Sei sulla pagina 1di 18

Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Contents lists available at ScienceDirect

Soil Dynamics and Earthquake Engineering


journal homepage: www.elsevier.com/locate/soildyn

Seismic microzoning of Belgrade MARK


a a, b d c c
V.W. Lee , M.D. Trifunac , B.. Bulaji , M.I. Mani , D. Herak , M. Herak
a
Dept. of Civil Eng., Univ. of Southern California, Los Angeles, CA 90089, USA
b
Dept. of Civil Eng. and Geodesy, Univ. of Novi Sad, 21000 Novi Sad, Republic of Serbia
c
Dept. of Geophysics, Faculty of Science, Univ. of Zagreb, Horvatovac bb, 10000 Zagreb, Croatia
d
Retired, Crvena skopska optina 4/1-1, Skopje, Macedonia

A R T I C L E I N F O A BS T RAC T

Keywords: Seismic microzonation maps for Belgrade (Serbia) and its surroundings are presented based on the uniform-
Microzonation maps, which include a hazard-spectrum (UHS) methodology. Such mapping must satisfy the guidelines for performance-based design
description of site geology and site soil (PBD), which at present requires the specication of two sets of spectral amplitudes, one in which the structure
properties would remain essentially linear, and one in which it would undergo nonlinear response. These requirements
Contribution of a large distant earthquake to
cannot be achieved by specifying the design spectra using only one (same) xed spectral shape, and such spectra
the design strong motion amplitudes
cannot be scaled by the peak ground acceleration alone. Another source of diculties in the selection of the
Computation of seismic hazard for earthquake
sources that follow dierent attenuation laws design amplitudes for PBD occurs when the standard spectrum shape is not capable of describing excitation by
large, distant earthquakes. Furthermore, scaling the site dependent design ground motion only via soil site
classication ignores the eects of site geology and thus leads to biased results. The maps we present in this
paper avoid these shortcomings and include the eects of near and distant large earthquakes, spatial
distribution of seismic activity, site geology, and site soil properties in a balanced way.

1. Introduction pipes, simultaneous action of surface faulting with strong shaking, and
many others that will be subsequently shown in maps), and ultimately
At the beginning of the twentieth century, engineers began to their probabilistic combination to determine the balanced outcome(s).
include the eects of earthquake shaking in the design of structures. At The rst systematic attempts to develop seismic microzoning maps
rst, this was done in terms of the equivalent static horizontal force, go back to the former Soviet Union [2] and Japan [37] in the 1930s.
and later by analyzis of the dynamic response via the response- Before the age of strong motion accelerographs, amplitudes of ground
spectrum method [104,115,24]. Investigations of earthquake damage, motion could be scaled only in terms of site intensity, which was then
and in particular of its irregular distribution in space, showed that for used to evaluate the design peak ground acceleration. On the basis of
comparable epicentral distances, these variations were related to the many observational studies following earthquakes, guidelines were
geologic and soil site conditions. To account for these variations, it was developed for the prediction of relative increase or decrease of site
proposed that city planners and earthquake engineers should be intensities (and then of the associated peaks of strong motion
provided with microzoning maps with coecients that characterize amplitudes) based on the nature of the site geology and surface soil
expected spatial variations in the amplitudes of shaking [37,61,77]. [23,61]. Many published seismic microzonation maps from that time
The equivalent horizontal earthquake force, and later, the response- resembled the spatial distribution of geological and soil deposits in the
spectrum amplitudes, were then increased or decreased according to area [38,59,61,77]. The local spatial variations were rst based
the values of the amplication coecients dened in the microzoning primarily on site geology [61,77] and then later expanded to include
maps. the eects of shallow sediments and local soils. Theoretical and
Preparation of seismic microzonation maps involves many inter- observational studies in Japan eventually evolved into methods that
mediate steps including description of seismic activity surrounding the aimed to include properties of the local site characteristics, determined
site, attenuation (from source to the site) of the quantity (peak through the measurement of microtremors [7981].
acceleration or velocity, site intensity, spectral amplitudes, duration Many engineering estimates of the expected site-specic strong
of strong shaking, power of strong shaking, energy required to initiate ground motion have been based on Kanais [35,36] interpretation,
liquefaction, peak strains for design of underground structures and which uses one-dimensional wave-propagation models with vertically


Corresponding author.
E-mail address: trifunac@usc.edu (M.D. Trifunac).

http://dx.doi.org/10.1016/j.soildyn.2017.02.002
Received 10 November 2016; Accepted 19 February 2017
0267-7261/ 2017 Elsevier Ltd. All rights reserved.
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

arriving seismic waves, which upon reection from free surfaces leads illustration, these erroneous predictions occur about 40% of the time.
to interference patterns, often referred to as site resonances. Multiple In spite of this evidence, many studies have continued to develop
recordings of weak and strong motion do conrm that such resonances scaling equations using only the soil-site classication variables (e.g.,
exist [20], but observations show that they do not occur with every [1,35,13,17,16]) as if all strong-motion data has been recorded under
earthquake and appear only rarelytypically less than about 20% identical geologic site conditions.
[114]. During large amplitudes of strong shaking, nonlinear soil The modern studies of site eects began in the late 1800s with the
responses rst shift these resonances toward longer periods, and for eld observations of Mallet and Milne in Japan, who provided an
very large and destructive amplitudes of ground motion, the peaks insightful analysis of recorded motions [117]. These studies continued
associated with the resonances usually disappear [103,110113]. with the work of Kanai and his co-workers, and then expanded into the
Before the 1960s, peak accelerations were read directly from analog second half of the twentieth century thanks to the rapid increase in the
instrument records without the possibility of instrument or baseline number of high-quality, recorded strong-motion accelerograms [78].
corrections [91,92], and frequency-dependent spectral characterization The purpose of this paper is to show how a model of seismic activity
of strong motion amplitudes was not available. After the earthquake in the region can be used to formulate microzoning maps of Belgrade,
magnitude scale was introduced in the 1930s, descriptions of shaking Serbia. The methodology, scaling equations, and descriptions of
levels gradually shifted to magnitude scale [26,27,75,76]. This resulted seismicity used in this paper are the same as those described by Lee
in the popular view that, because magnitude is an instrumental [42,43] and [53,54,56,57] and need not be repeated here. The new
measure, it should be used to describe the size of an earthquake. features in this work are (1) that the detailed spatial variations of the
Numerous empirical scaling equations, which determine the peaks and geologic site conditions are included directly in the calculation of
spectral amplitudes of strong motion, also started to favor the spectral amplitudes, and (2) the consequences of contributions from
magnitude scale. The diculty caused by this transition was that large distant earthquakes (from Vrancea in Romania) have been
seismic macrozonation and microzonation both require reliable de- included and quantied. The ways in which our results dier from
scriptions of earthquake occurrence rates, which, if available, were the old approach (based on probabilistic mapping of only peak ground
given in terms of intensity scales, while the coverage in terms of acceleration) will become apparent from what follows.
magnitudes was just starting. This problem was usually xed by A recent recommendation for peak ground accelerations to be used
converting the data on intensities to data on magnitudes, but the in the design of earthquake-resistant structures in Serbia (http://www.
price paid involved added uncertainties and multitudes of inconsistent seismo.gov.rs/) places Belgrade in the range from 0.02 to 0.08g (0.02
interpretations. We note that it was shown that, in terms of site 0.03g for the probability of being exceeded p=0.1 during an exposure
intensity scaling, seismic microzonation could be carried out with good period of Y =10 years; 0.05g for the probability of being exceeded p=0.
accuracy and without any use of magnitude scales [46,47,96]. Yet, most 1 during an exposure period of Y =50 years; and 0.08g for the
contemporary investigators continue to opt for scaling based on probability of being exceeded p=0.05 during an exposure period of Y
magnitudes. =50 years.). Fig. 1 shows a preliminary seismic hazard map of peak
Since the mid-1970s, after the rst direct empirical scaling equa- accelerations for the probability of being exceeded p=0.1 during an
tions of spectral amplitudes started to appear, it become possible to exposure period of Y =50 years.
formulate seismic zoning and microzoning in terms of more compre- Since 1970s, various projects have been organized with the aim of
hensive approaches. Such approaches could include the probabilities of reducing regional dierences in the formulation of earthquake-resis-
earthquake occurrence, the spatial distributions of earthquake sources, tant designs in the Western Balkan Countries (Albania, Bosnia and
the frequency-dependent attenuation of strong-motion amplitudes, Herzegovina, Croatia, Macedonia, Montenegro, and Serbia). These
and the site geologic and soil conditions [101,53,54,8389,98]. The projects have focused on a compilation of earthquake catalog data,
advantage of this new approach was that it considered simultaneously, seismic source modeling, determination of ground motion prediction
and in a balanced way, all factors that contributed to the end result. models, seismic hazard assessment, and development of seismic hazard
Comparisons with earthquake occurrence in southern California have maps. More recently, these projects have been intended to guide
conrmed the merits of this approach. For example, the seismic regional committees in their formulation of hazard maps for use with
microzonation maps based on the uniform hazard method (UHM) Eurocode 8 [19] (EC8) Type 1 and Type 2 spectra in earthquake-
[6,7] calculated and published in 1987 [47] for the Los Angeles resistant design. An example of such a map is shown in Fig. 1.
metropolitan area have not been contradicted by any of the earth- Lee and Trifunac [52] have described problems in the procedures
quakes that have occurred in the area since 1985 [106,99]. used in the development of hazard maps such as the one illustrated in
By the mid-1980s, a coherent picture started to emerge, in which Fig. 1. They noted that the values of design acceleration spectra (PSA)
observational analyses of site intensities and observed damage with site scaled by accelerations in these maps underestimate intermediate and
geologic and soil properties [37,61], regression analyses of peaks of long period PSA amplitudes.
recorded motions and their responses, and Fourier spectra could be Lee and Trifunac [52] map approximating peak accelerations for
combined into one, mutually consistent, whole [108,25,28,50,94]. Serbia, analogous to the map in Fig. 1, is shown in Fig. 2. It shows the
We note that, in terms of the geological site parameters and the soil amplitudes of PSA (T ) = 2PSV (T )/ T where PSA (T ) is the Pseudo
site parameters, both the derived scaling functions for site amplica- Absolute Acceleration spectrum, PSV (T ) is the Pseudo Relative
tion, as well as the corresponding parameters in the site database, are Velocity Spectrum, and T is the oscillator period. For T =0.04 s, this
correlated. This is to be expected because of the nature of the creation, gives an excellent upper bound for peak ground acceleration, since in
transport, and deposition of soil materials. In the data set used by the limit as T tends to zero, PSA (T ) tends to peak ground acceleration.
Trifunac [97], for example, although there were many (33%) deep-soil Although this is similar to what is shown in Fig. 1, working with upper
sites (SL =2) over sediments (s=0) and 10% rock-soil sites (SL =0) over bounds allows only an approximate comparison with Fig. 1.
basement rock (s=2), there were also many (27%) sti-soil sites (SL =1) Furthermore, the amplitudes in Fig. 1 are based on scaling for soil
over sediments (s=0) and 8% rock-soil sites (SL =0) over intermediate site condition A only, while the amplitudes in Fig. 2 are for sites on
geologic sites (s=1) [101]. Consequently, the use of regression models, geological basement rock (s=2), and on rock soil sites (SL =0), and use
which describe the site conditions in terms of only soil or only attenuation equations determined from strong motion recordings in
geological site parameters, averages out the dependence upon the site the former Yugoslavia [48,49].
parameter, which is not used in the analysis. This leads to an erroneous The approach implied in using the peak accelerations in Fig. 1 is
prediction of amplication by the local site conditions and, using the that the acceleration at rock sites (represented in Fig. 1 by site class
distribution of the site conditions in the study by Trifunac [97] as an A) can be modied to other soil site classes by an approximate one-

