ESTIMATION OF SEISMIC HAZARD ON A PROSPECTIVE NPP SITE IN SOUTHERN FINLAND

Pentti Varpasuo Fortum Nuclear Services Ltd Finland

1 Introduction

The purpose of the present work is the estimation of seismic hazard in the territory of the prospective nuclear power plant site in Southern Finland. Because there are no registered strong motion acceleration recordings of earthquakes in Finland, the earthquake recordings from Saguenay and Newcastle regions from Canada and Australia were taken as sources of initial data because of their geological and tectonical similarity to Fennoscandia. The probabilistic seismic hazard assessment consists of three parts: 1) source effects, 2) path effects, 3) site effects. The site effects phase of the study is not relevant to this study because the prospective target sites are located on solid bedrock.Theoretical bases of determination of seismic hazard, questions of seismicity of a southern part of Finland, initial data on earthquakes and techniques of their processing, are considered below.

2 Regional seismicity in Fennoscandia Finland is situated on the Baltic shield, which is the one of the seismically quietest areas in the world. According to the fault plane solutions of earthquakes the push from the North Atlantic Ridge in the NW­SE direction seems to be the major stress related to the seismicity of Finland. Other factors of the stress field, such as glacial rebound and local seismotectonics, are more local. Earthquake recurrence rates in Fennoscandia are very low if compared with plate boundary regions worldwide. Nonetheless, Fennoscandia is an active seismic region, albeit at low earthquake recurrence rates and with relatively low magnitudes.The earthquake catalogue for Northern Europe (FENCAT), maintained by the Institute of Seismology of the University of Helsinki, was used in this study [i]. This catalogue encompasses the whole of Fennoscandia and adjacent areas inside a window of about 55­80 oN and 10o W ­45 oE. The catalogue includes all documented earthquakes in the region since 1375. Although instrumental earthquake observations started in Finland in the 1920's, local short period recordings started in 1956 [ii]. The events in Finland and in Fennoscandia have been predominantly instrumentally located since the mid 1960’s [i]. The instrumental magnitudes are based on the Richter's classical local magnitude scale, ML, modified for the Fennoscandian region. The uncertainty of macroseismic magnitudes is assumed to be 10% at best [i]. Figure 2­1 shows that southern Finland, which is the target area of this study, is characterised by relatively low seismicity. The most active belts of seismicity close to it are the Swedish coast from the Bothnian Sea to the Bothnian Bay, western Lapland and the northern Bothnian Bay­Kuusamo region. Southern Finland and areas south and east of it are characterised generally by a lower seismic activity. However, two NW­ SE oriented belts of relatively high seismic activity run through the region [iii]. The northern zone of higher activity runs from the southern Bothnian Bay towards Ladoga (B­L). The other

1(34) active belt (Å­P­P) runs from the Åland archipelago to southeastern Estonia, where it extends to from Paldis to Pskov. The zones are distinguished from their surroundings particularly by the occurrence of relatively large earthquakes (Figure 2­2). However, the period of pronounced seismic activity from 1920 to 1941 brings out the same seismic belts in southern Finland, but elsewhere as well.

E 0 : n o ti a t n e 70 N s re p e th f o

r e rd o B

Lapland

SWEDEN Kuusamo

65 N BB Oulu­ järvi

FINLAND RUSSIA Bothnian Sea

60 N

200 km

ESTONIA M = 4.5­6.1

M = 3.5­4.4

M = 2.5­3.4

M = 1.5­2.4 20 E 25 E 30 E

55 N

Figure 2­1. Distribution of the earthquake epicenters in northern Europe since 1375 according to FENCAT.

The outlines of the seismic belts applied in this study are presented in Figure 2­2. Both of the seismically active belts of southern Finland are characterised by long NW­SE oriented fracture zones [iii]and [iv]. The more prominent seismotectonic zone in eastern coast of Sweden is associated with considerable (300­400 m) vertical displacements along the eastern coast of Sweden [vii] and [v]. This zone, over 200 km west of Olkiluoto, separates the inland Precambrian bedrock from the offshore Phanerozoic sedimentary rocks.

2(34) ))) ))) 333 )) )3)3)33 )))) ))) ))) ))))) )))333 ))))) ))) 333 333 333 333 333 )) ))) 333 333 3333 ))) 333 ))) 333 ))) White )))333 33) ) ) 333 ))) 333 Sea ))) ))) 333 33333 3 3 3 3 333 ))) 333 333 )) 333 ))3)33 3 3 3 3 3 333 333 333 ))) 3333 33 333 33 333 33 ))) 333 )))33 )) 333 333 3 3 3 Kuusamo SWEDEN ))) ))) 333 333 3 3 3 333 )) 333 ))) 333 333 ) ) )3 3 3 )))))3 3 3 )))3 3 333 )))3 3 3 3 )) ))) 333 )))) ) ) ))) 333 333 333 65 N ))) ))) 333 333 333 )3)3)333 )) ))) ))) ) ))33)3 ) ) ) ) ))) ))) Oulu­ ))) ))333 ))) järvi ))))) ))) ))) ))) ))) )))) CFQZ 333 333 ) ) B 333 333 ) )) ))) ))) )) 333 333 ) ))) ) ) 333 ))) )))3)3)3 )))) ) ) 333 ))) FINLAND NLAND ))))) 333 ))) 333 ))333 ))) 33 RUSSIA 333 333 ))) ))))) )))333 ))) 3 33 ) ) ) Bothnian )))333 ))) 3)3)3) ))) ))) Sea 33 ))) ))) Olkiluoto 333 %% SFQZ 33333 333 333 L ))) 333 ))) 60 N 333 %% 333 )))333 3 3 333 %% Å Hästholmen P 333 ))) ESTONIA 35 E ))) P ))) ))) )) 20 E 25 E 30 E Figure 2­2 Belts))) of higher seismic activity (shaded areas). Historical events (1375­1964) with magnitude M ≥ 3.5 and instrumentally located (1965­1995) events with magnitude M ≥ 3.0 are shown by open and filled black circles, respectively. Crosses denote earthquakes during the period of increased seismic activity 1920­1941. B­L Southern Bothnian Bay ­ Ladoga Zone, Å­ P­P = Åland Archipelago­ Paldis­Pskov Zone, CFQZ= Central Finland Quiet Zone, SFQZ = Southern Finland Quiet Zone.

The study utilises a similar size of subset of FENCAT for both of the nuclear power plant sites. The seismic characteristics of Finland and examinations of attenuation (Chapter 5 ) suggest that areas within a distance of 500 km from the nuclear power plant sites are considered large enough to include all significant seismic events.Within 500 km of the sites, most of the earthquakes are small (M<5.0) and well below the threshold of engineering concern. Only one event, which occurred in 1894 (M=5.1) in central Sweden can be classified as a moderate (M≥5.0) earthquake. The highest intensity at the Olkiluoto site is only IV (MSK­64 intensity scale) [ii and vii]. MSK­ 64 intensity IV is partially described as: felt indoors by many people, outdoors by few, some awakened, vibration is like that due to the passing of a heavily loaded truck; Windows, doors and dishes rattle; Noticeable in standing motor cars [vi]. In the Loviisa site the maximum observed intensity is higher: VI. Within a distance of 500 km from Olkiluoto, a clear majority (72%) of the seismic events has occurred in Sweden [vii]. The areas in southern Sweden (Telemark­ Vänern region) and the eastern coast of Sweden are well known belts of higher seismicity (see e.g. [i]). These belts are outside the Loviisa data set. Both of the data sets include a small part of the northern Bothnian Bay­Kuusamo seismic belt in the north.

3(34) 3 Theoretical bases of Probabilistic Seismic hazard Assessment (PSHA).

In the execution of the given work the seismic hazard assessment methodology described in reference viii and ix was used. The theory, on which the methodology of references viii and ix is based, has been developed for a number of years by Cornell x (1968) and Cornell xi (1974) Merz and Cornell xii (1973). The PSHA theory as presented in reference ix can be submitted in the basic form as the Theorem of total probability Equation 3­1 P [A] = ∫∫ P [A | S and R] fs (S) fr (R) dsdr. In Equation 3­1 P indicates the specified probability, A is the event whose probability is sought, S and R are continuos independent random variables that have an effect on the value of A. In other words the probability that A occurs can be determined by multiplying the conditional probability of A, when s and r are given, times the independent probabilities of s and r and integrating over all investigated of s and r. Variables s and r represent earthquake size in used measure and distance from the site of interest. Random size and random location of the events are taken into account in the method. At an estimation of continuous probability of logarithm of ground acceleration exceeding some value i at the given station, the normal distribution assumption of reference xiii The average size of continuous distribution of logarithm of ground acceleration is defined as

Equation 3­2 MI (S, R) = C1 + C2 * S + C3 * ln (R + h),

In Equation 3­2 C1 , C2, C3 are constant factors and S is the size of earthquake and R is epicentral distance in km. h is depth of a seismic source in km. This equation is called the attenuation equation of ground motion. The standard deviation of logarithm of ground acceleration σI is generally constant and independent from S and R. Applying normal distribution and Equation 3­2 , we have Equation 3­3 P [A | S and R] =

P [I > i | S and R] = Φ* ((i­C1 ­C2 * S ­ C3 * ln (R + h)) / σI) In Equation 3­3 Φ*­ cumulative standard normal distribution. Characteristic ground motion variable is generally assumed to be lognormally distributed. Thus the logarithms of this variable is normally distributed. This normality assumption is fundamental to the methodology. The number Nm of earthquakes having magnitude greater than m, occurring in the source area is assumed to be of the form xiv

Equation 3­4 Log10 Nm = a ­b m, In Equation 3­4 a and b are constants describing the seismicity of a particular source area. The constant b describes relative distribution of small and large magnitude events; larger values of b indicate relatively fewer large events, and vice versa. Assuming, that the sizes of consecutive earthquakes in the source area are independent, it follows from Equation 3­4, that the cumulative distribution magnitude for each event is given by the following equation

