IC/98/29

United Nations Educational Scientific and Cultural Organizationand

International Atomic Energy Agency

THE ABDUS SALAM INTERNATIONAL CENTRE FOR THEORETICAL PHYSICS

MONITORING OF THE PREPARATIONOF THE STRONG EARTHQUAKES

IN VRANCEA, ROMANIA, USING THE CN ALGORITHM

I. KuznetzovInternational Institute of Earthquake Prediction Theory and Mathematical Geophysics,

Academy of Sciences, Moscow, Russian Federation,

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

O.V. NovikovaInternational Institute of Earthquake Prediction Theory and Mathematical Geophysics,

Academy of Sciences. Moscow, Russian Federation,

G.F. PanzaDipartirnento di Scienze della Terra, Universita di Trieste, Trieste, Italy

andThe Abdus Salam International Centre for Theoretical Physics, SAND Group,

Trieste, Italy

and

I.A. VorobievaInternational Institute of Earthquake Prediction Theory and Mathematical Geophysics,

Academy of Sciences, Moscow, Russian Federation.

MIRAMARE - TRIESTE

March 1998

Abstract

We present the results of the monitoring of the current seismicity in Vrancea region(Romania) for the time interval from January 1, 1994, (1994.1) to December 31, 1997,(1997.12) using the updated catalog of the Romanian local network. The monitoring ofthe preparation of strong, intermediate-depth, seismic events is done on the intermediatetime scale, by applying the CN algorithm to the earthquake catalog obtained by mergingRomanian and USSR data. In the period of time considered (1936 - 1997.12) four out ofthe five strong earthquakes, with magnitude above the threshold Mo—6.4, are correctlyidentified. The total duration of the Time of Increased Probability (TIP) for the occur-rence of a strong earthquake occupies 20.4% of the time interval considered, that is about2.38 years for each strong earthquake.

Stability tests of the CN diagnosis are performed with respect to the variation of themain free parameters appearing in the algorithm. They are (1) the magnitude thresholdMo, (2) the depth, D, separating the intermediate-depth Vra,ncea events from the shallowones, (3) the radius, R, of the horizontal circle, and (4) the linear dimension, H in thevertical plane, around the main shock, used to declustcr the catalog from aftershocks.

1. Introduction

The Vrancea seismoactive region is characterized by a relatively high-level of seismic activity,mainly intermediate-depth events, that are localized in a small and well delimited area, 45.0° - 46.0° N,26.0° - 27.0° E. The shallow seismicity of Vrancea region covers a larger area, 44.5° - 46.5° N, 25.0° -28.0° E, During this century four catastrophic earthquakes occurred with magnitude M greater or equal to7 (Table 1). These earthquakes are responsible for the heaviest destruction experienced in Romania andmay seriously affect vulnerable high risk constructions (such as nuclear power plants, chemical plants, largedams, pipelines etc.) located on a wide territory, from Central Europe to Moscow. Therefore, for theseismic hazard assessment of the territory potentially shaken by the strong Vrancea intermediate-depthevents, the prediction of strong earthquakes is an extremely important goal to achieve.

Several attempts have been made to predict, on a long term basis (several years), the strongintermediate-depth events originating in the Vrancea region, ENESCU and ZAMARCA (1973) and ENESCUet al, (1974), on the base of the periodicity of the past strong earthquakes, state that "a high seismicactivity is expected in the time interval 1977-1990, with peaks of intensity Io = VE - Vlll and one peak ofIo = VII - DC". This forward long-term prediction is confirmed by the occurrence of the earthquakes ofMarch 4, 1977 (Io = Vfll - IX), August 30, 1986 (Io = VIII) and May 30, } 990 (Io = VII - VIII). In thesame papers other forward predictions are formulated: the occurrence of an earthquake of intensity Io =VU1, or two-three earthquakes of Io = VII or VTf-VHI is expected in the period from 2000 to 2008. Thesecond attempt to predict Vrancea earthquakes is made by ENESCU and TANAS (1975) by using Wienerpredictive filters. They predict that "the time interval 1981-1990 will be characterized by a high seismicactivity in Vrancea, with the maximum magnitude of 6.8-7.0". This long-term prediction is confirmed bythe earthquakes of August 30, 1986 and May 30-31, 1990, but failed to predict the major earthquake ofMarch 4, 1977.