396
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 1. A preliminary map of peak accelerations (g), at site A (Vs 800 m/s) for the
probability of 10% exceedance and an exposure period of Y =50 years (source http://
www.seismo.gov.rs/).

dimensional analysis of ground motion amplication. This approach, Fig. 2. The upper bound on peak accelerations in terms of PSA (T ) = (2 /T ) PSV (g)
computed from UHS at T =0.04 s and only for local seismicity within 175 km of the site
however, neglects to consider the geological variations surrounding the
[55]. The contributions from the Vrancea earthquakes source in Romania are not
site and hence necessitates the formulation of microzonation maps that
included. Contours are shown for probability of 10% exceedance and an exposure period
do consider site geology. The literature on the eects of site conditions of Y =50 years (redrawn from [52]).
on strong earthquake ground motion is too voluminous to cite here,
and to show examples of why the site characterization in terms of soil
site classication A, B, C, should not be used [51]; however, a simple which the site conditions are parameterized in terms of only one
pictorial illustration as shown in Fig. 3 may serve the purpose. categorical variable (V30 or A, B, C, and D, for example). The site
In this gure, linear describes models that are based on linear parameterization, which uses soil and geological site parameters SL and
wave motion, while nonlinear refers to the nonlinear models of soil s respectively, will typically include the top two layers: the top one for
and sediments and includes the equivalent linearization methods that SL , and the top two for s [109]. The site parameterization, which
are common in geotechnical engineering. Chaos refers to the cases requires all three layers and hence may extend to considerable depths
that involve very large nonlinear deformations such as lateral spread- below the surface, will be required for the modeling of coherencies in
ing, sliding of blocks, landslides, and the like. The size of the box in parallel-layered structures [18,90].
Fig. 3 suggests the volume wherein a particular model of site The above mentioned geological and soil site conditions represent a
conditions applies. This size depends on the frequency band under characterization of the recording site as a point on the ground surface
consideration. For frequencies between 0.2 and 25 Hz, for example, (see 1D boxes in the top two layers in Fig. 3) and ignore the horizontal
and the layered half space with a surface layer having a shear wave extent and geometry of those conditions. In a series of papers on the
velocity of, say, 100 m/s, and the deep layer having 2 km/s, the range duration of strong ground motion, [6267] introduced the additional
of the considered wave lengths would be from several meters to 10 km site- and earthquake-specic variables Ri, j ,ri , and j . Ri, j represents the
and hence the size of the box representing site conditions is large. eective horizontal distance (in km) from the site ( j ) to a basement
Figure boxes 1D, 2D, and 3D refer to the dimensions of the model outcrop (i ) that is at distance ri from the earthquake epicentre and is
adopted for the description of the wave motion in the site representa- capable of reecting strong-motion waves from the source back toward
tion. 1D includes all models with vertically propagating shear waves, the site ( j ), thus contributing to the prolongation of strong ground
while 2D represents models that aim to approximate three-dimensional motion (see 2D, the top two layers and linear models in Fig. 3). j is the
space by representative two-dimensional cross Section 3D will model angle containing those outcrops as seen from the recording station ( j )
the three-dimensional space surrounding the site. and is evaluated separately for each earthquake. Both Ri, j and j were
The three layers in Fig. 3 schematically show the extent to which a found to contribute signicantly to the duration of strong motion and
particular parameterization of the site extends below the ground were therefore adopted as new site-specic variables in the empirical
surface. The shaded shallow top layer corresponds to the models in scaling of the duration of strong shaking. Through prolongation of

397
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

the selected design spectra is small in the high-frequency range, but


increases progressively at intermediate and long periods. The same
applies to all other design spectra, and to all code spectra, which are
scaled by the design peak acceleration [102,93].
Beyond this, there are other concerns regarding the hazard maps
based on peak accelerations only. These concerns are related to the
contribution to hazard from large distant earthquakes, which due to
relatively faster attenuation of high frequency motions cannot be seen
in peak acceleration amplitudes at large distances. The reader will nd
further discussion about these concerns in Lee and Trifunac [52].

