Content uploaded by Snježana Markušić
Author content
All content in this area was uploaded by Snježana Markušić on Sep 10, 2018
Content may be subject to copyright.
HARMONIZED SEISMIC HAZARD MAPS FOR THE WESTERN
BALKAN COUNTRIES
Radmila SALIC1, Zeynep GULERCE2, Neki KUKA3, Snjezana MARKUSIC4, Jadranka
MIHALJEVIC5, Vladan KOVACEVIC6, Abdullah SANDIKAYA7, Zoran MILUTINOVIC8, Llambro
DUNI9, Davor STANKO10, Natasa KALUDJEROVIC11, Svetlana KOVACEVIC12
ABSTRACT
The Harmonization of Seismic Hazard Maps in the Western Balkan Countries Project (BSHAP) was funded for 7
years by NATO-Science for Peace Program to support the preparation of new seismic hazard maps of the Western
Balkan Region using modern scientific tools. One of the most important outputs of the BSHAP is an updated and
unified BSHAP earthquake catalogue that is compiled directly from the datasets of earthquake data providers of
the region. In the framework of BSHAP, the regional free field strong motion network capacity was increased
significantly by the purchased and installed recorders and the BSHAP strong motion database that includes both
pre-BSHAP (mostly analog) and post-BSHAP (all digital) recordings was compiled. The BSHAP strong motion
database is used for proper selection of the ground motion prediction equations (GMPEs) for the probabilistic
seismic hazard assessment (PSHA) by comparing the compiled strong ground motions with the predictions of
candidate global, European, and Euro-Med GMPEs in a systematic manner. BSHAP collected relevant knowledge
about the geological structure of southwestern Balkans provided a better understanding of the prevailing stress
regime in the region. The main output of BSHAP is the new probabilistic seismic hazard maps for Western
Balkans, obtained by implementation of the smoothed-gridded seismicity approach. The results are expressed in
terms of peak horizontal acceleration (PGA) for 95 and 475 years return periods aligned with Eurocode 8
requirements. The seismic hazard maps derived in this project are a good basis to characterize the seismic hazard
of Western Balkans. They will help the national authorities, public and private institutions, civil emergencies
agencies, etc., for urban planning, disaster preparedness, and seismic hazard mitigation.
Keywords: BSHAP, Seismic hazard, Western Balkan.
1Assist. Prof. Dr., Institute of Earthquake Engineering and Engineering Seismology (UKIM-IZIIS), Skopje,
Republic of Macedonia, r_salic@iziis.ukim.edu.mk
2Assoc. Prof. Dr., Middle East Technical University, Department of Civil Engineering, Ankara, Turkey,
zyilmaz@metu.edu.tr
3Prof. Dr., Institute of Geosciences, Energy, Water and Environment, Polytechnic University, Tirana, Albania,
nekikuka@live.com
4Prof. Dr., Faculty of Science, University of Zagreb, Zagreb, Croatia, markusic@irb.hr
5M.Sc., Institute of Hydrometeorology and Seismology, Podgorica, Montenegro, mihaljevic@seismo.co.me
6Eng., Seismological Survey of Serbia, Belgrade, Serbia, vladan.kovacevic@seismo.gov.rs
7Assoc. Prof. Dr., Hacettepe University, Civil Engineering Department, Ankara, Turkey,
abdullahsandikkaya@hacettepe.edu.tr
8Prof. Dr., Institute of Earthquake Engineering and Engineering Seismology (UKIM-IZIIS), Skopje, Republic of
Macedonia, zoran@iziis.ukim.edu.mk
9Prof. Dr., Institute of Geosciences, Energy, Water and Environment, Polytechnic University, Tirana, Albania,
llambroduni@yahoo.com
10Asst. M.Sc., Faculty of Geotechnical Engineering, University of Zagreb, Varazdin, Croatia,
stankodavor@gmail.com
11Eng., Institute of Hydrometeorology and Seismology, Podgorica, Montenegro, kaludjerovic@seismo.co.me
12Eng., Seismological Survey of Serbia, Belgrade, Serbia, svetlana.kovacevic@seismo.gov.rs
2
1. INTRODUCTION
Local seismic design code regulations, seismic risk estimation, risk management and seismic safety
improvements should be based on reliable hazard maps, especially for the seismically active regions.
The region of the Western Balkans is characterized by high earthquake hazard and risk when compared
to the rest of Europe. The fact that current seismic design provisions are dated from the early 1980-is
(practically in all of participating countries) underlines an evident need to upgrade these technical norms
with provisions harmonized with EU standards (Eurocode 8).
The Project: Harmonization of Seismic Hazard Maps for the Western Balkan Countries (BSHAP) was
funded by NATO Science for Peace and Security Program since 2007 and continued until the end of
2015 with the main objective of development of new regional seismic hazard maps as a necessary step
towards the seismic safety improvement and seismic risk management.
2. PROJECT ACCOMPLISHMENTS
Development of the seismic hazard maps as a final project goal presides an updated and unified
earthquake catalogue, strong motion database compilation, proper selection of the ground motion
prediction equations (GMPEs), compilation of all relevant regional geological knowledge and
development of seismo-tectonic model.
1.1 BSHAP Earthquake catalogue
As one of the most important project outputs is an updated and unified earthquake catalogue (Markušić
et al., 2016) assembled from the national earthquake catalogues of 12 countries in the region: Albania,
Bosnia and Herzegovina, Bulgaria, Croatia, Greece, Hungary, Italy, Montenegro, Macedonia, Romania,
Serbia and Slovenia. The compiled catalogue is enriched also with data from the global catalogues,
especially for the large magnitude events. One of the biggest efforts was to build a unified data format
under which were systemized large number of events with different data formats coming from the
different national and regional catalogues.