Based on a time variation model of the focal mechanism for Vrancea earthquakes, ENESCU (1983)makes the third forward long-term prediction that within four years from 1983, a major earthquake willoccur in the Vrancea region. The prediction is confirmed by the occurrence of the August 30, 1986 event.In agreement with the hypothesis of periodicity which was introduced on statistical grounds, the modelpredicts the occurrence of three different major earthquakes during the first fifty years of the 21st century:in2004±4,in2020±5, and in2040±5.

A recent paper by PANZA et al (1997) shows that, in the Vrancea region, from the simulation ofthe block-structure dynamics developed by GABRJELOV et al (1990), it is possible to generate a syntheticcatalog that has features similar to those of the real earthquake catalog. In the synthetic catalog, thatcovers a time interval of 7000 years, the distribution of the strong earthquakes (M > 6.8) is extremelyvariable with time and there is only one time interval of fifty years in which four strong events occur. Inmany cases the periodic occurrence of a single strong earthquake with the return period of about 100years, is typical, and for the remaining part of the synthetic catalog there is no periodicity in the occurrenceof the strong earthquakes. These results show that it is necessary to be careful when using the seismic cyclefor the prediction of the occurrence of a future strong earthquake, because the available observations coveronly a very short time interval, in comparison with the time scale of the tectonic processes, and theapplication of more formal prediction algorithms may be desirable.

One formal prediction algorithm is CN, originally developed considering the shallow seismicity ofthe California-Nevada region, and subsequently applied in many other parts of the World (KEILTS-BOROK

and ROTWAIN, 1990). The CN algorithm is based on the quantitative analysis of the premonitoryphenomena, which can be detected in the seismic flow preceding the occurrence of strong earthquakes, forthe determination of the Time of Increased Probability (TIP) of the strong events. The global result can besummarized as follows:1. In the retrospective analysis more than 80% of the strong earthquakes considered are preceded by

TIPs, and the duration of TIPs occupies about 25% of the time interval considered;2. In the actual prediction, the successes are more than 70% with TIPs occupying about 30% of the time

interval considered.

Since it makes use of normalized functions, the application of the CN algorithm to any seismicregion requires no adjustment of the parameters.

The first attempt to predict the strong, intermediate-depth, Vrancea events using the CN algorithmis made by DMITRI EVA et al (1987) and KEILIS-BOROK and ROTWAIN (1990) who used the catalog"Earthquakes in the USSR" (1965 - 1992).

The next application of the CN algorithm is made by NOVJKOVA etal, (1996) who considered theVrancea catalog, spanning from 1932 to 1993, compiled by merging the Romanian local earthquakecatalogs (RADU, 1979; CONSTANTINESCU andMARZA, 1980; TRIFU and RADULJAN, 1991) and thecatalog "Earthquakes in the USSR" (1965-1992). This catalog allowed NOVItLOVA et al (1996) tomonitor retrospectively the preparation process of the strong intermediate-depth events in the Vrancearegion. In fact, four out of the five strong events are identified and the total TIPs duration is 2] .7% of theperiod of time considered (1932 -1993).

The purpose of this work is:1. To extend the monitoring of the preparation process in the Vrancea region to the period from January

1, 1994 to December 31, 1997 (1994.1, 1997.12) using the updated Vrancea catalog (MOLDOITLANU

etal, 1995), and2. To perform stability tests of the CN diagnosis with respect to the variation of the main free parameters

of the algorithm.

