You are on page 1of 13

Journal of Seismology 1: 7385, 1997. c 1997 Kluwer Academic Publishers. Printed in Belgium.

73

Spatially smoothed seismicity modelling of seismic hazard in Slovenia


Janez K. Lapajne, Barbara Sket Motnikar, Bla z Zabukovec & Polona Zupan ci c
Ministry of the Environment and Physical Planning, Geophysical Survey of Slovenia, Kersnikova 3, SI-1000 Ljubljana
Received 17 September 1996; accepted in revised form 28 February 1997

Key words: peak ground acceleration map, seismic hazard, seismic energy, spatially smoothed activity, Slovenia

Abstract A progress report on the mapping effort for construction of a peak ground acceleration (PGA) map of Slovenia for a 475-year return period for rock and rm soil is presented. The methodology is similar to that recently applied in central and eastern United States. It is based on historical seismicity spatially smoothed to different length scales. The procedure is described by Frankel (1995). He uses the characteristic earthquake recurrence relationship and in his rst version, four different seismicity models. We also use four seismicity models, but instead of characteristic earthquake recurrence, we use the doubly truncated exponential magnitude-frequency relationship; no evidence of characteristic earthquakes in Slovenia has yet been found. Three of our models are similar to Frankels rst three models. Model 1 uses spatially smoothed activity rate based on magnitude 3.7 and above events since 1880. Model 2 deals with spatially smoothed activity rate based on magnitude 5.0 and above events since 1690. Model 3 smoothes the observed activity over the entire region; it represents a uniform seismicity zone. Frankel changed this model in his latest version (Frankel et al., 1996). In model 4, we introduce a new approach to calculating seismic activity rate taking into account released seismic energy. The ground motion attenuation model of Pugliese and Sabetta (1989) is used for all models. PGA maps for models 1, 2, 3 and 4 have been calculated, and a weighted mean map derived from them. A map of model 1 has been compared with the corresponding source zone map; the two maps do not differ signicantly. A worst-case map derived from all four models has also been produced. Introduction In recent years, a lot of effort has been put into the construction of new PGA maps in Slovenia. So far, the maps were based on delineation of seismic source zones (e.g., Lapajne et al., 1995, 1996), which leads to many uncertainties due to the lack of quality geological and seismotectonic data. Since the earthquake catalogue of mainly noninstrumental events is the only reliable source of data for seismic hazard assessment in Slovenia, we decided to use the methodology of spatially smoothed seismicity proposed by Frankel (1995), where no delineation of seismic sources is needed. Frankel applied this methodology to calculate trial hazard maps for central and eastern United States. In early 1996 the procedure and maps were also available on WWW. Frankels (1995), methodology still follows the basic approach laid out by Cornell (1968). In Frankel (1995), a four-model method was used. The rst model was based on spatially smoothed seismic activity derived from earthquakes of magnitude 3 to 7 since 1924. Spatial smoothing was done by Gaussian function with a subjectively chosen correlation distance of 50 km. It was assumed that magnitude 3 events are a reasonable guide to where moderate earthquakes are most likely to occur over the next 50 years. The second model used spatially smoothed seismic activity based on magnitude 5 to 7 events since 1700. The correlation distance was 75 km. This model assigned higher hazard in areas that have had moderate or large earthquakes in the past. Due to incompleteness, the rate of occurrence in the second model was normalized to the rst model. The third model smoothed the observed seismicity over the entire region. This model covers the possibility of moderate earthquakes in areas that have historically been quiescent. In the three mod-

