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

net/publication/258512677

On the seismic risk assessment of Tunis urban area: uncertainties

investigation and treatment using logic tree approach.

Conference Paper · May 2011

CITATION READS

1 1,448

2 authors:

Ahmed Ksentini Najla Bouden Romdhane

École Nationale d'Ingénieurs de Tunis University of Tunis El Manar
7 PUBLICATIONS 80 CITATIONS 15 PUBLICATIONS 208 CITATIONS

SEE PROFILE SEE PROFILE

All content following this page was uploaded by Ahmed Ksentini on 04 June 2014.

The user has requested enhancement of the downloaded file.

 Sixth International Conference of
Seismology and Earthquake Engineering
16-18 May 2011 Tehran, Iran

On the seismic risk assessment of Tunis urban area: uncertainties

investigation and treatment using logic tree approach.
A.Ksentini 1, N.B.Romdhane 2
1
Department of civil engineering, National School of Engineer of Tunis Tunisia ahmedksentini@yahoo.fr
2
Department of civil engineering, National School of Engineer of Tunis, Tunisia najlaromdhane@hotmail.fr

ABSTRACT

A recent seismic risk study is presented for the basin of Tunis in order to evaluate the probability
distribution of damage per building classes. Every step is discussed in the probabilistic hazard
assessment and the building stock response. Random and epistemic uncertainties are investigated in
order to measure the final model sensitivity to the input parameters. The latter concern essentially
recently developed predictive laws for the euro-Mediterranean and Middle East regions (Akkar and
Bommer 2010) and their combination with NGA (next generation of attenuation) models, seismic
source models which were proposed for Tunisia (Jordanovsky et al., 1991 and NASG 2006), seismic
recurrence estimation and amplification factors generated using global data on Vs30 velocity (Wald
and Allen 2009) for hazard assessment. Those parameters are used in logic tree approach (Kulkarni et
al., 1984) in order to give a median hazard map described in Peak ground acceleration and velocity
and acceleration in short and long periods to define the uniform hazard spectrum in the study region.
For buildings vulnerability, eight structure and occupancy classes are defined by site investigations
(Khalfet and Romdhane 2008) and capacity curves are studied for every class using modal pushover
analysis. All resulting parameters are finally used under GIS based FEMA-HAZUS loss estimation
tool using advanced engineering building module AEBM (based on CSM method, Freeman 1975) to
define probability distribution of different level of damage (from slight to complete) by building count
and class.

Keywords: Seismic risk, hazard, logic tree, uniform spectrum, building vulnerability, Hazus.

1. INTRODUCTION
The aim of this work is to assess the seismic risk in the city of Tunis (using a sample of more than
1000 buildings) while few models are proposed in the literature to estimate the direct and indirect
losses in this field (GSHAP 1999 for hazard estimation, some seismic regulations that are proposed by
the ministry of transportation and equipment in 2010). In this context, the seismicity of Tunisia is
considered as moderate according to the geometry and the depth of all known major geological
accident in the territory. Major historic events that were observed in Tunisia are probably those of
Utique 412, Kairouan 854 and Tunis 856 with an assigned moment magnitude of Mw=6.8
(Jordanovsky et al., 1991, Vogt, 1991). In the other hand, instrumental data processed by the National
Institute of Meteorology (INM) covers the period [1920-2010] and shows a maximum local magnitude
of ML=5.6.

SEE6 / 1 / IIEES
 Figure 1. INM earthquake catalog [412-2005]

This seismic activity is related to a complex pattern of crustal deformation along the Africa-Eurasia
plate boundary in the western Mediterranean region (Serpelloni et al., 2007). In this context, the
Tunisian territory is divided into several zones according to its position between two major domains:
the passive continental margin and the Cenozoic Alpine orogenic belt as shown in Figure 2 (Bouaziz
et al., 2002). It is now well established that Tunisia is composed of four principal geographic regions
(ETAP, 2003, Penã and abdesalam, 2006): the Saharan Platform (southern region), the Atlasic
Province (central region), the Eastern Tunisia Province (eastern region), and the Tellian Province
(northern region). Those data are used in this work in order to give a Probabilistic Seismic Hazard
Assessment (PSHA) for Tunisia using new intensity attenuation laws with a logic tree approach.