2. Belgrade and its surroundingsgeological site


classication

The geological site classication for computations of Uniform


Hazard Spectra for the city of Belgrade is based on our interpretation
of the geological maps of the Belgrade region. There are four basic
geological maps that cover the area dened by the Regional Spatial
Plan of the Administrative Area of Belgrade [68] and the General
Plan of Belgrade 2014 [69]. These maps include: Basic geological map
Belgrade [8]; Basic geological map Panevo [10]: Basic geological
map Obrenovac [9]; and Basic geological map Smederevo [11].
The Regional Spatial Plan of the Administrative Area of Belgrade
(Fig. 4) is divided into cells of 1515 s in geographic coordinates
between 44 20' N and 45 00' N, that is, between 20 00' E and 20 52'
E. For each of the 33,280 analyzed cells, the site geology is described by
considering the predominant lithostratigraphic formations and their
depths.
A General Plan of Belgrade (Fig. 5) is also divided into cells of
1515 s in geographic coordinates between 44 37' N and 44 56' N,
Fig. 3. A schematic representation of site conditions showing selected variables that
that is, between 20 13' E and 20 42' E. For each of the 8816 analyzed
have been used in the empirical scaling equations of strong ground motion (modied
from [108]).
cells, the site geology was described by the predominant lithostrati-
graphic formations and their depths.
To determine the geological site parameter, s, for the city of
shaking, these site parameters will also aect the spectral amplitudes of Belgrade, the description of the site geology that was compiled for
strong motion. each cell was interpreted following the method proposed by Trifunac
It should be clear from the above discussion that describing the site and Brady [109]. Each cell was then classied as either basement
conditions in terms of only surface soil properties (V30 or A, B, C, and D, rock (s=2), alluvial and sedimentary deposits (s=0), or as inter-
for example) cannot lead to reliable results and hence should not be mediate sites (s=1). The result is shown in the right half of Fig. 5.
used. In the following, we will work with the description of site
parameters that describe site geology (s=0, 1, or 2) and site soil 3. Mapping seismic hazard via uniform hazard spectra
properties beyond the depth of 30 m (SL =0 and 1).
The methodology for the development of hazard maps begins with Mapping the amplitudes of PSA (T ) = 2PSV (T )/ T where PSA (T ) is
denition of spatial distribution of seismic sources and their activity in the Pseudo Absolute Acceleration spectrum, PSV (T ) is the Pseudo
time. The description of this activity, as used in this paper, is briey Relative Velocity Spectrum, and T is the oscillator period for which
outlined in Appendix A. The next step involves the selection of T =0.04 s gives an excellent upper bound for peak ground acceleration
attenuation equations, which describe how the desired quantity (since in the limit, as T tends to zero PSA (T ) tends to peak ground
representing strong ground motion attenuates with distance. The last acceleration.
step combines the contribution of all earthquake sources surrounding Fig. 6a through d show this upper bound for peak ground
the site to compute the distribution functions of the quantity being acceleration for a 10% exceedance probability, exposure periods of
evaluated at the site and plotting maps showing its spatial distribution. Y =10 and 50 years, and for rock (SL =0) and sti (SL =1) soil-site
Our concern with use of the maps such as the one shown in Fig. 1 is conditions. The geological site parameters s are included in the hazard
that those will not provide correct design forces that are consistent with calculations. The spatial variations seen in these gures are dominated
performance-based design criteria. Peak ground acceleration is not by the distance to seismic activity south of Belgrade and by the
suitable for most applications requiring quantication of earthquake geological site condition parameters s =0, 1, and 2 (see Fig. 5).
hazard [100,102,93,95], particularly for performance-based earth- PSA (0.04) amplitudes in the northwest corner of these gures are
quake resistant design when code species xed shape response 0.50.7 of those in the southeast corner, which shows that the distance
spectra anchored to peak ground acceleration [107]. The reader can from contributing earthquakes is important factor determining these
peruse the cited papers for a discussion of physical reasons that lead to amplitudes. At T =0.04 s, b2(2) =0.049 for =0.0 (see Table A-1 in
this view. [57])i.e., PSA (0.04) is by factor of 1.12 smaller at sites s =2 (e.g., at
The shapes of all classical design spectra (Biot, Housner, Newmark Avala) than on sites s =0 (e.g. in Zemun) (see Appendix C). Thus, in this
and Reg. Guide 1.60 spectra, [14,15,116]) are xed (represent averages example, geological site conditions alone would result in 1.12 times
taken over the spectral amplitudes normalized to unit peak ground smaller peak accelerations at Avala than in Zemun, but the eect is
acceleration). Consequently, the spectral shape variations with respect small relative to the distance eect and so the motions at Avala are 1.3
to the site conditions, distance, and especially earthquake size (magni- 1.4 larger than in Zemun. For a 10% exceedance probability and an
tude or intensity), are all averaged out. As a result, the uncertainty of exposure time of Y =50 years, the average peak acceleration in Fig. 6b

398
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 4. The Regional Spatial Plan of the Administrative Area of the city of Belgrade.

is consistent with our overall regional estimate of 0.13 g (shown in frequencies attenuate fast, and by the time they travel from Vrancea
Fig. 2). For sites on SL =1 (sti soil) (Fig. 6d) peak accelerations are to Belgrade, their contribution is smaller than the contribution from
about two times larger. local events. Fig. 6a through d show that peak accelerations are larger
We note that the amplitudes of peak accelerations shown in Fig. 6a on basement rock (s =2) than on sediments (s =0) by about 6070%,
through d are not sensitive to the occurrence of distant earthquakes in for both rock soil sites (SL =0) and for sti soil sites (SL =1). It is seen
the Vrancea source zone in Romania. This is because the high that starting with peak acceleration of 0.05 g (e.g., as in Fig. 1), and

Fig. 5. A General plan of Belgrade (left) and the geological site parameters (s=0, 1 or 2) in the same area (right).

399
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 6. a. The upper bound for peak ground acceleration in Belgrade in terms of PSA (0.04) (with units of g ), at rock soil sites SL =0 for a 10% exceedance probability and an exposure
period of Y =10 years. b. The upper bound for peak ground acceleration in Belgrade in terms of PSA (0.04) (with units of g ), at rock soil sites, SL =0, for a 10% exceedance probability and
an exposure period of Y =50 years. c. The upper bound for peak ground acceleration in Belgrade in terms of PSA (0.04) (with units of g ), at sti soil sites, SL =1, for a 10% exceedance
probability and an exposure period of Y =10 years. d. The upper bound for peak ground acceleration in Belgrade in terms of PSA (0.04) (with units of g ), at sti soil sites, SL =1, for a 10%
exceedance probability and an exposure period of Y =50 years.

modifying it to represent peak acceleration at sites B or C, cannot Fig. 7a through d show contours of PSA (T = 1.0 s ) with contribu-
produce the spatial distribution of design values as in Fig. 6a through d. tions from Vrancea earthquakes. The Vrancea earthquakes contribute
signicantly at T = 1.0 s and this contribution progressively increases
4. Contribution of vrancea earthquakes to seismic hazard in Serbia as one moves east [58]. The contribution of Vrancea sources
is also larger where the local seismicity is relatively low (see Appendix
Vrancea earthquakes occur at large epicentral distances from Serbia A).
(Fig. B1). This results in the attenuation of high-frequency spectral A detailed comparison of Fig. 7a through d with Fig. 6a through d
amplitudes [56-58] so that the UHS for peak accelerations in Serbia, shows that the spatial distribution of large and small amplitudes is
for a typical range of exceedance probabilities, are dominated only by dierent. This is caused by the variable shape of UHS and by the
local seismicity. Consequently, for typical hazard mapping of peak contribution to long period spectral amplitudes from Vrancea sources
accelerations in Serbia, Vrancea earthquakes can be ignored. However, with strong motion waves arriving from the east. Fig. 8a and b show
this is dierent for intermediate and long-period spectral amplitudes, this in terms of the ratios of spectral amplitudes computed relative to
in which Vrancea earthquakes (M > 6.5) contribute progressively more the spectral amplitudes without contributions from Vrancea sources.
at long periods of motion. In Fig. 7a through d, amplitudes of PSA (T = 1.0 s ) are 1.51.6

400
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 7. a. Uniform Hazard Acceleration Spectra, PSA (T ) , with units in g, at T =1 s, at rock soil sites, SL =0, for a 10% exceedance probability and an exposure interval of Y =10 years. b.
Uniform Hazard Acceleration Spectra, PSA (T ), with units in g, at T =1 s, at rock soil sites, SL =0, for a 10% exceedance probability and an exposure interval of Y =50 years. c. Uniform
Hazard Acceleration Spectra, PSA (T ), with units in g, at T =1 s, at sti soil sites, SL =1, for a 10% exceedance probability and an exposure interval of Y =10 years. d. Uniform Hazard
Acceleration Spectra, PSA (T ), with units in g, at T =1 s, at sti soil sites, SL =1, for a 10% exceedance probability and an exposure interval of Y =50 years.

times smaller on the basement rock sites (s =2, e.g., Avala) than on all comparisons among the same geological site conditions (s=0, 1 and
sediments (s =0, e.g., Zemun). This dierence is dominated by the 2) and for the oscillator period shown (T = 1.00 s). However, the
regression coecient b2(2) = 0.212 for =0.05 damping, with a corre- amplitudes of PSA (T ) = (2 / T ) PSV for UHS at T = 1.00 s at sites on
sponding factor of 1.63 [57]. This happens because the Vrancea sources sediments (s=0) will be larger than at the intermediate geological sites
are at essentially the same distance from all points in Fig. 7a through d, (s=1) and will have the smallest amplitudes occurring at the basement
Vrancea sources dominate vis a vis local seismicity, and hence the rock sites (s=2). We show most results in this paper for the sites on
scaling dierences in spectral amplitudes at long periods, between s =2 geological basement rock (s=2) to facilitate qualitative comparisons
and s =0 dominate. Local seismicity still plays a role in bringing the with previously published results, which are often shown only for the
spectral amplitudes further upa fact that is seen from the ratios in type A sites.
Fig. 8a and b that show increasing values from south to north,
consistent with the local activity arriving to Belgrade from the south. 5. Practical determination of UHS
The relative amplication of spectral amplitudes due to contribu-
tions from Vrancea earthquakes, shown in Fig. 8a and b, is the same for UHS at a building site can be calculated for each site-specic