2. Seismotectonics of Vrancea region

FUCHS et al (1978) proposed a paleosubduction in the Eastern Carpathians oriented from NE toSW, the southeastern extremity of the plate Pannonian-Carpathian subplate being detached, at present,from the rest of the subplate. A variant of this model is given by ONCESCU (1984) who assumes that theintermediate-depth earthquakes in Vrancea are not generated inside the subducting lithospheric fragment,but in the zone of separation between the fragment and the rest of the subplate, which is roughly vertical.CONSTANTINESCU and ENESCU (1984) assume a palaeosubduction from SE to NW and describe theevolution, since the beginning of the consumption of the ancient oceanic plate, lying between the EurasianPlate, the African Plate and the Arabian Peninsula, of the region that now corresponds to the EasternCarpathians, with special reference to the Vrancea region.

ENESCU and ENESCU (1993) formulate the hypothesis of an active subduction on going in theregion, within the area of the Carpatians' continental - type arc. This process, started relatively recently (2-3 million years ago), was caused by a slow north-westward movement of the subcrustal lithosphere lyingbetween the Peceneaga - Camena and the Intramoesian faults, whereas the crustal lithosphere has beeninvolved in underthrusting motions. The slow movement of the lithosphere strip between the two majorfaults is likely to be caused by the Anatolian subplate thrusting on the Black Sea subplate. The upper partof the continental lithosphere, being lighter, does not subduct, and only its lower portion takes part into thesubduction process, with a velocity estimated to be about 5.0 cm/year. As a result of the stressesassociated to the subduction process, subcrustal earthquakes occur in the subducted lithosphere fragmentand in a surrounding area in the microplate under which the subduction takes place. The model issupported by an increased amount of data about the lithospheric structure, earthquake location and focalmechanism.

The effects of the viscous flow, phase transition and dehydration on the stress field of the relict slabare examined by ISMAIL-ZADEH et al (1996) who propose that a realistic mechanism for triggeringintermediate-depth events can be the dehydration of rocks, which makes fluid-assisted faulting possible,rather than the shear stress caused by the basalt-ecoglite phase transformation in an oceanic slab.

3. Seismicity of Vrancea and Input Data

The database obtained by merging the available catalogs for the Vrancea region (NOVIKOVA etal,1996) is updated by adding the Vrancea events located in the interval (1994.1,12, 1997.12.31) using the

Romanian local records (MOLDOVEANU et al., 1995, continuously updated). The catalog covers therectangle 44.8° - 48.4° N, 25.0° - 28.0° E, and it is representative for ML > 2,5. (The seismicity map of theregion for the period from 1932 to 1997.12 is presented in Figure 1.)

4. The CN Algorithm

The CN algorithm identifies the Time of Increased Probability (TIP) of strong earthquakes with M> Mo. The conditions defining Mo are:1, the average recurrence period of the strong events with M > Mo is approximately 7 years, and2. Mo is close to a minimum in the histogram of the number of earthquakes, occurred in the region, as a

function of magnitude.The algorithm is based on a set of empirical functions of time that evaluates the seismic activity,

seismic quiescence, space-time clustering of the seismic activity and spatial concentration of theearthquakes. The quantification of the seismicity patterns, by different temporal functions, is based on thesequence, within several sliding time windows, of the main shocks in the region analysed. The flow of theearthquakes is represented, at each time, t, by a vector formed by the values of the different functions,defining the algorithm CN. Some of the CN functions must be computed starting from ti - Dt (usually Dt =4 years) therefore the monitoring starts at t r - to + Dt (to is the beginning of the available catalog). Thefunctions are normalized, so that they can be applied to different territories, with different seismicity,without ad hoc adjustment of the parameters.

Analogous phenomena to those on which it is based the definition of the function for thequantification of seismicity are observed in many non-linear systems before collapse. In particular, theresponse to a perturbation (1) increases, (2) becomes more chaotic, and (3) acts at large distances. In ourcase, the non-linear system is the system of seismically active faults, whereas the small quakes are thesources of the perturbation to the system. Thus, before a strong earthquake, which represents the collapseof the system, we must observe:1. Increase of the seismic activity, clustering of the earthquakes in time and space, and spatial

concentration of sources; in other words, the increase of the response to the perturbation;2. increase of the variation of seismicity and its clustering, which reflects the chaotic response to the

perturbation,3. long-range interaction of earthquakes, which can be interpreted as an increase of the influence range of