The BSHAP earthquake catalogue consists of 26,100 earthquake events, related to the time span from
510 BC to 2012 and covers the area with geographic limits 38.0–48.0°N, 12.0–24.5°E (Figure 1).
Since the primary objective was to provide input to the PSHA and seismic hazard maps, the Mw proxy
is estimated for all entries in the BSHAP catalogue. Various magnitude scales, even from the same data
provider, were assigned to the events in the compiled catalogue; therefore, separate empirical
relationships between the local and other magnitude scales and Mw are developed for Albania, Croatia,
Macedonia, Montenegro and Serbia using errors-in-variables regression technique (Table 1).
The data pairs collected from regional and global catalogues are employed in the regression analysis.
The catalogue completeness thresholds are analyzed (Table 2) and incorporated into the seismotectonic
model developed within the BSHAP Project. The unified and updated BSHAP catalogue, as
demonstrated, is fully compatible with the current well-established European and worldwide catalogues.
2.2. BSHAP Fault plane solution (FPS) database
The territories of the BSHAP countries mostly lie in the Adria-Eurasia collision zone which is
characterized by a great differences in seismicity rate, present-day stress direction, strain rate, and
consequently in fault slip rate among neighboring regions.
With the main purpose to study the dominant tectonic regime, under the BSHAP framework was
compiled FPS database covering area bounded by 38-48° N and 12-24.5° E. The database includes 714
FPS for M≥4.0 earthquakes occurred between 1909 and 2015 (Figure 2). For this purpose 332
earthquake events were analyzed in the framework of the project, gathered by a partner institutions (An
updated Croatian FPS Database first described by Herak et al. (1995) – current version of which is
presented in Herak et al. (2016), the updated Montenegrin FPS database (Kaludjerovic, 2015), and the
FPS databases of Seismological Survey of Serbia, and Institute of Geosciences, Energy, Water and
Environment, Albania) or collected from global resources.
3
Table 1. Empirical relationships between moment
magnitude Mw and local magnitude ML (Markušić
et al., 2016)
Agency
Regression
equation
Mw=b0+b1ML
Number of
events
Determination
coefficient, R2
SD of
regression, se
Tirana Mw=1.22+0.813ML
a (0.25) (0.056)
96 0.715 0.256
Podgorica Mw=-0.01+1.028ML
a (0.16) (0.033)
75 0.930 0.184
Zagreb Mw=-0.11+1.011ML
a (0.38) (0.080)
31 0.852 0.229
Belgrade Mw=0.70+0.858ML
a (0.21) (0.049)
50 0.953 0.182
Skopje Mw=0.56+0.913ML
a (0.48) (0.101)
28 0.773 0.267
a In the second rows, in parenthesis are given the standard
errors of regression coefficients.
Table 2. Catalogue completeness intervals of
BSHAP catalogue (Markušić et al., 2016)
Figure 1. Spatial distribution of the earthquakes in the
proposed BSHAP catalogue (earthquakes M ≥ 3.5 in the
period 510BC–1969, and earthquakes M ≥ 3.0 in the
period 1970–2012 (Markušić et al., 2016)
Mw ≥
3.0
3.5
4.0
4.5
5.0
5.5
6.0
6.5
Complete
From year
1982
1978
1965
1945
1900
1850
1605
1280
Analysis of the FPS indicates that the majority of the earthquakes observed along the coastlines of
Croatia, Montenegro and Albania have reverse mechanism, correlated to the thrusting in the most part
of the External Dinarides and Albanides. Tectonic compressions are directed in SW–NE direction in the
southern and eastern parts, and in S–N direction in the northern and western parts of the coastline. In
the continental part the faults are active as strike-slip to oblique strike-slip or even as reverse faults. This
distribution reflects the counter-clockwise motions of Adria and its compression against the Dinarides.
In the Albanides, the boundary between normal faulting to east and thrust faulting to west runs through
central Albania. The extension is observed in eastern Albania and Macedonia
2.3 BSHAP Seismic source characterization (SSC)
Since hazard results are sensitive to seismic source modeling, the broader area 12-24.5E and 38-48N
have been considered in SSC modelling. Development of SSC models was based on uniform and
updated BSHAP earthquake catalogue (Markušić et al., 2016), relevant knowledge about the geological
and seismotectonic structure of Western Balkans as well as the stress information indicated by the
BSHAP FPS database. Additionally, sets of sensitivity analyses were performed to support final
estimates of some models’ parameters (Mihaljević et al., 2017).
In order to avoid undue fluctuations in the recurrence model parameters (b-value, mean annual rate of
earthquake occurrence, etc.) that are commonly present when addressing smaller areas - particularly in
the zones of low seismicity, the super zone model (SZM) (Figure 3) was proposed. The SZM model
consists of seven larger and two smaller zones that were delineated based on the seismotectonic
characteristics.
Since the BSHAP earthquake catalogue - for the out of BSHAP countries (Figure 3) zones strongly
depends on the global catalogues, the completeness levels of the catalogue show a large temporal and
spatial variability. Completeness time intervals of the BSHAP earthquake catalogue in each super zone
are estimated for different magnitude of completeness Mc. Considering that spatial variability of Mc-
the smallest value of magnitude at which the catalogue is thought to have included all seismic events,
can lead to erroneous results of the seismicity recurrence parameters, the minimum magnitude (mmin) is
set to 4.0 for all super zones.
4
Figure 2. BSHAP FPS database, insert for the
geographical area 13-23.5°E, 39-47°N. Color of symbols
marks the FPS for different mechanisms: the blue, red and
green symbols stand for the reverse, normal and strike-slip
events, respectively (Mihaljević et al., 2017).
Figure 3. Super zone model (SZM) is defined for
influence area covering 12-24.5°E and 38-48°N.
(Mihaljević et al., 2017).