Equation 3­5 FM (m) = k [1­exp (­β (m­m0))]; m0 < m < m1,

In Equation 3­5 m0 is the lower­bound magnitude, m1 is the upper bound­magnitude, which can be developed from properties of the source area, constants β and k are given by

­1 Equation 3­6 β = b * ln10; k = [1­exp (­β (m1­mo))] From the Equation 3­5 follows that the density function on magnitude is given by

Equation 3­7 fM (m) = β  k * exp (­β  (m­mo)); mo < m < m1

4(34) By substituting Equation 3­3 and Equation 3­7 into Equation 3­7 and equating s to m the exceedance probability of logarithmic acceleration i is at the site is obtained as m1  Equation 3­8 P[I>i]= ∫ ∫ Φ ((i­C1­C2*m­C3*ln(R+h))/σI)* β*k*exp(­β*(m­ R mo mo)*fR(r)dmdr With the aid of some algebraic manipulation, the integration on magnitude in Equation 3­8 may be performed analytically in closed form and resulting equation will be as follows

R * * βC3/C2 Equation 3­9 P[I>i] = {(1­k)Φ (z/σI) + k*Φ (z'/σI)+k*(r+h) *exp(­ ∫0 2 2 2 * 2 * 2 iβ/C2+β*C1/C2+β*m0+β σI /2/C2 )*[Φ ((z­βσI /C2)/σI)­Φ ((z'­βσI /C2)/σI)]}fR(r)dr In Equation 3­9 constants z and z' are defined in reference xii. The theory given in above formulas Equation 3­1­ Equation 3­9 is only given as reference material. The actual numerical calculations are carried out with the aid of EQRISK­program [ix] and with the aid of SEISRISK III ­program [viii]. In the work reported in this document reference [ix] was only used for calibrating and verifying the input parameter values given to algorithm of reference [viii]. This was carried out in the beginning of the work in order to validate the obtained results. Later on reference viii was solely used to carry out the numerical computations reported in this document. In the following section the most important features of the algorithm used in reference viii are described. The probabilistic seismic hazard assessment (PSHA) procedure described here is based on USGS Open File Report 82­293 [viii]. The method uses the concept of seismic source zones but allows earthquakes within a zone to be normally rather than uniformly distributed. This allows the.seismicity to vary smoothly across the boundaries of the source zone. enables smoother acceleration variation at sites near source zone boundary. Probabilistic models are used in PSHA because sizes and locations of future earthquakes cannot be predicted precisely. In PSHA source zones are typically regarded as seismically homogeneous. Each point within a source zone is assumed to have the same probability of being the epicenter of a future earthquake, because there is no reason to assume that an earthquake is more likely to occur at one point than at another within the zone. Catalogs of recorded earthquakes, and whatever geologic, geophysical or other information is available, are used to determine seismic source zones. The evaluation and interpretation of the various types of information available for delineating depends strongly on individual judgement or opinion. Earthquake location uncertainty inside source zones means that each point within the zone is regarded as the mean or most likely location of a future earthquake. The locations of actual earthquakes are normally distributed with standard deviation σ about their mean locations. This permits earthquake rates to vary smoothly near a source zone boundary, without being significantly different in the center of the region from the rates assumed in the uniform­seismicity model. As σ increases, a higher percentage of the earthquakes that were expected to occur within the source zone when no location uncertainty (σ=0) is assumed, occur instead beyond the boundaries. In the Figure 1 [viii] the differences between uniform seismicity source model and earthquake location uncertainty source model are depicted graphically. In this report, we use acceleration as a generic measure of ground­motion. By acceleration occurrences in an interval aj­1

5(34) Figure 3­1 Earthquake rates using varous assumptions of location uncertainty within gridded areas (10 km x 10 km). In A earthquakes occur only within two well­defined seismically homogeneous source zones σ =0. In B. earthquake locations are normally distributed with standard deviation σ = 10 km; in C, σ = 25 km: in D. σ = 50 km [viii].

Stated more precisely, in this model, if σ is the standard deviation in location, and if the expected (mean) location of an earthquake is at (Xe, Ye), the probability that the earthquake will occur instead in a small area A surrounding (Xe + ∆x, Ye+∆y) is given by 2 2 2 2 Equation 3­10 pA(∆x,∆y) = A/2/π/σ exp(­(∆x +∆y )/2/σ )) Figure 2 (A­C) [viii] shows acceleration levels calculated as having a 90­percent probability of not being exceeded during exposure times of 10, 50, 250, and 1000 years. Two variants are considered: 1) no variability in earthquake location (σ = 0) is assumed, and 2) σ =10 and 20 km. The attenuation function of Joyner and Boore [xv] with σa = 0.5 in loge acceleration was used.

The sites are located 5 km apart on a line through the centers of two contiguous source zones. Each zone is a square 60 km wide; the second zone has an earthquake rate 10 times higher than the first. Unless σ is a substantial fraction of the length or width of the source zone, the acceleration levels calculated at sites near the center and at sites on the boundaries of the source zone are nearly the same for various values of σ. For a site outside the source and in case when σ is assumed to be different from zero, some of the earthquakes will occur closer to the site than when σ =0 is assumed. In this case these closer earthquakes will produce higher accelerations at the site than will any earthquakes of the same magnitude occurring within the fixed source zone.

Figure 2D [viii] shows the difference in acceleration level calculated for the case of 1000­year exposure time at sites 5 km apart on the line when σ = 0, 10, and 20 km. For example, a change

6(34) in acceleration of +0.05 g at the site at X km means that the acceleration level calculated at the site at. X+5 km is 0.05 g higher than the level calculated at X. To explain how acceleration levels are calculated when earthquake­location variability is taken into account, we first observe that, if earthquakes are normally distributed, those that occur at. distance (∆x,∆y) from their mean locations, in effect, create a new source zone translated by (∆x,∆y) from the original zone. This means that the possible locations of the source zone may be regarded as having a normal distribution. More explicitly, the source zone is said to be located at (∆x,∆y) if an arbitrary point in the zone that would have been at (Xb, Yb) if the zone were at its mean or most likely location, lies instead at (Xb+∆x, Yb+∆y). The probability that the zone will be located at (∆x,∆y), is proportional to exp(­(∆x2 + ∆y2 ) /2/σ2 )), where σ is now the standard deviation in source location. Furthermore, the acceleration occurrences in an interval aj­1

The occurrences in the interval aj­1

In Equation 3­12 ρ[X1+k1dx,Y1+k2dy2),j]=the acceleration occurrences calculated in the interval aj­1

7(34) Figure 3­2 (A)­(C): Accelerations calculated to have a 90 percent probability of not being exceeded during various time periods, at sites on a line through the centers of two seismic sources. Source II has an earthquake rate 10 times higher than Source I. Earthquake locations are normally distributed with A, σ=0 km: B, σ=10 km: C, σ=20 km. D, Gradient curve of accelerations for various location uncertainty assumptions calculated for 1000­year return period at sites 5 km apart on the line of sites [viii].

In this section, we briefly describe the model before incorporating earthquake location uncertainty. Earthquake occurrences are assumed to have a Poisson distribution, and rates that remain constant during the time periods of interest. Mean ground­motion from an earthquake is an increasing function of magnitude and decreasing function of site­to­source distance. Earthquakes occur as points within quadrilateral source zones. Earthquakes in a given zone or fault are restricted to occur within a specified magnitude range; the range may be different for different zones. More specifically, for each zone a maximum magnitude mmax, a minimum magnitude m0, and a magnitude interval ∆m are assumed. All magnitudes occurring within a magnitude interval are grouped at the center of the interval, that is, at a set of n distinct magnitudes mj: Equation 3­13 mj=m0+(j+1/2)∆m, 0

8(34) for each magnitude interval does not change with time. A table of mean ground­motion values for the desired attenuation function is entered into the program for a set of earthquake magnitudes and distances. The program assumes that ground motion increases with magnitude and decreases with site­to­source distance; ground motions for various magnitudes and distances are obtained as needed by interpolating values in the table. Given the source zones, earthquake rates and attenuation function, an average yearly rate of accelerations within each acceleration interval is constructed for each site. Acceleration occurrences in the range ai­1 < a < ai are accumulated into variable reg(i), the accumulator for the i th’ interval, as described below. For the given attenuation function, let dm(i) be the distance at which a magnitude m earthquake produces mean acceleration ai at a site. The probability that an acceleration in the range ai­1

In Equation 3­14 Nm=rate of magnitude m earthquakes for the entire zone, A = total area of zone. The program determines A[rn(i)J for each magnitude and adds the corresponding rate p,,, (i) into the ~ accumulator, reg(i). The total rate accumulated for accelerations in the range ai­1

9(34) occurrences accumulated in reg(j) to regs(k) as follows. The entry in reg(j) is assumed to be concentrated at acceleration a where Equation 3­16 loge (a) = (loge[aj­1]+loge[aj])/2

In a log­normal distribution, given 1oge a, the probability of an acceleration in the range a1

In Equation 3­18 ρ(j)=occurrences of median accelerations in the range aj­1

Figure 3­3 Gradient curve of accelerations [viii] calculated for 1000­year period at sites 5 km. apart, assuming σa=0 and assuming σa = 0.5 in loge­acceleration in the attenuation function of Joyner and Boore [xv]; no uncertainty (σ = 0) in earthquake location

Given that average annual acceleration occurrence rates in the interval aj­1aj, is simply N Equation 3­19 Ex[aj] = Σi=j+1 ρ*(i)

In Equation 3­19 ρ*(i) = accelerations in the range ai­1

10(34) The probability of no event during the same time interval is, therefore, Equation 3­21 P(0,t) = exp(­µt) Setting µ=Ex(a), the probability of not exceeding a in t years is Equation 3­22 P(0,t) = exp[­Ex(a)t] Taking the natural logarithm of both sides of Equation 3­22 gives Equation 3­23 Ex(a) = ­(loge [P(0,t)])/t The value of a for which Equation 3­23 is satisfied is the acceleration, which has probability p = P(0,t) of not being exceeded during the time interval t. Equivalently, a has probability q = 1­P(0,t) of being exceeded one or more times during time t. By specifying any two of the three parameters p, a and t, one can determine the remaining parameter. In the procedure described here a single probability level p and several values of return periods, usually in the order from one hundred to one million years, are specified and the corresponding values of a are obtained. A pair of latitude and longitude values is selected to be two points on a great circle that becomes the equator in a transformed coordinate system. The great circle is chosen to pass through the region of interest, and all subsequent input coordinates for example, end points of source areas are transformed relative to the new equator. Furthermore, the subsequent calculations are done assuming a rectangular, rather than a spherical coordinate system; in the new system the equator becomes the x­axis. For points in the vicinity of the equator, calculations done using a rectangular coordinate system give a good approximation to those obtained using spherical coordinates. Because of convergence of parallels of latitude, the error could be substantial if the original coordinates are treated as being on a flat surface. This error is minimized when the coordinates are transformed to lie in the vicinity of a new equator. For example, within a square region of 12 degrees width and centered on the equator, the maximum error introduced by using the straight line distance between two points (xi, yi), (xj, yj) instead of the great circle distance is less than 0.6 percent. Sites at which accelerations are calculated lie on a rectangular m by n grid in which the site (xi, yj) is

xi = ck (longitude of row i), 1 < i < m yj = ck (latitude of col j), 1 < j < n where ck = conversion factor for degrees to kilometers at the equator.