the perturbation.In the CN analysis of the earthquakes flow, along the time axis 3 different time interval categories

are identified: D (Dangerous), N (Non-dangerous) and X (Undetermined). The D intervals extend for 2years before each strong event. Intervals X extend for 3 years after each strong event; and if a strongearthquake occurs within an X interval, the interval becomes a D interval. The remaining time intervals areN intervals. The division of the temporal axis into three types of time intervals is used at the stage ofpattern recognition to choose the objects for learning. The intervals X (3 years after a strong event) are notused in the pattern recognition stage because they follow a strong shock.

The functions are discretized by defining the thresholds small, medium and large, on the basis ofthe quantiles levels 1/3 and 2/3. Then the algorithm estimates the combinations of the different discretizedfunctions which are more typical for intervals D and N. Following the procedure of pattern recognition,features D are defined by the condition that they occur during most of the intervals D, and just in a fewcases during the intervals N. Features N are defined by the opposite condition. Each feature corresponds toa discretized value of the function or to a combination of such values for two or three functions. At thestage of voting each time t is tested regardless of its position with respect to the occurrence time of astrong earthquake.

A TIP for a strong earthquake is declared at the time t and for one year if: (1) the differencebetween the number of D and N features is greater or equal to a constant V, and (2) the total source areaof earthquakes occurred during the last 3 years before the time t is less than a constant E.

These two conditions mean that there are many D features at the time t, and that the releasedseismic energy is not high. Consecutive TIPs may overlap and originate an alarm period exceeding 1 year.If during the TIP no strong event occurs - the alarm is false, while the occurrence of a strong event outsidethe TIP is a failure to predict.

5. Retrospective Application of the CN Algorithm to Vrancea Region

One of the basic problems in the application of the CN algorithm is the choice of theregionalization to be used, as shown, for instance by COSTA et al. (1994; 1996; 1997) for the Italianterritory. In this respect, the Vrancea region offers a special opportunity since a single zone (region) can beconsidered, quite well defined by the seismic activity (Figure 1).

The results obtained in NOVTKOVA et al. (1996) can be summarised as follows. When consideringonly the intermediate-depth earthquakes (D > 60 km) from 1962 and defining Mo=6.4 the earthquakes tobe predicted are the last three listed in Table 1. The result of the CN diagnosis shows that TIPs precede allthree strong earthquakes, their duration occupying 27.9% of the total time interval considered.

The predictive power of the CN algorithm is verified by its application to the time interval from1932 to 1962, period for which the Vrancea catalog is complete only for magnitudes above 4. In feet, inanalogy with what was done by COSTA et al. (1994) in the Italian region, NOVIKOVA et al (1996) do notuse the 1932 - 1962 data for the determination of the parameters ml, m2, m3 and of the thresholds ofdiscretisation of functions, but they apply the algorithm CN, using the parameters determined for theperiod (1962, 1993), to the part of the catalogue from 1932 to 1962. There are two strong shocks, withM> Mo, in this period: one of magnitude 7.2, in 1940, and another of magnitude 6.4, in 1945. The TIP ispresent before the earthquake of 1940, while the earthquake of 1945 is a failure to predict.

6. Stability Tests

To investigate the stability of these results we consider here the variation of the main freeparameters entering the CN algorithm: the magnitude threshold, Mo, the depth threshold, D, separatingthe intermediate-depth events from the shallow ones, the radius, R, of the horizontal circle and the lineardimension in the vertical plane, H, around the main shock, used for declustering the catalog fromaftershocks. The results are analyzed in comparison with those obtained by NOVIKOVA et al. (1996).

From the main shock catalogs obtained using different values for the declustering parameters(R,H) and considering either all the events, or only the intermediate depth events, we observe that part ofthe shallow events are aftershocks of the intermediate-depth events and viceversa. Taking intoconsideration this possible correlation between shallow and intermediate depth events, a better way fordeclustering the catalog could be:1. To compile the main shock catalog with no limitation on the focal depth of the events, and2. To impose the depth restriction, D > 60 km, only after declustering.Unfortunately this procedure cannot be applied because, from 1980 to 1993, the available catalog containsonly intermediate-depth events.