401
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 8. a. Ratio of Uniform Hazard Acceleration Spectra, PSA (T ) , with and PSA (T ) computed without contribution from Vrancea earthquakes at T =1 s, at rock soil sites, SL =0, for a
10% exceedance probability and an exposure interval of Y =10 years. b. Ratio of Uniform Hazard Acceleration Spectra, PSA (T ) , with and PSA (T ) computed without contributions from
Vrancea earthquakes at T =1 s, at rock soil sites, SL =0, for a 10% exceedance probability and an exposure interval of Y =50 years.

condition, but these calculations are time consuming and require 6. Conclusions
detailed knowledge to select the required scaling parameters. A simple
alternative for engineering applications, which we introduced in the We have shown that the regional variations of UHS amplitudes over
mid-1980s, is to prepare UHS maps for given response periods, site a large city area can be considerable. In the examples shown for
conditions, and probabilities of exceedance and exposure time, and Belgrade, three sources of these variations are (1) the geological and
then to read the spectral amplitudes from the contours in the maps. soil site parameters, (2) the local seismicity mainly south of Belgrade,
Our report entitled, Microzonation of a Metropolitan Area, describes and (3) the contributions from large Vrancea earthquakes at a distance
this procedure and presents examples of how it can be executed [47]. of about 500 km east, in Romania. The relative contributions from
Fig. 9a through d show examples of such maps for log10 PSV (T ) these sources will dier for dierent cities, depending on the degree to
amplitudes at 12 periods ranging from 0.04 s to 2.00 s, for 5% which geologic and soil site conditions vary and the relative strength of
damping, horizontal motions at rock and sti soil sites (SL =0 and local versus distant earthquakes.
1), and for a 10% exceedance probability and exposure periods of Y =10 Almost everywhere in Serbia, peak accelerations will be dominated
and 50 years. By reading the spectral amplitudes at the given location, by the local seismic activity, and the contribution from Vrancea
the UHS of PSV can be constructed by interpolating the values read earthquakes can be neglected. Along the eastern borders of Serbia,
from the 12 periods. Examples of such an outcome are illustrated in Vrancea earthquakes will just begin to contribute to peak ground
Figs. 10a through 12b, at three sites with dierent geological site accelerations. However, in all areas in Serbia, the intermediate and
parameters (AvalaSpomenik Neznanom Junaku, with s =2 and SL =0; long-period spectral amplitudes are dominated by contributions from
Beogradjanka, with s =1 and SL =0; and Zemun P+20 building site large Vrancea earthquakes. This contribution is small in western Serbia
with s =0 and SL =1). and strongest in eastern parts of the country along the borders with
Figs. 10a through 12b all show monotonically increasing contribu- Romania and Bulgaria [52,58].
tions from Vrancea sources as the oscillator periods become longer. At Throughout this paper, in all calculations, we have used empirical
T =1 s, these increases range from 1.55 (for Avala) to 1.80 (Zemun) for scaling equations, which were developed from the strong motion data
p =0.1, and Y =50 years, for example. Overall, long-period spectral recorded in the former Yugoslavia [41,45,48,49,57]. This data had only
amplitudes tend to be larger on s =0 and SL =1 (ZemunFig. 12a,b) than few recordings on deep soil sites (SL =2, [101,105]) and hence our
on s =2 and SL =0 (AvalaFig. 10a,b), provided the sites are at empirical scaling equations for PSV amplitudes do not include the
comparable distances from large Vrancea sources. scaling in terms of SL =2. Therefore, examples including SL =2 are not
All results in this paper have been illustrated only for horizontal included in this paper. For future engineering design at SL =2 sites,
components of ground motion, but can easily be converted to represent additional analyses will be required to modify the UHS presented in
verical motions. This is because all empirical scaling equations we use this paper, to approximate the amplitudes expected at SL =2 sites. This
are formulated as logarithms of scaled quantities, so that all terms in can be based on several other scaling equations that include SL =2
these equations are like correction factors (when used on a linear scaling factors in California and elsewhere [44]. Deep soil sites appear
scale), which depend only on the period of motion. Thus, it is possible in the area of Novi Beograd (the northwest corner in Fig. 5), but an
to convert the results computed for horizontal motions to the results analysis of UHS for SL =2 sites is beyond the scope of this paper. This
for vertical motions by using the appropriate scaling factors. This will be addressed in our future work.
procedure and the required scaling functions for such conversions are It is noted that all results presented in this paper are only of
described in detail in Lee and Trifunac [47] and again in [56,57] and preliminary nature. Our scaling models for strong ground motion in the
will not be repeated here. former Yugoslavia are more than 20 years old and based on strong

402
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 9. a. An example of seismic microzoning map for Belgrade, for Pseudo Spectral Velocity (PSV), for a 5% fraction of critical damping, and computed with the UHS method for
local seismicity and contributions from Vrancea earthquakes combined, at the rock soil sites (SL =0) at 12 periods ranging from 0.04 to 2.00 s, for a probability of exceedance p=0.10
and an exposure time of Y =10 years. b. An example of a seismic microzoning map for Belgrade, for Pseudo Spectral Velocity (PSV), for a 5% fraction of critical damping, and
computed with the UHS method for local seismicity and contribution from Vrancea earthquakes combined, at the rock soil sites (SL =0), at 12 periods ranging from 0.04 to 2.00 s,
for a probability of exceedance p=0.10 and an exposure time of Y =50 years. c. An example of a seismic microzoning map for Belgrade, for Pseudo Spectral Velocity (PSV), for a

403
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 9. (continued)
5% fraction of critical damping, and computed with the UHS method for local seismicity and contributions from Vrancea earthquakes combined, at the sti soil sites (SL =1) at 12 periods
ranging from 0.04 to 2.0 s, for the probability of exceedance p=0.10 and an exposure time of Y =10 years. d. An example of a seismic microzoning map for Belgrade, for Pseudo Spectral
Velocity (PSV), for a 5% fraction of critical damping, and computed with the UHS method for local seismicity and contributions from Vrancea earthquakes combined, at the sti soil sites
(SL =1) at 12 periods ranging from 0.04 to 2.00 s, for the probability of exceedance p=0.10 and an exposure time of Y =50 years.

404
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 10. a. A comparison of Uniform Hazard Spectra of PSV at Avala (Spomenik Neznanom Junaku), for seismic activity with (dashed lines) and without (solid lines) contributions from
Vrancea earthquakes in Romania, for an exposure time of Y =10 years. b. A comparison of Uniform Hazard Spectra of PSV at Avala (Spomenik Neznanom Junaku), for seismic activity
with (dashed lines) and without (solid lines) contributions from Vrancea earthquakes in Romania, for an exposure time of Y =50 years.

Fig. 11. a. A comparison of Uniform Hazard Spectra of PSV at the site of building Beogradjanka, in the city center of Belgrade, for seismic activity with (dashed lines) and without (solid
lines) contributions from Vrancea earthquakes in Romania, for an exposure time of Y =10 years. b. A comparison of Uniform Hazard Spectra of PSV at the site of building Beogradjanka,
in the city center of Belgrade, for seismic activity with (dashed lines) and without (solid lines) contributions from Vrancea earthquakes in Romania, for an exposure time of Y =50 years.

405
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. 12. a. A comparison of Uniform Hazard Spectra of PSV at the site of building P+20 in Zemun, for seismic activity with (dashed lines) and without (solid lines) contributions from
Vrancea earthquakes in Romania, for an exposure time of Y =10 years. b. Comparison of Uniform Hazard Spectra of PSV at the site of building P+20, in Zemun, for seismic activity
with (dashed lines) and without (solid lines) contributions from Vrancea earthquakes in Romania, for an exposure time of Y =10 years.

motion data that was recorded up to the early 1980s [34]. Our model classications in terms of V30 or A, B, C, and D, for example, and that
for the scaling of PSV spectra in Serbia for earthquakes in the Vrancea researchers and practicing engineers will realize that scaling design
source zone was published recently, but it includes only the data for spectra by peak acceleration and xed shape spectra is not always
large Vrancea earthquakes. A more detailed and comprehensive conservative.
analysis will be possible only when new and abundant recordings of
local and Vrancea earthquakes become available in Serbia. It is hoped Acknowledgments
that the current analysis will help motivate and guide observational
programs to contribute such needed data. Research work of the third author was supported in part by the
We also hope that our profession will abandon outdated and ill- Serbian Ministry of Education, Science, and Technology Development
founded methods of scaling strong motion amplitudes, as in site Grant No. 36043.

