IC/98/26

United Nations Educational Scientific and Cultural Organizationand

International Atomic Energy Agency

THE ABDUS SALAM INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS

MODELLING, FOR MICROZONATION PURPOSES,OF THE SEISMIC GROUND MOTION IN BUCHAREST,

DUE TO THE VRANCEA EARTHQUAKE OF MAY 30, 1990

C.L. MoldoveanuNational Institute for Earth Physics, Bucharest. Romania

and

G.F. PanzaDipartimento di Scienze delta Terra, Universitd degli Studi di Trieste,

Trieste, Italyand

The Abdus Salam International Centre for Theoretical Physics, SAND Group,Trieste, Italy.

MIRAMARE - TRIESTE

March 1998

ABSTRACT

The Vrancea seismoactive region, characterized by intermediate-depth earthquakes, isthe quake source that has to be taken into account for microzonation purposes ofBucharest that could suffer serious damage also because of the severe local site effects.The strong seismic events originating in Vrancea have caused the most destructivedamage experienced on the Romanian territory and may seriously affect vulnerable highrisk constructions (such as nuclear power plants, chemical plants, large dams, pipelinesetc.) located on a wide area, from Central Europe to Moscow.

Realistic numerical simulation, describing the propagation of the seismicwavefield generated by a given quake in a complex geological structure, is a powerfultool, that may be efficiently used to estimate the ground motion for microzonation of thewhole Bucharest area.

The realistic modelling of ground motion is carried out by means of asophisticated hybrid technique that combines modal summation (Panza, 1985; Vaccariet al., 1989; Florsch et al., 1991; Panza, 1993; Romanelli et al., 1996) and finitedifference (Fah 1991; Fah and Panza, 1994; Fah et al., 1994). The input data necessaryfor computations are the laterally variable anelastic structural model and the focalmechanism of the seismic source.

The medium is modelled with a regional layered structure (bedrock structure),containing the seismic source and assumed to be representative of the path from thesource to Bucharest, and a local structure, that is a NE20°SW oriented cross section,describing the local structure of Bucharest, along the studied path. The seismic source isdescribed as a double-couple, buried in a layered medium, and corresponds to the focalmechanism of the May 30, 1990 Vrancea earthquake. The upper frequency limitconsidered in the computations is 1.0 Hz, and this allows us the modelling of seismicinput appropriate for ten storeys and higher buildings.

The simulated signals are satisfactorily compared with the available instrumentalrecords from Magurele station (44.347°N, 26.030°E), and stability tests are performedwith respect to the variation of focal mechanism, regional and local structure.

1. Introduction

Bucharest, the capital of Romania, is strongly affected by the intermediate-depthVrancea earthquakes. Four destructive events in this century, and the presence of morethan two million inhabitants, together with a remarkable number of high seismic riskvulnerable buildings and infrastructures, makes the microzonation of the city a goal ofmain importance, as an essential step toward the mitigation of the local seismic risk.Two main approaches can be considered for mapping of the seismic ground motion formicrozonation purposes: (a) collection and extended use (for engineering purposes) ofthe recorded strong motion data, and (b) advanced modelling techniques that allow thecomputation of a realistic seismic input which can compensate the lack of strong motionrecords. The ideal situation is represented by the possibility to follow both ways and tocalibrate modelling with the available recordings. In practice, strong motion data arevery scarce and correspond to events occurred during the last 20-30 years.

Bucharest represents a typical case where the complementary use of modellingand data processing may allow us to obtain quite useful predictions of the expectedground motion since a few strong motion records of the last three strong Vrancea events(1977, 1986, 1990) are available. Using the accumulated information about seismicsources, sampled medium and local soil conditions, together with realistic groundmotion modelling techniques, it is now possible to estimate for microzonation purposesthe local behavior of a given site. Whenever possible, the complementary use of the twoapproaches should be followed because of (1) the high installation and operation cost ofa dense permanent seismic network, and (2) the necessity to calibrate with observationsthe synthetic signals, obtained using the geological and geotechnical knowledgeaccumulated for the investigated region.

2. Seismicity of Vrancea Region

The Vrancea seismic region, situated beneath the bending of Eastern Carpathian Arc, isresponsible for the most destructive effects experienced in Romania in historical time.The strong seismic events originating in this seismogenic area may seriously affectvulnerable high risk constructions such as nuclear power and chemical plants, largedams, pipelines, high buildings etc., located on a wide territory, from Central Europe toMoscow, making this intermediate-depth source a regional danger.