Using the maximum likelihood estimation (MLE) for each super zone b-value was calculated
considering unequal completeness intervals for different magnitude ranges (Weichert, 1980), and the
MLE procedure developed by Kijko and Sellevoll (1992, KS-92). The magnitude recurrence parameters
(b, mmax) estimated by the KS-92 procedure were compared to the relevant values from the MLE
approach defined by Weichert (1980), as well as to the relevant estimates obtained in the framework of
the SHARE project for the areal source model (AS Model, Basili et al. 2013).
Two alternative estimates for the b-value are used for each source zone: (1) the average value of the
relevant estimates derived using the super zones sub-catalogs, considering the variable magnitude
completeness MLE (Weichert, 1980) and the estimates obtained using the Kijko-Sellevoll (1992)
approach on the same sub-catalogs; and (2) b=1.0 for all the source zones, as comparable to the AS
model of the SHARE project (Giardini et al., 2013).
While the super zone model has been implemented with the purpose of estimating statistically-stable b-
values, other parameters of the SSC model (mmax, dominant style of faulting and fault directions) were
estimated for smaller areas, delineated within two alternative zonation models. Seismic source 1- SSM1
and Seismic source 2 - SSM2 (Figure 4) representing the local tectonic features provided input data for
the two-stage (circular and elliptical) smoothing procedure (Lapajne et al., 1997).
In Western Balkans, SSM1 and SSM2 were delineated considering a detailed analysis of tectonic
settings, known active faults, activity rates, observed magnitudes, and foci depths. Zones covering the
neighboring (out-of-BSHAP) region are preserved in both models and were delineated considering
SHARE project (Basili et al., 2013, Giardini et al., 2013), and according to Vamvarakis et al., 2013.
Borders of the source zones are mostly consistent with the borders of the super zones since the b-value
estimated for the corresponding super zone is directly implemented for the zones of SSM1 and SSM2.
Each zone is attributed by a zone ID, covered area, maximum observed magnitude, average foci depth,
and sets of weighted parameters: b-value, mmax, style of faulting and fault strike angle. To assign the
weights related to tectonic information, faults were grouped based on the mechanism and the median
strike azimuth. Their weights were calculated based on measured length of the (grouped) faults (Poljak
et al., 2000). In SSM1 and SSM2, the mmax for each source zone was chosen by considering the largest
observed magnitude in the zone. Taking into account the uncertainties related to this parameter - two
alternative estimates of mmax are included by adding 0.25 and 0.5 magnitude units to the largest observed
magnitude in each zone. We assumed that the minimum mmax value in any zone cannot be lower than
Mw=6.0, even if the largest observed magnitude is much smaller.
5
a) SSM1 b) SSM2
Figure 4. Geographical partition of SSM1’s and SSM2’s zones and their position vs. super zone model.
(Mihaljević et. al, 2017)
For the adopted BSHAP seismic source characterization models incorporated epistemic uncertainties
associated with construction of the seismic source models (uncertainty of the b-value, choice of SSM,
maximum magnitude), type of smoothing and ground motion prediction equations are provided in the
logic tree scheme (Figure 6).
2.4. BSHAP Strong ground motion database
During the implementation of BSHAP project, 46 in total strong motion stations were deployed in the
territories of participating countries (Albania, Croatia, Macedonia, Montenegro and Serbia) with the
main purpose of enhancement the regional strong motion networks capacity. Also one of the main goals
of this project was to provide a reliable and uniformly processed strong motion database that contains
all of the available regional free field strong motion data. Therefore, all free field strong motion data
found in the archives of Albanian, Croatian, Macedonian, Montenegrin, Serbian, and Slovenian
seismological networks were collected (Table 3a and Table 3b). In total BSHAP strong motion database
consists of 672 three-component accelerograms from 358 earthquakes recorded at 121 strong motion
stations. After the initial strong motion database was compiled, the waveform quality of all ground
motion recordings was checked and some recordings were eliminated from the dataset due to non-
standard errors (Douglas, 2003). Remaining recordings in the database were associated to the
earthquakes in the BSHAP earthquake catalogue (Markušić et al., 2016) and the moment magnitude
(Mw) values from the catalogue were directly adopted.
After the strong motions were collected, earthquake metadata information of the strong motion database
was enriched with the fault plane parameters based on the fault plane solutions (FPS) of the earthquakes
gathered either from the relevant global resources (e.g. Global Centroid Moment Tensor Project) or the
FPS analysis performed by the project participants (Mihaljević et al., 2017) using different softwares
(developed within the institutions or commercial). Unfortunately, for the majority of the data (64 %),
any information regarding the fault plane parameters still does not exist in the BSHAP database. Using
the Mw values and double-couple solutions, the style-of-faulting of each event was determined by P-
and T- angle definitions given in Boore and Atkinson (2007) and the source-to-site distance metrics
(epicentral distance: Repi; hypocentral distance: Rhyp; Joyner–Boore distance: RJB; rupture distance:
Rrup) were computed using the procedure described for Reference Database for Seismic Ground-motion
in Europe (RESORCE; Akkar et al. 2014a).
All digital records in the BSHAP database were processed by uniform processing technique for the
6
elimination of high and low frequency noise. Processing of the digital records was performed using the
USDP (Utility Software for Data Processing) (Boore et al. 2011) and the analog waveforms were
processed using the standard CALTECH procedure, modified due to use of different type of data
processing equipment (Petrovski and Naumovski 1979; Petrovski et al. 1982). Details are presented in
Salic et al. (2016).
Table 3a. General statistics of the digital portion of
the BSHAP strong motion database (Salic et al.,
2016).
Table 3b. General statistics of the analogue portion of
the BSHAP strong motion database (Salic et al.,
2016).