74 els, magnitudes were presumed to be distributed by log-linear GutenbergRichter equation. In the fourth model, events with moment magnitudes greater than 7 were taken into account as characteristic earthquakes, generated by individual faults. Frankel subjectively assigned weights 0.5, 0.25 and 0.25 to the rst, second and third model, respectively. To the weighted mean he added the accelerations derived from the fourth model. He calculated the PGA (peak ground acceleration) map for a 475year return period, corresponding to a 10% probability of exceedance in 50 years. Frankel also presented the worst-case map, showing the highest PGA value from the four models at each location. He used the attenuation relation for hard rock sites in the eastern United States (Atkinson and Boore 1995). Following the basic idea of Frankel, we derived a four-model weighted mean PGA map and worstcase map of Slovenia for the same return period. However, we have modied the original method in many ways and adapted it to local seismic properties (i.e., magnitude-frequency relationship, relatively small active zones). We also made some modications in order to increase the accuracy of computations. In Frankels method, the grid cells for the computation were grouped into rings within a certain distance increment from the site. That is, since we are dealing with cells instead of points, rather inaccurate. We have performed some numerical examples on our data and found that inaccuracies in that case are too high. Therefore, we considered every grid cell separately in our computations. This is more time consuming, but we have increased the accuracy, which we believe was necessary. Theoretical assumptions The calculation of hazard is similar to that described in Frankel (1995). Thus, we strongly advise the reader to refer to that work, because all our basic assumptions are based on that paper. Nevertheless, some modications of the procedure were indispensable. In Slovenia, no large earthquakes of magnitude 7 or above have been observed, and no evidence of characteristic earthquakes has yet been found. We found (Lapajne and Sket Motnikar, 1996) that in the area studied, the doubly truncated, exponential recurrence relationship (Youngs and Coppersmith, 1985; Reiter, 1990)

N (m) = N (m0 ) 10

bm m0 
1

10 bmu m0  (1) b  10 mu m0 

ts the observed data very well. N m is the cumulative number of earthquakes per year equal to or greater than magnitude, m, N m0 is the total number of earthquakes per year equal to or greater than the lower bound magnitude, m0 , mu is the upper bound magnitude and b is the decay rate. The activity rate N m0 is determined by simply counting the events or by calculation according to a theoretical distribution. Different approaches are used in different models.

( )

( )

( )

Alternative activity rate determination Due to very scarce data in the older part (pre 1880) of the earthquake catalogue (see seismological database), only the last part of it (18801994) can be used in statistical calculations. To take into account the hazard from some large reported historical earthquakes which occurred centuries ago, we introduce an approach to estimating activity rates which is not simply based on counting. We calculate an equivalent activity rate from the released seismic energy in the whole history of the recorded earthquakes, also adding the energy of foreshocks and aftershocks, although their contribution is very small. The seismic energy Ek has a simple relation to the magnitude mk of an earthquake (Gutenberg and Richter, 1954, 1956; Richter, 1958; B ath, 1973; Willmore, 1979) logEk

= Amk + B:

(2)

Summing all N earthquakes in the catalogue we obtain the total released energy E by the formula

N X B E = 10 10Amk : k

(3)

Knowing the total released energy E, we can compute the number N m0 of earthquakes which ideally ts the doubly truncated exponential law (1). Note that we can compute the total number of earthquakes by this method for a small area (cell of a grid) or for the entire region. The former case is possible in the case of constant b. The released energy may be written as a product of the number of earthquakes N m0 and the average energy E per earthquake

( )

( )

 E = N (m0 )E:

(4)

75

Figure 1. Epicentral map. Open circles represent earthquakes since 1880, while lled circles represent earthquakes from 1690 to 1880. Smaller circles show earthquakes of magnitude 3.75 and bigger circles earthquakes above magnitude 5. The inuence area of 40 km from the borders of Slovenia, which is used as the area of normalization and as the uniform seismicity zone in model 3 (see text), is also shown.

 is the energy that would be released, if there were E


only one, of course only theoretically, earthquake distributed by Equation 1. On the other hand, the average energy is

where m is the abbreviation for 1 m = A log(

 = 10B E ( )

Zm
m0

X Am ); 10
k

10Am pm

(m)dm;

(8)

(5)

where pm m is the probability density function of the magnitude obtained from Equation 1, and equals ln10  10 pm (m) = b1 : 10 bm m 
u
0

and denotes the magnitude of an earthquake which has released all the energy. Ground motion attenuation model The choice of a ground motion attenuation model is of great importance because attenuation has proven to be a highly inuential factor of seismic hazard. As in a previous study (Lapajne et al., 1996), we use the Pugliese and Sabetta (1989) or Sabetta and Pugliese (1996) attenuation equation. This equation is valid for distances up to 100 km.

bm m0 

(6)

From Equations 3, 4, 5 and 6 we have

Am m  (1 10 bm m  N (m0 ) = (A b)b 10 ; (10A bm m  1)


0

(7)

76

Figure 2. Flow-chart of four-model method procedure used to produce PGA maps.

Figure 3. Spatially smoothed activity rates N(4.5) of model 1 for a correlation distance of 17 km. Unit is number of earthquakes per 115 years (complete part of the catalogue).