Figure 3­4 Four points define transformation and target window for PSHA computations [viii] Great circle through (λ1,φ1),(λ2,φ2) becomes new equator. (λ1,φ1) → (d,0). ((λ2,φ2) → (0,0). (λ3,φ3) and (λ4,φ4) become end points of rectangular area containing target window (λ3,φ3) → (x3,y3).(λ4,φ4) →(x4,y4).

Coordinates of two opposite corners of target window in the initial system of spherical coordinates and the grid spacing are input data for the procedure. Sites in the transformed flat

11(34) rectangular coordinate system are located at even increments in ∆x, ∆y (∆y = ∆x) at intersections of grid lines At the end of the computation, sites are transformed to the coordinates they would have in the original spherical system. Sites may also be points evenly spaced on a number of lines; in this case, the two end points of each line, number of sites per line, and number of lines are specified. A table of N acceleration levels ak, such that ak < ak+1 for 1 < k < N, is constructed. A scale factor allows the original range of acceleration values in the table to be expanded or contracted. For a scale factor of one, 50 of the acceleration levels are in the range 0.02­1.0 g at increments of 0.02 g. Four acceleration levels are less than 0.02 g; accelerations at these levels are saved for calculations that are done when acceleration variability is taken into account; these accelerations are not printed in the output and, hence, are invisible to the analyst. A scale factor of 0.5 would permit values from 0.01 g­0.5 g. Given the table of acceleration levels ak, a corresponding array reg(k) is set aside to accumulate fractional acceleration occurrences, with occurrences in the range aj­1 < a < aj being placed in reg(j). Accelerations less than a(1) are placed in reg(1) and those exceeding the highest entry a(max) are placed into reg(max + 1). A seismic source zone is defined as a seismically homogeneous area enclosed by arbitrary quadrilaterals, connected or disjoint. A typical source is shown in Figure 8

Figure 3­5 Seismic source zone consisting of two sets quadrilaterals [viii] Input: For each quadrilateral set (1) jseg, ifr, tot jseg=number+1 of quadrilaterals in this set ifr identifies current set. 1 < ifr < tot tot=total number of sets of quadrilaterals (2) pairs of quadrilateral corner points. In this example, input is 4, 1, 2 λ1, φ1 λ2, φ2 λ3, φ3 λ4, φ4 λ5,φ5 λ6, φ6 λ7,φ7 λ8, φ8 2, 2, 2 λ9, φ9 λ10, φ10 λ11, φ11 λ12,φ12

12(34) Recall that an acceleration range represents occurrences of accelerations in some interval aj­1 < a < aj. For a specified magnitude, acceleration range boundaries [aj­1, aj] correspond to earthquakes at distances in the belt r(j) < r< r(j­1) from the site. The contribution to accelerations aj­1 < a < aj, produced by magnitude m earthquakes in a seismically homogeneous quadrilateral is equal to the rate per unit area within the quadrilateral, multiplied by the area within the quadrilateral at distances r(j) < r < r(j ­ 1) from the site. The area of the quadrilateral within the distance range r(j) < r < r(j ­ 1) is approximately the area contained within a ring of radius r and width dr = [r(j ­ 1) ­ r(j)] centered at the site. The geometry of integration over distances is shown in Figure 3­6:

Figure 3­6 Arcs of circles with centers at the site are used to approximate the area of a quadrilateral. Area = φrdr in D. E. and F. Shaded area in E and F is the difference between the quadrilateral area and arc area. In E errors nearly cancel; in F the arc area is an overestimate [viii] Arcs of circles with centers at the site are used to approximate the area of a quadrilateral. Area = 4)rdr in D. E. and F. Shaded area in E and F is the difference between the quadrilateral area and arc area. In E errors nearly cancel; in F the arc area is an overestimate

The distance range over which earthquakes produce accelerations in the interval aj­1 < a < aj depends on magnitude. The accelerations from earthquakes of each magnitude in the appropriate distance range within a source area are summed to obtain the total contribution from these earthquakes added into reg(j). The procedure is repeated for all acceleration intervals, and subsequently for all sources. A table of distance versus arc­length for a predetermined set of radii is established. Interpolation is used to obtain areas at desired distances between the predetermined radii. In this case

Equation 3­24 Area = φrdr

In Equation 3­24 φ = arc length in radians of that portion of the circle of radius r = rj that lies wholly within the quadrilateral; dr = interval width. The quantity φr is tabulated as a function of r for seven evenly spaced radii r per quadrilateral and for two to eight additional radii to vertices and edges, depending upon the location of the site relative to the source. The accuracy of the calculated areas depends on the geometry of the source, location of the site, and on the number of radii used in time calculations. The actual input value used for standard deviation of event location inside source areas is 20 km.

13(34) Initially the calculation of expected acceleration occurrences at each site on a grid is done assuming that seismicity is uniform throughout a source zone. The location uncertainty is included by the use of weighted average of acceleration occurrences calculated at various sites assuming uniform seismicity. The fractional contribution or weight assigned to the accelerations calculated at a site at (Xi, Yk) depends on σ and on the distance between (X1, Y1) and (Xi, Yk).See Equation 3­11. The grid spacing or distance between sites is of major importance in these calculations. Accelerations at sites at distances greater than 2.0σ ­ 2.5σ from (X1, Y1) have low weights and little effect on the weighted accelerations at (X1, Y1). This means that if the spacing between sites is large relative to σ, the only accelerations that will have much weight at (X1, Y1) are those originally calculated at (X1, Y1) and, in this case, the weighted results will give a poor approximation to the true effects of earthquake location uncertainty. The maximum acceptable grid spacing is up to one σ., where is the standard deviation of the location uncertainty distribution. Ground­motion is computed by interpolating in a table of acceleration as a function of magnitude and distance. Interpolations are linear in distance and linear in log acceleration and magnitude.The table of ground­motions is read in for up to eight magnitudes and twenty distances. If the ground­motion values read in are in a range different from 0.02 to 1.0, the appropriate scale factor must be input to make the range levels compatible with the desired values. The earthquake rates must be entered for each source at equally spaced magnitude intervals. The magnitudes that are entered are assumed to be values at the centers of the intervals. A division of each magnitude interval into two parts and performing the calculations for magnitudes at the midpoints of the halved intervals can be carried by parameter selection. Earthquake rates are calculated for each interval from the total earthquakes assigned to the original intervals using a Gutenberg­Richter b­value based on the number of earthquakes in the first two original magnitude intervals. The yearly exceedance rate for ground motion a at the site is Equation 3­25 Ex(a) = loge (p)/t The solution a that corresponds to the specified (p, t) may be found by interpolating between the calculated values Rates that have been accumulated in each acceleration interval for each site are saved at the end of a run. These values may be read back for a subsequent run. Doing successive runs makes it possible, for example, to accumulate accelerations from sources having different attenuation functions. The equator, maximum σ, gridded region, and rows and columns containing sites of interest should not be changed from their original values in a run continuation.

4 Division of the source area to source zones Intra­plate earthquakes tend to occur on old zones of crustal weakness reactivated by present stress field. Given the low earthquake recurrence rates and small magnitudes of the Fennoscandian shield region, the earthquake data is not likely to be complete and completeness varies geographically. Local high gain seismic recording stations would be needed to record the small events that characterise most of the regional seismicity.

Figure 4­1 shows the frequency of Finnish earthquakes observed per decade. Since 1750, the number of felt earthquakes in each decade fluctuates but shows a certain continuity of reporting. The beginning of systematic macroseismic mapping in the 1880’s, installation of regional seismic network since 1950's are clearly seen in the histogram. The utilisation of digital instruments in 1980's caused the latest stepwise increase of the number of observations [xvi]. The same facts can be distinguished in the Loviisa data set (Figure 4­2).

14(34) Figure 4­1. Numbers of earthquakes in Finland reported per decade [xvi]. Presumably mainly due to the Second World War, there are no reported earthquake observations in Finland from 1942 ­ 1951 (see also Figure 4­1 and Figure 4­2). This period is disregarded in calculations of annual seismic activity.

Cumulative number of events (500 km from Loviisa)

350

300

250

s t n

e 200 M1.5­ v e f M4­ o r

e M2­3 b

m 150 M3­4 u N

100

50

0 1600 1625 1650 1675 1700 1725 1750 1775 1800 1825 1850 1875 1900 1925 1950 1975 2000

year Figure 4­2. Cumulative number of seismic events of the Loviisa data set in four groups: M ≥ 1.5, M = 2 ­ 3, M = 3 ­ 4 and M ≥ 4.

