Earthquake hazard assessment in the North Sea

Earthquake hazard assessment in the North Sea

Physics of the Earth and Planetary Interiors, 44 (1986) 201—210 Elsevier Science Publishers B.V., Amsterdam — Printed in The Netherlands 201 Earthqu...

1MB Sizes 3 Downloads 20 Views

Physics of the Earth and Planetary Interiors, 44 (1986) 201—210 Elsevier Science Publishers B.V., Amsterdam — Printed in The Netherlands

201

Earthquake hazard assessment in the North Sea H. Bungum and P.H. Swearingen NTNF/NORSAR, P.O.B. 51, N-2007 Kjeller (Norway)

G. Woo Principia Mechanica Lid., 50 Vineyard Path, East Sheen, London SWI4 8ET (Gt. Britain) (Received September 24, 1985; revision accepted December 2, 1985)

Bungum, H., Woo, G. and Swearingen, P.H., 1986. Earthquake hazard assessment in theNorth Sea. Phys. Earth Planet. Inter., 44: 201—210. The assessment of seismic hazard in regions such as the North Sea where the activity rate is comparatively low requires the adoption of a methodology designed to maximize the use of available sources of information, including historical and instrumental seismicity, geology, tectonics and local soil properties. In a series of site-specific studies of seismic hazard in the Norwegian sector of the North Sea, a procedure has been developed which reduces uncertainties through emphasis on primary data. Thus the sizes and locations of felt extents have been reassessed from original historical documents; the epicentres of early twentieth century events have been relocated using instrumental data, and a correlation between felt areas and M, has been developed which allows magnitudes to be assigned to historical earthquakes. Recently recorded instrumental data have been used in delineating the finer details of the current seismicity patterns. An essential part of the adopted methodology is a thorough geological study based on seismic reflection and other geophysical data, with the main purpose to delineate possible active faults. The computationof seismic hazard includes the modelling of tectonic regions as derived from a seismotectonic synthesis of both seismological and geological data, as well as the modelling of individual active faults where also non-planar geometries (such as listricities) are accounted for. A reliable hazard assessment also requires much emphasis on strong-motion attenuation models, where new relationships now have been derived, specifically applicable to North Sea conditions. A seismic hazard assessment finally requires a comprehensive analysis of uncertainties, both because a total uncertainty estimate for the computed loads is needed, and because it is essential to know the relative importance of the different sources of uncertainty. For this purpose, a. method has been adopted, based essentially on ,a Bayesian analysis in combination with a Monte Carlo technique.

1. Introduction Until the recent exploration and development of North Sea oil and gas resources, the degree of NORSAR contribution No. 360. This paper was part of Symposium No. 1, Earthquake Hazard Assessment and Prediction presented at the 23rd General Assembly of the International Association of Seismology and Physics of the Earth’s Interior held ~n August 19—30, 1985, Tokyo, Japan. 0031-9201/86/$03.50

© 1986 Elsevier Science Publishers B.V.

attention given to the seismicity of the area was limited by the low level of academic interest. Individual national catalogues of locally felt earthquakes were maintained by the various seismological institutes of Scaildinavia and Britain, but no pressing reason existed for a homogeneous regional catalogue to be produced, since the perceived benefit of such a catalogue was outweighed by the amount of work. involved. In the past 20 years, large investments have been made in recovering oil and gas from the

202

depths of the North Sea, which has made the region into one of the world’s most important economic zones. To safeguard this investment against environmental hazards, stringent precautions are taken to meet wave and wind-loading on offshore platforms, and increasingly rigorous measures are taken to investigate earthquake loading, The Norwegian Petroleum Directorate (1983) now requires seismic hazard analyses to be conducted for all offshore oil production facilities. For several major offshore sites in the Norwegian sector of the North Sea, this requirement has led Norwegian oil companies to commission state-ofthe-art studies from a joint Norwegian—British consortium consisting of NORSAR, Norwegian Geotechnical Institute, and Principia. These studies have established a new foundation for the determination of engineering seismic hazard in the region, one which places emphasis on scientific methods of data~acquisition and analysis (NORSAR et a!., 1984, 1985; NGI et a!., 1985). These methods are associated with a range of subjects including historical and instrumental seismicity, attenuation of ground motion, structural geology and tectonics and soil mechanics. In each of these areas research has been conducted to review and expand the data base of technical information and to interpret the data in a systematic and scientific fashion so that the resulting computations of seismic hazard are soundly based. The purpose of the present paper is to describe these methods in more detail, with emphasis on seismology and seismic hazard computations. The historical seismicity in this area is discussed in more detail by Muir Wood et a!. (1985) and the recent seismicity by Bungum et al. (1986b), while the geology, the local soil properties and the engineering aspects will be covered more thoroughly in a forthcoming paper (Bungum et a!., 1986a); 2. Historical seismicity Just as today, the areas of Europe most poorly covered instrumentally are those on national boundaries (Adams, 1985), so in the past the same has been true of earthquake documentation. The North Sea is bordered by Norway, Denmark,

.