Figure 2. Tectonic pattern of the western Mediterranean domain (Bouaziz et al., 2002)

2. SEISMIC HAZARD
As mentioned for this study, a probabilistic seismic hazard assessment is adopted to describe ground
motion for short and long periods according to A.Cornell, Mc-Guire (1968, 1976) method assuming a
poissonian distribution for observed events behavior: events are not dependant on separating time. For
this task, earthquake bulletins provided from different sources (INM [412-2005], International
Seismological Centre ISC [1900-1998] and Euro Mediterranean Seismological Centre EMSC [1998-

SEE6 / 2 / IIEES
2003]) are homogenized into moment magnitude Mw using empirical relationships and similar events
are deleted using Bltn_shk code (ISC, 2006). The obtained catalog is then declustered from foreshocks
and aftershocks using many codes: Reasenberg (1985), Gardner & Konopoff (1974) and Dominique et
al. (2001). At this step, several time and space windows (Gruenthal, Urhammed 1986, Gardner &
Konopoff 1974) were used to come out with all possible clusters among instrumental data using
ZMAP mahtlab codes.

Figure3. Time and space windows (Gruenthal, Urhammed 1986, Gardner&Konopoff 1974)

It was established that the catalog presents discrepancies between the two periods: [1920-1975] and
[1975-2005]. This is due to seismic network updates in the 1975-1976 and 1990 and lower magnitude
bands were therefore easily recorded. Seismicity rate changes in the adopted period according to
Habermann algorithm is shown in figure 3 where vertical bars represent significant seismicity rate
changes. It maintains a high changes in low magnitude classes: [3.5-4.0] and [4.0-4.5]. That clearly
confirms the moderate seismicity for Tunisia associated to its crustal active faults with limited lengths.

Figure 4. Seismicity rate changes by bins magnitude classes [MW=3.5 to 5.8]

At this stage, frequency-magnitude distribution, according to Gutenberg Richter law (1944), is plotted
for the whole territory using declustered catalog and covering the period [1975-2005]. That allowed to
define the magnitude of complete reporting Mc=3.0 and global b-value=0.735.

SEE6 / 3 / IIEES
 Figure 5. Frequency-magnitude distribution for the declustered catalog

For the seismic sources, two models are adopted in this work. The first one is the Jordanovsky et al.,
1991 model which was updated in the GSHAP (global seismic hazard assessment program). It was
performed using the INM catalog (1920-1991) combined with historical data (Vogt, 1991). All events
are covered by seven seismic sources. The second model, with 5 sources, was defined by the North
African Seismological Group studies (actually NAGET: North African Group for Earthquake and
Tsunami) in collaboration with scientifics from the Italian Abdus Salam International Centre for
Theoretical Physics ICTP, Algeria in 2006. The same seismic data as the first model was used in
addition to instrumental data covering [1991-2005] provided by INM. We take in that both models are
composed of surface sources and no major faults are defined as linear sources.

Figure 6. Seismic source models for Tunisia: I. Jordanovsky (left) and II. NASG (right)

The last parameter that was used for probabilistic hazard assessment is the predictive laws for seismic
intensity attenuation. Recently developed law is used in this context: Akkar & Bommer 2010 to
describe pseudo-spectral acceleration “PSA”:
logሺPSAሻ = bଵ + bଶ M + bଷ M² + ሺbସ + bହ MሻlogටRଶ୨ୠ + bଶ଺ + b଻ Sୱ + b଼ S୅ + bଽ F୒ + bଵ଴ Fୖ + εσ
Where SS and SA are function of VS (30) velocity, FN and FR depend on fault type, ε is the number of
standard deviation σ and bi are constants.