Appendix A. Seismic activity and seismicity model

To illustrate computations for the strong motion earthquake hazard, we briey summarize the data on seismic activity surrounding investigated locations
and procedures to derive the relevant seismicity model. The earthquake activity in the region is assumed to be well represented by a catalog compiled by
merging records listed in the following catalogs: the BSHAP2 catalog (510 BCE2012) that was compiled during the BSHAP2 NATO-funded project by
seismologists from Albania, Croatia, Macedonia, Montenegro, Serbia, and Turkey [60]; the ISC-catalog (20132014) (http://www.isc.ac.uk).
If not reported as such, all magnitudes were converted to moment magnitudes (Mw) using regionally adjusted regressions between Mw on one
side and MS, ML, or mb on the other [60]. The catalog was declustered using time-space windows the size of which depend on the mainshocks
magnitude as described, for example, in Herak et al. [30], thus removing dependent events (foreshocks and aftershocks). In estimating recurrence
parameters, only mainshocks with magnitude exceeding 3.4 were considered. Fig. A1 shows the geographical distribution of earthquake epicenters.
Seismic activity is described by the earthquake occurrence rate in terms of moment magnitude (M=Mw), assuming the validity of the truncated
Gutenberg-Richter recurrence relation:
10 a bM M M M
N (M ) = min max ,
0 otherwise (A.1)
where N(M) is the number of events with magnitudes greater than or equal to M, and Mmin M Mmax is the allowable range of magnitudes. Mmin
varies in space and time according to the completeness of the contributing catalog(s), and the distribution of Mmax (Fig. A2) is assumed by taking
into account the magnitudes and intensities of the largest historical earthquakes and the lengths of the known major fault segments. The spatial and
temporal completeness of the catalog (see Fig. A3 for examples) was estimated as proposed by Herak et al. [30] (for details, see also [31]).

406
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. A1. Earthquake epicenters (main shocks only, M 3.4). The colors indicate focal depths according to the color scale on the right. The size of the symbols scales according to
magnitude.

The seismicity of the region was modeled using a variant of the distributed smoothed seismicity approach [e.g., Frankel [21] and Frankel et al.
[22]; see also Lapajne et al. [40] applied to model the seismicity of Slovenia]. This method was also used to compile the earthquake hazard maps for
Croatia [29], which are adopted as base maps in the National Annex to EC8 [32] For computations, the region is divided into a mosaic of rectangular
cells (0.1 0.1, or approximately 11.111.1=123 km2). For each cell, parameters a and b in Eq. (A.1), along with their uncertainties, are calculated
taking the magnitude completeness thresholds into account. Parameter b is estimated using the maximum-likelihood algorithm of Weichert [119],
which considers only earthquakes above their respective completeness thresholds within the smallest circle, with their centers in each of the cells
that hold at least 40 such events. The resulting spatial distribution of the b-value is shown in Fig. A2. a is assessed by counting the number of events
N1=N(M 3.5), N2=N(M 3.8), N3=N(M 4.2) within the circle that occurred after the corresponding onset of complete reporting. For each Ni,
N0i(b) =10a is estimated using Eq. (A.1), and representative a is obtained by taking the logarithm of the average. The seismicity rates thus obtained
are normalized to one year and to an area of 10,000 km2. After a-values were assigned to each of the grid-cells, the resulting spatial distribution is
smoothed using a bivariate, normal-elliptical smoothing kernel (for an example, see [40]), with the major axis directed along the predominant strike
of faults within the corresponding source zone whenever it is known. Here we extracted such information from the database of seismogenic fault,
which is one of the products of the SHARE-project ([12], http://diss.rm.ingv.it/share-edsf/), and also lists the corresponding predominant style of
faulting (normal, reverse, strike-slip). If the predominant strike was unknown, circular Gaussian smoothing was applied. The widths of Gaussian
distribution (standard deviations) along the major and minor axes are scaled to the expected maximum fault length and to the width of the surface
projection of the fault plane, respectively, estimated for the corresponding Mmax in the cell by relations of Wells and Coppersmith [120]. In this way,
the seismogenic potential estimated on the basis of past earthquakes that occurred in the neighborhood of each of the grid-cells is distributed along
the known fault systems. The seismicity model is then dened for each of the cells by the following parameters:

geographical coordinates of the cell center,


a-value, and its standard deviation,
b-value, and its standard deviation,
maximum moment magnitude, Mmax,
average focal depth (km), and its standard deviation,
predominant strike of the seismogenic faults (), and its standard deviation,
predominant style of faulting (unknown, normal, reverse, strike-slip).

407
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. A2. Parameter spatial distributions of the seismicity model. Top left: the a-value in the Gutenberg-Richter relation (normalized to 10,000 km2 and one year); Top right: the b-value
in the Gutenberg-Richter relation; Bottom left: the activity rate for M 3.5 (number of events normalized to one year and 10,000 km2); Bottom right: maximum magnitude, Mmax.

Our seismicity model is mostly based on the past seismicity record. It does not explicitly consider fault sources, as we feel that necessary data on
positions of seismogenic faults, their segmentation, geometry, Quaternary activity rates, and so forth, are still far from reliably known and complete.
Nevertheless, as noted above, the model includes some of the fault-specic data (predominant strike, the style of faulting, lengths of some known
segments) when such data were available.

408
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. A3. Examples of the catalogs completeness analyses. Left: the magnitude completeness threshold for the year 1980; Right: the beginning year of complete reporting for
magnitudes 3.9.

Appendix B. Seismicity of the Vrancea source zone

The Vrancea earthquakes in Romania occur near a sharp bend of the southeastern Carpathians [33]. The seismicity is concentrated in a high-
velocity focal volume in the depth range from about 60200 km [82]. At shallower depths (060 km), earthquakes occur sporadically and are not
large, with magnitude typically below 5.5. Source properties of large Vrancea earthquakes and the geology of the surrounding areas generally lead to
elongation of the intensity contours toward the southwest and Serbia. This is illustrated in Fig. B1 by the intensity contours of the 1977 earthquake
[39]. Larger cities in Serbia are at considerable epicentral distances from the Vrancea earthquake zone, ranging from 330 to 550 km. In the four
earthquakes (March 4, 1977; August 30, 1986; May 30, 1990; and May 31, 1991, which contributed strong motion data for the study by [56,57])
these distances were respectively: Belgrade (486524 km), Kraljevo (509552 km), and Negotin (353397 km). As Fig. B1 shows, for example,

Fig. B1. The Vrancea source zone is centered near 45.880 N and 26.980 E in Romania. Intermediate and large earthquakes occur at depths between 60 and 200 km. Also shown are the
MSK intensity contours for the March 4, 1977 earthquake (extracted from [39]) with Mw =7.4, H =98 km, and I0 = IX. The three strong motion stations, two in Serbia and one in
Bucharest, that recorded this event are shown by solid black circles.

409
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Fig. B2. Number of earthquakes greater or equal to M , per year, in the Vrancea source zone in Romania.

during the 1977 event, Belgrade, Smederevo, Kragujevac, Kraljevo, and Ni were in zone V (MSK scale), while Zajear was in zone VI, and Negotin
and Kladovo were in zone VII.
Historical data on earthquake occurrence in Vrancea have been compiled by Radu [74] and Purcaru [72]. The corresponding Gittenberg-Richter
trend log10 N = a bM , which describes the number of earthquakes N per year that are greater or equal to a magnitude M , has been studied by many
authors. Here we mention Radulian et al. [73], who give a = 4.77 0.24 and b = 0.89 0.04 ; Wenzel et al. [118] and Oncescu et al. [70] who give
a = 4.10 and b = 0.78; and Paskaleva [71], who gives a = 4.21 and b = 0.80 . The associated estimates of the largest possible magnitude for Vrancea
events range from 7.8 to 8.1. In Fig. B2, we also reproduce the observed values of N versus M for the period between 1980 and 1997 (redrawn from
[70]). For the calculations in this paper we assumed a = 4.10 0.24 , b = 0.78 0.04 , and that Mmax is uniformly distributed between 7.9 and 8.1.

Appendix C. Jumps in the microzonztion contours

While comparing the spectral amplitudes at dierent locations in the maps we prsesent in this paper, we noted, for example, that PSA (T = 0.04)
is smaller at sites s =2 (e.g. at Avala) than on sites s =0 (e.g. in Zemun), by a factor of 1.12. This can be explained as follows. In Lee et al. [57], the
scaling equation for PSV (T ) takes the form
log10 (PSV (T )) = M + Att (, M , T ) + b1 (T ) M
+b2(1) (T ) S (1) + b2(2) (T ) S (2) + b3 (T ) v
+b4 (T ) + b5(1) (T ) SL(1) + b5(2) (T ) SL(2) + b6 (T ) M2

In Table A-1, of the same paper, at T=0.04 b2(2) (T ) = 0.049 for s = 2 (when S (2) = 1), so that
log10 (PSV (T )) s =2 = log10 (PSV (T )) s =0 0.049
PSV (T )
or log10 PSV (T ) s =2 = 0.049
s =0

PSV (T ) 1
which gives PSV (T ) s =2 = 100.049 = 1.12
s =0
conrming that PSV (0.04) is smaller by a factor of 1.12 at sites s=2 than on sites s=0.

