)
must be developed for El Salvador and neighboring
Central American countries taking into account the
effects of surface geology on the seismic motion.
It is expected that this information will contribute to
the mitigation of the seismic risk in the country if new

1041
and proper seismic hazard methodologies sustain the
tectonics features and the seismicity patterns observed
in the catalogue.
Acknowledgments
Jon de Cortina at Central American University UCA
made everything possible to publish the first version of
the catalogue on 2003 with M
s
magnitude for the period
1898-2001 and made possible to retrieve information
from the Indian Sevilla Archives through Agustn
Udas in Spain. Dimitri Papastamatiou and Julian
Bommer worked with the corresponding author in the
initial ideas of depuration for the catalogue in
1996-1997 while working in El Salvador at UCA and at
the Imperial College of London, England. Nicolas
Ambraseys from the Imperial College provided us first
order material from his historical archives in Central
America. Claire McQueen provided in 1996 an
electronic version of the IPGH catalogue of Tanner and
Shepherd [19].
Manuel Daz from General Directorate of
Environmental Observatory DGOA (El Salvador),
Carlos Redondo and Lepolt Lnkimer from the CASC
(Central American Seismic Centre) kindly provided us
the local and regional catalogues for El Salvador and
Central America (1994-2001 and updates) Isthmus
during the Ph.D. course of the corresponding author
and already used in Ref. [28]. Rodolfo Torres from
the seismology section of DGOA in El Salvador
provided the preliminary focal mechanism for the
event in the Ipala Graben in December 2006 and the
correspondent seismic swarm information. Carlos
Caas-Dinarte sent to the corresponding author the
first draft of his historical and macroseismic
earthquake compilation in 2002. Wilfried Strauch,
from INETER, assisted in hypocentral determinations
for Nicaraguan earthquakes in the volcanic chain.
Hector Monzn suggested key literature and
inconsistencies in historical earthquakes among the
different sources of information especially for
Guatemala seismicity.
Thanks must be expressed to Gordon Woo (RMS,
London) to allow us to use his KERFRACT Fortran
Program for Seismic Hazard Calculations. Maps have
been prepared using ESRI ArcMap 10.1 (Arc View)
Geographic Information System and Surfer 8.0.
Thanks also to Tony Soeller (University of California,
Irvine US) kindly provide and give advice to us
regarding with his visual basic script for plotting
moment tensor solutions using ARCGIS. Cesar Elas
and Celina Cruz through OPAMSS and the Spanish
Cooperation Agency supported the updated seismicity
evaluation based on Salazar [28].
The computer program EZ-FRISK 7.62, a
software for earthquake ground-motion estimation,
developed by Risk Engineering Inc., has been used for
the Cornell-McGuire seismic hazard computations.
This catalogue has been developed since 1995 at
several institutions to which we greatly thank for their
unconditional support: Complutense University and
Geographical National Institute in Madrid, Spain
(1995), Central American University UCA in El
Salvador (1996-1998,) Imperial College of London,
England (1996-1998), Tokyo Institute of Technology,
Japan (2001-2003) and Seismic Research Centre, UWI,
Trinidad, West Indies (2010-2013).
References
[1] M. Martnez, Cronologa ssmica y eruptiva de la
Repblica de El Salador a partir de 1520, Departamento de
Sismologa, CIG (Centro de Investigaciones Geotcnicas),
Ministerio de Obras Pblicas MOP, San Salvador, 1980.
[2] S.T. Algermissen, S.L. Hansen, P.C. Thenhaus, Seismic
Hazard Evaluation for El Salvador, Report for the US
Agency for International Development, 1988, p. 21.
[3] C.S. Alfaro, A.S. Kiremidjian, R.A. White, Seismic
Zoning and Ground Motion Parameters for El Salvador,
Report No. 93, The John A. Blume Center for Earthquake
Engineering Research, Stanford University, 1990.
[4] R. White, Catalogue of Historic Seismicity in the Vicinity
of the Chixoy-Polochic and Motagua Faults, USGS Open
File report 84-88, Guatemala, 1984.
[5] S.K. Singh, C. Gutirrez, J. Arboleda, M. Ordaz, Peligro
ssmico en El Salvador, Universidad Autnoma de Mxico,
Mxico, 1993.
[6] Resesco, Reglamento para la Seguridad Estructural de las