Country
No. of
Records
Mw
Hypocentra
l depth
(km)
Repi
(km)
Period
(year from-
to)
No. of
records
used in the
analysis
Albania 163
3.70–
6.10
0.0–
24.0
2.2–
500.0
2003–
2015 155
Croatia 3
4.73–
5.10
5.2–
21.9
55.8–
62.1
2003–
2005 3
Macedonia 37
3.00–
5.20
0.0–
21.0
11.4–
301.3
2013–
2014 0
Montenegro 50
3.07–
5.50
0.0–
33.0
15.9–
180.1
2009–
2012 33
Serbia 24
3.50–
5.50
1.0–
15.0
16.9–
137.9
2006–
2013 3
Slovenia a 21
2.92–
5.64
7.0–
18.3
4.0–
103.5
1998–
2011 1
Total 298
2.92–
6.10
0.0–
33.00
2.2–
500.0
1998–
2015 195
a Slovenia was not directly part of NATO SfP-983054 (BSHAP-1) and
NATO SfP-984374 (BSHAP-2) project. Data provided was courtesy of
ARSO (Dr. Mladen Zivcic).
Country
No. of
Records
Mw
Hypocentra
l depth
(km)
Repi
(km)
Period
(year from-
to)
No. of
records
used in the
analysis
Albania 1 5.91 5.0 14.8 1998 1
Croatia 36
3.20–
5.99
1.1–
16.4
4.6–
157.2
1986–
1996 19
Macedonia a 337
2.10–
7.40
0.0–
94.0
0.9–
527.4
1975–
1994 235
Total 374 2.10–
7.40
0.0–
94.0
0.9–
527.4
1975–
1998 255
a Provided data from Macedonia are related to the territories of all
former Yugoslavian Countries (By alphabetic order: Bosnia and
Herzegovina, Croatia, Macedonia, Montenegro, Serbia and Slovenia.
2.5. Selection of ground motion models
The established strong motion database is used for selection of the ground motion prediction equations
(GMPEs) to be employed in the probabilistic seismic hazard assessment by comparing the compiled
strong ground motions with the predictions of candidate global and Euro-Mediterranean GMPEs in a
systematic manner (Salic et al., 2016).
Guidance provided by the recent world-wide projects (such as SHARE-Seismic Harmonization in
Europe and GEM-Global Earthquake Model) was followed in addition to the well-known criteria
proposed by Cotton et al. (2006) and Bommer et al. (2010) for building the list of candidate models.
Therefore, the set of candidate models chosen by the BSHAP working group (Figure 5) was quite similar
to the candidate list of Delavaud et al. (2012) and Stewart et al. (2015) except that the updated versions
of the NGA-W2 and European models were included in this study (Abrahamson et al., 2014; Boore et
al., 2014; Campbell and Bozorgnia, 2014; Chiou and Youngs, 2014; Akkar et al., 2014; Bindi et al.,
2014 and Cauzzi et al., 2015).
On the other hand, provided is a comprehensive methodology for testing the applicability of candidate
GMPEs for the PSHA studies in the Western Balkan area based on the collected BSHAP ground motion
database. Our approach combines the residual analysis methods, evaluation of the median predictions
for the scaling and functional form of candidate GMPEs, and recently published quantitative model-data
comparison methods (Scasserra et al., 2009 and Gülerce et al., 2016). Four GMPEs (2 global NGA-
West2 models and 2 recently published European models) are selected based on the behavioral analysis
i.e. their satisfactory fit in the BSHAP strong motion dataset with these approaches. Accordingly
selected are: [BSSA14] Boore et al. (2014), [CY14] Chiou and Youngs (2014), [Aetal14] Akkar et al.
(2014a, c,) and [Betal14] Bindi et al. (2014).
In table 7 are presented calculated and combined (EDR, LLH) weighting factors as well as the adopted
ones employed in PSHA calculations. The procedure used, allows that weighting factors are calculated
based on either EDR (Kale and Akkar, 2013) or LLH (Scherbaum et al., 2009) method, in which cases
either LLHi=1 or EDRi=1, respectively.
The calculated logic tree weights (Table 4) are much more in favor of EDR than LLH scoring scheme,
attributing larger weights of 0.25 - 0.28 to CY14 and BSSA14, respectively, than to Aetal14 and Betal14
being of ~0.23. While the subsequent PSHA analyses could incorporate the epistemic uncertainty by
7
using the calculated logic tree weights, the consensus decision of the BSHAP team was to additionally
stipulate the CY14 and BSSA14 GMPEs attributing them the equal weights of 0.3 in respect to Aetal14
and Betal14, being de-stipulated to weights of 0.2.
Table 4. EDR and LLH scores of the selected GMPEs and
calculated/adopted logic tree weights (Salic et al., 2016).
GMPE
Period
EDR
Score
LLH
Score
Rank
(EDR)
Rank
(LLH)
Logic
Tree
Weights
(w)
SR
GR
SR
GR
Calculated
Adopted
BSSA1
4
PGA 1.485 2.647 2
(1)
1
(1)
0.28199
0.30 0.2s 1.500 2.687 1 1
1.0s 1.275 2.249 1 2
CY14
PGA 1.425 3.320 1
(1)
4
(4)
0.25022
0.30 0.2s 1.439 3.038 1 4
1.0s 1.256 2.469 1 4
Aetal14
PGA 1.644 3.227 4
(4)
4
(3)
0.23516
0.20 0.2s 1.543 2.878 2 3
1.0s 1.415 2.289 4 2
Betal14
PGA 1.641 3.120 4
(4)
3
(3)
0.23262
0.20 0.2s 1.777 3.008 4 4
1.0s 1.261 2.111 1 1
Figure 5. Comparison of the median predictions of
the candidate GMPEs for EC site class A (VS30 =
800 m/s, Faulting style: SS, Earthquake size: Mw =
7.5 (Salic et al., 2016)
2.6. Seismic Hazard Assessment
Because too little information on active faults and their corresponding slip rates is known in the BSHAP
region, it was impossible to define a reliable fault-based source model. Therefore, for seismic hazard
assessment we decided to use the background-gridded source models, which account for crustal
earthquakes not occurring on modeled faults. The overall method for modeling of background-gridded
seismicity is based on the spatial smoothing approach, whereby the rate of past earthquakes and a
regionally consistent MFD are used to forecast the rate of future earthquakes. The method accounts for
the spatial variability of seismicity rate, and is used for areas where faults are not known or cannot be
parameterized. Smoothed seismicity models assume that future, large-magnitude events are likely to
occur near the locations of smaller earthquakes.