410
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

References 1956;9:115.
[28] Gutenberg B. Eects of ground on earthquake motion. Bull Seism Soc Am
1957;47:22150.
[1] Abrahamson NA, Silva WJ. Empirical response spectral attenuation relations for [29] Herak M et al., Republika hrvatska, Karta potresnih podruja, http://seizkarta.
shallow crustal earthquakes. Seism Res Lett 1997;68(1):94127. gfz.hr . ;2011.
[2] Akademia Nauk, SSSR. Seismicheskoe Reionirovanie Teritorii SSSR, Izdatelstvo [30] Herak D, Herak M, Tomljenovi B. Seismicity and focal mechanisms in North-
Nauka, Moskva; 1980. Western Croatia. Tectonophysics 2009;465:21220. http://dx.doi.org/10.1016/
[3] Akkar S, Bommer JJ. Empirical equations for the prediction of PGA, PGV, and j.tect.2008.12.005.
Spectral accelerations in Europe, the Mediterranean Region, and Middle East. [31] Herak M, Herak D. Analyses of seismicity as input for earthquake hazard studies
Seismol Res Lett 2010;81(2):195206. in Bosnia and Herzegovina, Chapter 1, in selected topics in earthquake engi-
[4] Ambraseys NN, Douglas J, Sarma SK, Smit PM. Equations for the estimation of neering from earthquake source to seismic design and hazard mitigation,
strong ground motions from shallow crustal earthquakes using data from europe published by ZIBL, Banja Luka, Republic of Srpska, Oct. 2009, 124; 2009.
and the middle east: Horizontal peak ground acceleration and spectral accelera- [32] Hrvatski zavod za norme. HRN EN 1998-1:2011/NA:2011, Eurokod 8:
tion. Bull Earthq Eng 2005;3:153. Projektiranje potresne otpornosti konstrukcija 1. dio: Opa pravila, potresna
[5] Ambraseys NN, Douglas J, Sarma SK, Smit PM. Equations for the estimation of djelovanja i pravila za zgrade Nacionalni dodatak (Eurocode 8: Design of
strong ground motions from shallow crustal earthquakes using data from europe structures for earthquake resistance Part 1: General rules, seismic actions and
and the middle east: Vertical peak ground acceleration and spectral acceleration. rules for buildings National Annex); 2011.
Bull Earthq Eng 2005;3:5573. [33] Ismail-Zadeh AT, Panza GF, Naimark BM. Stress in the descending relic slab
[6] Anderson JG, Trifunac MD. Uniform risk functionals for characterization of strong beneath the Vrancea Region, Romania. Pure Appl Geophys 2000;157:11130.
earthquake ground motion. Bull, Seism Soc Am 1978;68:20518. [34] Jordanovski LR, Lee VW, Mani MI, Olumeva T, Sinadinovski C, Todorovska MI,
[7] Anderson JG, Trifunac MD. Application of seismic risk procedures to problems in Trifunac MD. * Strong earthquake ground motion data in EQINFOS: Yugoslavia,
Microzonation, In: Proceedings of the second international conference on micro- Part I, Civil Eng. Report CE 87-05, Univ. of Southern California, Los Angeles, CA;
zonation, San Francisco; 1978b, p. 559-569. 1987.
[8] Basic geological map Belgrade. Basic geological map Belgrade, L34-113 (in [35] Kanai K. Semi-empirical formula for the seismic characteristics of the ground. Bull
Serbian: Osnovna geoloka karta SFRJ, BEOGRAD, L34-113), Geological Earthq Res Inst Tokyo 1957;35:30925.
Institute, Belgrade (RO Geoloki institut, Beograd), editorial and publication of the [36] Kanai K. The requisite conditions for predominant vibration of ground. Bull
Federal Geological Institute of Belgrade (Savezni geoloki zavod, Beograd); 1985. Earthq Res Inst Tokyo 1957;35:45771.
[9] Basic geological map Obrenovac. Basic geological map Obrenovac, L34-125 (in [37] Kanai K. Engineering Seismology. Tokyo, Japan: Univ. of Tokyo Press; 1983.
Serbian: Osnovna geoloka karta SFRJ, OBRENOVAC, L34-125), Institute for [38] Karnik VMicrozoning Programme within the UNDP-UNESCO Survey of
geological, hydro-geological, geophysical and geotechnical researches Geozavod, Seismicity of Balkan Region.In; Proceedings Microzonation Conference, Seattle,
Belgrade (Zavod za geoloka, hidrogeoloka, geozika i geotehnika istraivanja Washington, I; 1972, p.213-215.
Geozavod Beograd, editorial and publication of the federal geological institute of [39] Kronrod T, Radulian M, Panza G, Popa M, Paskaleva I, Radovanovich S,
Belgrade (Savezni geoloki zavod, Beograd); 1980. Gribovszki K, Sandu I, Pekevski L. Integrated transnational macroseismic data set
[10] Basic geological map Panevo. Basic geological map Panevo, L34-114 (in for the strongest earthquakes of Vrancea (Romania). Tectonophysics
Serbian: Osnovna geoloka karta SFRJ, PANEVO, L34-114), Institute for 2013;590:123.
Geological and Geophysical Researches, Belgrade (Zavod za geoloka i geozika [40] Lapajne J, ket Motnikar B, Zupani P. Probabilistic seismic hazard assessment
istraivanja, Beograd), editorial and publication of the Federal Geological Institute methodology for distributed seismicity. Bull Seism Soc Am 2003;93:250215.
of Belgrade (Savezni geoloki zavod, Beograd); 1975. [41] Lee VW. Pseudo relative velocity spectra in former Yugoslavia. Eur J Earthq Eng
[11] Basic geological map Smederevo. Basic geological map Smederevo, L34-126 1995;VII(1):1222.
(in Serbian: Osnovna geoloka karta SFRJ, SMEDEREVO, L34-126), Institute for [42] Lee VW. Discussion: prediction of horizontal response spectra in Europe. Int J
Geological, Hydro-geological, Geophysical and Geotechnical Researches Earthq Eng Struct Dyn 1997;26(2):28993.
Geozavod, Belgrade (Zavod za geoloka, hidrogeoloka, geozika i geotehnika [43] Lee VW. Empirical scaling of strong earthquake ground motion: Part I: attenua-
istraivanja Geozavod Beograd, editorial and publication of the Federal tion and scaling of response spectra []. ISET J Earthq Technol