SEE6 / 4 / IIEES
This law was developed using data from Europe, Middle East and Maghreb region and is one of the
most appropriate laws that can be used for this context. This law is plotted in figure7 with US-NGA
(Next Generation of Attenuation) project laws: Abrahamson and Silva AS2008; Boore and Atkinson
BA2008; Campbell and Bozorgnia CB2008; Chiou and Youngs CY2008. While Akkar & Bommer
law is limited to MDOF system short periods (3 seconds) and magnitudes larger than Mw=5.0 and
distances less than 100 km, authors advice to use it in combination in a logic tree approach with one or
two laws from NGA which were verified for use for the Euro-Mediterranean crustal seismicity regions
(Stafford, Strasser, Bommer 2008) for larger distances and lower magnitudes reaching 10 sec for
MDOF system periods.We see also, on figure8, the impact of the magnitude and soil type described in
shear velocities for a source-site distance of 10km and the impact of source-site distance for two
different magnitudes (MW=5.0 and 6.0) on the spectral acceleration. We verified that soil type is very
dominant vis-à-vis the horizontal acceleration generated at the base of the structure. Indeed, we note
that an earthquake of magnitude Mw = 6.0 will result (at 10 km from the site) a ground acceleration
with S-wave velocity Vs (30) less than 360 m/s, higher than that generated by an earthquake with
magnitude Mw = 7.0 at the same distance but with a source site Vs (30) velocity exceeding 360 m/s for
the spectral periods range of 0.4 to 1.6 seconds. It is also notable that the source-site greatly affects the
resultant horizontal acceleration. In particular, for distances below 50 km, the attenuation of the
acceleration is clear on the right curves of figure8. It shows that the peak ground acceleration
generated by an earthquake of magnitude Mw = 6.0 at 50 km from the site is doubled for an earthquake
of magnitude Mw = 5.0 at 10 km and is equivalent to an earthquake of magnitude Mw = 5.0 at 20 km.
As foreseeable, it follows that the calculation of the hazard probability is very sensitive to the adopted
attenuation laws. The choice of a single predictive law now looks like a design flaw. The designer
must therefore judge the reliability of the laws he uses to assign realistic probabilities in the logic tree
simulation to avoid model redundancy and to make it more exhaustive.

Figure 7. Predictive laws: Chiou & Youngs (CY 2008) NGA, Abrahamson & Silva (AS 2008) NGA, Campell &
Bozorgnia (CB 2008) NGA, Boore & Atkinson (BA 2008) NGA and Akkar & Bommer (AB 2010) Europe.

Figure 8. Predictive law of Akkar and Bommer (2010), spectral acceleration for different magnitudes and shear
wave velocities (left) and magnitudes and distances (right)

SEE6 / 5 / IIEES
For hazard assessment, A.Cornell-McGuire (1968, 1976) probabilistic method is used. The hazard
level depends on many parameters such as adopted attenuation law, seismic source models, b-
values…Those errors that can more or less influence the resulting intensity are mentioned as aleatory
or epistemic. The former can be avoided by more site investigation and data collection, while the latter
are related to the lack of knowledge about some parameters like events depths, active faults parameters
or appropriate attenuation law for the Maghreb region. In this filed, it is actually considered that logic
tree approaches (Kulkarni et al., 1984) are highly recommended when a robust seismic model is not
developed like most regions in the world except some ones like California or Japan where plentiful
data are used to best describe seismicity process. At this step, we developed a logic tree to capture the
epistemic potential errors. The branches of the tree are mainly: the seismic source models, the sources
depths and the attenuation laws.

Figure 9. Logic tree for probabilistic seismic hazard assessment for the Tunisian context

For the seismic source model, a 60% probability of activity is associated to the NASG model, 2006.
This choice is justified by the use of more rich data about seismicity and recent geological studies.
Another error is related to seismic sources depths. Most known faults in Tunisia are shallow and are
limited to 15 km in depth. That was taken into account in the logic tree by equal probability of
occurrence (33%) between three levels: 5km, 10km and 15km. For the attenuation laws, we combined
maintained law of Akkar and Bommer 2010 with a probability of prediction of 40% and two NGA
laws: Abrahamson and Silva 2008 and Chiou and Youngs 2008 (30% of prediction).
This multi site analysis (using Risk engineering-EZ-Frisk software) allowed to define the necessary
parameters in calculating the elastic acceleration spectra, namely the acceleration levels in different
spectral periods (peak acceleration PGA is reported according to SA = 0.01 sec) and peak ground
velocity PGV. Those maps are then used in the seismic risk assessment for Tunisia. Especially, for
Tunis urban area, we found that PGA for a 475 years return period reaches 0.24g, a considerable
intensity level.