77

Figure 4. Magnitude-frequency relationship graphs for models 1, 2, and 4. The activity rates N(4.5) of models 2 and 4 are normalized to the activity rate N(4.5) of model 1.

Hazard calculation We divided the entire observed area (44.547.5N, 12.517.5E) into a grid with spacing of 0.090 in latitude and 0.129 in longitude (about 10 km per side; Frankel (1995) grid cells were about 11  11 km2 ). In each cell i of the grid, we counted the number ni of earthquakes of magnitude equal to or greater than the lower bound magnitude m0 (its value depends on the model). We call ni the activity rate for cell i. These values are spatially smoothed to the values ni (Figure 3) using the formula (Frankel, 1995)

GutenbergRichter relationship (used by Frankel), we calculated the incremental values by subtracting cumulative smoothed values ni at magnitude m m=2 from those at magnitude m m=2, using Equation 1. Unlike Frankel (1995), we use integral instead of the sum over magnitude classes. Thus, we obtain for the annual rate of exceedance of ground motion value the following formula

+

1 (u > u0 ) = T

X
i

n ~i

P  =c j nj e n ~i = P  =c ;
2

ij

je

ij

(9)

Zm

where c is the correlation distance, and ij is the distance between the ith and jth cell. To calculate the annual rate  u > u0 of events exceeding ground motion u0 at a specic site, we need the incremental instead of the cumulative number of events. Because of the changed magnitudefrequency relationship, we also changed the conversion from cumulative to incremental values. Like Hermann (1977), who gave the conversion for the log-linear

 )

mmin

P [u > u0 jdi ; m]pm(m)dm;

(10)

where T means the number of years for which the activity rate is counted. The summation index i goes over all grid cells, and probability density function pm m is dened by Equation 6. The expression P u > u0jdi ; mj stands for the conditional probability that ground motion value u0 would be exceeded, if the distance from the site to the center of grid cell i was di and magnitude was m. This can be calculated

( )

78 from lnu(di ; mj ) p ;  2 (11) where lnu is the attenuation function,  is the standard

P [u > u0 jdi ; mj ] = 1  2

lnu

deviation of uctiation of ln u, and  represents the complementary error function. Equation 10 is a simplied version of a well known formula for calculating seismic hazard (e.g., Reiter, 1990; risk Engineering, 1988) for point sources, represented here by centers of grid cells. After calculating values  for several values of u0 , we can obtain the ground motion value for a given probability of exceedance by interpolating values of . Seismological database As a seismological database we used the earthquake catalogue prepared within the Copernicus international programme for the project Quantitative Seismic Zon ci ing of the Circum Pannonian Region (Ziv c 1996). For the territory of Slovenia, this catalogue is based on the catalogue of Ribari c (1982). The catalogue covers the period 5671994 AD and an area of approximately 100 000 km2 . The limits of the catalogue are approximately 100 km away from the international borders of Slovenia (Figure 1). The magnitude MLH dened by K arn k (1968) was selected as a uniform measure of earthquake size. It is based on surface waves. Magnitudes of most historical earthquakes were obtained ci from intensities using the equation (Ziv c, 1992)

observed magnitude is 6.3. The epicenter distribution of these two parts of the catalogue is shown in Figure 1. Magnitudes in the catalogue are rounded to one decimal place and represent the centers of magnitude classes of width 0.1. For example, the lower magnitude 3.7 in the complete catalogue represents the class [3.65, 3.75]. This grouping into magnitude classes is also taken into account in the determination of parameters.

Seismicity modelling We used four different seismicity models for seismic hazard assessment, denoted as models 1 to 4. Each model assumes a different spatial distribution of seismic activity. Figure 2 shows how hazard maps are derived from these models. Models 1 to 3 are similar to the rst three models proposed by Frankel (1995). In model 4, Frankel used the characteristic earthquake recurrence. As already mentioned, no evidence of characteristic earthquakes in Slovenia has yet been found. To take into account large historic earthquakes not included in models 1 to 3, we developed a special model, 4. Models 2, 3 and 4 are normalized to model 1 (see Figure 4), so that the total activity rate in a chosen area, called the inuence area, is the same in all models. The four models differ only in the spatial distribution of seismic activity. The inuence area is a region whose earthquakes signicantly inuence the computation of seismic hazard in the area of interest (in our case Slovenia), which lies inside of it. Earthquakes from the inuence area are used to estimate the GutenbergRichter decay rate b and the normalization factors for models 2 and 4. The inuence area is also used as a uniform seismicity zone in model 3. The choice of such an area is not an easy task. We dene the following criteria that an inuence area should meet: a. The borders of the inuence area should be as close as possible to the borders of the area of interest to obtain more representative seismic parameters; b. The borders of the inuence area should be far enough from the area of interest, so that the error from the contribution of earthquakes outside the inuence area to the hazard inside the area of interest, which is caused by inappropriate parameters, is negligible (accuracy to 2 decimal places); c. The borders of the inuence area should surround the area of interest as uniformly as possible.