Development of the background-gridded source model consists in the following steps: specification of
a magnitude-frequency distribution (MFD), development of a model for the maximum magnitude,
estimation of earthquake rates, and specification of locations and source zone geometries. For all
smoothed seismicity models, we assumed the Gutenberg-Richter (GR) relationship between earthquake
magnitude and frequency:
log
(1)
where n(m) is the number of earthquakes with M ≥ m, and a and b are the GR parameters controlling the
seismicity rate and the relative proportion of earthquakes with different magnitudes, respectively.
BSHAP-SSC employs a truncated form of the GR relation whereby the earthquake magnitudes are
constrained to the range, mmin ≤ M ≤mmax:
min
min max min
max min
exp[ ( )] exp[ ( )]
1 exp[ ( )]
mm
mm m m
mm
bb
ll b
-- -- -
=⋅
-- -
, mmin < m < mmax (2)
in which
m is the mean annual number of earthquakes with M ≥ m, and
min
m
lis the mean annual number
of earthquakes with M ≥ m
min; mmin is the minimum magnitude capable of producing structural damage,
and mmax represents the largest considered magnitude that can occur within a defined source zone. For
all crustal sources in the BSHAP region, we accepted mmin=4.0. So, for description of the recurrence of
seismic sources given the truncated exponential model (2), three parameters are required: the rate of
earthquake activity
min
m
l
,
the b-value and Mmax.
8
Development of smoothed seismicity models requires calculation of regional b-values and the
completeness levels from the earthquake catalogs, which defines the magnitude above which all
earthquakes have been included. The completeness levels for the BSHAP-SSC are defined regionally
(Table 2) and consist of sets of minimum magnitude of completeness (Mc) and corresponding
completeness years. For each source super zone, two alternative estimates for the b-value are obtained
using the declustered BSHAP earthquake catalog and the relevant completeness levels (Table 2), as
described in the section 2.3.
BSHAP smoothed seismicity models are defined on a 10km x 10km grid for the latitude range 38.0°-
48.0°N and the longitude range 12.0°-24.5°E. The areal seismic sources are modelled as set of the grid
points included within the relevant seismic source zones. Earthquakes with magnitude greater than or
equal to mmin that passed the completeness test of BSHAP catalogue (Table 2) are counted in each grid
cell. Then, the annual rate of earthquakes occurrence is computed by a maximum-likelihood method
(Weichert, 1980), using the b-value from the corresponding super zone and the number of the events in
each grid-cell - adjusted to account for the magnitude completeness levels. The smoothed seismicity
models are obtained by smoothing earthquake rates to produce a spatially varying estimate of seismicity
rates (Frankel, 1995; Lapajne et al., 2003).
To evaluate the effects of spatial smoothing on the seismic hazard, two alternative seismicity smoothing
methods are investigated. At first, a two-dimensional isotropic Gaussian smoothing (Frankel, 1995),
hereinafter circular smoothing (CS), is applied to smooth the annual rates of earthquake occurrence (λ-
grid) in each grid cell. Based on the sensitivity analysis, a fixed correlation distance of 30 km is applied
for the circular smoothing. In the second alternative, referred as CES in the following, aside from the
circular smoothing with 30 km correlation distance, an anisotropic fault-oriented smoothing (Lapajne et
al., 2003), hereinafter elliptic smoothing (ES), is also applied on the grid of circular smoothed seismicity
rates. The elliptic smoothing considers the rupture directions and the respective lengths (estimated by
Wells & Coppersmith magnitude scaling relationships) of the main tectonic structures within the seismic
source zones, provided by the BSHAP seismotectonic database. Spatial smoothing is considered as a
branch in the logic-tree structure, to account for the epistemic uncertainties associated with construction
of the seismic source model. We assigned the same weight (0.5) to both CS and CES smoothing
approaches. These smoothed seismicity models are therefore stored as grids of the annual rate of
earthquakes (MW≥mmin), and are used later in the hazard calculations.
It is evident the prehistoric and historical earthquakes in each source zone provide a lower bound on the
maximum considered magnitude, mmax. That is, mmax must be at least as large as the largest observed
earthquake. We cannot know, however, if the largest observed earthquake is the largest possible
earthquake. We represent the uncertainty in mmax by assigning weights to two alternative mmax estimates.
For each source zone, two alternative estimates for mmax are considered adding an increment of 0.25 and
0.50 magnitude units, respectively, to the maximum magnitude observed in the zone. To the both
branches is assigned the same weight, 0.5.
A logic tree with 64 branches (Figure 6) has been designed to combine models in the hazard analysis
and to derive the hazard maps for the Western Balkans, using the background-gridded source model.
Each node in the logic tree defines alternative models or ‘branches’ in the logic-tree, with weights that
sum to one. The nodes in this logic tree includes: (1) two background-gridded source models (SSM1,
and SSM2), (2) two alternative estimates for the b-value, (3) two alternative estimates of the maximum
considered magnitude for each source zone, (4) two alternative algorithms for smoothing of the
seismicity rates (CS with 30 km correlation length, and CS+ES), and (5) four GMPEs (Aetal14, Betal14,
BSSA14 and CY14) for ground motion prediction.