Owing to the long recurrence interval, the current and future seismic hazard assessments will rely strongly on historical seismicity data. The Finnish earthquake data is considered satisfactory after 1880 (see [ii]). This time window is applied to the magnitude range 24.0) this period is coincided too short in relation to the recurrence time of events.

15(34) During the historical era, the interpretation is more accurate on land than in the water areas, when relatively small earthquakes are concerned. However, there is a relatively long history of dense population in southern Finland, Estonia and Central Sweden, especially in the coastal areas. It has been shown [iii] that the radius of perceptivity of 100 km of a Finnish earthquake is related to magnitude of the order of M = 3.8 ­ 4.0. Therefore it can be concluded, that it is very unlikely, that continental or submarine events larger than M=4.0 have remained undetected for the last two or three centuries. Based on the above judgement and the Figures above, the earthquake catalogue is considered quite representative for magnitudes M> 4.0, since 1750. In addition to the magnitude frequency distributions based on source area delineation, alternative distributions are calculated. With the scarce data in FENCAT it is not possible to establish statistically reliable parameters for every target area. Therefore the magnitude­frequency distribution of the entire Loviisa data set (Figure 4­3) is scaled to represent the regional seismic belts in proportion to the surface area. These distributions are applied in areas of insufficient statistics and as a branch in the decision tree (Chapter 6). In eastern Sweden the results presented by Kijko et al. [xvii] are used as a reference distribution. The seismicity parameters within the target areas are presented in Table 4­1.

Loviisa region (w ithin a distance of 500 km)

1,00E+01

1,00E+00

r LogN = 2.003 ­ 0.818M e mb u n l a u n An

1,00E­01

1,00E­02 2,00 2,25 2,50 2,75 3,00 3,25 3,50 3,75 4,00 4,25 4,50 4,75 Magnitude

Figure 4­3. Magnitude frequency distribution of the Loviisa data set.

Table 4­1. Seismicity parameters [Equation 3­4] for regional source areas (see e.g. Figure 2­2).

n Subregion bn an 1 Central Finland Quiet Zone (CFQZ) 1) 0.829 ± 0.035 1.121 ± 0.099 + Southern Kuusamo 2 Southern Bothnian Bay­ Ladoga Zone 1) (B­L) 0.782 ± 0.028 1.459 ± 0.090 3 Åland Archipelago ­ Paldis ­ Pskov Zone 1) (Å­P­P) 0.731 ± 0.033 1.097 ± 0.101 4 Southern Finland Quiet Zone 1) (SFQZ) 1.166 ± 0.066 2.065 ± 0.179 5 Eastern Sweden 2) 1.173 ± 0.059 3.460 ± 0.213 (part of the Olkiluoto data set) 6 Gulf of Bothnia, Swedish coast 3) 1.26 ± 0.06 3.40 ± 0.2 (comparison values for the decision tree) 7 Regional seismicity 2) 0.818 ± 0.016 2.003 ± 0.052 (within a distance of 500 km from Loviisa) 1) [xviii ]; 2) This study; 3) [xvii]

The data of seismicity of territory around Loviisa with radius 500 km are investigated. This source area is divided into six source zones presented in Figure 4­4 and Table 4­2.

16(34) The annual event rates are used directly for seismic hazard analysis, as well the greatest magnitudes registered in the used catalogue for each source zone. Each zone was divided into sub­zones, each of which in its turn represents the quadrilateral area (e.g. 5.1 ... 5.5 in Figure 4­4). The vertices of these quadrilateral areas are given with global longitudinal and latitudinal coordinates. These coordinates formed the geometrical initial data for the analysis done by the program SEISRISK III [viii]. 44.3.3.3 44.2.2.2 44.1.1.1 44.2.2.2 44.4.4.4 33.1.1.1 11.1.1.1 33.2.2.2 11.2.2.2 33.3.3.3

33.4.4.4 11.3.3.3 55.1.1.1 55.3.3.3 33.5.5.5 55.2.2.2 55.3.3.3 33.6.6.6

55.4.4.4 11.4.4.4 66.1.1.1 22.2.2.2 11.4.4.4 22.1.1.1 22.2.2.2 66.2.2.2 22.3.3.3 22.3.3.3 Häässthththoolllmeenn 55.5.5.5 55.6.6.6 Earthquakes: (1375­1995) 66.3.3.3 22.4.4.4 M=1.5­2.4 22.5.5.5 11.5.5.5 M=2.5­3.4

66.4.4.4 MM=3.5­=3.5­4.4.4 4 M=4.5­ 66.5.5.5 11.6.6.6 100 km

500 km

Figure 4­4. The source zone division for the Loviisa site embedded on the Fennoscandian epicentral map for time window 1375­1995 and for magnitudes greater than 1.5. Schematic model for the program SEISRISK III [viii].

Table 4­2. Seismicity parameters for the source zones of the Loviisa region. Parameters a and b are based on Table 4­1. Parameters arel are scaled to the regional seismicity (a7 in Table 4­1) in proportion relative surface area of the source zone in question. The b­value applied with arel is 0.818 (i.e. brel = b7 in Table 4­1). Mmax = observed maximum magnitude.

Source zone b a arel Mmax 1. Russia 0.818 * 1.515 * 1.515 2.9 2. Å­P­P Zone 0.731 1.097 1.040 4.9 3. B­L Zone 0.782 1.459 1.213 4.6 4. Bothnian Bay ­S. Kuusamo 0.829 1.121 0.383 4.7 5.SFQZ 1.166 2.065 1.229 3.2 6. Latvia 0.818 * 1.329 * 1.329 3.5 * Lack of data: a=arel and b = brel .

17(34) The zones that have greatest seismicity are the zones 2 and 3. The zone 3 (B­L Zone) extends from the southern Bothnian Bay to Lake Ladoga and its length in NW­SE direction is about 500 km. The total number of events greater than two in magnitude for this zone is 70. The above mentioned periods (increase of observations since 1880 and lack of data related to the Second World War) can be seen clearly in Figure 4­5. The magnitude frequency distribution based on this date is shown in Figure 4­6.

The zone that has the smallest seismicity is the zone 1 that has only one event. The zone six is named Latvian zone and its occupies the eastern Baltic see and part of Estonia and Latvia includes only 14 observations with M≥2. In these source zones, the regional magnitude frequency distribution is applied in the both branches of the decision three.

Cumulative number of events

80

70

1958 60 1941

50

M2­3 40 M3­4 M4­

30

20 N=19

10 N=6

0 1675 1700 1725 1750 1775 1800 1825 1850 1875 1900 1925 1950 1975 2000 Years (Region B­L) Figure 4­5. Cumulative number of seismic events in source zone 3 (B­L) grouped in three magnitude ranges [xviii].

B­L Seismic Zone

1,0E+00

LogN=1.459­0.782M

1,0E­01

er b m u n al u n An

1,0E­02

1,0E­03 1,5 2,0 2,5 3,0 3,5 4,0 4,5 5,0 Magnitude

Figure 4­6. The Richter magnitude­frequency diagram for source zone 3 [xviii].

18(34) The data of seismicity of territory around Olkiluoto with radius 500km are investigated. This source area is divided into six source zones presented in Figure 2­1 and Table 4­3.

11...11 4.4.1 1 11...22 4.4.2 2 EEararttthhququakesakes::: 4.4.3 3 EEararttthhququakesakes::: 11...33 33...11 m 11...33 (1(1(1337755­1­1­1999955) ) )) k 0 4.4.4 4 0 5 3.3.2 2 4.4.5 5

11...44 4.4.6 6

55...11 55...33 11...55 55...22 55...33 3.3.3 3 OlOlkikillluuootototo OlOlkikillluuootototo 3.3.4 4 %% 5.5.4 4 3.3.4 4 11...66 22...11 3.3.5 55 22...11 2.2.2 2 11...77 22...33 5.5.5 5 5.5.6 6 5.5.7 7 66...11 2.2.4 4 5.5.7 77 11...88 66...11

M=1,5­2,4

6.6.2 2 M=2,5­3,4 11...99 11...99 2.2.5 5 M=3,5­4,4

11...1100 6.6.3 3 M=4,5­

6.6.4 4

Figure 4­7 The source zone division for Olkiluoto site embedded on the Fennoscandian epicentral for time window 1375­1995 and for magnitudes greater than 1.5. Schematic model for the program SEISRISK III [viii] Except the source zone 1 (eastern Sweden), the source division resembles division presented for the Loviisa data set. As mentioned, over 70 % of the events in the Olkiluoto data set are within the source zone 1. It is obviously completely different to the other seismic source zones of this study. However, the completeness analysis of the Finnish data seems to be valid in eastern Sweden as well (see Figure 4­8). Therefore, the same time windows have been applied when the annual magnitude frequency relation was estimated (Figure 4­9). The b­value obtained (1.173) fit well the reference value for Sweden (b6 =1.26 in Table 4­1). Table 4­3. Seismicity parameters for the source zones of the Olkiluoto data set. Parameters a and b are based on Table 4­1. Usually parameters arel and brel are scaled as for the Loviisa data set, but arel for Sweden is scaled to correspond the surface area of zone 1 and brel = b6 is in Table 4­1. Mmax = observed maximum magnitude.

Source zone b a arel Mmax 1. Eastern Sweden 1.173 3.460 3.642 1) 5.1 2. Å­P­P Zone 0.731 1.097 1.073 4.9 3. B­L Zone 0.782 1.459 1.192 4.6 4. Bothnian Bay ­ S. Kuusamo 0.829 1.121 0.810 4.7 5.SFQZ 1.166 2.065 1.230 3.2 6. Latvia 0.818 * 1.033 * 1.033 3.3 * 1) Lack of data: a = arel. and b = brel ; [xvii]

19(34) S W EDEN (easte rn coast)

700

600

500

s t n

e 400 M 1.5­ v e

f M 4­ o

r M 2­3 e b 300 M 3­4 m u N

200

100

0 1350 14 50 1550 1650 1750 1 850 195 0

year Figure 4­8. Cumulative number of seismic events of the source zone 1 (eastern Sweden) in four groups: M ≥ 1.5, M = 2 ­ 3, M = 3 ­ 4 and M ≥ 4.