MLH = 0:09 + 0:494I + 1:27logh;

(12)

where I and h are earthquake intensity and depth respectively. The maximum observed historical event was evaluated to be of magnitude 6.8. Foreshocks and aftershocks were removed from the catalogue in order to follow the Poissonian probability model. The catalogue may be considered roughly complete from 1880 for m equal to or greater than 3.7. In the period 1880 1994, the maximum observed magnitude was 6.1. In model 4 (see section 4), we used the original version of this catalogue including foreshocks and aftershocks. The complete part of the catalogue 18801994 is considered as the most representative, hence, seismic activities of all models (see section 4) are normalized to this part of the catalogue. There are 433 events. For magnitudes m  5, the subcatalogue since 1690 is considered. There are 37 such events, and the maximum

79

Figure 5. PGA maps related to four seismicity models for 475-year return period.

Sensitivity analysis showed that a distance of approximately 40 km from the borders of Slovenia meets the above criteria (see Figure 1). Note that seismicity in all models is normalized only inside this area, although the seismicity of a wider area (100 km from the borders of Slovenia) is taken into account in the hazard calculations. Earthquakes outside the inuence area also contribute to hazard in the area of interest, but the error which is caused by inappropriate normalization (with the normalization factor which is valid for the inuence area) is negligible. For each model we determine the magnitude range for which seismic activity and hazard are calculated. The lower bound magnitude m0 is chosen as 3.7 for models 1, 3 and 4, and 5.0 for model 2. The maximum observed magnitude since 1690 is 6.3. As recommend ed by Lapajne and Sket Motnikar (1996), this value should be increased by 0.2 to obtain the upper bound magnitude mu = 6.5. We used the same upper bound magnitude for the rst three models, although the max-

imum value 6.3 corresponds only to a subcatalogue of model 2. In the period covered by model 4 we have the highest magnitude of 6.8, hence mu = 7.0 was chosen. The magnitude range [3.7, 6.5] (except in model 4) was used for the determination of seismic activity, while the minimum magnitude mmin for which the hazard is calculated was 4.5, as proposed by Frankel (1995). This is based on the fact that earthquakes below magnitude 4.5 do not usually cause damage to buildings. A uniform decay rate b = 0.9 was determined by the maximum likelihood estimator, described in Weichert (1980), for models 1, 2 and 3. Due to different upper bound magnitude we have b = 0.91 for model 4. The decay rate was calculated from the complete part of the catalogue (used for model 1) within the inuence area. In order to obtain the proper correlation distance c for models 1, 2 and 4, we computed many maps of smoothed activities for each model. Comparing maps of activity rate N 4:5 , we dened the following two

( )

80

Figure 6. The weighted mean PGA map with 0.3, 0.3, 0.2, and 0.2 for models 1 to 4.

qualitative criteria for the determination of the correlation distance: a. Relative ranking of activity rates in different regions and for different correlation distances should remain unchanged after smoothing. For example: if the activity in the cell i was higher than in the cell j before smoothing, this should remain so after smoothing. b. Maximum activity rate should not vary too much from model to model. It was 20% in our case. But this is, of course, a matter for discussion. It should be mentioned that correlation distance is a very sensitive parameter, so caution is required in choosing its value. Figure 3 shows the smoothed activity rate map for model 1 for a correlation distance of 17 km. Similar maps were produced for the other three models. Model 1 Model 1 assumes that future earthquakes may occur in areas where they have occurred in the recent past, regardless of their magnitude. Thus, in model 1 the

complete part of the catalogue 18801994 for magnitudes greater than or equal to 3.7 is considered. The activity rate was counted in every grid cell and was then spatially smoothed according to Equation 9, and the chosen correlation distance c = 17 km (Figure 3). This may seem fairly low, but we should note that we are dealing with a seismotectonically very heterogeneous region (Ravnik et al., 1995). Model 2 In model 2 we assume that the probability of earthquake occurrence is greater where larger earthquakes have occurred in the past. In this model, we obtain the activity rate by counting earthquakes of magnitude m  5 in the time interval 16901994. To normalize the actual total number of earthquakes to model 1, we multiply it by a factor of completeness. From the magnitude-frequency relationship (1) and the total number of earthquakes above magnitude m 3:7, we compute the expected number of earthquakes above magnitude m 5 for the complete part of the catalogue (115 years). The factor of completeness used for