The CN algorithm is applied to the four main shock catalogs obtained from each pair of values (R,H), when R and H can assume the values 50 km and 75 km, and the parameter D, the upper depth limit forVrancea intermediate-depth events, is equal to 60 km. Table 2 gives the values of the spatial parametersused for declustering the catalog, and the TIPs corresponding to the four obtained catalogs. In addition weconsider the case R=50 km, H=50 km and D=100 km: the corresponding results are shown in Table7. Themain shock catalog and, consequently, the TIPs turn out to be strongly dependent upon the values (R, H)used for the identification of the aftershocks. The results of the CN algorithm depend upon the parameterH not directly, but through the parameter D. In fact, there is a high probability that the precursors of theVrancea strong events occur at depth larger than 100 km and the aftershocks of these earthquakes arelocalized at depths between 60 km and 100 km.

5

The intermediate-depth Vrancea earthquakes have a spatial distribution strongly dependent ondepth. The total number of the main shocks obtained by declustering the catalog underline that theidentification of the aftershocks is stable with the variation of the spatial parameter in the horizontal plane(R) from 75 km to 50 km, while it is unstable with the variation of the spatial parameter in the verticalplane (H), in the same distance range (from 75 km to 50 km). Thus we can conclude that the main shockcatalog slightly depends on the value assigned to the radius R but, at the same time, depends strongly onthe selection of the value of H.

The analysis of the variation of the main shock catalog with the variation of the spatial parameters(R, H) is naturally followed by the test of the stability of the results. The results obtained by CN algorithm,using these different main shock catalogs, are presented in Tables 3 to 7. Table 8 contains the failure topredict and the corresponding TIPs. The comparison can be made considering the two quantities:n% - the number of failures to predict, in percent of the total number of strong earthquakes,/% - time of increased probability (TIP), in percent of the total period of time considered.This couple of values has been selected because the prediction problem, as a stochastic point process, isevaluated in terms of a loss function. Losses depend on prediction errors that are: the failure to predict, n,and the fraction of space-time alarm, t. Therefore, the couple in, f) expresses the prediction capability andit is very important in evaluating the results of prediction algorithms (MOLCIIAN, 1990).

The modification of the magnitude threshold, Mo, from 6.4 to 6.2 has no influence on the TIPdiagnosis, as we could expect from the gap existing in Vrancea seismicity for the magnitude range (6.2,6.4). Some effect can be seen when increasing Mo from 6.4 to 6.5: we obtain a slight extension of the totalduration of TIPs to about 23% (Table 9), due to the presence of a long false alarm associated to theoccurrence of the M = 6.4, 1945, earthquake, but all the four strong events are preceded by TIPs.

7. The Monitoring of the Vrancea Seismicity in the Period 1994 -1997

We consider the Vrancea seismicity data updated till December 1997, and the values of all CNparameters as determined by NOVIKOVA el al (1996). No strong earthquakes occurred during the timeinterval from 1994 to 1997, and the CN algorithm does not declare any TIP (Table 10).

S.Conclusions

Even if the Vrancea seismicity is mainly formed by intermediate-depth events, spatiallyconcentrated in a very narrow zone, the CN algorithm can be successfully applied to predict theintermediate-depth strong earthquakes. The results of the stability tests of the retrospective TIP prognosiswith respect to changes of the key free parameters of the algorithm indicate the importance of thecontinuous monitoring of the intermediate-depth seismicity in Vrancea. This is now made possible by theuse of Romanian local seismic catalog, continuously updated.

Acknowledgments

This research has been made possible by the NATO Linkage Grant ENV1RLG. 931206, by RFF1grants N 96-05-65710 and 96-05-64800, by MURST (40% and 60%), by Copernicus ProjectERBCIPACT 940238 and UNESCO-IGCP project 414 "Realistic Modelling of Seismic Input forMegacities and Large Urban Areas".