SEE6 / 6 / IIEES
 Figure 8. Intensity map for PGA, return period = 475 years (left) Annual Frequency of exceedance of PGA for
the site of Tunis (right)

3. BUILDING STOCK VULNERABILITY

In order to estimate the seismic risk, we choose the study region of the capital Tunis and we focused
the direct loss estimation on buildings that are located in the down city (Habib Bourguiba Avenue) and
the old city (Medina). For this, we used street collected data (Khalfet and Romdhane 2008), such as
building type, occupancy classes, and number of stories…etc. Those buildings are plotted on a GIS
system for use in risk evaluation.
The new city of Tunis is characterized by mid to high rise buildings (4 to 10 stories). All structures
were built during the last fifty to sixty years according to French “CCBA68”, “BAEL91” and “CM66”
buildings codes and are generally composed of concrete frames with unreinforced masonry infill walls
or concrete shear walls. Some of them were subject to rehabilitation and sometimes steel frame stories
were added. Whereas for the old city, buildings are characterized by unreinforced masonry (URM)
stone walls and are low rise (maximum of 3 stories). Eight building types were therefore proposed in
table 1 (Khalfet and Romdhane 2008).

Figure 9. Left: study region delineation; right: general building stock map under GIS

SEE6 / 7 / IIEES
 Table1. Eight building types for the Tunisian building scheme (khalfet and Romdhane 2008, modified).

Type Structural system Slabs Classification

Masonry Masonry walls - steel beams rubble

- wood diaphragms
Manufactured stones
- concrete slabs
Massive stones
Reinforced Reinforced Concrete Concrete slabs Reinforced concrete frame
concrete frame with infill masonry walls
frame(RC frame) Reinforced conrete Shear
walls
Steel frame Steel light frame Steel deck diaphragms Steel frame with infill
buildings masonry walls

Wood frame Wood frame Wood diaphragms Wood frame with infill
buildings masonry walls

In order to describe building vulnerability, capacity curves can be defined for each class using recent
static pushover methods like the displacement coefficient method of ASCE-41, the ATC-40 capacity
spectrum method and the N2 method of Eurocode 8. Modal pushover analysis is applied here for
capacity spectrum method using ATC-40 for concrete buildings and FEMA 273 for steel moment
frames. The well known capacity spectrum method or « CSM » (Blume et al. 1961, Freeman 1978,
Mahaney et al. 1993) assume that the vulnerability of every building is related to the displacement of
the top of the structure induced by the acceleration levels at spectral periods. After defining building
section frames and reinforcement for colmuns and beams for every class, the dead load case is
transformed into static nonlinear load case and plastic hinges are defined and assigned according to the
FEMA 356 steel and concrete beams and columns tables. Monotonically increased loads are applied
on the structure height and force-displacement curves are calculated at every step of the analysis.
Plastic hinges will then control the building behavior in nonlinear domain and the final plot of the
pushover capacity curves. This method was applied to some representative existing buildings in the
study region according to their as built drafts from archives and using CSI-SAP2000 and ETABS
finite element programs.