Sweden (eastern coast)

1,0E+01

1,0E+00 LogN = 3,460 ­ 1,173M

r be m

nu 1,0E­01 l ua nn A

1,0E­02

1,0E­03 2,0 2,5 3,0 3,5 4,0 4,5 5,0 Magnitude

Figure 4­9. The Richter magnitude­frequency diagram for eastern Sweden (source zone 1).

5 Attenuation of ground acceleration The prediction of seismic ground motions in prospective target sites is an essential ingredient in any probabilistic seismic hazard assessment. Its importance is even greater in procedures whose aim is the assessment of the probability of exceedance of ground motion in terms of response spectra for seismic load definition purposes. In opposite to the traditional approach, where a standard normalized design spectrum is anchored to peak ground acceleration at the high frequency asymptote of the spectrum, the uniform risk spectra approach is adopted in this study. This way of estimating response spectra has the advantage of yielding results of equal exceedance probability for all frequencies. In spite of the fact that this method of uniform hazard

20(34) risk spectra has been available for a long time and that peak ground acceleration for long time has been known to be insufficient to estimate the damage potential and to characterize the seismic load for buildings and structures, the amount of published attenuation equations for other frequencies than peak ground acceleration is scarce. This is true even for plate margin areas and even more so of course for intraplate areas. There are different ways in which strong­motion attenuation relationships can be established: • Methods, assuming that there is a relationship between intensity and the ground motion parameter to be estimated. • Methods that are based either on attenuation relationships developed for parallel regions, on relationships developed for parallel regions but adjusted to accommodate local information like source scaling and inelastic attenuation, or on relationships developed from theory, calibrated by available information in terms of local strong­motion data. • Methods utilizing stochastic models based on random vibration theory in simulating time series or peak parameters such as amplitudes. • Methods based on conventional regression analysis of observed strong­motion data from the area within which the predictions are intended to be applied. Fennoscandia, where seismicity is moderate, is such area where some kind of semi­empirical approach must be taken in the development of attenuation laws. In earlier earthquake hazard assessments for Fennoscandia, a variety of such approaches have been used. This study concentrates on intra­plate areas that are tectonically stable and geologically more uniform than plate boundary areas. A large number of intra­plate strong­motion acceleration recordings have recently become available from several intra­plate regions including eastern North America, Europe, China and Australia, thus providing the opportunity to derive attenuation models by regression analysis of spectral estimates of ground motion. In the following sections the derivation of an intra­plate attenuation relationship is described following the presentation given in reference [xix]. The study is restricted on modelling results of response spectral estimates for solid bedrock site conditions. In any statistical analysis of sparse data, it is important to consider critically the selection criteria for choosing the data. The collection of intra­plate strong­motion data presented in reference [xix] is reproduced in the following tables:

Table 5­1 Strong­motion data bank of reference [xix] Date Event Site Соmр Dist (km) Mag Type Ref

14/6/1972 Ancona, Italy Rocca, Italy NS 6.0 4.б M s (15)

17/9/1972 Cephalonia, Greece Argostoli, Greece EW 38.0 6.4 Ms (15)

30/10/1972 Cephalonia, Greece Argostoli, Greece EW 26.0 5.0 Ms (15)

23/5/1974 lmotski, Yugoslavia Imotski, Yugoslavia NS 15.0 4.7 M b (a)

06/5/1976 Friuli, Italy Tolmeizzo Italy NS Зб.0 4.5 M L (a)

06/5/1976 Friuli, Italy Tolmezzo Italy NS 36.0 6.5 Ms (b)

07/5/1976 Friuli, Italy Tolmezzo Italy NS 45.0 4.7 M s (15)

07/5/1976 Friuli, Italy Tolmczzo Italy EW 28.0 4.2 ML (a)

11/5/1976 Friuli, Italy Tolmezzo Italy EW 22.5 4.7 M s (15)

11/5/1976 Friuli, Italy Tolmezzo Italy EW 22.5 4.7 M s (15)

18/5/1976 Friuli, Italy S. Rocco, Italy EW 10.3 4.0 M L (b)

01/6/1976 Friuli, Italy Tolmezzo, Italy EW 14.5 4.2 M L (48) 09/6/1976 Friuli, Italy Tolmezzo, Italy NS 18.4 4.0 ML (48)

09/6/1976 Friuli. Italy Tolmezzo, Italy MS 18.4 4.0 M L (48)

11/6/1976 Friuli, Italy Tolmezzo, Italy NS 22.0 4.6 M L (48)

11/6/1976 Friuli, Italy Tolmezzo, Italy NS 22.0 4.6 M L (48)

28/7/1976 Tangshan, China Hongshan, Hebei EW 385.5 7.8 Ms (39)

2/8/1976 Tanashan, China Qiainan Lanhe SN 17.2 4.1 M s (39)

3/8/1976 Tangshan, China Qianan Lanhe SN 14.4 3.7 M s (39) 3/8/1976 Tangshan, China Qianan Lanhe SN 12.0 4.3 M s (39) 5/8/1976 Tangshan, China Qiaiian Lanhe SN 16.7 3.7 Ms (39)

21(34) 7/8/1976 Tangshan, Сhinа Qianan Lanhe EW 12.5 3.6 Ms (39) 8/8/1976 Tangshan, China Tangshan Cement SN 38.0 5.1 M s (39) 8/8/1976 Tangshan, China Qianan Lanhe SN 28.0 5.1 M s (39) 8/8/1976 Tangshan, China Changli, China EW 69.1 5.1 Ms (39) 9/8/1976 Tangshan, China Tangshan Cement EW 64.7 5.4 M s (39) 9/8/1976 Tangshan, China Qianan Lanhe SN 11.9 5.4 Ms (39) 9/8/1976 Tangshan, China Changli, China SN 32.5 5.4 M s (39) 14/8/1976 Tangshan, China Qianan Lanhe SN 20.0 З.6 M s (39) 14/8/1976 Tangshan, China Qianan Lanhe SN 17.2 4.1 Ms (39) 14/8/1976 Tangshan, China Qianan Lanhe SN 17.6 4.0 M s (39) 15/8/1976 Tangshan, China Qianan Lanhe SN 16.3 4.6 M s (39) 15/8/1976 Tangshan, China Tangshan Cement EW 21.9 4.3 M s (39)

Table 5­2 Strong­motion data bank of reference [xix] (continuation)

Date Event Site Comp Dist Mag Type Ref

15/8/1976 Tangshan, China Tangshan Cement SN 27.0 4.3 Ms (39)

16/8/1976 Tangshan, China Qianan Lanhe EW 25.0 4.3 Мs (39)

18/8/1976 Tangshan, China Tangshan Cement EW 19.9 3.9 М s (39) 23/8/1976 Tangshan, China Qianan Lanhe SN 12.1 3.0 М s (39) 26/8/1976 Tangshan, China Tangshan Cement SN 12.2 4.5 М s (39) 26/8/1976 Tangshan, China Qianan Lanhe SN 17.7 2.9 М s (39) 28/8/1976 Taagshan, China Qianan Lanhe SN 27.2 3.7 Ms (39) 28/8/1976 Tangshaa, China Qianan Lanhe SN 21.6 4.6 М s (39) 31/8/1976 Tangshan, China Qianan Lanhe SN №2 5.5 Мs (39) 31/8/1976 Tangshan, China Changli, China SN 37.2 5.5 M s (39) 31/8/1976 Tangshan. China Qianan Lanhe SN 18.0 5.4 M s (39) 31/8/1976 Taagshan, China Changli, China EW 24.2 5.4 Ms (39) 31/8/1976 Tangshan, China Qianau Lanhe SN 22.0 4.7 M s (39) 31/8/1976 Tangshan, China Qianan Lanhe SN 17.4 3.7 М s (39) 31/8/1976 Tangshan, China Qianan Lanhe EW 13.8 3.8 Мs (39) 2/9/1976 Tangshan, China Qianan Lanhe EW 20.6 4.6 M s (39) 7/9/1976 Tangshaa, China Qianan Lanhe SN 37.0 4.6 М s (39) 7/9/1976 Tangshan, China Qianan Lanhe SN 30.9 4.0 Мs (39) 11/9/1976 Friuli, Italy Bregini, Italy EW 15.0 5.5 М s (b) 25/9/1976 Tangshan, China Qianan Lanhe EW 39.1 5.0 Мs (39) 29/9/1976 Tangshan, China Qianan Lanhe SN 26.0 4.9 Мs (39) 1/10/1976 Tangshan, China Qianan Lanhe EW 31.2 4.6 M s (39) 4/7/1977 Oolong, Australia Oolong, Australia N 35.0 4.8 ML (b) 16/1/1978 Swabische Alb, Germany Albstadt, Gernany EW 11.0 4.5 M L (A) 9/4/1979 Montenegro, Yugoslavia Ulcinj.2,Yugoslavia WE 18.0 5.2 Ms (15) 23/11/1980 Campania, Italy Bagnoli, Italy EW 38.0 6.9 М s (b) 23/11/1980 Campania, Italy Calitri, Italy EW 10.8 6.9 Мs (b) 24/11/1980 Iirpina, Italy Calitri, Italy EW 11.0 4.8 M s (15) 19/1/19S2 New Hampshire, USA Franklin Falls 315 10.0 4.7 CL (а) 19/1/1982 Mew Hampshire, USA Union Village 245 60.0 4.7 CL (а) 19/1/1982 New Hampshire. USA North Hartland 15 61.0 4.7 CL (а) 23/12/1985 Nahanni, Canada Site 1, N.W.T, 280 20.0 6.9 Мs (b) 23/12/1985 Nahanni, Canada Site 1, N.W.T. 10 9.7 5.9 М s (с) 08/08/1988 More Basin, Norway Loviisa, Finland E 1300.0 5.1 Ms М 08/08/1988 More Basin, Norway Molde, Norway EW 280.0 5.1 Мs (с) 08/08/1988 Мorе Basin, Norway Sulen, Norway NS 300.0 5.1 М s (с) 25/11/1988 Saguenay, Canada Site 1, Canada 0 116.0 5.8 Мs (b) 25/11/1988 Saguenay, Canada Site 2, Canada 321 152.0 5.8 Мs (b) 25/11/1988 Saguenay, Canada Site 8, Canada 63 96.0 5.8 Мs (Ь) 25/11/1988 Saguenay, Canada Site 5, Canada 97 113.0 5.8 М s (b) 25/11/1988 Saguenay, Caiiada Site 9. Canada 270 125.0 5.8 М s (b) 25/11/198S Sagucnay, Canada Site 10, Canada 270 117.0 5.8 Мs (b) 25/11/19$$ Saguenay, Canada Site 14, Canada 270 100.0 5.8 Мs (b) 25/11/1988 Saguenay, Canada Site 16, Canada 214 51.0 5.8 М s (b) 25/11/1988 Saguenay, Canada Site 17, Canada 0 70.0 5.8 Мs (b) 25/11/1988 Saguenay, Canada Site 20, Canada 0 94.0 5.8 Мs (b) 25/11/1988 Saguenay, Canada EMME, East USA 65 471.6 5.8 М s (b) 25/11/1988 Saguenay, Canada NEWC, East USA 105 525.3 5.8 Мs (Ь) 25/11/1988 Saguenay, Canada PAL, East USA 90 821.2 5.8 М s (Ь)