Geological Institute of Belgrade (Savezni geoloki zavod, Beograd); 1980. 2002;39(4):21954.
[12] Basili R, Kastelic V, Demircioglu MB, Garcia Moreno D, Nemser ES, Petricca P, [44] Lee VW. Empirical scaling and regression methods for earthquake strong-motion
Sboras SP, Besana-Ostman GM, Cabral J, Camelbeeck T, Caputo R, Danciu L, spectraA review []. ISET J 2007;44(1):3969.
Domac H, Fonseca J, Garca-Mayordomo J, Giardini D, Glavatovic B, Gulen L, [45] Lee VW, Mani MI. Empirical scaling and regression methods for strong earth-
Ince Y, Pavlides S, Sesetyan K, Tarabusi G, Tiberti MM, Utkucu M, Valensise G, quake ground motions in YugoslaviaA review. Izgradnja 2009;2009(5
Vanneste K, Vilanova S, Wssner J. The European Database of Seismogenic Faults 6):23457.
(EDSF) compiled in the framework of the Project SHARE.http://diss.rm.ingv.it/ [46] Lee VW, Trifunac MD. *. Uniform Risk Spectra of Strong Earthquake Ground
share-edsf/, doi: http://dx.doi.org/10.6092/INGV.IT-SHARE-EDSF; 2013. Motion [85-05]Department of Civil Engineering, Report CE. Los Angeles,
[13] Bindi D, Luzi L, Pacor F, Sabetta F, Massa M. Towards a new reference ground California: University of Southern Calif.; 1985.
motion prediction equation for Italy: update of the Sabetat-Pugliese (1996). Bull [47] Lee VW, Trifunac MD. *. Microzonation of a Metropolitan Area [87-02]
Earthq Eng 2009;7(3):591608. Department of Civil Engineering, Report CE. , Los Angeles, California: University
[14] Biot MA. A mechanical Analyzer for the prediction of earthquake stresses. Bull of Southern California; 1987.
Seism Soc Am 1941;31:15171. [48] Lee VW, Trifunac MD. Frequency dependent attenuation of Strong earthquake
[15] Biot MA. Analytical and experimental methods in engineering Seismology. ASCE ground motion in Yugoslavia. Eur Earthq Eng 1992;VI(1):313.
Trans 1942;108:365408. [49] Lee VW, Trifunac MD. Empirical scaling of Fourier amplitude spectra in former
[16] Boore DM, Atkinson GM. Ground-motion prediction Equations for the Average Yugoslavia, with V.W. Lee. Eur Earthq Eng 1993;VII(2):4761.
Horizontal Component of PGA, PGV, and 5% - Damped PSA at Spectral Periods [50] Lee VW, Trifunac MD. Frequency dependent attenuation function and fourier
between 0.01 s and 10.0 s. Earthq Spectra 2008;24(1):99138. amplitude spectra of strong earthquake ground motion in California. Dept. of Civil
[17] Boore DM, Joyner WB, Fumal T. Equations for estimating horizontal response Eng. Report No. CE 9503, Univ. of Southern Cal., Los Angeles, CA; 1995.
spectra and peak acceleration from western North American earthquakes: a [51] Lee VW, Trifunac MD. Should average shear wave velocity in the top 30 m of soil
summary of recent Work. Seism Res Lett 1997;68(1):12853. be the only local site parameter used to describe seismic amplication?. Soil Dyn
[18] Ding H, Trifunac MD, Todorovska MI, Orbovi N. Corerence of dispersed strong Earthq Eng 2010;30(11):12508.
earthquake ground motion at small separation distances. Soil Dyn Earthq Eng [52] Lee VW, Trifunac MD. Seismic hazard maps in Serbia, submitted for publication;
2015;70:110. 2017.
[19] Eurocode 8. Design provisions of structures for earthquake resistance Part 1: [53] Lee VW, Herak M, Herak D, Trifunac MD. Uniform hazard spectra in north-
General rules, seismic actions and rules for buildings (EN1998-1) / CEN Techn. western Bosna and Hercegovina. Izgradnja 2010;64(56):282304.
Comm. 250 - SC8 , CEN, Brussels; 2005. [54] Lee VW, Trifunac MD, Herak M, Herak D. Uniform hazard earthquake accelera-
[20] Faik-Kara H, Trifunac MD. Two-dimensional earthquake vibrations in sedimen- tion spectra in Kraljevo contributions from local seismicity. Izgradnja
tary basins SH waves. Soil Dyn Earthq Eng 2014;63:6982. 2011;65(56):22735.
[21] Frankel A. Mapping seismic hazard in the central and eastern United States. Seism [55] Lee VW, Trifunac MD, Herak M, Herak D. Minimum radius of seismic activity for
Res Lett 1995;66:821. earthquake hazard analyses. Izgradnja 2011;65(56):21926.
[22] Frankel A, Mueller C, Barnhard T, Leyendecker E, Wesson R, Harmsen S, Klein F, [56] Lee VW, Trifunac MD, Bulaji B, Mani M. A preliminary empirical model for
Perkins D, Dickamn N, Hanson S, Hopper M. USGS national seismic hazard maps. frequency-dependent attenuation of Fourier amplitude spectra in Serbia from the
Earthq Spectra 2000;16:120. Vrancea earthquakes. Soil Dyn Earthq Eng 2016;83:16779.
[23] Gubin IE. On some questions of seismic regionalization, Akad. Nauk SSSR, Tr [57] Lee VW, Trifunac MD, Bulaji B, Mani M. Preliminary empirical scaling of
Geoz Inst 1954;25(152):3673. pseudo-relative velocity spectra in Serbia from the Vrancea earthquakes. Soil Dyn
[24] Gupta ID, Trifunac MD. Order statistics of peaks in earthquake response. J Eng Earthq Eng 2016;86:4154.
Mech, ASCE, EMD 1988;114(10):160527. [58] Lee VW, Trifunac MD, Herak D, Herak M, Bulaji B, Mani M. Problems with
[25] Gutenberg B. Eects of ground on shaking in earthquakes. EOS scaling design spectra by peak acceleration the contribution of Vrancea earth-
1956;37(6):75760. quakes to seismic hazard in Serbia. Izgradnja 2016;70:26576.
[26] Gutenberg B, Richter CF. Earthquake magnitude, intensity, energy, and accel- [59] Liao ZP. Seismic Microzonation in China. Instt Eng Mech, State Seismol Bur
eration. Bull Seism Soc Am 1956;46(2):10545. 1985:137.
[27] Gutenberg B, Richter CF. Magnitude and energy of earthquakes. Ann Geos [60] Markui S, Glerce Z, Kuka N, Duni L, Ivani I, Radovanovi S, Glavatovi B,