SEE6 / 8 / IIEES
 Figure10. Building drafts and models as defined under CSI-sap 2000 and ETABS finite element programs

4. LOSS ESTIMATION ON TUNIS URBAN AREA USING FEMA-HAZUS

SOFTWARE
In order to assess the seismic risk for Tunis urban area, the well known “Federal Emergency
Management Agency FEMA-Hazus” earthquake loss estimation tool is used to calculate direct
damage on buildings. That allows benefiting from an existing methodology proposed by FEMA and
running on GIS to focus on other important parameters such buildings characteristics and ground
motion in regional and local scales. Starting from defining a user defined deterministic or probabilistic
scenario, the damage probabilities for building classes are given for different damage levels (from
slight to complete). Two methodologies were adopted in our works: the first one consists in populating
geodatabases with building occupancy classes given by count and by floor area for census tracts. Then
a user defined building scheme (building type-occupancy classes relationship) will be applied to
generate building type by count and calculate the building displacement and damage probabilities.
Results of such calculation are not presented here. The second methodology is based on damage
calculation for every building using Advanced Engineering Building Module AEBM in Hazus.

Figure 11. AEBM Flowchart

SEE6 / 9 / IIEES
This module is composed of procedures that are extension of the more general FEMA/NIBS
earthquake loss estimation methodology. More accurate results are given then by building for use by
experienced structural engineer. This module uses pushover analysis to calculate demand and damage
level by mean of fragility curves that are defined in Hazus software.
At this step, Capacity curves are updated according to the Tunisian building scheme which is different
from the default one proposed in Hazus databases and spectral acceleration are defined at 0.01 sec (for
PGA), 0.3 sec and 1.0 sec to plot uniform hazard spectrum.
An inventory of more than 1000 buildings from street collected data is defined under AEBM with 23
profiles that describe building type, occupancy classes and spectral displacement and acceleration at
yielding and ultimate bilinear capacity curves points. Figure 12 and 13 show the probabilities
distribution for moderate damage state for every structure in the study region:

Figure 12. Moderate structural damage probabilities distribution for the investigated building stock

Figure 13. Moderate structural damage probabilities by count and by profiles

While extensive and complete damage probabilities are low, we noticed through AEBM scenario that
moderate damage can’t be ignored. In fact, most moderate damage probabilities are between 18% and

SEE6 / 10 / IIEES
23% and concern more than 900 buildings in the study region. The most affected profiles are
com_masonry and mix_masonry for low and moderate rises. Those profiles are related to small
commercial occupancy class and diverse activities masonry buildings. For the concrete and steel
buildings, the change in the default capacity curves does not affect considerably the damage
probabilities since they do not exceed almost 5% in both cases.

5. LOSS ESTIMATION MODEL SENSITIVITY

The developed model uses ground motion parameters that results from probabilistic analysis. During
hazard estimation, we deleted major historical earthquakes in order to avoid distorting the estimation
of recurrence curves with GR model. To measure the sensitivity of the risk model with regard to the
considered hazard, we conducted different types of scenarios based on the deterministic hazard. It
consists of studying the ground motion level generated by an arbitrarily chosen event. It can be given
by magnitude and distances or picked from the catalog. The ground acceleration is then calculated
using appropriate attenuation law. Previous work (Romdhane & Ksentini 2004) was used to study the
acceleration levels generated by different earthquakes from the INM catalog to retain those inducing
peak accelerations at the study region.
The earthquake that resulted in the maximum acceleration is located at (N°=10.24, E°=36.84) with a
moment magnitude of 6.20. Site effects also come with this type of scenario. The amplification factors
of the seismic signal depend empirically on the shear waves velocities Vs (30) over the last 30 meters.
These velocities can be measured or simulated. The Wald and Allen (2007) model was established on
a correlation between the topography and velocities Vs (30) on a large global data. This correlation
was used to generate empirical relationships between topography slopes and expected Vs (30) for a
given site. Online simulations are available on the United States Geological Survey (USGS) web site
for global analysis. Comparing the results of this model with measured data gave satisfactory results.
However, authors specify that this simplified model cannot be precise for damage estimation and may
have deficiencies in certain areas. In this context, previous results: Romdhane et al., 2003, 2008 relate
measures of seismic background noise on the site of Tunis for different profiles in order to determine
the ground natural periods and patterns of S-wave velocities. These results are more representative of
the basin of Tunis lithology and are best suited for the seismic scenario given the precision they offer.
However, in this work, we measured the sensitivity of the loss model vis-à-vis the consideration of the
site effects for a deterministic scenario and amplification of elastic response spectra through the Wald
and Allen (2007) model, see figure14. This choice is due to the lack of measurements in the whole
study region which requires appropriate equipment and a lot of time in order to sweep the entire basin
of Tunis. The results presented in this paper are not a definitive model of damage estimation at the site
of Tunis but a mean to assess the impact of amplification of the response spectrum on the considered
building stock.