22(34) 25/11/198S Saguenay, Canada MIME, East USA NS6E 360.3 5.8 Мs (b) 25/11/1988 Saguenay, Canada LYON, East USA N35W 432.3 5.8 M s (b) 35/11/1988 Saguenay, Canada ISFL, East USA 267 323.6 5.8 M s (b) 25/11/1988 Saguenay, Canada DCKY, East USA 0 200.1 5.8 M s (b) 29/01/1989 Etoe, Western Norway Sulen, Norway E­W 173.0 4.6 М s (с)

Intra­plate strong­motion recordings are very limited. Moreover recordings on solid bedrock are even more scarce. The Dahle regression model [xix] represents a intra­plate strong­motion model, applicable to areas where the lack of local strong motion recordings prevents the use of appropriate region specific strong ground motion models. Table 5­1and Table 5­1summarize the recordings selected for the development of Dahle model. The data set listed in Table 5­1and Table 5­2 consists of 87 recordings from 56 different earthquakes, out of which the Tangshan earthquake of July 28, 1976, and several smaller aftershocks comprise a significant part. The Friuli earthquakes, also occurring in 1976, comprise a second large subset. Of particular importance for Dahle model are the many recordings achieved from the November 25, 1988, Saguenay earthquake in Quebec, Canada, with many good recordings at large distances. The remaining data listed Table 5­1and Table 5­2 emerge from various regions in eastern North America, Canada, Australia and Europe. The Dahle [xix] attenuation function form with minor modifications is adopted for the attenuation relationship used in this work. In the following sections the development of the functional form of Dahle attenuation relationship is explained. The determination of significant site­source distance is important in probabilistic seismic hazard analysis. The hypo­central distance is therefore defined as Equation 5­1 R = (d2 + h2 ) 1/2 In Equation 5­1d is epicentral distance and h is depth of focus. The epi­central distances and the depths of focus are obtained from data set presented in Table 5­1and Table 5­2. Surface wave magnitudes, Ms, were chosen for the magnitude definition for development of Dahle attenuation model. In Dahle model the maximum of the two horizontal components was selected to represent the ground motion parameters to be modelled. The following parametric expression for the size of motion for a harmonic wave in infinite elastic half­space is adopted for the regression problem: b Equation 5­2 A = A0/R exp(aM+qR) In Equation 5­2 A is the observed ground motion amplitude, R is the hypo­central distance, M is magnitude and A0, a, b and q are constants to be determined. Linearization of Equation 5­2 then yields Equation 5­3 lnA = lnA0+aM­blnR+qR The Dahle attenuation model adopts a geometrical spreading model for ground motion cut­off in the vicinity of source as follows: ­1 Equation 5­4 G(R,R0) = R for RR0

The recommended value for R0 is 100 km. as a distance at which the spherical spreading for S waves is overtaken by cylindrical spreading for Lg waves. Using the geometric spreading model adopted may conveniently be written as Equation 5­5 ln A – ln G(R,R0) = c1 + c2M +c4R where c1, c2 and c4 are coefficients to be determined. Coefficients c1,c2 and c4 have been estimated with the aid of non­linear regression on the basis of the data given in Table 5­1 and Table 5­2 for the geometric spreading model described in Equation 5­5, with R0 = 100 km. The units of ground motion were selected in terms of meters and seconds. Table 5­3 Coefficients for Dahle attenuation model for spectral amplitudes at 5% damping for principal horizontal component.

23(34) Spectral Freq. C1 C2 C4 σlnA estimate (Hz) PSV(m/s) 0.25 ­7.507 1.181 0.0 1.01 PSV(m/s) 0.50 ­8.200 1.386 ­0.00046 0.81 PSV(m/s) 1.00 ­7.557 1.330 ­0.00108 0.86 PSV(m/s) 2.00 ­6.726 1.247 ­0.00173 1.01 PSV(m/s) 5.00 ­5.442 1.025 ­0.00412 0.95 PSV(m/s) 10.0 ­4.520 0.769 ­0.00454 0.90 PGA(m/s 2) 40.0 ­1.471 0­849 ­0.00418 0.83

The sigma value, the standard deviation of the regression residuals, plays an important role in any probabilistic assessment of earthquake hazard. This quantity deserves therefore considerable attention in any development of strong ground motion models. In Figure 5­1 we show the ratio between predicted and observed ground motion values versus magnitude and distance, for frequencies between 0.25 and 40 Hz. The average sigma value estimated from the predicted/observed distribution at all frequencies is shown as horizontal straight lies. The estimated sigma values are given in terms of natural logarithms in Table 5­3. The average sigma value is 0.92, while the value for PGA at 40 Hz is 0.83 and for PSV at 1.0 Hz 0.86. Compared with regional assessments of spectral estimates, the sigma values are high twice the values usually adopted in probabilistic seismic hazard studies for specific sites or restricted target areas.

Figure 5­1 Distribution of predicted vs. observed response spectral accelerations and velocities from Dahle attenuation model [xix] at frequencies 0.25,0.5,1.0,2.0,5.0,10.0 and 40 Hz, 5 per cent damping as a function of magnitude (left) and as a function of distance (right). The main reason why the Dahle attenuation model was not accepted as such in this study was large standard deviation values over twice the commonly used values. Second reason was that the spectral attenuations were developed for pseudo­velocity spectrum and not for acceleration response spectrum, which was needed for the purpose of current task. Thirdly, the frequency band used in Dahle model was not sufficiently wide for the purposes of current study. It could be expected that that spectral accelerations over 10 Hz would be of interest in developing the design spectra for geologic conditions where the bedrock outcrops were the expected level of grade at prospective site and the Precambrian rock had very high stiffness qualities. On the basis of the work and considerations given in the Section Error! Reference source not found. and Section Error! Reference source not found. ground motions were estimated from an attenuation relationships of the following form,

24(34) 1/2 ­1 1/2 Equation 5­6 ln (y) = c1 + c2M +c3ln ((R+h) ) +c4 ((R+h) ) for RR0

In Equation 5­6 y is the strong motion parameter of interest; M is earthquake magnitude, R is the distance from the earthquake epicenter to the site; h is the depth of earthquake focus; and c1, c2, c3 and c4 are regionally dependent coefficients. This form is a mix of the Dahle model and a widely used model that is in described reference [ix]. The R0 used in Equation 5­6 was chosen to be 100 km. Attenuation of strong shaking is not only dependent on distance to the source and earthquake magnitude, but is also dependent on the type of earthquake source and of geologic site conditions as was explained in the Sections Error! Reference source not found.and Error! Reference source not found.. In this study, the ground acceleration registrations used for evaluating coefficients c1, c2, c3 and c4 were selected from those geological and tectonical regions that were judged to be similar to the investigated area. The second principle for choosing these areas was the availability of registrations. By use this procedure the Saguenay region from Eastern Canada and the Newcastle region from Australia were chosen. These are both moderate seismicity, intraplate regions and the registrations were observed on the bedrock. In case of Saguenay, the bedrock was of Precambrian formations, similar to Fennoscandia, but in case of Newcastle the rock formations were sedimentary rocks. This difference in rock formation was the weakness of Newcastle data in respect of its similarity to Fennoscandia, but in other respects also this area was similar to Fennoscandia. The reason for selecting these similar areas as the source of basic data for attenuation is that there are no strong motion acceleration recordings available from Fennoscandia. In Fennoscandia, the earthquakes occur mainly at depths from 5 km to 20 km. The (80%) majority of the hypocenters are in the depth range of 10­20 km [i]. In spite of the small amount of shallow earthquakes they are important when seismic risk is concerned. Therefore the Newcastle events (depths <5 km) complete essentially the data of Eastern Canada with depths from 10 km to 30 km. Table 5­4 The Saguenay­Newcastle registration set used for evaluating coefficients c1, c2, c3¸and c4 in Equation 5­6.