81

Figure 7. The worst-case PGA map derived from models 1 to 4.

normalization is dened as the ratio of the expected and actual annual rates. Note that normalization is performed in the entire inuence area, not in every grid cell separately. The normalization scheme is shown in Figure 4. In our case, the factor of completeness is 2.02. The appropriate correlation distance c for model 2 was found to be 25 km. The correlation distance for model 2 is greater than that for model 1 because of the smaller (average) density of most inuencing grid cells (cells with largest activity rates). Larger correlation distance suits larger location uncertainty of older events, and partly larger rupture areas of larger events. Model 3 In model 3 the observed seismicity 18801994 in the inuence area is uniformly smoothed, converting all this area into a uniform seismicity zone. This model covers the possibility of having earthquakes of magnitude 3.7 to 6.5 in areas that have been historically quiescent.

Model 4 In model 4 seismic activity is estimated from the total released seismic energy since the year 567 using Equation 7. As MLH is based on surface waves (K arn ik, 1968), we have taken the parameter A equal to 1.5 (Richter, 1958; Willmore, 1979; Reiter, 1990) for surface wave magnitude. Taking into account smoothing and normalization, the resulting activity rate map is not very sensitive to modest changes of parameter A, which cover some differences in denitions of magnitudes. After we have computed the total number of earthquakes according to the total released energy, we normalize this number to the number of earthquakes in model 1 (Figure 4). Having thus dened the normalized activity rates, we can follow the same procedure of computation as in the previous three models. The energy of the majority of earthquakes is negligible in comparison to the energy of the greatest events. Thus, only a small number of events contribute to the activity rate map. The result of very small density of inuencing grid cells is a correlation distance c = 35 km for this model. The largest correlation distance is in accordance

82

Figure 8. Seismic source areas of model A of the Lapajne et al. (1996) study. Source number 0 represents the background source, where only the events that are not in any other source area are considered.

with the largest location uncertainties and the largest rupture areas. This model was chosen to include in the hazard assessment some great historical earthquakes which were not considered in previous models.

Peak ground acceleration maps With the procedure described above, we computed for each of the four models a PGA map for the 475-year return period for rock and rm soil (Figure 5). The model 1 map clearly shows a region of high hazard in the southeastern part of the country, where a large number of small events were recorded, but, contrary to the distribution function, very few large events. Hazard in the model 2 map follows the distribution of larger events in the period 16901994. From these two maps, it can be seen that the most active regions are in the southeastern and far western parts of the country, but there is an active region also in the central part of the

country, which spreads north-south and includes the capital Ljubljana. This is the region where a relatively large number of strong events have been recorded in comparison to small ones. The model 3 map is a map of uniform hazard. The model 4 map shows very high peak ground accelerations in the western part of the country. This is due to an earthquake of magnitude 6.8 (the largest observed earthquake on the territory of Slovenia since 567) which occurred in 1511. This event contributed the greatest amount of energy and hence, the largest equivalent seismic activity. By assigning subjectively determined weights to the maps of the four models, we have obtained a weighted mean map (Figure 6). We have chosen a weight of 0.3 for models 1 and 2, and a weight of 0.2 for models 3 and 4. A larger weight is given to models 1 and 2 because they are based on more reliable data and, presumably, better represent the real seismic activity. Model 3 deals with a weak assumption

83

Figure 9. PGA map based on seismic source areas of model A of the Lapajne et al. (1996) study and the parameters of model 1.

that earthquakes with magnitudes 3.76.5 are equally probable everywhere in Slovenia, while for model 4 there is great uncertainty in the data used. We have also produced the worst-case map, which is given in Figure 7. In this map only the highest values of the four maps (shown in Figure 5) are taken into account.