REFERENCES

CONSTANTINESCU, h., and ENESCU, D. (1984), A Tentative Approach to Possibly Explaining theOccurrence of the Vrancea Earthquakes, Rev. Roum. Geol. Geogr., Geophys., 28, 19-32.

CONSTANTINESCU, L , and MARZA, V. (1980), A computer-compiled and computer-oriented catalogueof Romania's earthquakes during a millennium (984-1979), Rev. Roum. Geol. Geogr., Geophys., 24,193-234.

COSTA, G., PANZA, G.F, and ROTWAIN, I.M. (1994), Stability of premonitory seismicity pattern andintermediate-term earthquake prediction in Central Italy, Pure Appl. Geophys., in press.

COSTA, G., STANISKOVA, I., ROTWAIN, I.M., and PANZA, G. F, (1996), Regionalization and Stabilityof CN algorithm: the case of Italy, Pure Appl. Geophys, 147, 119-130.

COSTA, G., PERESAN, A., OROZOVA, 1., PANZA, G. F. and ROTWAIN, I. M., (1997). CN algorithm inItaly: Intermediate-term earthquake prediction and seismotectonic model validation, Proc. 30 Int'l.Geol. Congr. (Ye Hong editor), 5, 193-201, VSP Utrecht.

DMTTRIEVA, O. E., KEILIS-BOROK, V. I., KOSSOBOKOV, V. G, KUZNETSOV, I. V., LEVSH1NA, T.A.,MIRZOEV, K.M, NEGMATULLAEV, S.Kh., PISARENKO, V.F., ROTWAIN, I.M., SCHREIDER, S.Yu.(1987), Identification of the periods of increased probability of strong earthquakes in seismoactiveregions of USSR and other countries, Computational seismology, vol. 20, Nauka, Moscow, (1987).

Earthquakes in the USSR. 1962-1990, (Nauka, Moscow, 1965-1992).ENESCU, D,, and ZAMARCA, L (1973), A brief report concerning the Vrancea earthquake prediction,

Report of the Institute for Geology and Geophysics, (IGG) 1973; paper presented to the ReportsSession oflGG on May 17, 1973.

ENESCU, D., MARZA, V., and ZAMARCA, I. (1974), Contributions to the statistical prediction ofVrancea earthquakes, Rev. Roum. Geophys., 18, 67-79.

ENESCU, D., and IANAS, M. (1975), Attempts at predicting earthquakes in Vrancea for the periods 1976-1980 and 1981-1990, Rev. Roum. Geophys., 19, 27-35.

ENESCU, D,(1983), New data regarding the periodicity of Vrancea earthquakes and attempts to give atectonophysical explanation of this periodicity (in Romnian), Studies and Research in Geophysis., 21,24-30.

ENESCU, D., and ENESCU, B.D. (1993), A new model regarding the subductionprocess in the Vranceazone, Romanian Journal of Physics, Vol.38, N o J , 321 - 328.

FUCHS, K., BONJER, K. P , BOCK, G., RADU, C, ENESCU, D., JTANU, D., NOURESCU, A., MERKLER,G, MOLDOVEANU, T., and TUDORACHE, G. (1978), The Romanian earthquake of March 4, 1977.U. Aftershocks and migration of seismic activity, Tectonophysics, 53, 225-247.

GABRIELOV, A.M., LEVSHINA, T. A., and ROTWAIN, I.M (1990), Block model of earthquake sequence,Phys. Earth and Planet. Inter,, 61, 18-28,

ISMAIL^ZAHED, A.T., PANZA, G.F., and NAIMARK, B,M. (1996), Stress in the descending slab beneathVrancea, Romania, ICTP preprint, IC/96/93, Trieste, Italy.

KEILIS-BOROK, V.I, and ROTWAIN, I.M. (1990), Diagnosis of Time of Increased Probability of strongearthquakes in different regions of the World: Algorithm CN, Phys. Earth Planet. Int., 61, 57-72.

MOLCHAN, G.M. (1990), Strategies in strong earthquake prediction, Phys. Earth and Planet. Inter., 61,84-98.