Date Event Site Comp Epic.Dist Mag Type Depth (km) (km)

25/11/1988 Saguenay, Canada Site 4, Canada L,T 32 5.8 Мs 29

25/11/1988 Saguenay, Canada Site 9, Canada L,T 51 5.8 М s 29

25/11/1988 Saguenay, Canada Site 10, Canada L,T 70 5.8 М s 29

25/11/1988 Saguenay, Canada Site 11, Canada L,T 94 5.8 Мs 29

25/11/1988 Saguenay, Caiiada Site 5. Canada L,T 96 5.8 М s 29 25/11/1988 Sagucnay, Canada Site 8, Canada L,T 100 5.8 М s 29 25/11/1988 Saguenay, Canada Site 3, Canada L,T 113 5.8 Мs 29

25/11/1988 Saguenay, Canada Site 1, Canada L,T 116 5.8 М s 29

25/11/1988 Saguenay, Canada Site 7, Canada L,T 117 5.8 Ms 29

25/11/1988 Saguenay, Canada Site 6, Canada L,T 125 5.8 М s 29

25/11/1988 Saguenay, Canada Site 2, Canada L,T 152 5.8 Мs 29

06/05/1982 Miramichi, Canada Site 1, Canada L,T 6 4.0 MN 18

31/11/1982 Miramichi, Canada Site 2, Canada L,T 7 4.8 M N 18

06/08/1994 Ellalong, Australia Site 1, Australia L,T 39 5.3 ML 1.4

06/08/1994 Ellalong, Australia Site 2, Australia L,T 47 5.3 M L 1.4

06/08/1994 Ellalong, Australia Site 3, Australia L,T 95 5.3 ML 1.4

06/08/1994 Ellalong, Australia Site 4, Australia L,T 101 5.3 M L 1.4

28/12/1989 Kelunji, Australia Site 1, Australia L,T 10 5.6 M L 5

25(34) All magnitude scales are designed to be as compatible as possible with the original local magnitude (ML). Unfortunately, this is rarely the case. In addition, different regions have their own modified ML­scale. In Fennoscandia, for example, the ML estimations vary 0.1 ­ 0.5 unit between different regional formulae applied, on the average. In addition, in most of the local magnitude scales, the standard deviation of individual station magnitudes is about 0.2 units. The uncertainties related to magnitude estimates are taken into account by adding 0.1 and 0.5 units to the values of maximum magnitude estimated in Section 4. For the determination of coefficients, which are included in Equation 5­6, the method of optimization described in reference [xx] is used. The method implemented in the algorithm of reference [xx] consists in the nonlinear curve­fitting problem that is solved with the aid of the least­squares method. The initial data for Saguenay events consist of digitized, three component acceleration registrations for eleven Saguenay events, four Miramichi events and some additional registrations for Nahanni1 and Nahanni2 events. The digital SMA recordings for Newcastle events consist of four three component recordings. These events were converted to g units from SMA recordings by conversion coefficient described in reference [xxi]. Synthesized recording for Kelunji earthquake complemented the instrumentally recorded Newcastle events. The Kelunji synthesized recording was amplified to magnitude level of 5.6 from the instrumentally recorded event the magnitude 2.3 by methods described in the reference [xxii]. The processing of these materials were carried out by Microsoft Excel software. First, the acceleration histories were plotted on the basis of the recordings. From these plots we choose the ten­second time windows corresponding maximum acceleration peaks. From these acceleration plots the response spectra for 5% damping were computed with the aid of software from reference [xxiii]. The spectral ordinates are calculated for longitudinal and transversal earthquake components. The investigated frequency values were 0.3,1, 2, 5, 7, 10, 15, 20, 25 and 97 Hz. The magnitude value for the Saguenay event was 5.8 and the depth focus was 29 km. The Kelunji synthesized recording were used for spectral calculations after the scaling of the magnitude to the value of 5.6 and the distance to the value of 10 km. The magnitude scaling was carried out according to the relations given in reference [xxiv] and the distance scaling according to the relations given in reference [xxi]. The Miramichi recordings used for spectral calculations were scaled for same magnitude and same depth value as the recordings of Saguenay event with the aid of relations in the reference [xxiv]. The Kelunji recordings were used to supplement the four Ellalong recordings that are called together the Newcastle recordings in this report because of their vicinity to town of Newcastle in New South Wales, Australia. The two recordings of the Miramichi event were used to supplement the ten Saguenay recordings that are called Saguenay recordings in this report after the name of the main event. In the following figures the attenuation relations for the logarithm of peak ground acceleration or various spectral accelerations are given as well as the attenuation for the peak ground acceleration acceleration ordinates. The figures are for longitudinal and transversal components of Saguenay and Newcastle event, which for the purposes of this study are regarded as independent recordings. For reliable determination of the coefficient c2 the magnitude data represented in Saguenay­Newcastle data set was too scarce. For that reason the c2 coefficient was determined on the basis of reference [xxv]. The attenuation curves for Saguenay and Newcastle datasets are given next:

26(34) Figure 5­2 Spectral attenuation fit for Saguenay longitudinal component. Damping 5%. Magnitude 5.8.

Figure 5­3 Spectral attenuation fit for Saguenay transversal component. Damping 5%.

27(34) Magnitude 5.8.

Figure 5­4 Spectral attenuation fit for Newcastle longitudinal component. Damping 5%. Magnitude 5.8.

28(34) Figure 5­5 Spectral attenuation fit for Newcastle transversal component. Damping 5%. Magnitude 5.8.

Table 5­5 The coefficients c1, c2, c3, c4 and the logarithmic standard deviation of the attenuation model.

Spectral Freq. c1 c2 c3 c4 σlnA estimate (m/s2 ) (Hz)

PSA R0<100km 0.30 ­11.83 1.30 ­0.234 ­0.03688 0.68 PSA R0>100km 0.30 0.2262 1.30 ­3.217 0.0 0.68 PSA R0<100km 1.00 ­11.69 1.30 ­0.679 ­0.04827 0.59 PSA R0>100km 1.00 3.966 1.30 ­3.805 0.0 0.59 PSA R0<100km 2.00 ­7.475 1.20 0.4457 ­0.02088 0.64 PSA R0>100km 2.00 ­4.520 1.20 ­2.451 0.0 0.64

PSA R0<100km 5.00 ­3.941 1.00 0.6711 ­0.02049 0.62 PSA R0>100km 5.00 ­0.435 1.00 ­2.652 0.0 0.62

PSA R0<100km 7.00 ­6.518 1.00 ­0.324 ­0.04112 0.54 PSA R0>100km 7.00 6.910 1.00 ­2.652 0.0 0.54

PSA R0<100km 10.0 ­5.965 1.00 ­0.070 ­0.03213 0.49 PSA R0>100km 10.0 4.5924 1.00 ­2.948 0.0 0.49

PSA R0<100km 15.0 1.6144 1.20 2.7021 0.008361 0.57 PSA R0>100km 15.0 ­0.167 1.20 ­2.149 0.0 0.57

PSA R0<100km 20.0 8.0727 1.20 4.7478 0.03467 0.72 PSA R0>100km 20.0 ­1.639 1.20 ­1.896 0.0 0.72

PSA R0<100km 25.0 13.780 1.20 6.6452 0.061186 0.72 PSA R0>100km 25.0 ­3.982 1.20 ­1.463 0.0 0.72

PGA R0<100km 97.0 7.9296 1.20 4.8645 0.036029 0.49 PGA R0>100km 97.0 ­2.188 1.20 ­1.895 0.0 0.49

The attenuation is used as input for SEISRISK III [viii] in table form. The parameter in these attenuation tables is the earthquake magnitude and the argument is the hypo­central distance R. All attenuation relationships are developed for solid bedrock and all site effects are excluded from this investigation because the feasible target sites are solid rock sites. The depth of the source information is included in the Saguenay­Newcastle data set. The source depth varies from 1.4 to 29 km, which covers the range of expected source depths in the Olkiluoto3 earthquake catalog source area circles. The total amount of acceleration records used in the study is 36. This number is considered quite adequate for the purposes of the study. The chosen records were compared to those used in Dahle study [xix] and EPRI guidelines [xxvi]. The aim in choosing the available intraplate recordings was not obtain largest possible number but the best possible relevance to the target regions geological and seismological conditions: pre­cambrian solid rock, moderate seismicity about the same level to be expected in Fennoscandia, intra­plate conditions

6 Decision tree for the treatment of uncertainties The code basis for the ground motion estimation in probabilistic seismic hazard studies stipulates the median spectra for mean return period of 100 000 years [xxvii]. The decision three used in the treatment of uncertainties in this study was as follows:

29(34) SAG U ENA Y LO N GITU D INA L(0.3) D L observed a,b SAG U ENA Y TR AN SV ER SAL(0.3) O m ax. E by N EW C ASTL E LO N GITU D IN AL(0.2) V m agn. S sour N EW C ASTL E TR A NSV ER SA L(0.2) ­ +0.1 (0.5) I I ces I (0.6) observed SAG U ENA Y LO N GITU D INA L(0.3) G SAG U ENA Y TR AN SV ER SAL(0.3) S m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) N N EW C ASTL E TR AN SV ER SA L(0.2) A +0.5 (0.5) C A a,b observed SAG U ENA Y LO NG ITU DINA L(0.3) SAG U ENA Y TR AN SV ER SAL(0.3) T by m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) +0.1 (0.5) N EW C ASTL E TR AN SV ER SA L(0.2) A cata­ logs L (0.4) observed SAG U ENA Y LO NG ITU DINA L(0.3) O SAG U ENA Y TR AN SV ER SAL(0.3) m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) G N EW C ASTL E TR AN SV ER SA L(0.2) G (0.5) +0.5 (0.5) R O U O SAG U ENA Y LO NG ITU DINA L(0.3) N a,b observed L SAG U ENA Y TR AN SV ER SAL(0.3) D by m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) K N EW C ASTL E TR AN SV ER SA L(0.2) sour­ +0.1 (0.5) I ces L (0.6) observed SAG U ENA Y LO NG ITU DINA L(0.3) SAG U ENA Y TR AN SV ER SAL(0.3) U m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) O +0.5 (0.5) N EW C ASTL E TR AN SV ER SA L(0.2) T O observed SAG U ENA Y LO NG ITU DINA L(0.3) a,b SAG U ENA Y TR AN SV ER SAL(0.3) C m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) by N EW C ASTL E TR AN SV ER SA L(0.2) A +0.1 (0.5) cata­ T logs A M (0.4) O L observed SAG U ENA Y LO NG ITU DINA L(0.3) SAG U ENA Y TR AN SV ER SAL(0.3) T O m ax. m agn. N EW C ASTL E LO N GITU D IN AL(0.2) I +0.5 (0.5) N EW C ASTL E TR AN SV ER SA L(0.2) O G N (0.5)