Comparison with the seismic source model approach As already mentioned, the rst computations of the PGA map were based on the delineation of seismic source zones (Lapajne et al., 1995, 1996) using the previous version of the earthquake catalogue of Slove ci nia and surrounding areas (Ziv c, 1992), which covers a slightly smaller area and contains fewer events than ci the new version (Ziv c, 1996). Three seismic source models were developed in these studies, all of them based on the complete part of the catalogue. As none of the three models took into account some large historical earthquakes, there are signicant differences between the old weighted mean map and the weighted mean map of this study. This is particularly pronounced in those parts of Slovenia that have been relatively quiescent in the period covered by the recent complete

part of the catalogue. In previous studies, the values of seismic parameters were also slightly different from the values of the corresponding parameters dened in this study. To reasonably compare the spatially smoothed seismicity modelling of seismic hazard with seismic source modeling, some determinations and input data should be unied. An adequate comparison could be done for instance between the PGA maps related to model 1 of this study and to the seismic source model A of the previous study (Lapajne et al., 1996). The seismic sources considered in model A are shown in Figure 8. For both approaches, we used the new version of the catalogue ci (Ziv c, 1996). Therefore, the activity rates for model A were determined directly from the complete part of the new catalogue counting earthquakes in each source. For other seismic source parameters, we used uniform values dened in this study: mmin = 4.5, mu = 6.5, b = 0.9. As in previous studies, the PGA map of model A was computed by computer program Seisrisk III (Bender and Perkins, 1987). The corresponding PGA map is shown in Figure 9. It can be seen that the PGA map of model A is quite similar to the map of model 1 (Figure 5) of this study. To check the validity of our computations we also compared the model 3 map with PGA values computed by the seismic source model approach. The cor-

84
Table 1. The PGA value of model 3 compared with the PGA values of the background seismic source model computed with computer programs FRISK88 and Seisrisk III, and with the average PGA of model 1 inside the borders of Slovenia. Model 3 FRISK88 Seisrisk III Average of Model 1 0.144 g

PGA

0.148 g

0.149 g

0.147 g

suggestions on the methodology presented here. Elastic response spectra and design ground acceleration maps for rock and rm soil produced using the method described above, as well as the maps for the other soil types dened in Eurocode 8 (CEN 1994), have been already computed, but are not shown here. In future, we intend to further improve this method by taking into account the heterogeneity of the studied area. We will also try some other attenuation models.

responding seismic source model has only one background seismic source that equals the uniform seismicity zone for model 3. Table 1 shows that the results are practically the same. In addition, the last column shows the average PGA value of the model 1 inside the borders of Slovenia.

References
Ambraseys, N. N., Simpson, K. A. and Boomer, J. K., 1996, Prediction of horizontal response spectra in Europe, Bull. Seism. Soc. Am. 58, 15831606. Atkinson, G. M. and Boore, D. M., 1995, Ground motion relations for eastern North America, Bull. Seism. Soc. Am. 85, 1730. B ath, M., 1973, Introduction to Seismology, Birkh auser Verlag, Basel, Stuttgart, 395 pp. Bender, B. and Perkins, D. M., 1987, Seisrisk III: A Computer Program for Seismic Hazard Estimation, US Geological Survey, Washington, 48 pp. Cornell, C. A., 1968, Engineering seismic risk analysis, Bull. Seism. Soc. Am. 58, 15831606. CEN - European Committee for Standardization, 1994, Eurocode 8 Design provisions for earthquake resistance of structures Part 11: General rules Seismic actions and general requirements for structures, European Prestandard, ENV 1988-1-1, 25 pp. Frankel, A., 1995, Mapping seismic hazard in the central and eastern United States, Seismological Research Letters, Vol. 66, No. 4, JulyAugust 1995, 821. Frankel, A., Mueller, C., Leyendecker, E. V., Barnhard, T., Harmsen, S., Perkins, D., Dickman, N., Hanson, S., Hopper, M., 1996, http://gldage.cr.usgs.gov/eq/. Gutenberg, B.; Richter, C. F., 1954, Seismicity of the Earth and Associated Phenomena, Princeton University Press, Princeton, 310 pp. Gutenberg, B., Richter, C. F., 1956, Earthquake magnitude, intensity, energy and acceleration. Bull.Seism. Soc. Am. 46, 105145. Herrmann, R. B., 1977, Recurrence relations, Earthquake Notes, Vol. 48, Nos. 12. JanuaryJune 1977, 4749. K arn k, V., 1968, Seismicity of the European Area, Part 1,2 Academia, Czechoslovak Academy of Sciences, Prague. Lapajne, J. K., Sket Motnikar, B., 1996, Estimation of upper-bound magnitude in earthquake hazard assessment, In: V. Schenk (ed.), Earthquake Hazard and Risk, Kluwer Academic Publishers, Dordrecht, 3948. Lapajne, J. K., Sket Motnikar, B., Zupan ci c, P., 1995, Delineation of seismic hazard areas in Slovenia, Proceedings of the Fifth International Conference on Seismic Zonation, Nice, France, 429436. Lapajne, J. K., Sket Motnikar, B., Zupan ci c, P., 1996, Preliminary seismic hazard maps of Slovenia, Natural Hazards (in press). Lee, V. W., 1995, Pseudo relative velocity spectra in former Yugoslavia, European Earthquake Engineering 1, 1222. Pugliese, A., Sabetta, F., 1989, Stima di spettri di risposta da registrazioni di forti terremoti Italiani, Ingegneria Seismica VI/2, 314. ci Ravnik, D., Rajver, D., Poljak, M., Ziv c, M., 1995, Overview of the geothermal eld of Slovenia in the area between the Alps, the