Figure 14. VS (30) correlation with slope topography for Tunisia (Wald and Allen 2007)

SEE6 / 11 / IIEES
Using the adopted scenario, amplification factors are generated on the base of Vs (30) and calculated
peak ground acceleration in the basin of Tunis with NEHRP soil classification. Figure15 shows
induced peak ground acceleration with amplification impact on seismic intensity in the basin of Tunis
using Boore and Atkinson 2008 predictive law. Visual exam of those maps confirms the proposed
correlation between topography slopes and acceleration amplification in the Wald and Allen (2007)
model.

Figure 15. Intensity maps using “Boore and Atkinson 2008 NGA” for M6.20 event: PGA without amplification
(left) with amplification (right)

Finally, three scenarios were peformed in order to meausure the model sensitivity to: (1) adopted
hazard level and type, (2) site effects and spectrum amplification and (3) updated capacity curves for
buildings according to the maintained structure types. Those entire scenarios were performed on
Hazus software which allows updating its default parameters according to local context. Figure16
summarize the results of those scenarios by building types and occupancy classes:

The first is for probabilistic hazard with default capacity curves,

The second is with the same hazard but with updated capacity curves,

The third is with deterministic hazard, updated capacity curves and with site effects.
We identified significant variation in damage rates depending on the selected scenario. For more than
1000 buildings of the street survey, the damage rate is allocated as follows:
- 1stScenario: 51 buildings present at least moderate damage (more than 5%)
- 2ndScenario: 215 buildings present at least moderate damage (over 20%)
- 3rdScenario: 808 buildings present at least moderate damage (over 74%).
The shape of adopted capacity curves is significant. This parameter is very important in losses
estimation. The shape of these curves is strongly influenced by the considered finite element model,
applied loads (dead, live, seismic…) and plastic hinges behavior ... and the resulting error is
considered as random. It can be reduced by analyzing a larger sample of buildings and spending more
time for analysis. Some parameters can be ignored in the development of such curves like the exact
characteristics of building materials or the actual dimensions of the supporting elements of the
building. However, the applied loads, reinforcement, the height of the building, its natural period ...
are fundamental parameters in the development of capacity curves. Some new methods are proposed
to overcome this type of error and we retain for future work:
- The measurement of background noise of buildings to determine the mode shapes and natural
periods of the structure that is used to calibrate the capacity curves in linear domain;
- Using the displacement method to generate the capacity curves on a random sample generated
according to known characteristics of the structures of the same class (Borzy, Pinho, and Crowley
2008 [8]).
The strong change in the rate of damages observed in the use of deterministic hazard is very important.
The calculation resulted in 189 buildings that are completely damaged and cannot be repaired.
Although this rate is not representative, since the uncertainty that we have reported in adopted

SEE6 / 12 / IIEES
amplification factors, but it helps highlighting the importance of the development of the deterministic
earthquake scenarios for more investigation on loss estimation for a given site.

Figure 16. Building count by damage, building types and occupancy classes according to the three adopted
scenarios.

SEE6 / 13 / IIEES
6. CONCLUSION:
In this paper, various components of seismic risk are studied in the context of Tunis urban area. From
the regional hazard, the site effects and the vulnerability of the building stock, we highlighted the
uncertainties and their relative impact on the expected damages estimation. It is clear that existing
models for estimating the seismic risk based on a single scenario tend to influence the damage curves
and annual frequencies of exceedance. This leads us to justify that this type of analysis cannot be
based on isolated events or limited samples of structures but rather on simulation with logic tree
approaches, covering most known events by encompassing different risk parameters and their standard
deviations or random sample based on the observed variability of those parameters, either on-site
measurements as well as more detailed investigations leading to eliminate epistemic and random
errors on the expected losses calculation.