1042
Construcciones, Ministerio de Obras Pblicas, El Salvador,
1994.
[7] D. Muoz, A. Udas, The earthquake of San Salvador,
Central America, of 21 April 1594: The first questionaries
on the damage of an earthquake in the western Hemisphere,
Bull. Seismol. Soc. Am. 86 (2006) 1538-1544.
[8] M. Guzmn-Speziale, C. Valds-Gonzlez, E. Molina, J.
Gmez, Seismic activity along the Central American
volcanic arc: Is it related to the subducted cocos plate?,
Tectonophysics 400 (2005) 241-254.
[9] R. White, J. Ligorra, I. Cifuentes, Seismic history of the
middle America subduction zone along El Salvador,
Guatemala and Chiapas, Mexico: 1526-2000, Geological
Society of America Papers 375 (2004) 379-396.
[10] D. Leeds, Catalog of Nicaraguan earthquakes, Bull.
Seismol. Soc. Am. 64 (1974) 1135-1158.
[11] National Geophysical Data Center/World Data Service
(NGDC/WDS) Significant Earthquake Database, Boulder,
CO, USA, http://www.ngdc.noaa.gov/nndc/struts/
form?t=101650&s=1&d=1 (accessed Sep. 1, 2010).
[12] N. Ambraseys, R. Adams, Large Central American
earthquakes 1898-1994, Geophys. J. Int. 127 (1996) 665-692.
[13] N. Ambraseys, Magnitudes of Central American earthquakes
1898-1930, Geophys. J. Int. 121 (1995) 545-556.
[14] R. White, D. Harlow, Destructive upper-crustal
earthquakes of Central America, Bull. Seismol. Soc. Am.
82 (1993) 1306-1349.
[15] P. Molnar, L. Sykes, Tectonics of the Caribbean and
middle America regions from focal mechanism and
seismicity, Geological Society of America Bulletin 80
(1969) 1639-1684.
[16] ISC (International Seismological Centre), Berkshire,
England, http://www.isc.ac.uk/ (accessed Sep. 1, 2010).
[17] NEIC (National Earthquake Information Center)
Centennial, http://earthquake.usgs.gov/research/data/
centennial.php (accessed Sep. 1, 2010).
[18] NGDC (National Geophysical Data Center),
http://www.ngdc.noaa.gov/ (accessed Sep. 1, 2010).
[19] J.G. Tanner, J.B. Shepherd, Seismic Hazard in Latin
America and the Caribbean, Project Catalogue and
Seismic Hazard Maps, IDRC (International Development
Research Center), Ottawa, 1997, p. 143.
[20] CMT (Centroid Moment Tensor Solutions), http://www.
globalcmt.org/CMTsearch.html (accessed Sep. 1, 2010).
[21] L. Feldman, Mountains of Fire, Lands That Shak:
Earthquakes and Volcanic Eruptions in Central America
(1505-1899), Labyrinthos, USA, 1993.
[22] C. Lindholm, W. Rojas, H. Bungum, A. Dahle, E.
Camacho, H. Cowan, et al., New regional seismic
zonation for Central America, in: Proceeding of 5th Int.
Conf. Seismic Zonation, Nice, 1995, pp. 437-444.
[23] W. Rojas, H. Bungum, C. Lindolm, A catalogue of
historical and recent earthquakes in Central America,
NORSAR (Norwegian Seismic Array), 1993.
[24] M. Benito, C. Lindholm, E. Camacho, A. Climent, G.
Marroqun, E. Molina, et al., A new evaluation of seismic
hazard for the Central American Region, Bull. Seismol.
Soc. Am. 102 (2) (2012) 504-523.
[25] J. Grases, Terremotos Destructores del Caribe 1502-1990:
UNESCO-RELACIS, Regional Office of Science and
Technology of UNESCO for Latin America and the
Caribbean, 1994, p. 132.
[26] W.M. Salazar, N.N. Ambraseys, J.J. Bommer,
Compilacin de un catlogo ssmico para El Salvador y
zonas aledeas, in: Proceedings of the Seminar on the
Evaluation and Mitigation of the Seismic Risk in Central
America Area, Universidad Centroamericana Jos Simen
Caas, San Salvador, El Salvador, Sep. 22-27, 1997, pp.
67-76.
[27] J. Bommer, C. McQueen, W. Salazar, S. Scott, G. Woo, A
case of study of the spatial distribution of seismic hazard
(El Salvador), Natural Hazards 18 (1998) 145-166.
[28] W. Salazar, Seismic Catalogue for El Salvador 1898-2001,
UCA Editores, Coleccin Estructuras y Procesos, Serie
Mayor, 2003, p. 41.
[29] M. Carr, L. Patino, M. Feigenson, Petrology and
geochemistry of lavas, in: J. Bundchuh, G.E. Alvarado
(Eds.), Central America: Geology, Resources and Hazards,
Taylor and Francis, London, 2007.
[30] V. Rapprich, W. Hernndez, V. Erban, Geology of
Metapan volcanic field NW El Salvador, Revista
Geolgica de Amrica Central 33 (2005) 61-74.
[31] P.S. Osiecki, Estimated intensities and probable tectonic
sources of historic (pre-1898) Honduran earthquakes, Bull.
Seismol. Soc. Am. 71 (3) (1981) 865-881.
[32] P. la Femina,, T. Dixon, W. Strauch, Bookshelf faulting in
Nicaragua, Geology 30 (8) (2002) 751-754.
[33] J. Martnez-Daz, J.J. Alvarez-Gmez, B. Benito, D.
Hernndez, Triggering of destructive earthquakes in El
Salvador, Geological Society of America 32 (1) (2004)
65-68.
[34] G. Corti, E. Carminati, F. Mazzarani, M. Garca, Active
strike-slip fault in El Salvador, Geology 33 (12) (2005)
989-992.
[35] J. Marshall, The geomorphology and physiographic
provinces of Central America, in: J. Bundchuh, G.E.
Alvarado (Eds.), Central America: Geology, Resources
and Hazards, UK, 2007.
[36] R. Rogers, H. Krason, R. van der Hilst, Epeirogenic uplift
above a detached slab in northern Central America,
Geological Society of America 30 (11) (2002) 1031-1034.
[37] W. Salazar, K. Seo, Earthquake disasters of January 13th
and February 13th, 2001, El Salvador, Seismological
Research Letters 74 (4) (2003) 420-439.