The main feature of the Vrancea region is the subcrustal (intermediate-depth)persistent background seismic activity that consists of about 10-15 events per month(2 ,5<ML<5.5) and three to five strong events (Mw - 7.0) per century. This seismicactivity is confined in a very well delimited parallelepiped about 100 km long, 40 kmwide, with a vertical extension from 50-60 km to 160-170 km of depth. The epicentralarea of the subcrustal seismic activity, oriented NE-SW, is of about 3000 km^ and it ispartly overlapped by the epicentral area of the crustal activity (Radulian et al., 1998). Aseismic gap is observed between the crustal and subcrustal activity, its depth rangespanning from about 40 km to about 60 km. The four strongest earthquakes thatoccurred in Vrancea (Mo>64019N-m, Mw>7.1) during this century (in 1908, 1940, 1977and 1986) caused many casualties and large damage, not only on the Romanian territory,

but also in other parts of Europe (Oncescu et al., this issue). The last twenty years werequite active for this region, the major events occurring on: March 4, 1977 (Mw=7.4),August 30, 1986 ( M ^ . l ) , May 30, 1990 (Mw=6.9), May 31, and 1990 (Mw=6.4).

A reverse faulting mechanism with the T-axis almost vertical and the P-axisalmost horizontal characterizes the major intermediate-depth Vrancea events. The samemechanism is observed for more than 90% of the studied events, regardless of theirmagnitude (Enescu, 1980; Enescu and Zugravescu, 1990, Oncescu and Trifu, 1987).The fault plane orientation can be divided into two main groups mainly oriented on a:(1) NE-SW direction, with the P-axis perpendicular to the Carpathian mountain arc (e.g.the Mw=7.4 event of March 4, 1977, Mw=7.1 event of August 30, 1986, and Mw=6.9event of May 30, 1990); and (2) NW-SE direction, with the P-axis parallel to themountain arc (e.g. the Mw=6.4 event of May 31, 1990).

Several models have been proposed to explain some aspects of the tectonicprocesses in Vrancea, but the driving mechanism of the intermediate-depth seismicity isnot yet completely understood.

Accordingly with McKenzie (1970, 1972), the subcrustal seismicity occurs in avertical relic slab sinking into the asthenosphere and now overlaid by continental crust.The origin of this body can be the rapid south-west motion of the plate containing theCarpathian Arc and the surrounding regions, relative to the Black Sea plate, or thegravitational sinking of an oceanic slab, detached from the continental lithosphere, intothe asthenosphere (Fuchs et al., 1978).

The double subduction model, proposed by Oncescu (1984) and Oncescu et al.(1984), suggests that the decoupling of the sinking slab could be caused by the north-west push of the Black Sea subplate, the slip being caused by the gravitational force(Oncescu and Trifu, 1987). The seismicity is taking place in a cold relic slab, denser andmore rigid than the surrounding mantle that sinks due to the gravity. The hydrostaticbuoyancy forces help the slab to subduct, but the viscous and frictional forces act as aresistance to its descent. At intermediate depth these forces produce an internal stresswith the principal axis directed downward, and earthquakes occur in response to thisstress.

Enescu and Enescu (1993) propose as an explanation of the Vrancea subcrustalseismicity an active continental subduction which has a smaller velocity than any otherprocess of this kind observed on the earth up to the present.

Finally, the strong earthquakes at intermediate-depth account for high pressurefaulting processes, which could be facilitated by the stress produced by heterogeneity involume change, due to basalt-eclogite phase transition (Ismail-Zadeh et al., 1996).

3. Ground Motion Modelling

Recently, complex methods combining theoretical and computational progresses havebeen developed to model, at a given site, the seismic ground motion which is the resultof the contribution of three main factors: source, traveled path, and local site conditions.These factors describe how the earthquake source controls the generation of seismicwaves, the effect of the earth on these waves as they travel from the source to aparticular location, and the effect of the uppermost rocks and soils, together with the sitetopography, on the resultant ground motion, respectively.