REFERENCES
[1] Anil K. Chopra, Rakesh K. Goel, A Modal Pushover Analysis Procedure to Estimate Seismic
Demands for Buildings: Theory and Preliminary Evaluation, PEER Report 2001/03 Pacific
Earthquake Engineering Research Center, College of Engineering University of California Berkeley
[2] ATC-40: California seismic safety commission, Seismic evaluation and retrofit of concrete
building, November 1996
[3] ATILLA ANSAL, RECENT ADVANCES IN EARTHQUAKE GEOTECHNICAL
ENGINEERING AND MICROZONATION, ©2004 Springer Science + Business Media, Inc. ISBN:
1-4020-2528-9
[4] Boore, D.M., and Atkinson G. M. (2008), "Ground-Motion Prediction Equations for the Average
Horizontal Component of PGA, PGV, and 5%-Damped PSA at Spectral Periods between 0.01 s and
10.0 s", Earthquake Spectra, Volume 24, No. 1, pages 99–138, February 2008; © 2008, Earthquake
Engineering Research Institute
[5] Bouden-Romdhane N, Mechler P, Etude de bruit de fond sismique en vue d'un microzonage
sismique de la ville de Tunis", Bulletin des laboratoires des Ponts et Chaussées -213- janvier-février
1998 – REF. 4180 – PP. 43 – 57.
[6] Brian S.-J. Chiou, and Robert R. Youngs, "A NGA Model for the Average Horizontal
Component of Peak Ground Motion and Response Spectra", Earthquake Spectra, Volume 24, No. 1,
pages 173–215, February 2008; © 2008, Earthquake Engineering Research Institute
[7] Campbell, K.W., Bozorgnia, Y., "NGA Ground Motion Model for the Geometric Mean
Horizontal Component of PGA, PGV, PGD and 5% Damped Linear Elastic Response Spectra for
Periods Ranging from 0.01 to 10 s", Earthquake Spectra, Volume 24, No. 1, pages 139–171, February
2008; © 2008, Earthquake Engineering Research Institute
[8] Computers and structures INC. Berkeley, California, USA Concrete Frame Design Manual
Eurocode 2-2004 For SAP2000®
[9] Computers and Structures, Inc. Berkeley, California, USA SAP2000 Web Tutorial 1
DETAILED TUTORIAL INCLUDING PUSHOVER ANALYSIS
[10] Eurocode 8 (EN 2004): Design of structures for earthquake resistance, European standard,
December 2003
[11] EZ-FRISK 7.37 user manual, 2009
[12] Fabio Sabetta, Antonio Lucantoni, Hilmar Bungum, Julian J. Bommer, Sensitivity of PSHA
results to ground motion prediction relations and logic-tree weights, Soil Dynamics and Earthquake
Engineering 25 (2005) 317–329
[13] FEMA-356: Prestandard and Commentary for the Seismic Rehabilitation of Buildings”, National
Earthquake Hazards Reduction Program, Reston, VA. Federal Emergency Management Agency,
American Society of Civil Engineers, 2000
[14] FEMA-Hazus MH MR-1 technical and user manuals.
[15]Frank Scherbaum, Julian J. Bommer, Hilmar Bungum, Fabrice Cotton, and Norm A.
Abrahamson, Composite Ground-Motion Models and Logic Trees: Methodology, Sensitivities, and