MOLDOVEANU, C.L., NOVIKOVA, O.V., VOROBIEVA, I.A, and POPA, M. (1995), The updatedVrancea seismoactive region catalog, ICTP preprint, 1C/95/104, Trieste, Ttaly.

NOVIKOVA, O.V., VOROBIEVA, I.A., ENESCU, D., RADULIAN, M., KUZNETSOV, I., and PANZA, G.F.(1996), Prediction of the Strong Earthquakes in Vrancea, Romania, Using the CN Algorithm,PAGEOPH, vol.147, No. / , 99-118.

ONCESCU, M.C. (1984), Deep structure of the Vrancea region, Romania, inferred from simultaneous

inversion for hypocenters and. 3-D velocity structure, Ann. Geophys., 2, 23-28.PANZA, G.F., SOLOVEV, A.A., and VOROBIEVA, I A. (1997), Numerical modeling of block structure

dynamics: application to the Vrancea region, PAGEOPH, 149, 313-336.PANZA, G. F.; PERESAN, A. and COSTA, G. (1997), Zone sismogenetiche e previsions a medio termine

dei terremoti in Italia, II Quaternario: Italian J. of Quaternary Sciences, 10, in press.RADU, C. (1979), Catalogue of strong earthquakes originated on the Romanian territory, Part II: 1901-

1979, in "Seismological Researches on the Earthquake of March 4, 1977" - Monograph (Eds. 1. Corneaand C.Radu, Central Institute of Physics, Bucharest).

TR1FU, C.-I, and RADULIAN, M. (1991), A depth-magnitude catalog of Vrancea intermediate depthmicroearthquakes, Rev. Roum.Geophys., 35, 35-45.

Table 1Strong earthquakes in Vrancea since 1932

Date

yy

1940

1977

1986

1990

mm

11

3

8

5

dd

10

4

30

30

Time

Hh

1

19

21

10

Mm

39

21

28

40

Hypocenter

lat.

45.SON

45.78N

45.5 IN

45.83N

Lon.

26.70E

26.80E

26.47E

26.74E

Depth (km)

133

110

138

90

M

7,2

7.1

7.0

7.0

Table 2Values of the spatial parameters used for declustering the catalog,

and the TIPs corresponding to the four obtained catalogs.

Nr.crt

i—i

2.

3.

4,

H(km)

75

50

75

50

R(km)

75

75

50

50

D(km)

60

60

60

60

N

434

527

472

571

CN

Table 3

Table 4

Table 5

Table 6

H - linear dimension in the vertical plane, around the main shock; R - radius of the horizontal circle; D - depththreshold, separating the intermediate-depth events from the shallow ones; N - total number of events in the mainshock catalog; CN - table containing the results of the CN diagnosis.

Table.3Results of the CN algorithm applied to the main shock catalog

obtained with the pair of parameters (H-75 km, R-75 km), Mo^-6.4

Start of TIP

1.11.1940

1.11.19451. 3,19481.9,19741. 1.19811.11.19821.9.19851.9.19891.3.1994

Strong earthquakes

Date

10.11.19407.9.1945

4.3,1977

30.8.198630.5.1990

M

7.46.4

7.2

7.07.0

End of false

Alarm

1.11.19461.5.1951

1.9.19821.7.1985

1.3,1995

Duration of TIP

(months)

0.3failure to predict

12.038.030.120.032.012.09.012.0

Total number of strong earthquakes = 54 earthquakes are predictedTotal duration of TIPs = 165,3 months (23.4% of total time)

Table 4Results of the CN algorithm applied to the main shock catalog

obtained with the pair of parameters (H=50 km, R~75 km), Mo-—6.4

Start of TIP

1.11.19451. 9.19741,3.1981

Strong earthquakes

Date

10.11.19407.9.1945

4.3.197730. 8.198630. 5.1990

M

7.46.4

7.27.07.0

End of false

Alarm

1.5.1951

Duration of TIP

(months)

failure to predictfailure to predict

66,030.166.0

failure to predict

Total number of strong earthquakes = 52 earthquakes are predictedTotal duration of TIPs =162.1 months (22.9% of total time)