1043
[38] C. Caas-Dinarte, El Salvador: Cronologa de una Tierra
Danzarina, http://www.bvs.edu.sv/desastres/internas/ historia
_terremotos_el_salvador.htm (accessed Jan. 15, 2009).
[39] T. Hanks, H. Kanamori, A moment magnitude scale, J.
Geophys. Res. 84 (1979) 2348-2350.
[40] T. Hanks, D. Boore, Moment-magnitude relation in theory
and practice, J. Geophys. Res. 89 (1984) 6229-6235.
[41] M. Pasyanos, D. Dreger, B. Romanowicz, Toward
real-time estimation of regional moment tensors, Bull.
Seismol. Soc. Am. 86 (1996) 1255-1269.
[42] K. Chen, Y. Tsai, A catalog of Taiwan earthquakes
(1900-2006) with homogenized M
w
magnitudes, Bull.
Seismol. Soc. Am. 98 (2008) 483-489.
[43] E.M. Scordilis, Empirical global relations converting MS
and mb to moment magnitude, Journal of Seismology 10
(2006) 225-236.
[44] F. Bozzoni, M. Corigliano, C. Lai, W. Salazar, L.
Scandella, E. Zuccolo, et al., Probabilistic seismic hazard
assessment at the eastern Caribbean islands, Bull.
Seismol. Soc. Am. 101 (2011) 2499-2521.
[45] G. Burbach, C. Frohlic, T. Fumal, Seismicity and
tectonics of the subducted Cocos plate, J. Geophys. Res.
89 (1984) 7719-7735.
[46] J.E. Case, T. Holcombe, R. Martin, Map of geological
provinces in the Caribbean region, Geol. Soc. Am. 87
(1984) 825-829.
[47] B. Dean, C. Drake, Focal mechanism solutions and
solutions and tectonics of the middle America Arc, Journal
of Geology 86 (1978) 111-128.
[48] J. Dewey, G. Surez, Seismotectonics of the middle
America, Geol. Soc. Am. 1 (1991) 323-338.
[49] R. Stoiber, M. Carr, Geological setting destructive
earthquakes in Central America, Geological Society of
America Bulletin 88 (1977) 151-156.
[50] J. Dewey, R. White, D. Hernndez, Seismicity and
tectonics of El Salvador, Geological Society of America
375 (2004) 363-378.
[51] R. Weyl, Geology of Central America, 2nd ed., Gerburger
Borntraeger, 1980, p. 371.
[52] H. Williams, A.R. McBirney, G. Dengo, Geologic
reconnaissance in southern Guatemala, Univ. California
Publ. Geol. Sci. 50 (1964) 1-56.
[53] J. Alvarez-Gmez, P. Meijer, J. Martnez-Daz, R. Capote,
Constraints from finite element modeling on the active
tectonics of northern Central America and the middle
America trench, Tectonics 27 (1) (2008).
[54] C. Canora, J. Martnez-Daz, P. Villamor, K. Berryman, J.
Alvarez-Gmez, C. Pullinger, et al., Geological and
seismological analysis of the February 13, 2001 Mw 6.6 El
Salvador earthquake: Evidence of surface rupture and
implications for seismic hazard, Bull. Seismol. Soc. Am.
10 (2010) 2873-2890.
[55] B. Cailleau, P. La Femina, T. Dixon, Stress accumulation
between volcanoes: An explanation for intra-arc
earthquakes in Nicaragua, Geophys. J. Int. 169 (3) (2007)
1132-1138.
[56] T. Algermissen, J. Dewey, C. Langer, H. Dillinger, The
Managua, Nicaragua, earthquake of December 23, 1972:
Location, focal mechanism and intensity distribution, Bull.
Seismol. Soc. Am. 64 (1974) 993-1004.
[57] H. Lyon-Caen, E. Barrier, C. Lasserre, A. Franco, I. Arzu,
L. Chiquin, et al., Kinematics of the North
America-Caribbean-Cocos plate in Central America from
new GPS measurements across the Polochic-Motagua fault
system, Geophysical Research Letters 33 (19) (2006) 1-5.
[58] C. Canora, P. Villamor, J. Martnez-Daz, K. Berryman, J.
lvarez-Gmez, R. Capote, Paleoseismological analysis
of the San Vicente segment of the El Salvador Fault Zone,
El Salvador, Central America, Geologica Acta 10 (2)
(2012) 103-123.
[59] W. Salazar, V. Sardina, J. Cortina, A hybrid inversion
technique for the evaluation of source, path and site effects
employing S-wave spectra for subduction and
upper-crustal earthquakes in El Salvador, Bull. Seismol.
Soc. Am. 97 (1) (2007) 208-221.
[60] G. Hayes, D. Wald, R. Johnson, A three-dimensional
model of global subduction zone geometries, J. Geophys.
Res[Online], 2012, B01302, http://earthquake.usgs.gov/
research/data/slab/ (accessed Jan. 15, 2013).
[61] J. Pindell, L. Kennan, Tectonic evolution of the Gulf of
Mexico, Caribbean and northern South America in the
mantle reference frame: An update, Geological Society,
London, Special Publications 328 (2009) 1-55.
[62] A. Franco, E. Molina, H. Lyon-Caen, J. Vergne, T.
Monfret, A. Nercessian, et al., Seismicity and crustal
structure of the Polochic-Motagua Fault System Area
(Guatemala), Seismological Research Letters 80 (6) (2009)
977-984.
[63] C. DeMets, P. Jansma, G. Mattioli, T. Dixon, G. Farina, R.
Bilman, et al., GPS geodetic constraints on
Caribbean-North America plate motion, Geophys. Res.
Lett. 27 (2000) 437-440.
[64] M. Fernndez, E. Molina, J. Havskov, K. Atakan,
Tsunamis and tsunami hazards in Central America,
Natural Hazards 22 (2000) 91-116.
[65] W. Hernndez, D. Escobar, Informe de la inspeccin
realizada a los cantones El Chaval y Lomas de Alarcn
afectados por el enjambre ssmico localizado en la zona de
Atiquizaya, Turn y San Lorenzo, Departamento de
Ahuachapn. General Directorate of Environmental
Observatory, El Salvador, 2006.
[66] SNET (Servicio Nacional de Estudios Territoriales),
Enjambre ssmico en el Departamento de Ahuachapn,
zona occidental de El Salvador, Internal report, 2006, p. 25.