The simulation of the ground motion is performed using the hybrid method (Fah,

1991; Fan and Panza, 1994; Fah et al., 1994), that combines the modal summationtechnique (Panza, 1985; Vaccari et al., 1989; Florsch et al., 1991; Romanelli et al.,1996), used to describe the SH and P-SV-wave propagation in the anelastic bedrockstructure, with the finite difference technique (Alterman and Karal, 1968; Boore, 1972;Kelly et al., 1976) used for computation of wave propagation in the anelastic, laterallyvarying sedimentary area of Bucharest, located about 170 km away from the epicentralarea.

For the ground motion simulation in Bucharest area, due to Vranceaintermediate-depth seismic events, we consider the bedrock structure shown in Figure 1,the local structure given in Figure 2, and the source of the May 30, 1990 event, asdescribed by CMT catalogue (Dziewonski et al., 1991): Lat.=45.92°N, Long,=26.81°E,depth=74±16 km, dip=63°, rake=101°, strike=236°, seismic moment Mo=3.04019Nm,and ^ ^ 6 . 9 . The bedrock structure, shown in Figure 1 down to a depth of 250 km,represents an averaged regional model for Vrancea-Bucharest path. This model iscompiled by Radulian et al. (1986) considering: (a) for the crust the velocity model usedfor event location with the Romanian telemetered observatories, and (b) for the deeperstructure a low-velocity channel from 90 to 190 km with standard Q values. Below thedepth of 250 km, an average continental model is adopted. To investigate the influenceof Vs and Q variations within resonable limits, four variants of the bedrock structuregiven in Figure 1 have been considered. Vs changes affect significantly only arrivaltimes of the signals, and Q variations do not produce relevant changes in the simulatedwaveforms.

Istrita

3 1=0

• ' \: ^

1 1 1

O

5O

1OO

15O

20O

25O

1

0

50

1O0

150

200

250 .1 xO 1 2 3D.n.ity (g/tm1)

Q 2 4-(km/,)

500 1OPD 1500

Figure 1. Bedrock structure (Istrita). Variations with depth of density, P- and S-wavephase velocity, quality factor, Q, for P- and S-wave, in the uppermost 250 km.

The presence of unconsolidated sediments with irregular geothechnicalcharacteristics and distribution in space for the Bucharest area has been detected by theGeological Prospecting Enterprise during the prospecting work made for theconstruction of the Bucharest subway, hi this framework more than 2000 bore-holeswere analyzed, and the seismic wave velocity was measured by seismic refraction inmore than 200 points. We use the synthesis of these results, given by Mandrescu andRadulian (1998), for constructing the simplified local structure shown in Figure 2. Thelocal structure, formed by alluvium, loess like, gravel, sand, clay and sandy marl,corresponds to the cross section oriented NE 20° SW, marked on the city sketchpresented in Figure 3. The quality factors, Qs and Qp, are evaluated from empiricalcorrelations with geology, and from similar data published in the literature.

The earthquake triggered the strong motion instruments SMA-1 in Bucharest -Magurele station (44.347°N, 26.030°E), and we use this three-component record,Figure 4 (a,b), to validate our computations, performed with a minimum period of 1.0 s,comparable with the natural period of oscillation of 10 and more storeys buildings. Theavailable SMA-1 records are band-pass filtered from 0.1 to 15 Hz (Figure 4a), and havebeen low-pass filtered with a cut-off frequency of 1.0 Hz (Figure 4b). The signal-to-noise ratio for the three components of this signal is at least 10.

4. Results

The epicentral distance considered for the local structure ranges from 164 km to 185km. Magurele station is located in the south-western part of the town, at one end of thecross section, shown in the city sketch presented in Figure 3. The following stabilitytests are performed considering the bedrock structure and the epicentral distance ofMagurele station.

Epicentral distance [km]

0.65

p = 1.9 g/cm3Vp=1.00kro/s;Qp=66Vs=0.37 km/s; Qs=30

p = 2.15 g/cm3Vp=2.00 km/s; Qp=66Vs=0.90 km/s: Os=30

Vp=1.20bn/s;Qp=66Vs-0.44km/s; Qs=30

p = 2.2 g/cm3Vp=2.30 km/s; Qp=110Vs=1.10km/s; Qs=50

p = 2.1 g/cm3Vp=1.90km/s;Qp=66Vs=0.78 km/s; Qs=3C

A Magureie

Figure 2. Simplified local structure used for the ground motion modellinginduced by the May 30,1990, Vrancea earthquake, in Bucharest.