Table 5Results of the CN algorithm applied to the main

obtainedwith the pair of parameters (H=75km, R=shock catalog-50 km), Mo-6.4

Start of TIP

1.3.19371.11.1940

1.11.19451. 9.19741.11.1980

1.9.1994

Strong earthquakes

Date

10,11.19407.9.1945

4. 3.197730.8.198630.5.1990

M

7.46.4

7.27,07,0

End of false

Alarm

1.3.1938

1.5,1951

1.3.1995

Duration of TIP

(months)

12.00.3

failure to predict66.030.170.0

failure to predict6.0

Total number of strong earthquakes = 53 earthquakes arc predictedTotal duration of UPs = 184.4 months (26.0% of total time)

Table 6Results of the CN algorithm applied to the main

obtained with the pair of parameters (H-50 km, R-shock catalog-50 km), Mo-6.4

Start of TIP

1.3.19371.11.1940

1. 7.19481.9.1974

1.5.19801. 1.1986

Strong earthquakes

Date

10.11.19407,9.1945

4.3,1977

30. 8.J98630.5.1990

M

7.46.4

7.2

7.07.0

End of false

alarm

1.3.1938

1. 1.1951

1.7.1985

Duration of TIP

(months)

12.00.3

failure to predict30.030.162.08.0

failure to predict

Total number of strong earthquakes = 43 earthquakes are predictedTotal duration of TIPs = 142.4 months (20.1% of total time)

10

Table 7Results of the CN algorithm applied to the main shock catalog

obtained with the pair of parameters (H=50 km, R=50km), D^lOOKm, Mo=6.4

Start of TIP

1.11.19401.7.19481. 9.19751. 1.19811. 1.1985

Strong earthquakes

Date

10.11.1940

4.3.1977

30.8.1986

M

7.4

7,2

7.0

End of false

alarm

1.5.1951

1.7,1985

Duration of TIP

(months)

0,334.018.154.010.0

Total number of strong earthquakes = 33 earthquakes arc predictedTotal duration of TIPs = 116.4 months (16.4% of total time)

Table 8Failure to predict and corresponding

H(km)

75

50

75

50

50

R(km)

75

75

50

50

50

D(km)

60

60

60

60

100

n (%)

20

60

40

40

0

TIPs

t (%)

23,4

22.9

26

20,1

16.4

Table 9Results of the CN algorithm applied to the

complete catalogue (1936-1997.12), Mo=6.5

Start of TIP

1.7.19401.11.1945

1. 9.19741. 1.19811.11.19821.11.19851.9.1989

Strong earthquakes

Date

10.11.1940

4.3.1977

30. 8.198630.5.1990

M

7.4

7,2

7.07.0

End of false

Alarm

1, 1.1951

1.9.19821.7.1985

Duration of TIP

(months)

4.362.030.120.032,010.09.0

Total number of strong earthquakes = 44 earthquakes are predictedTotal duration of TIPs = 167,3 months (22.6% of total time)

11

Table 10Results of the CN algorithm applied to the

complete catalogue (1936-1997.12), Mo-6.4

Start of TIP

1.7,1940

1.11.19451.3.19481. 9.19741. 1.19811.11.19821.11.19851.9.1989

Strong earthquakes

Date

10,11.19407.9.1945

4.3.1977

30. 8.198630.5.1990

M

7.46.4

7.1

7.07.0

End of false

alarm

1.11.19461. 1.1951

1.9.19821.7.1985

Duration of TIP

(months)

4.3failure to predict

12.034.030,120.032.0JO.O9.0

Total number of strong eartliquakes = 54 earthquakes arc predictedTotal duration of TIPs = 151.3 months (20.4% of total time)

Vrancea, 1932 -1997,12

«N -

41HZOE 21E

WM3HITU0ESUmax •

Sw t̂on

3

tHL 1. 1-)H

s

S.J - B.fl

01 . 1 1

Figure 1. Map ofVrancea seismicity, 1932- 1997,12,

12