SEE6 / 14 / IIEES
 Uncertainties, Bulletin of the Seismological Society of America, Vol. 95, No. 5, pp. 1575–1593,
October 2005, doi: 10.1785/0120040229
[16] HELEN CROWLEY, JULIAN J. BOMMER, and PETER J. STAFFORD, Recent
developments in the Treatment of Ground-Motion Variability in Earthquake Loss Models, Journal of
Earthquake Engineering, 12(S2):71–80, 2008
[17] Helen Crowley, Julian J. Bommer, Rui Pinho and Juliet Bird, The impact of epistemic
uncertainty on an earthquake loss model, EARTHQUAKE ENGINEERING AND STRUCTURAL
DYNAMICS Earthquake Engng Struct. Dyn. 2005; 34:1653–1685
[18] Julian J. Bommer and Norman A. Abrahamson, Why Do Modern Probabilistic Seismic-
Hazard Analyses Often Lead to Increased Hazard Estimates? Bulletin of the Seismological Society of
America, Vol. 96, No. 6, pp. 1967–1977, December 2006, doi: 10.1785/0120060043
[19] Julian J. Bommer, Frank Scherbaum, Hilmar Bungum, Fabrice Cotton, Fabio Sabetta, and
Norman A. Abrahamson, On the Use of Logic Trees for Ground-Motion Prediction Equations
[20] Julian J. Bommer, M.EERI, and Frank Scherbaum, The Use and Misuse of Logic-Trees in
PSHA, ENGINEERING PRACTICE
[21] Julian J. Bommer, Peter J. Stafford, John E. Alarcón, and Sinan Akkar, The Influence of
Magnitude Range on Empirical Ground-Motion Prediction, Bulletin of the Seismological Society of
America, Vol. 97, No. 6, pp. 2152–2170, December 2007, doi: 10.1785/0120070081
[22] Laurent Pasticier, Claudio Amadio and Massimo Fragiacomo, Non-linear seismic analysis
and vulnerability evaluation of a masonry building by means of the SAP2000 V.10 code, Earthquake
Engng Struct. Dyn. 2008; 37:467–485
[23] Mario Ordaz, Some Notes about the Use of Logic Trees in Probabilistic Seismic Hazard
Analysis, Crisis 2007 papers
[24] MAX WYSS, STEFAN WIEMER & RAMÓN ZÚÑIGA, ZMAP A TOOL FOR ANALYSES
OF SEISMICITY PATTERNS, TYPICAL APPLICATIONS AND USES: A COOKBOOK
[25] Najla Bouden Romdhane and Afef Khalfet, Paper “the topology of Tunisian building in
appraisal of its seismic vulnerability”, 2008
[26] Norman A. Abrahamson, and Julian J. Bommer, Probability and Uncertainty in Seismic
Hazard Analysis, OPINION PAPER DOI: 10.1193/1.1899158
[27] Norman Abrahamson, and Walter Silva, Summary of the Abrahamson & Silva NGA Ground-
Motion Relations Earthquake Spectra, Volume 24, No. 1, pages 67–97, February 2008; © 2008,
Earthquake Engineering Research Institute
[28] Peter J. Stafford, Fleur O. Strasser, Julian J. Bommer, An evaluation of the applicability of
the NGA models to ground-motion prediction in the Euro-Mediterranean region, Bull Earthquake Eng
(2008) 6:149–177 DOI 10.1007/s10518-007-9053-2
[29] Report IZIIS 500-91-61, Investigation for elaboration of seismotectonic map and draft seismic
design code of Tunisia, Volume II: Seismic hazard analysis and seismic zoning map of Tunisia.
[30] Rich Hansen & Douglas Bausch, A GIS Based methodology for exporting the hazard U.S.
(HAZUS) earthquake model for global application, RELEMR workshop in Malta, April 2006.
[31] Sinan Akkar and Julian J. Bommer, Empirical Equations for the Prediction of PGA , PGV, and
Spectral Accelerations in Europe, the Mediterranean Region, and the Middle East, Seismological
Research Letters Volume 81, Number 2 March/April 2010
[32] Srinivasan Chandrasekaran, Luciano Nunziante, Giorgio Serino, Federico Carannante,
Seismic Design Aids for non linear analysis of reinforced concrete structures, CRC press: ISBN: 978-
1-4398091-4-3, July 2009, 258 pp.
[33] Steven L.Kramer, Geotechnical Earthquake Engineering, © 1996 by Prentice-Hall INC. ISBN:
0-13-374943-6

SEE6 / 15 / IIEES

View publication stats