4.1 TESTS OF STABILITY

A large change is observed when the source depth (H) is modified. The accelerationtime series for SH and P-SV waves of an epicentral distance of 181.7 km (Magurelestation), are computed for the earthquake mechanism of the May 30, 1990, Vranceaevent when H varies from 100 to 10 km, with a step of 10 km, using the modalsummation method (Figure 5). The amplitudes of the signals are related to a seismic

moment of 1013N-m, and the First trace, from the top, corresponds to the focal depth of100 km. By imposing the constraint that the ratio between the simulated components ofthe motion is as close as possible to the one between the observed components ofmotion, than H has to be around 60 km. This choice is well compatible, within the errorlimit, with the CMT depth determination.

The shape, peak ground acceleration (PGA) and ratios between the componentsof the simulated signal are almost unchanged when the rake angle varies from 91° to111°. Significant variations of the shape and PGA are caused by the variation of the dipangle in the range from 53° to 73°, the most sensible component to this sourceparameter variation being the radial one. As a result of the two tests, the rake and dipangles are taken 101° and 63°, respectively, as given by the CMT catalogue(Dziewonski et al., 1991).

Figure 3. Bucharest city sketch and the cross section considered for computations. Bore-holes(ticks) along the profile (thick line); the position of Magurele station is indicated by a triangle.

The introduction of a seismic source with finite dimensions may cause a strongmodification of the main features of the synthetic signals - shape, PGA, duration, ratiobetween components (e.g. Panza and Suhadolc, 1987). Since we do not have anyreliable estimate of the parameters necessary to describe an extended source, tominimize the number of the free parameters, in all the following computations a double-couple point-source is considered. As we will see, in spite of such a strongsimplification of the source process, the synthetic signals reproduce most of the mainfeatures of the observations, which are relevant for seismic engineering.

The assumption of a simplified local structure (Figure 2) for the Bucharest cross-section, when the upper 15 m of very low velocity (S-wave velocity is in the range 0.1 -0.2 km/s) unconsolidated material are omitted, is justified by the frequency range of oursimulation - up to 1 Hz. In fact, the frequency of the mechanical resonance of a 15 mthick layer with a shear-wave velocity of 0.1 km/s is about 1.6 Hz.

4,2 HYBRID METHOD RESULTS

The accelerograms (radial, R, vertical, V, and transverse, T, components) obtained along

50-

(a)

0 10 20 30 40 50

10-

(b)

10 20 30 40 50 3

Figure 4, May 30, 1990 Vrancea earthquake. Three-component records at Magurele station in cm/s2;a - band pass filtered from 0.1 to 15 Hz , b - low pass Buterworth filtered, cut-off frequency 1.0 Hz.Radial component (R), vertical component (V), and transverse component (T). The origin of recordscorresponds to the triggering time,

the cross section shown in Figure 2, for a set of equally spaced sites, are presented inFigure 6 (a, b). The signals are computed for a seismic moment Mo=3.0-1019N'm and acorner frequency of 0.41 Hz estimated from the available acceleration records. The firstand the last traces (from top to bottom) correspond to the epicental distances of 166.7km and 184.7 km, respectively, the distance between two successive receivers is 3km.The comparison of the two sets of synthetic signals corresponding to the laterally

varying model made by means of the hybrid method (Figure 6 a) along the local profilein Bucharest, and the simulation made considering the bedrock structure (Figure 6 b),

2 5

O-175E-&4

a. ii i&-e*

- * »

7 5 125 25 1 2 5

Figure 5. The three-component accelerograms simulated for the bedrock structure at the epicentraldistance of 181.7 km. Focal depth varies from 100 to 10 km (from top to bottom), with a step of 10 km;acceleration - in cm/s2 (peak values are given in the right part), time - in seconds; the amplitudescorrespond to a seismic moment of 1013N-m.

shows that the local effects, in terms of PGA, affect mainly the T and V components(amplifications up to 2), while for the R component the amplification is not larger than1.3. The presence of the sedimentary cover modifies, as it should be expected, the shapeof the signals, local surface waves being excited, especially in the radial and verticalcomponents. For this reason, the total duration of the accelerograms in the laterallyvarying model is about 5 seconds longer with respect to the homologue ones obtainedfor the bedrock structure. The epicentral distance of the sixth trace from the top, Figures6a and 6b, corresponds to the location of Bucharest-Magurele station (44.347°N,26.030°E). The time series simulated for this particular receiver, shown in detail inFigure 7, are in good agreement with the observed ones, low-pass filtered with a cut-offfrequency of 1.0 Hz (Figure 4, b).