Figure 6­1 Logic three structure for treating uncertainties In the logic tree the long vertical bar at far left of the diagram summarizes the task at hand, namely, generation of design ground motion for the target window. It branches in two subsequent vertical bars named Loviisa catalog and Olkiluoto catalog. The meaning of these bars are the 500 km radius source regions around both prospective sites as described in Chapter 4. These bars are called catalogs because they include that particular subset of the parent catalog

30(34) FENCAT that is located inside each particular source circle. Each catalog in its turn is divided to source areas also described in Chapter 4. Both Loviisa and Olkiluoto catalogs have six source areas, which contiguously fill whole source circle drawn around Loviisa, and Olkiluoto3 sites. The source term seismicity felt at particular site is originated in the source areas. The seismicity of source areas is quantified by Gutenberg­Richter magnitude recurrence relationship determined by two parameters: a and b. Consequently, the four following vertical branch bars in the decision tree characterize the seismicity of source areas. Two different hypotheses are adopted to characterize the source areas, namely, characterization by areas itself or characterization by respective parent catalogs. Characterization by areas means that Richter's b parameters are estimated on the basis of events occurring inside each particular source area. Characterization by catalogs means that Richter's b parameters are calculated on the basis of events occurring inside each particular catalog. Next eight branches in the decision tree describe various assumptions concerning the maximum possible magnitude inside each source. The determination of maximum magnitudes attributed to source areas is probably the most difficult and controversial aspect of probabilistic seismic hazard analysis. In this study after the suggestion of reviewers the maximum observed magnitude inside each source area was added by 0.1 magnitude units and 0.5 magnitude units meaning that the maximum possible magnitude was twice as large as the maximum observed or five times as large as the maximum observed. Both these hypotheses were assigned equal weights. Also both catalogs used in the decision tree were assigned equal weights and the weights for hypotheses for Richter parameters were assigned as follows: 0.6 for determination by source areas and 0.4 for determination by respective catalogs. Each box in the fourth level of the decision tree branches to four branches meaning the assumptions concerning the attenuation of earthquake ground motion. The adopted four variants of the attenuation relationships were models evaluated based on Saguenay events longitudinal component recordings, Saguenay transversal component recordings, Newcastle longitudinal and Newcastle transversal recordings, respectively. The weights for Saguenay were 0.3 and for Newcastle 0.2. Each branch end node in logic­tree of Figure 6­1 characterize the credible alternative inputs to probabilistic seismic hazard analysis and their likelihoods. The end node likelihood can be calculated by multiplying the branch likelihoods leading to end node. The sum of end node likelihoods as well as branch likelihoods at each level must be one.

7 Results By preparing the initial data according to the previous section, the further analysis was carried out on Fortran Computer Program for Seismic Risk Analysis described in reference [viii]. The resulting raw site hazard curves are given in following figure:

31(34) Olkiluoto3 raw seismic hazard curves in PGA; Seisrisk II analysis

1.E­01

1.E­02 ce n

a 1.E­03 d e 1.E­04 xce e

f

o 1.E­05 cy n

e 1.E­06 u q re f

1.E­07 l a u

n 1.E­08 n A 1.E­09

1.E­10 0.0001 0.001 0.01 0.1 1 PGA amplitude (g)

L0_1ln L0_1ls L0_1tn L0_1ts L0_1Rln L0_1Rls L0_1Rtn L0_1Rts L0_5ln L0_5ls L0_5tn L0_5ts L0_5Rln L0_5Rls L0_5Rtn L0_5Rts O0_1ln O0_1ls O0_1tn O0_1ts O0_1Rln O0_1Rls O0_1Rtn O0_1Rts O0_5ln O0_5ls O0_5tn O0_5ts O0_5Rln O0_5Rls O0_5Rtn O0_5Rts

Figure 7­1 Seismic Hazard in PGA for Olkiluoto3 site. Presented are the 32 raw hazard curves forming the hazard distribution at site according to the decision tree presentation of Figure 6­1.

Distribution of hazard curves for Olkiluoto site; Seisrisk II analysis

1.E­02

1.E­03

e

c 1.E­04 n a d e e

c 1.E­05 x e f o

y 1.E­06 c n e u q e

r 1.E­07 f

l a u n 1.E­08 An

1.E­09

1.E­10 0.0001 0.001 0.01 0.1 1 PGA amplitude (g)

5%fractile median 95%fractile Figure 7­2 Hazard curves with confidence bounds (5%, median, 95%) for Olkiluoto site 8 References

[i] Ahjos, T. And Uski, M. 1992. Earthquakes in northern Europe in 1375­1989. Tectonophysics, 207: 1­23.

32(34) [ii] Ahjos, T., Saari, J., Penttilä, E. and Korhonen, H. 1984. Earthquakes and seismic hazard in Finland. Engineering Geology, 20:1­12. [iii] Saari, J. 1998b. Regional and Local Seismotectonic Characteristics of the Area Surrounding Loviisa Nuclear Power Plant in SE Finland. Institute of Seismology, University of Helsinki. Report 1998. [iv] Saari, J. 1998c. Seismicity in the Kivetty area (in Finnish with an English abstract). Posiva Oy, 36p. Working Report 98­43. [v] Axberg, S. 1981. Seismic stratigraphy and bedrock geology of the Bothnian Sea, Northern Baltic. Acta Universitatis Stocholmiensis. Stockholm Contributions in Geology. Vol. XXXVI, no 3, pp.154­212. [vi] Barosh, P.J. 1969. Use of seismic logarithm of ground acceleration data to predict the effects of earthquakes and underground nuclear explosions in various geologic settings. U.S.Geological survey bulletin, 1279: 93. [vii] Saari J. 1998a. Seismicity in the Olkiluoto area area (in Finnish with an English abstract). Posiva Oy, 37 p.Working Report 97­61. [viii] Bender B. , Perkins David M. 1987. SEISRISK III: A Computer Program for Seismic Hazard Estimation, U. S. Geological Survey Bulletin 1772, United States Governement Printing Office; Washington. [ix] McCuire, R.K. 1976. Fortran computer program for seismic risk analysis, USGS Open­File Report 4­76, pp. 1­90. [x] Cornell, C.A. 1968. Engineering seismic risk analysis. Seismol. Soc. America Bull., v.58, no.5: 1503­1606. [xi] Cornell, C.A. 1974. Seismic risk analysis program support documents. Cambridge, Mass., 110p. [xii] Merz, H.A. And Cornell, C.A. 1973. Seismic risk analysis based on a quadratic magnitude­ frequency law. Seismol. Sos. America Bull., v. 63, no.6, pt. 1: 1999­2006. [xiii] Esteva, L. 1970. Seismic risk and seismic decisions, in Hansen, R.J., ed., Seismic design for nuclear power plants. Cambridge, Massachusetts Inst. Technology Press: 142­182. [xiv] Richter, C.F. 1958. Elementary seismology. San Francisco, W. H. Freeman and Co.: 768 p. [xv] Joyner, W. B., and Boore, D. M., 1981. Peak horizontal acceleration and velocity from strong motion records including records from the 1979 Imperial Valley, California earthquake. Bulletin of Seismological Society of America v. 71, p. 2011­2038. [xvi] Mäntyniemi, P. and Ahjos, T., 1990. A Catalog of Finnish Earthquakes in 1610 ­ 1990. Geophysica, 26(2): 17­35. [xvii] Kijko, A., Skordas, E., Wahlström R. and Mäntyniemi P., 1993. Maximum Likelihood Estimation of Seismic Hazard for Sweden. Natural Hazards, Vol. 7, 41­57. [xviii] Saari, J., 2000. Seismic activity parameter of the Finnish potential repository sites for the calculations of bedrock displacements induced by earthquakes. Posiva Oy. Working report (manuscript). [xix] Dahle A., Bungum H., Kvamme L. B., Attenuation Models inferred from Intraplate Earthquake Recordings, Earthquake Engineering and Structural Dynamics, vol 19, 1125­ 1141(1990). [xx] Branch, M.A., Grace, A. 1990. Optimization Toolbox, for use with Matlab. [xxi] McCue, K., Dent, V. and Jones T. The characteristics of Australian strong ground motion. Pacific Conference on Earthquake Engineering, Australia, 20­22 Nov. 1995: 71­80. [xxii] Sinadinovski, C., McCue K. F., Somerville M. Strong ground motion simulation of Australian intra­plate earthquakes, 17th European Conference on Earthquake Engineering, 1998, Balkema, Rotterdam. [xxiii] Xu J.,Philippacopoulas A. J., Miller, C. A., Constantino C. J., Cares (Computer Analysis for Rapid Evaluation of Structures 1.0, NUREG/CR­5588,BNL­NUREG­52241,Vol. 1­3, Brookhaven National Laboratory, May 1990.

33(34) [xxiv] Protection of Nuclear Power Plants against Seismic Effects Reference Ground Motion: Practice Followed in European Countries (Synthesis Report), prepared by Prof. Ludwig Ahorner, Universität Köln, European Applied Research Reports, Nuclear Science and Technology, Vol 4, No. 6, (1983) pp.1325­1422. [xxv] Ahorner L., Protection of Nuclear Power Plants Against Seismic Effects, Reference Ground Motion Practice Followed in European Countries, European Applied Research reports, Nuclear Science and Technology, Vol. 4, No. 6 (1983), pp. 1325­1422. [xxvi] Guidelines for Determining Design Basis Ground Motions, Volume 1: Method and Guidelines for Estimating Eathquake Ground Motion in Eastern North America, EPRI TR­ 102293, November 1993. [xxvii] Draft Regulatory Guide DG­1015, Identification and Characterization of Seismic sources, Deterministic Source Earthquakes, and Ground Motion, U.S. Nuclear Regulatory Commission Office of NuclearRegulatory Research, November 1992.

34(34)