Hazard calculation for each branch of logic tree are performed using the computer code OHAZ
(Zabukovec et al., 2007), jointly developed by ARSO (Slovenian Environment Agency) and IGEWE
(Institute of GeoSciences, Energy, Water and Environment, Albania), and recently upgraded and
improved by Kuka (OHAZ 2015) to fulfil requirements of the BSHAP project. Hazard assessment is
applied for firm rock conditions, with 800 m/sec shear-wave velocity in the upper 30 m of the soil section
(CEN, 2004). Based on the hazard results obtained according to the above mentioned procedure, the
probabilistic seismic hazard maps that characterize the spatial variability of maximum horizontal
acceleration (PGA) were derived. In compliance with EC8 standards, the hazard was calculated for two
characteristic return periods: 95- (Figure 7a) and 475-years (Figure 7b), which correspond to the
exceedance probabilities of 10% in 10 years and 50 years, respectively.
9
3. CONCLUSIONS
Seismic Source
Model (SSM)
→
b-value
→
Mmax
→
Smoothing
method
→
GMPE
SSM1
(0.50)
→
MLE 1
(0.50)
→
Obs+0.25
(0.60)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
Obs+0.50
(0.40)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
SHARE 2
(0.50)
→
Obs+0.25
(0.60)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
Obs+0.50
(0.40)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
SSM12
(0.50)
→
MLE 1
(0.50)
→
Obs+0.25
(0.60)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
Obs+0.50
(0.40)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
SHARE 2
(0.50)
→
Obs+0.25
(0.60)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
Obs+0.50
(0.40)
→
CS
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
→
CES
(0.50)
→ Betal14 (0.20)
→ Aetal14 (0.20)
→ BSSA14 (0.30)
→ CY14 (0.30)
One of the most important outputs of the
BSHAP is an updated and unified
earthquake catalogue that is compiled
directly from the datasets of earthquake
data providers of the region. New
magnitude conversion equations for
various local magnitude scales of the data
providers are developed with the aim of
having homogeneous moment magnitude
estimates.
The BSHAP strong motion database is
used for proper selection of the ground
motion prediction equations (GMPEs) for
the probabilistic seismic hazard
assessment (PSHA) by comparing the
compiled strong ground motions with the
predictions of candidate global,
European, and Euro-Med GMPEs in a
systematic manner.
The main output of BSHAP are the new
probabilistic seismic hazard maps for
Western Balkans (Figure 7), obtained by
implementation of the smoothed-gridded
seismicity approach. They are prepared
based on the BSHAP earthquake
catalogue, selected GMPEs and
developed seismotectonic model. Hazard
calculations are carried out following a
logic-tree structure with 64 branches,
which describes the epistemic
uncertainties associated with construction
of the seismic source model, and of the
GMPEs selected for ground motion
prediction.
Completeness time intervals for the
catalogue data are provided as inputs to
the seismic source models for updated
seismic hazard of Western Balkan
Region. The unified and updated BSHAP
catalogue is found to be compatible with
the current well-established European
and world-wide catalogues and represents
a sound basis for analysis of the
seismicity of this region.
BSHAP collected relevant knowledge
about the geological structure of
southwestern Balkans. Database of focal
plane solutions, held by BSHAP partners,
provided a better understanding of the
prevailing stress regime in the region.
Entire influence area covered by
earthquake catalogue data is represented
Figure 6. Logic-tree for the seismicity based background
source model in the Western Balkans. Assigned branch
weights shown in parentheses.
1) Average value of Weichert and KS-92 MLE, 2) AS
denotes Areal Source model of SHARE (Kuka et al.,
2017).
10
by logic - tree branches of seismic source models: each model is composed of full set of seismogenic
zones (groups of cells with the same attributes that are grouped into regions). The weighted b-values are
adopted from the Super zone model and using MLE (Weichert 1980) and Kijko-Sellevoll (1992)
approaches and in accordance to estimates of SHARE Area Source model (Basili et al., 2013). Presented
Seismic source models SSM1 and SSM2 - represented by their geometry, seismicity and seismotectonic
information is provided as the input to perform the spatial smoothing of seismic activity rates for the
hazard calculations.
In the framework of BSHAP, the regional free field strong motion network capacity was increased
significantly by the purchased and installed recorders and the BSHAP strong motion database that
includes both pre-BSHAP (mostly analog) and post-BSHAP (all digital) recordings was compiled.
The BSHAP harmonized strong motion database includes the uniformly processed strong motions along
with the related earthquake metadata and station information; therefore, it provides a solid background
for the ground motion characterization studies in the surrounding region.
The hazard results are expressed in terms of peak horizontal acceleration (PGA) for 95 and 475 years
return periods. The assessment has been carried out for rock conditions with average velocity of shear
waves VS≥800 m/sec in the upper 30 meters of soil section (classified as soil type A according to
Eurocode 8 soil definitions). Thus, obtained results are in full agreement with the Eurocode 8 standard
for seismic zonation and aseismic design. The seismic hazard maps derived in this project are a good
basis to characterize the seismic hazard of Western Balkans. They will help the national authorities,
public and private institutions, civil emergencies agencies, etc., for urban planning, disaster
preparedness, and seismic hazard mitigation.
a) Seismic hazard map of Western Balkans showing
peak ground acceleration for 10-percent probability of
exceedance in 10 years (RP 95 years).
b) Seismic hazard map of Western Balkans showing
peak ground acceleration for 10-percent probability of
exceedance in 50 years (RP 475 years).
Figure 7. Seismic hazard maps of Western Balkans showing peak ground acceleration for VS30 site condition of
800 meters per second (Kuka et. al, 2018).
4. ACKNOWLEDGMENTS
This work is funded by NATO SPS Program under: “Improvements in the Harmonized Seismic Hazard
Maps for the Western Balkan Countries Project“(NATO SfP Award Number: 984374). BSHAP
participants are indebted to the institutions/data providers for positive cooperative attitude in data releasing
and continuous support. We, the BSHAP Projects teams and authors of this paper, do expect that NATO,
Public Diplomacy Divisions’ expectations and policies are meet, in terms of technical and synergic
achievements, and that BSHAP’s results justify granting that we highly acknowledge. The part of the work
has been supported by Croatian Science Foundation under the project HRZZ IP-2014-09-9666.