The synthetic time series can be processed in the same way as the realseismograms. Typical ground motion related quantities, computed in seismicengineering, are the PGA and the quantity W defined as:

where x(t) is the time series describing the ground displacement. The Arias Intensity isequal to (juW)/(2g), where g is the acceleration of gravity. To discuss the site effects, itis convenient to consider the spatial distribution of the relative peak ground

acceleration, that is PGA(2D)/PGA(1D), and of the relative Arias Intensity, that is

«a

11

1

1

1

1

1

10

w

u

a

11

to

„ ' J '

1JJ« v s * ^ f l o ^

" • 1 • I30 40 SO

i » i • i " :

i i ' i -

i • i • i _

t • i • i J

1 i • ( • i

60 70 HI

W(2D)/W(1D), where 2D indicates the computations for the laterally varying model,

Figure 6a. Acceleration time series (R, V, and T components) at an array of 7 receivers equally spaced at3 km, on the local profile considered in die simulations. The epicentral distance of the first trace from thetop is 166,7 km, and of the last one it is 184.7 km; the 6th trace from the top corresponds to Magurelelocation. Acceleration is given in cm/s2 and time in seconds. The amplitudes arc computed for a seismicmoment of 3.01019Nmandacorner frequency of 0.41 Hz.

-l[ -II —

« » a TO1 1 ' A

1i

ID

11

1l

ID

1i

1i

]i

1(

1114

1110

H

• MIWJ

1 • 1 ' 1 •

-

EC- JO &J

Figure 6b. Acceleration time series (R, V, and T components) at an array of 7 receivers equally spaced at3 km, on the bedrock profile considered in the simulations. The epicentral distance of the first trace fromthe top is 166.7 km, and of the last one it is 184.7 km, respectively; the 6th trace from the top correspondsto Magurele location. Acceleration is given in cm/s2 and time in seconds. The amplitudes are computedfor a seismic moment of 3.01019Nm and a corner frequency of 0.41.

while ID represents the computations for the bedrock model. Both quantities for thethree ground motion components along the profile considered for Bucharest sedimentarystructure, are presented in Figure 8. A small increase can be observed for the relativePGA of the radial component (maximum 1.3), while the vertical component reachesvalues as large as 2.2 and the transversal component attains values of relativeamplification up to 2. For the relative Arias Intensity, the largest values are observed inthe transversal component (maximum 7.5) and the smallest values in the radialcomponent (maximum 2.75), this quantity being more sensitive than the relative PGA tothe presence of sediments characterized by low values both for the seismic wavevelocities and quality factors.

In addition we can compute the relative response spectra, i.e., the ratio betweenthe response spectra for the laterally varying model, Sa(2D), and the response spectra forthe bedrock model, Sa(lD), and the spectral ratio, i.e., the ratio between the Fouriertransform for the laterally varying model, FT(2D), and for the bedrock model, FT(1D).For illustration, Figure 9 show these quantities for a site located in the central part of thetown, at an epicentral distance of 176.1 km. The spectral amplifications at sites in thelaterally heterogeneous model, normalized to the corresponding response spectrum ofthe bedrock model, are computed for zero damping. The Fourier spectra of the signalscomputed at the sites in the laterally heterogeneous model normalized to the Fourierspectrum of the corresponding signals computed for the reference model are smoothedwith a frequency window of 0.025 Hz.

5 —

0- 5

- I D6420

- 2- 4- 620

10

0

- 10

T J

30 40 50 60 70 80

Figure 7. The three-component simulated accelerogram corresponding toMagurele location: (H-60km, epicentral distance=181.7km, Mo=3.0-1019N-m,corner frequency 0.41 Hz).