1044
[67] C. Langer, G. Bullinger, Secondary faulting near the
terminus of a seismogenic strike-slip fault: Aftershocks of
the 1976 Guatemala earthquake, Bull. Seismol. Soc. Am.
69 (1979) 427-444.
[68] M. Rodrguez, C. DeMets, R. Rogers, C. Tenorio, D.
Hernandez, A GPS and modelling study of deformation in
northern Central America, Geophys. J. Int. 178 (2009)
1733-1754.
[69] D. Panagiotopoulos, Long-term earthquake prediction in
Central America and Caribbean Sea based on the time and
magnitude-predictable model, Bull. Seismol. Soc. Am. 85
(1995) 1190-1201.
[70] E. Hauksson, Earthquakes, faulting and stress in the Los
Angeles Basin, J. Geophys. Res. 95 (1990) 365-394.
[71] S. Bilek, T. Lay, Variation of interplate fault zone
properties with depth in the Japan Subduction Zone,
Science 281 (1998) 1175-1178.
[72] J. Woessner, S. Wiemer, Assessing the quality of
earthquake catalogues: Estimating the magnitude
completness and its uncertainty, Bull. Seismol. Soc. Am.
95 (1995) 684-698.
[73] J.K. Gardner, L. Knopoff, Is the sequence of earthquakes
in southern California with aftershocks removed,
Poissonian?, Bull. Seismol. Soc. Am. 64 (1974) 1363-1367.
[74] S. Tinti, F. Mulargia, Completeness analysis of a seismic
catalogue, Ann. Geofis 3 (1985) 407-414.
[75] H. Shah, C. Mortgat, P. Kiremidjian, T. Zsutty, A Study of
Seismic Risk for Nicaragua, Part I, Report 11, The John A.
Blume Earthquake Engineering Center, Stanford
University, California, 1975.
[76] D. Monterroso, O. Kulhnek, Spatial variations of
b-values in the subduction zone of Central America,
Geofsica Internacional 42 (4) (2003) 575-587.
[77] A. Ghosh, Earthquake frequency-magnitude distribution
and interface locking at the middle American subduction
zone near Nicoya Peninsula, Costa Rica, Master Thesis,
Georgia Institute of Technology, 2007, p. 59.
[78] C. Cornell, Engineering seismic risk analysis, Bull.
Seismol. Soc. Am. 58 (1968) 1583-1606.
[79] R.K. McGuire, FORTRAN Computer Program for
Seismic Risk Analysis, U.S. Geological Survey, Open-File
report, 1976, pp. 76-67.
[80] G. Woo, Kernel estimation methods for seismic hazard
area source modelling, Bull. Seismol. Soc. Am. 86 (2)
(1996) 353-362.
[81] T. Kanno, A. Narita, N. Morikawa, H. Fujiwara, Y.
Fukushima, A new attenuation relation for strong motion
in Japan based on recorded data, Bull. Seismol. Soc. Am.
96 (2006) 879-897.
[82] R. McGuire, J. Arabasz, An introduction to probabilistic
seismic hazard analysis, Geotechnical and Environmental
Geophysics 1 (1990) 333-353.
[83] IBC (International Building Code), IBC Inc., 2012.
[84] Minimum Design Loads for Buildings and Other
Structures, ASCE 7-10, American Society of Civil
Engineers, 2010, p. 658.
[85] C. Cramer, R. Wheeler, C. Mueller, Uncertainty analysis
for seismic hazard in the southern Illinois basin,
Seismological Research Letters 73 (2002) 792-805.
[86] W. Salazar, K. Madrid, G. Moreno, D. Saracay, Una
resea histrica de los terremotos destructores del siglo
XX en El Salvador, Desde la Ciencia 3 (1) (1998) 25-34.