11
5. REFERENCES
Abrahamson NA, Silva WJ, Kamai R (2014) Summary of the ASK14 ground motion relation for active crustal regions.
Earthq Spectra 30(3):1025–1055.
Akkar S, Sandikkaya MA, Senyurt M, Sisi AA, Ay BO, Traversa P, Douglas J, Cotton F, Luzi L, Hernandez B, Godey
S (2014a) Reference database for seismic ground-motion in Europe (RESORCE), B. Earthq Eng 12:311–339.
Basili, R., V. Kastelic, M.B. Demircioglu, D. Garcia Moreno, E.S. Nemser, P. Petricca, S.P. Sboras, G.M. Besana-
Ostman, J. Cabral, T. Camelbeeck, R. Caputo, L. Danciu, H. Domac, J. Fonseca, J. García-Mayordomo, D. Giardini, B.
Glavatovic, L. Gulen, Y. Ince, S. Pavlides, K. Sesetyan, G. Tarabusi, M. M. Tiberti, M. Utkucu, G. Valensise, K.
Vanneste, S. Vilanova and J. Wössner (2013). The European Database of Seismogenic Faults (EDSF) compiled in the
framework of the Project SHARE. http://diss.rm.ingv.it/share-edsf/, DOI: 10.6092/INGV.IT-SHARE-EDSF.
Bindi D, Massa M, Luzi L, Ameri G, Pacor F, Puglia R, Augliera P (2014) Pan-European ground-motion prediction
equations for the average horizontal component of PGA, PGV, and 5 %-damped PSA at spectral periods up to 3.0 s
using the RESORCE dataset. Bull Earthq Eng 12(1):391–430. Doi: 10.1007/s10518-013-9525-5.
Bommer JJ, Douglas J, Scherbaum F, Cotton F, Bungum H, Fah D (2010) On the selection of ground-motion prediction
equations for seismic hazard analysis. Seismol Res Lett 81:783–793.
Boore DM, Kale O, Akkar S (2011) Utility software for data processing (USDP). http://web.boun.edu.tr/
sinan.akkar/usdp1.html.
Boore DM, Stewart JP, Seyhan E, Atkinson GM (2014) NGA-West 2 equations for predicting PGA, PGV, and 5 %-
damped PSA for shallow crustal Earthquakes. Earthq Spectra 30(3):1057–1085. Doi: 10.1193/ 070113EQS184M.
Boore M, Atkinson GM (2007) Boore–Atkinson NGA ground motion relations for the geometric mean horizontal
component of peak and spectral ground motion parameters. In: PEER Report 2007/01. Pacific Earthquake Engineering
Research Center, University of California, Berkeley.
Campbell KW, Bozorgnia Y (2014) NGA-West2 ground motion model for the average horizontal components of PGA,
PGV, and 5 %-damped linear acceleration response spectra. Earthq Spectra 30(3):1087–1115. Doi:
10.1193/062913EQS175M.
Cauzzi C, Faccioli E, Vanini M, Bianchini A (2015) Updated predictive equations for broadband (0.01–10 s) horizontal
response spectra and peak ground motions, based on a global dataset of digital acceleration records. Bull Earthq Eng
13(6):1587–1612. Doi: 10.1007/s10518-_014-_9685-_y.
CEN (2004). EN-1998-1:2004 - Eurocode 8: Design of Structures for Earthquake Resistance – Part 1: General rules,
seismic actions and rules for buildings, European Committee for Standardization, Brussels.
Chiou BS-J, Youngs RR (2014) Update of the Chiou and Youngs NGA model for the average horizontal component of
peak ground motion and response spectra. Earthq Spectra 30(3):1117–1153. Doi: 10.1193/072813EQS219M.
Cotton F, Scherbaum F, Bommer JJ, Bungum H (2006) Criteria for selecting and adjusting ground-motion models for
specifc target regions: application to Central Europe and rock sites. J Seismol 10:137–156.
Delavaud E, Cotton F, Akkar S, Scherbaum F, Danciu L, Beauval C, Drouet S, Douglas J, Basili R, Sandikkaya A,
Segou M, Faccioli E, Theodoulidis N (2012) Toward a ground-motion logic tree for probabilistic seismic hazard
assessment in Europe. J Seismol 16(3):451–473.
Douglas J (2003) Earthquake ground motion estimation using strong-motion records: a review of equations for the
estimation of peak ground acceleration and response spectral ordinates. Earth Sci Rev 61:43–140.
Frankel, A. (1995). Mapping seismic hazard in the central and eastern United States, Seismological Research Letters,
Vol. 66, No. 4, July–August 1995, 8–21.
Giardini, D., J. Woessner, L. Danciu, H. Crowley, F. Cotton, G. Grünthal, R. Pinho, G. Valensise, S. Akkar, R.
Arvidsson, R. Basili, T. Cameelbeeck, A. Campos-Costa, J. Douglas, M.B. Demircioglu, M. Erdik, J. Fonseca, B.
Glavatovic, C. Lindholm, K. Makropoulos, C. Meletti, R. Musson, K. Pitilakis, K. Sesetyan, D. Stromeyer, M. Stucchi
and A. Rovida (2013). Seismic Hazard Harmonization in Europe (SHARE): Online Data Resource, DOI:
10.12686/SED-00000001-SHARE.
Gülerce Z, Kargioglu B, Abrahamson NA (2016) Turkey-adjusted NGA-W1 horizontal ground motion prediction
models. Earthq Spectra 32(1):75–100. Doi: 10.1193/022714EQS034M.
Herak M., D. Herak and S. Markušić (1995). Fault-plane solutions for earthquakes (1956-1995) in Croatia and
neighboring regions. Geofizika, 12, 43-56.