The position of the peaks is different for the different components. For example,the large excitation of the radial component (2.5 times) for the frequency around 0.35

10

Hz is not seen in the vertical component that has mainly four peaks of relative valuesgreater than 2.0 for the frequencies around 0.53, 0.65, 0.75 and 0.95 Hz, and in thetransverse component that has two peaks of relative values greater than 2.5, around 0.42and 0.9 Hz, respectively. The resonance frequency of the sedimentary layers mightexplain these spectral amplifications at the considered site. For example, the peakaround 0.4 Hz can be due to the resonance of the upper 400m, while the peak of 0.9 Hzcan be due to the resonance of the uppermost 270m.

6. Conclusions

The availability of realistic numerical simulations allows us significant progresses inground motion mapping. This powerful tool enables us to estimate the amplificationeffects in complex structures exploiting the available geophysical parameters,geotechnical, lithological, tectonic, historical, paleoseismological data, topography ofthe medium, and seismotectonic models. The technique applied in this study proves that

FGA(2DypOA(lD)W(2D)/W(1)

a160 170 ISO

2 -

0190 160

' !iiii

jf*- '\>

ff

—

rj

PGA{2DyPGA{lD)

WC2DVW<1D)

\\

— PGA(2DypGA(lD)W(2DVW(1D)

I \

170 180 190 160 170 ISO 190

Figure 8. Spatial distribution of the PGA (PGA(2D)/PGA(1D)) - continuous line, and W(W(2D)/W(1D)) - dotted line, relative values for the three components of motion (R,V,T)along the cross-section of Bucharest considered in the computations; 2D stands for the localsedimentary structure, while 1D stands for the bedrock structure.

it is possible to investigate the local effects at large epicentral distances, taking intoaccount both the seismic source and the propagation path effects.

The synthetic signals computed for the May 30, 1990, Mw=6.9, Vrancea event,are successfully compared with the available instrumental records obtained at Magurelestation, even if a relatively simple local structure and seismic source have beenconsidered.

For the complete microzonation of Bucharest the modelling will be extended to aset of representative cross sections, that span the entire area of the city, and to a set ofsource parameters, typical for the strong Vrancea earthquakes.

11

0.0 0.2 0.4 0.6 0.8 1.0 0.0 0.2 0,4 0.6 0-B

Figure 9. Relative response spectra Sa(2D)/Sa(lD) for 0% damping - continues line, andspectral ratio for 0.025 Hz smoothing FT(2D)/FT(1D) - dotted line, obtained for the threecomponents of the synthetic signal (R,T,V) at an epicentral distance of 176.1km.

Acknowledgements

The authors have been supported by: (1) University of Trieste, Italy, (Italian Law19/1991 "Aree di confine"), (2) UNESCO-IGCP project 414 "Seismic Ground Motionin Large Urban Areas", (3) NATO linkage grant AS.12-2-02 (ENVIR.LG 960916)677(96) LVdC "Microzonation of Bucharest, Russe and Varna in connection withVrancea earthquakes", (4) Copernicus project CIPA-CT94-0238 "Quantitative SeismicZoning of the Circum-Pannonian region". The authors thank Prof. G. Marmureanu, Dr.M. Radulian and Dr. N. Mandrescu for many fruitful discussions, Dr. F. Marrara and Dr.F. Vaccari for helping with the hardware and software facilities at the DST of Universityof Trieste, and an anonymous referee for his very careful review.

12

References

Altennan, Z.S., Karal, F.C. (1968), Propagation of elastic waves in layered media by finite differencemethods, Bull. Seism. Soc. Am., 58, 367-398.

Boore, D.M. (1972), Finite difference methods for seismic wave propagation in heterogeneous matherials,Methods in Computational Physics, Vol. 11, B.A.BoId, ed., New York, Academic Press, 1-37.

Dziewonski, A.M., Ekstrom, G., Woodhouse, J.H., Zwart, G. (1991), Centroid-moment tensor solutionsfor April-June 1990; Physics of the Earth and Planetary Interiors, 66,133-143.

Enescu, D., (1980), Contribution to the knowledge of the focal mechanism of the Vrancea strongearthquake of March 4, 1977, Rev.Roum.G6oi, Geophys., Geogr., Ser, Geophys., 24, 3-18.

Enescu, D., Zugravescu, D. (1990), Geodynamic consideration regarding the eastern Carpathians arcbend, base on studies on Vrancea earthquakes, Rev.Roum.Ge'ophysique, 34, 17-34.

Enescu, D., Enescu, B.D. (1993), Contributions to the knowledge of the genesis of the Vrancea (Romania)earthquakes, Romanian Reports in Physics, 45, 777-796.