Appendix: Format and Catalogue Excerpt.
The format of the earthquake catalogue is as follows:
Column Description
1-4 Year
5 Blank space
6-7 Month
8 Blank space
9-10 Day
11 Blank space
12-13 Hour
14 Blank space
15-16 Minutes
17 Blank space
18-21 Seconds
22 Blank space
23-29 Epicentral coordinate in west longitude (degrees)
30 Blank space
31-36 Epicentral coordinate in north latitude (degrees)
37 Blank space
38 Source information code for the origin time and epicentral coordinates determinations (Table 1)

1045
39 Blank space
40-42 Focal depth (km)
43 Blank space
44 Source information code for the depth determination (Table 1)
45 Blank space
46-49 Magnitude M
W

50 Blank space
51 Source information code for the magnitude determination (Table 1)
52 Blank space
53-55
Magnitude used to convert to M
W
: *Ms: from surface wave magnitude; *mb: from body wave
magnitude; *MD: from Codaduration magnitude. Mw without a star means that the original value of
moment magnitude was taken directly when available
56 Blank space
57-61
Focal mechanism classification according to the slip , LL: left lateral (-44
o
44
o
); RL: right
lateral (136
-136
); T1: thrust (45
135
) and 20
within the strike of the Middle American

Trench N1: normal (-135
-45
) and 20
within the strike of the Middle American Trench T:

thrust (45
135
), N: normal (-135
-45
)
62 Blank space
63 Related phenomena: V: volcanic eruption; T: tsunami; E: eclipse