12
Herak, M., D. Herak, I. Dasović (2016). Fault-plane solutions and stress orientation in the greater region of Northern
and Central Dinarides. 35th General Assembly of the European Seismological Commission. Trieste, Italy, 4-11
September 2016.
Kaluđerovic N., (2015): Focal mechanisms and tectonic stress in Montenegro. 4th NATO Project No 984374 Meeting.
Podgorica, Montenegro, 26-27 March 2015.
Kale O, Akkar S (2013) A new perspective for selecting and ranking ground-motion prediction equations (GMPEs): the
euclidian distance-based ranking method. Bull Seismol Soc Am 103(2A):1069–1084.
Kijko A. and M.A. Sellevoll (1992). Estimation of earthquake hazard parameters from incomplete data files. Part II:
incorporation of magnitude heterogeneity. Bull Seismol Soc Am, 82(1):120–134.
Kuka, N., Gülerce Z., Milutinović Z., Mihaljević J., Salic, R., Duni Ll, Markušić, S., Kovačević S., Probabilistic Seismic
Hazard Assessment for Western Balkans, in preparation for Bulletin of Earthquake Engineering.
Lapajne, J.K., B. Šket Motnikar and P. Zupančič (2003). Probabilistic Seismic Hazard Assessment Methodology for
Distributed Seismicity. Bulletin of the Seismological Society of America, 93, No. 6, 2502-2515.
Lapajne, J. K., B. Sket Motnikar, B. Zabukovec & P. Zupancic (1997). Spatially-smoothed seismicity modelling of
seismic hazard in Slovenia“, Journal of Seismology, Vol. 1, No. 1, 73-85.
Markušić S, Gülerce Z, Kuka N, Duni L, Ivancic I, Radovanovic S, Glavatovic B, Milutinović Z, Akkar S, Kovačević
S, Mihaljević J, Salic R (2016) An updated and unified earthquake catalogue for the Western Balkan Region. Bull Earthq
Eng 14(2):321–343. Doi: 10.1007/s10518-015-9833-z.
Mihaljević, J., P. Zupancic, N. Kuka, N. Kaludjerovic, R. Koci, S. Markušić, R. Salic, E. Dushi, E. Begu, Ll. Duni, M.
Zivcic, S. Kovačević, I. Ivancic, V. Kovačević, Z. Milutinović, M. Vakilinezhad, T. Fikret and Z. Gülerce (2017).
BSHAP Seismic Source Characterization Models for the Western Balkan Region, Bulletin of Earthquake Engineering,
Bulletin of Earthquake Engineering (Accepted for publication).
Petrovski D, Naumovski N (1979) Processing of strong motion accelerograms, Part I-Analytical methods. IZIIS
Publication No. 66. Skopje, November 1979.
Petrovski D, Naumovski N, Stamatovska S (1982) Processing of strong motion accelerograms, Part IIComputer
programs. IZIIS Publication No. 69. Skopje, April 1982.
Poljak M., P. Zupančič, J.K. Lapajne and B. Šket Motnikar (2000). Seismotectonic Input for Spatially Smoothed
Seismicity Approach. Proceedings of Seismicity Modeling in seismic Hazard Mapping Workshop, Poljče, Slovenia.
Salic, R., Sandıkkaya, M.A., Milutinović, Z., Gülerce, Z., Duni, Ll., Kovačević, V., Markušić, S., Mihaljević, J., Kuka,
N., Kaludjerovic, N., Kotur, N., Krmpotic, S., Kuk, K., and Stanko, D. (2016). BSHAP Project Strong Ground Motion
Database and Selection of Suitable Ground Motion Models for the Western Balkan Region, Bulletin of Earthquake
Engineering, DOI 10.1007/s10518-016-9950-3.
Salic, R., Sandıkkaya, M.A., Milutinović, Z., Gülerce, Z., Duni, Ll., Kovačević, V., Markušić, S., Mihaljević, J., Kuka,
N., Kaludjerovic, N., Kotur, N., Krmpotic, S., Kuk, K., and Stanko, D. (2017). Reply to ‘‘Comment to BSHAP project
strong ground motion database and selection of suitable ground motion models for the Western Balkan Region’’ by
Carlo Cauzzi and Ezio Faccioli, Bulletin of Earthquake Engineering, DOI 10.1007/s10518-017-0095-9.
Scassera G, Stewart J, Bazzurro P, Lanzo G, Mollaioli F (2009) A comparison of NGA ground-motion prediction
equations to Italian data. Bull Seismol Soc Am 99(5):2961–2978.
Scherbaum F, Delavaud E, Riggelsen E (2009) Model selection in seismic hazard analysis: an information theoretic
perspective. Bull Seismol Soc Am 99(6):3234–3247.
Stewart JP, Douglas J, Javanbarg M, Abrahamson NA, Bozorgnia Y, Boore DM, Campbell KW, Delavaud E, Erdik M,
Stafford PJ (2015) Selection of ground motion prediction equations for the Global Earthquake Model. Earthq Spectra
31(1):19–45. Doi: 10.1193/013013EQS017M.
Vamvakaris, D. A., C.B. Papazachos, C Papaioannou, E.M. Scordilis and G.F. Karakaisis (2013). A detailed seismic
zonation model for shallow earthquakes in the broader Aegean area. Nat. Hazards Earth Syst. Sci. Discuss., 1, 6719–
6784, DOI: 10.5194/nhessd-1-6719-2013.
Weichert, D.H. (1980). Estimation of the earthquake recurrence parameters for unequal observation periods for different
magnitudes, Bull. Seism. Soc. Am. 70, 1337-1346.
Zabukovec, B., Kuka, N., Sostaric, M., Motnikar, B. S., Suler, T. (2007). OHAZ: Computer Program for Seismic Hazard
Calculation, User Manual, Environmental Agency of Slovenia and Institute of Seismology of Albania, 65p.