Fah, D. (1991), Stima del moto sismico del suoto in bacini sedimentari, Tesi di dottorato, tutor: G.F.Panza, Trieste University.

Fah, D., Panza, G.F. (1994), Realistic modelling of observed seismic motion in complex sedimentarybasins, Annali di Geofisica, Vol.XXXVII, No. 6, 1771-1796.

Fah, D., Suhadolc, P., Miiller, St., Panza, G.F, (1994), A hybrid method for the estimation of the groundmotion in sedimentary basins: quantitative modelling for Mexico city, Bull. Seismol. Soc. Am., 84,No.2, 383-399.

Florsch, N., Fah, D., Suhadolc, P., Panza, G.F. (1991), Complete synthetic seismograms for high-frequency multimode SH-waves, PAGEOPH, 136, 529-560.

Fuchs, K., Bonjer, K.P., Bock, G., Radu, C , Enescu, D., Jianu, D., Nourescu, A., Merkler, G.,Moldoveanu, T., Tudorache, G. (1978), The Romanian earthquake of March 4, 1977. II. Aftershocksand migration of seismic activity, Tectonophysics, 53, 225-247.

Ismail-Zadeh, A.T., Panza, G.F., Naimark, B.M. (1996), Stress in the descending relict slab beneathVrancea, Romania, ICTP preprint, K796/93, Trieste, Italy.

Kelly, K.R., Ward, R.W., Treitel, S., Alford, R.M. (1976), Synthetic seismograms: A finite differenceapproach, Geophysics, 41, 2-27.

Mandrescu, N, Radulian, M, (1998), Seismic microzoning of Bucharest (Romania): a critical review;Natural Hazards, this issue.

McKenzie, D.P. (5970), Plate tectonics of Mediterranean region, Nature, 226, 239-242.McKenzie, D.P. (1972), Active tectonics of the Mediterranean region; Geophys., J.R.A.S., 39,109-185.Oncescu, M.C. (1984), Deep structure of the Vrancea region , Romania, inferred from the simultaneous

inversion for hypocenters and 3-D velocity structure, Ann.Geophys., 2, 23-28.Oncescu, M.C, Burlacu, V., Anghel, M., Smalberger, V. (1984), Three dimensional P-wave velocity

image under the Carpathian Arc, Tectonophysics, 106, 305-319.Oncescu, M.C, Trifu, C-I. (1987), Depth variation of moment tensor principal axes in Vrancea (Romania)

seismic region, Ann.Geophysicae, Ser.B, 5, 149-154.Oncescu, M.C, Trifu, C.-I. (1987), Depth variation of the seismic moment tensor principal axes in

Vrancea (Romania) seismic region, Ann.Geophysicae, SB, 149-154.Oncescu, M.C, Marza, V.I., Rizescu, M., Popa, M. (1998), The Romanian earthquake catalogue between

984 - 1996, Natural Hazards, this issue.Panza, G.F. (1985), Synthetic seismograms: the Rayleigh waves modal summation, J.Geophys., 58, 125-

145.Panza, G.F., Suhadolc, P. (1987), Complete strong motion synthetics, Seismic Strong Motion Synthetics,

Academic Press Inc., London, 153-204.Panza, G.F. (1993), Synthetic seismograms for multimode summation - theory and computational aspects,

Acta Geod.Geoph.Mont.Hyng., Vol.28 (1-2), 197-247.Radulian, M., Mandrescu, N., Popescu, E,, Utale, A., Panza, G.F.(1998), Seismic activity, stress field and

seismogenic zones in Romania, Terra Nova, in press.Radulian, M., Ardeleanu, L., Campus, P., Sileny, L, Panza, G.F. (1986), Waveform inversion of weak

Vrancea (Romania) earthquakes, Studia Geoph, et Geod., 40, 367-380.Romanelli, F., Bing. Z., Vaccari, F., Panza, G.F., (1996), Analytica! computations of reflection and

transmission coupling coefficients for Love waves, Geophys.J.lnt., 125,132-138.Vaccari, F., Gregersen, S., Furlan, M., Panza, G.F. (1989), Synthetics seismograms in laterally

heterogeneous, anelastic media by modal summation of the P-SV waves, Geophys.J.lnt, 99, 285-295.

13