Excerpt of Earthquake Catalogue for Zone 1: Volcanic Chain

1 2 3 4 5 6
123456789012345678901234567890123456789012345678901234567890123

1917 06 08 00 51 51.0 -89.310 13.820 M 10 M 6.69 M *Ms V
1917 06 08 02 54 0.0 -89.500 13.770 M 10 M 5.73 M *Ms V
..
2001 01 14 16 22 21.7 -89.565 13.662 R 16 R 5.01 I *Ms
2001 01 14 18 17 31.8 -89.425 13.690 I 5 I 5.06 I *Ms
2001 01 16 11 53 22.9 -89.655 13.960 I 1 I 5.09 I *Ms
2001 01 17 06 51 10.1 -86.400 12.205 R 3 R 5.12 I *Ms
2001 02 13 14 22 5.6 -88.845 13.620 L 7 L 6.50 C Mw LL RL
2001 02 17 20 25 15.7 -89.249 13.661 L 5 L 5.56 L *Ms
2001 02 23 16 22 8.0 -89.200 13.715 I 12 I 5.01 I *Ms
2001 02 25 08 28 15.5 -89.200 13.708 I 13 I 5.03 I *Ms
2001 03 08 06 17 32.8 -87.861 13.339 R 5 R 5.03 I *Ms
2001 04 10 21 46 58.9 -88.722 13.641 L 8 L 5.12 I *Ms
2001 05 08 18 02 17.3 -88.706 13.615 L 9 L 5.70 C Mw LL RL
2001 05 08 18 15 47.0 -88.682 13.608 L 11 L 5.52 I *mb
2001 05 08 18 40 21.4 -88.662 13.616 L 7 L 5.31 I *Ms
2001 05 08 19 52 12.0 -88.715 13.592 L 3 L 5.09 I *Ms

Excerpt of Earthquake Catalogue for Zone 4: Interface Subduction

1 2 3 4 5 6
123456789012345678901234567890123456789012345678901234567890123

1609 00 00 00 00 0.0 -86.550 12.420 E 0 7.19 E *ML T
1648 00 00 00 00 0.0 -86.800 12.500 E 0 7.19 E *ML T
1663 00 00 00 00 0.0 -86.550 12.420 E 0 7.19 E *ML T
1719 03 05 00 00 0.0 -89.500 13.000 G 0 7.23 G *Ms T E
1752 05 07 00 00 0.0 -87.500 12.300 G 0 6.74 G *Ms T
1765 10 24 00 00 0.0 -91.917 15.000 B 0 7.66 W *Ms T
1773 07 29 00 00 0.0 -91.167 14.667 B 0 7.55 W *Ms T
1773 12 13 00 00 0.0 -91.500 14.583 B 0 7.12 W *Ms T
1776 05 30 00 00 0.0 -90.080 13.180 G 0 7.23 G *Ms T
1815 08 20 00 00 0.0 -89.000 12.750 G 0 7.23 G *Ms T
1831 02 07 00 00 0.0 -89.700 13.200 G 0 7.02 G *Ms T
1859 12 08 00 00 0.0 -90.000 13.200 G 0 7.02 G *Ms T T
1867 06 30 00 00 0.0 -89.160 13.200 G 0 6.56 G *Ms T
1869 03 01 00 00 0.0 -90.000 13.000 G 0 7.02 G *Ms T
1870 06 12 00 00 0.0 -90.167 14.167 B 0 7.23 W *Ms T
1874 09 03 00 00 0.0 -90.833 14.667 B 0 7.12 W *Ms T
1885 10 12 00 00 0.0 -87.030 12.080 G 0 6.74 G *Ms T
1900 11 09 16 08 0.0 -90.000 13.000 A 30 A 6.93 A *Ms
1901 10 08 02 16 0.0 -86.500 11.000 A 30 A 7.02 A *Ms T
1915 09 07 01 20 0.0 -89.600 13.900 A 60 A 7.83 A *Ms T T

Issue 8, 2013 Journal of Civil Engineering and Architecture

Diunggah oleh

Informasi Dokumen

Deskripsi Asli:

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

Issue 8, 2013 Journal of Civil Engineering and Architecture

Diunggah oleh

Hak Cipta:

Format Tersedia

Journal of Civil Engineering

(). In the above

(b) and may be considered constant. Therefore,

(b) E() (33)

(b) can be neglected.

W), the lakes

W which cover inland

azimuth of the Central American

of the strike of the Middle

of the strike of the Middle

); T1: thrust (45

within the strike of the Middle American

within the strike of the Middle American Trench T:

Anda mungkin juga menyukai