411
V.W. Lee et al. Soil Dynamics and Earthquake Engineering 97 (2017) 395412

Milutinovi Z, Akkar S, Kovaevi S, Mihaljevi J iR ali. An updated and unied site condition. Soil Dyn Earthq Eng 2015;79:25364.
earthquake catalogue for the Western Balkan region. Bull Earthq Eng [91] Trifunac MD. Zero baseline correction of Strong-motion Accelerograms. Bull
2016;14(2):32143. Seism Soc Am 1971;61(5):120111.
[61] Medvedev SV. Inzenernaya Seizmologiya, Akademia Nauk SSSR, Inst. Fiziki [92] Trifunac MD. A Note on correction of strong-motion accelerograms for instrument
Zemli, Moskva; 1962. response. Bull Seism Soc Am 1972;62(1):4019.
[62] Novikova EI, Trifunac MD. Modied mercalli intensity and the geometry of the [93] Trifunac MDCharacterization of response spectra by parameters governing the
sedimentary basin as the scaling parameters of the frequency dependent duration gross nature of earthquake source mechanism. In: Proceedings of the Fifth World
of strong ground motion. Soil Dyn Earthq Eng 1993;12(4):20925. Conf.Conference on earthquake Engineering, Rome, Italy, Vol. I, 1973, p 701-704.
[63] Novikova EI, Trifunac MD. Duration of Strong Earthquake Ground Motion: [94] Trifunac MD. Preliminary analysis of the peaks of strong earthquake ground
Physical Basis and Empirical Equations, Dept. of Civil Eng. Report No. CE 9302, motion dependence of peaks on earthquake magnitude, epicentral distance and
Univ. Southern California, Los Angeles, CA; 1993b. recording site conditions. Bull Seism Soc Am 1976;66:189219.
[64] Novikova EI, Trifunac MD. Duration of strong ground motion in terms of [95] Trifunac MD. Forecasting the Spectral Amplitudes of Strong Earthquake Ground
earthquake magnitude epicentral distance, site conditions and site geometry. Motion, In: Proceedings of the sixth world conference earthquake engineering,
Earthq Eng Struct Dyn 1994;23(6):102343. New Delhi, India, Vol. I, 139-152; 1977.
[65] Novikova EI, Trifunac MD. Inuence of geometry of sedimentary basins on the [96] Trifunac MD. Preliminary empirical model for scaling Fourier amplitude spectra
frequency dependent duration of strong ground motion. Earthq Eng Eng Vibration of strong motion acceleration in terms of modied mercalli intensity and geologic
1994;14(2):744. site conditions. Earthq Eng Struct Dyn 1979;7:6374.
[66] Novikova EI, Trifunac MD. Empirical Models of the Duration of Strong [97] Trifunac MD. Inuence of local soil & geologic site conditions on Fourier
Earthquake Ground Motion Based on the Modied Mercalli Intensity, Dept. of spectrum amplitudes of recorded strong motion accelerations. Report No. CE 87
Civil Eng. Report CE 94-01, Univ. Southern California, Los Angeles, California; 04, Dept. of Civil Engineering, Univ. of Southern California, Los Angeles, CA;
1994c. 1987.
[67] Novikova EI, Trifunac MD. Frequency dependent duration of strong earthquake [98] Trifunac MD. Seismic Microzonation Mapping Via Uniform Risk Spectra, In:
ground motion: updated empirical equations, Dept. of Civil Eng. Report No. CE Proceedings of the 9th World Conference Earthquake Engineering, VII, 75-80,
95-01, Univ. Southern California, Los Angeles, California; 1995. Tokyo-Kyoto, Japan; 1988.
[68] Ocial Gazette of the City of Belgrade, Edition for the September 22, 2011. The [99] Trifunac MD. Threshold magnitudes, which exceed the expected ground motion
City of Belgrade Ocial Gazette, Vol. 55, No. 38; 2011. during the next 50 years in a metropolitan area. Geozika 1989;6:112.
[69] Ocial Gazette of the City of Belgrade, Edition for the September 18, 2014. The [100] Trifunac MD. Dependence of fourier spectrum amplitudes of recorded Strong
City of Belgrade Ocial Gazette, Vol. 58, No. 70; 2014. earthquake accelerations on magnitude, local soil conditions and On depth of
[70] Oncescu MC, Marza VI, Rizescu M, Popa M. The Romanian earthquake catalogue sediments. Int J Earthq Eng Struct Dyn 1989;18(7):9991016.
between 19841997. In: Wenzel F, Lungu D, Novak O, editors. Vrancea [101] Trifunac MD. How to model amplication of strong earthquake ground motions
Earthquakes: Tectonics, Hazard and Risk Mitigation. Dordrecht: Kluwer; 1999. p. by local soil & geologic site conditions. Earthq Eng Struct Dyn
437. 1990;19(6):83346.
[71] Paskaleva I. Personal communication; 2016. [102] Trifunac MD. Should peak accelerations be used to scale design spectrum
[72] Purcaru G. He Vrancea, Romania, earthquake of march 4, 1977 a quite amplitudes? Proc. 10th world Conf.. Earthq Eng Madr, Spain 1992;10:581722.
successful prediction. Phys Earth Planet Inter 1979;18:27487. [103] Trifunac MD.Nonlinear Soil Response as a Natural Passive Isolation Mechanism,
[73] Radulian M, Mandrescu N, Panza GF, Popescu E, Utale A. Characterization of Paper II The 1933, Long Beach, California Earthquake, Soil Dynamics and
seismic zones in Romania. Pure Appl Geophys 2000;157:5777. Earthquake Eng. 23(7); 2003, 549-562.
[74] Radu C. Contribution a letude de la seismicite de la Romanie at comparaison avec [104] Trifunac MD. 75th Anniversary of the response spectrum method A historical
la seismicite de sud-est de la France [Ph.D. Thesis]. Universite de Strasbourg; review. , Soil Dyn Earthq Eng 2008;28(9):67685, [2008].
1974. p. 404. [105] Trifunac MD. The Nature of Site Response During Earthquakes, Proc. NATO ARW
[75] Richter CF. An instrumental earthquake magnitude scale. Bull Seism Soc Am workshop in Borovec, 30 August 3 September 2008, Bulgaria, published in
1936;25:132. Coupled Site and Soil-Structure Interaction eects with Applications to Seismic
[76] Richter CH. Elementary Seismology. San Francisco: Freeman and Co.; 1958. Risk Mitigation. In: Schantz T, Iankov R, editors. NATO Science for Piece and
[77] Richter CF. Seismic Regionalization, bull. Seismol Soc Am 1959;49:12362. Security Series C: Environmental Security. Springer Science + Business Media, B.
[78] Snchez-Sesma FJ, Crouse CB. Eects of site geology on seismic ground motion: V.; 2008. p. 331, [2009].
early history. Earthq Eng Struct Dyn 2015;44(7):1099113. [106] Trifunac MD. Seismic Microzonation and Near-Field Eects, In: Proceedings of
[79] Stojkovi MB.Seismic microzoning of the city of Banja Luka area, Proceedings of the 9th International Workshop on Seismic Microzoning and Risk Reduction
the International conference on earthquake engineering, October 26-28, 2009, (9IWSMRR), Cuernavaca, Morelos, Mexico, 21-24 February 2010; 2010.
Banja Luka, Book 4 - Seismology and engineering seismology; 2009a, p.345-357. [107] Trifunac MD. Earthquake response spectra for performance based design A
[80] Stojkovi MB. Characteristic damage degrees of masonry residential structures critical review. Soil Dyn Earthq Eng 2012;37(6):7383.
due to the 1969 earthquake which struck the Banja Luka area, Proceedings of the [108] Trifunac MD. Site conditions and earthquake ground motion A review. Soil Dyn
International conference on earthquake engineering, October 26-28, 2009, Banja Earthq Eng 2016;11(4):22941.
Luka, Book 4 - Seismology and engineering seismology, ; 2009b, p. 359-371. [109] Trifunac MD, Brady AG. On the correlation of seismic intensity scales with the
[81] Stojkovi M, Kirijas T, Mihailov V, Petrovski J, Paskalov T, Bikovski B, Galbov S, peaks of recording strong ground motion. Bull Seismol Soc Am 1975;90:88100.
Kapsarov R. Seismic microzonation of Banja Luke, Report No. S-21/71, Institute [110] Trifunac MD, Ivanovi SS. Reocurrence of site specic response in former
of Earthquake Engineering and Engineering Seismology (IZIIS), Skopje; 1971. Yugoslavia part I: Montenegro. Soil Dyn Earthq Eng 2003;23(8):63761.
[82] Telesca L, Alcaz V, Sandu I. Analysis of 19782008 crustal and sub-crustal [111] Trifunac MD, Ivanovi SS. Reoccurrence of site specic response in former
earthquake catalog of Vrancea region, natural hazards and earth. Syst Sci Yugoslavia Part II: Friuli, Banja Luka and Kopaonik. Soil Dyn Earthq Eng
2013;12:13215. 2003;23(8):66381.
[83] Todorovska MI, Trifunac MD. The role of earthquake hazard maps in loss [112] Trifunac MD, Todorovska MI. Can aftershock studies predict site amplication?
estimation: a study of the Northridge earthquake, Discussion of paper by R.B. Northridge, CA, earthquake of 17 January 1994. Soil Dyn Earthq Engrg
Olshansky. Earthq Spectra 1998;14(3):55763. 2000;19(4):23351.
[84] Todorovska MI, Trifunac MD. A seismic hazard model for peak strains in soils [113] Trifunac MD, Todorovska MI. Long period microtremors, microseisms and
During Strong earthquake Shaking. Earthq Eng Eng Vib 1996;16:112. earthquake damage: Northridge, CA, earthquake of 17 January 1994. Soil Dyn
[85] Todorovska MI, Trifunac MD. Hazard mapping of normalized peak strain in soil Earthq Engrg 2000;19(4):25367.
During earthquakes - microzonation of a metropolitan area. Soil Dyn Earthq Eng [114] Trifunac MD, Hao TY, Todorovska MI. On reoccurrence of site specic response.
1996;15(5):3219. Soil Dyn Earthq Eng 1999;18(8):56992.
[86] Todorovska MI, Trifunac MD. Discussion of "The role of earthquake hazard maps [115] Udwadia FE, Trifunac MD. Characterization of response spectra through the
in loss estimation: a study of the Northridge Earthquake," by R.B. Olshansky. statistics of oscillator response. Bull Seism Soc Am 1974;64(1):20519.
Earthq Spectra 1998;14(3):55763. [116] United States Atomic Energy Commission . Design response spectra for seismic
[87] Todorovska MI, Gupta ID, Gupta VK, Lee VW, Trifunac MD. *. Selected topics in design of nuclear power plants, Regulatory Guide No1.60, Washington, D.C. U.S:
probabilistic seismic hazard analysis, Dept. of Civil Eng. Report No. CE 95-08, Atomic Energy Commission; 1973.
Univ. of Southern California, Los Angeles, California; 1995. [117] Walker B. Earthquake, Time-Life Books, Alexandria, Virginia; 1982.
[88] Todorovska MI, Lee VW, Trifunac MD. Shaking hazard compatible methodology [118] Wenzel F, Lorenz FP, Sperner B, Oncescu MC. Seismotectonics of the Romanian
for probabilistic assessment of permanent ground displacement across earthquake Vrancea Area, in vrancea earthquakes :tectonics, hazard and risk mitigation,
faults. Soil Dyn Earthq Eng 2007;27(6):58697. (edited by Wenzel et al.); 1999, p. 15-25.
[89] Todorovska MI, Trifunac MD, Lee VW. Probabilistic Assessment of Permanent [119] Weichert DH. Estimation of the earthquake recurrence parameters for unequal
Ground Displacement Across Earthquake Faults, Proceedings of Earthquake observation periods for dierent magnitudes. Bull Seism Soc Am
Engineering in the 21st Century to mark 40th anniversary of IZIIS - Skopje, 1980;70:133746.
August 28-Sptember 1, 2005, Skopje and Ohrid, Macedonia; 2005. [120] Wells DL, Coppersmith JK. New empirical relations among magnitude, rupture
[90] Todorovska MI, Trifunac MD, Ding H, Orbovi N. Coherency of dispersed width, rupture area, and surface displacement. Bull Seism Soc Am
synthetic earthquake ground motion at small separation distances: dependence on 1994;84:9741002.

412

Potrebbero piacerti anche