Conclusions A simple, four-model method, that produces probabilistic hazard maps without the use of any other data but the earthquake catalogue, is presented here. We feel that this method is more convenient than seismic source modelling when no reliable geological and (seismo)tectonic data are available. The method avoids certain problems connected with the delineation of seismic sources. In the seismic source model approach, different models express different interpretations of the same database. Subjective judgment in the delineation of seismic sources and in weighting alternative parameters and models is difcult to control, and modelling is not repeatable by other independent interpreters. On the other hand, in the spatially smoothed seismicity modelling, different models are based on different subcatalogues and express different spatial distributions of seismic activity. Weights of models are based on the reliability of the subcatalogues used, and at least relative weighting is not problematic. Subjective judgment in the modelling is mainly expressed in the determination of the correlation coefcient c, and is easy to control. The modelling is transparent and can easily be repeated by other interpreters. Besides, the maps do not differ signicantly from those produced by seismic source modelling in the case of the complete database. However, a proper evaluation of the quality of this method is still to be done. Maps presented in this paper are for illustrative purposes only and they are not intended to be used in any application. We welcome any comments, criticisms or

85
Dinarides and the Pannonian basin, Tectonophysics 250, 135 149. Reiter, L., 1990, Earthquake Hazard Analysis, Columbia University Press, New York, 254 pp. Ribari c, V., 1982, Seismicity of Slovenia Catalogue of earthquakes, Seismological Survey of Slovenia, Ljubljana, 694 pp. Richter, C. F., 1958, Elementary Seismology, W. H. Freeman and Company, San Francisco, 768 pp. Risk Engineering, 1988, FRISK88 Users manual, Golden, Colorado, 163 pp. Sabetta, F., Pugliese, A., 1996, Estimation of response spectra and simulation of nonstationary earthquake ground motions, Bull. Seism. Soc. Am. 86, 337352. Theodulidis, N. P., Papazachos, B. C., 1994, Dependence of strong ground motion on magnitude-distance, site geology and macroseismic intensity for shallow earthquakes in Greece: II, Horizontal pseudovelocity, Soil Dynamics and Earthquake Engineering 13, 317343. Weichert, D. H., 1980, Estimation of the earthquake recurrence parameters for unequal observation periods for different magnitudes, Bull. Seism. Soc. Am. 70, 13371346. Willmore, P. L. (editor), 1979, Manual of Seismological Observatory Practice, World Data Center A for Solid Earth Geophysics, Washington, 165 pp. Youngs, R. R., Coppersmith, K. J., 1985, Implications of fault slip rates and earthquake recurrence models to probabilistic seismic hazard estimates, Bull. Seism. Soc. Am. 75, 939964. ci Ziv c, M., 1992, Catalogue of earthquakes, Supplement to: Institute of Structural and Earthquake Engineering: 1994, Probabilistic assessement of seismic hazard at Kr sko Nuclear Power Plant, Revision 1, Final Report , University of Ljubljana, Department of Civil Engineering, Ljubljana, 92 pp. ci Ziv c, M. (coordinator), 1996, Quantitative Seismic Zoning of the Circum Pannonian Region, QZEZ-CIPAR, Geophysical Survey of Slovenia, Scientic report on task A of project CIPA CT 940238 of Commision of European Communities.

You might also like