Germany, Holland, Belgium and the British Isles, and traditionally the reporting and collation of reports has been a strictly national enterprise. Thus for events originating in and around the Skagerrak, which have been felt in both Denmark and Norway, few attempts have been made to combine the macroseismic data into single event maps. Since the primary function of a national seismological institute was narrowly defined in a parochial way, its concern was the documentation of earthquake effects felt within its national boundaries rather than the elucidation of a more complete picture. For example, the Swedish catalogue lists a small earth tremor at Varpnas, Warmland, on January 24, 1927 at 6.15 a.m., with no reference to the large North Sea earthquake coincident with it, which was felt in the Shetlands and northeastern Scotland. The absence of synoptic intensity and isoseismal maps of historical earthquakes has for a long time been an important shortcoming of North Sea seismicity coverage, and the production of a uniform set of new macroseismic maps has been given due priority in the reassessment of regional earthquake activity. As a first step towards this objective, a review of historical sources of earthquake documentation has been undertaken, which traces the changing pattern of earthquake observations from the eleventh century to the present. A graphical illustration of the fluctuation in regional earthquake documentation is shown in Fig. 1. For Norway,~Sweden, England and Scotland the number of earthquakes reported per 20 years is charted from 1650 to 1900. The first surge of interest in earthquake phenomena followed in the wake of the great Lisbon earthquake of 1755. However, it was not until the latter half of the nineteenth century that earthquake documentation received the impetus of attention from Davison in Britain and Reusch in Norway, who both increased substantially the volume of earthquake information through soliciting the assistance of :the general public. Before their contribution to archive research, the documentation of earthquakes was unsystematic, thus the task of unravelling the diverse strands of evidence for an early event today re-

203 2

EVENTS MSS.5 A~CASOVE

i

63.0

‘7.c~

••,‘

x Scotland ~

~.

+

England

V

Sweden

.

_________

~ Norway

,,

.

~S9.O-

1850

1700

1750

1800

1850

.

~

-6.0

Fig. 1. Number of earthquakes reported per decade in Norway, Sweden, England and Scotland from 1650 to 1910.

quires painstaking work in libraries and civic records offices. Indeed, even for events of the past 100 years this is true, given the practice of cataloguers such as Davison to publish their own interpretation of earthquake information, and not the information itse!f. In view of the need for basic historical research, a program of information retrieval from libraries and record offices in Norway, Denmark, Sweden and Britain has been instigated. This has resulted in the acquisition of new information on many known events, the identification of event duplications and meteor-induced tremors, and the discovery of some new events. In the course of this programme, data have been exchanged with Ambraseys who has lately made his research public in the same area (Ambraseys, 1985). With this new data base of primary earthquake data (see Fig. 2), a uniform set of intensity and isoseismal maps has been produced which can be compared with each other. For example, the comparison of maps from different periods allows the known instrumentally determined epicentres of recent events to be used to estimate the epicentres of earlier events with similar intensity distributions (Muir Wood et al., 1985). This kind of

/

~

-2.0

)

i 2.0

I

/

I 6.0

40.0

~ ~ I T~ ~ E C C~ G E Fig. 2. Map of reported earthquakes in the North Sea region from 1884 to 1983, including events with an M, magnitude of 3.5 or above. The figure also includes a tectonic regionalization, or area source model, for the North Sea and surrounding regions. The areas with the highest activity levels are those along the coast of western Norway and in the Viking Graben (at around 2°E).This source model is very rough and regional; it is intended primarily for illustration purposes, and not . .. normally applicable for specific analysis of particular sites. Offshore tectonic information from Hamar (1979).

comparison has previously been very difficult since the maps were produced by different people using a variety of methods. The use of recent instrumental data to infer seismological parameters of historica! events is a technique which can be applied also to magnitude determination. For each felt earthquake with measurable felt areas within the outer isoseismals III and/or IV, an estimate of surface wave magnitude can be obtained via a correlation using twentieth century macroseismic and instrumental data. The form of this correlation is explained in the next section on instrumental data.

3. Instrumental seismicity data To establish a means for quantifying the sizes of historical earthquakes, the surface wave magni-

204

tudes (M,) for a set of significant regiona! events of this century have been recalculated and correlated with macroseismic felt areas. M, is used rather than an alternative definition of magnitude because of the greater quantity of data readings for early events. The form of the derived correlation expresses M, in terms of the areas within isoseismal contours III and IV in the following way M, = —0.36 + 1.00 log A (III) (i) M5

=

0.91 + 0.82 log A(IV)

(2)

In any region the instrumental record of seismicity provides a quantitative picture of the earthquake activity during the period of actual observation. Where the seismicity is high and the observation thresho!d low, as in California, this picture in itself can delineate the broad pattern of local activity. However, in less active regions such as the North Sea the instrumental data must be viewed in conjunction with the historical data if the proper perspective on seismicity is to be gained, In the past the pattern of seismicity in the North Sea inferred from historical sources has not been directly comparab!e with that inferred from instrumental observations because both macroseismic and instrumental data sets have been ridden with extraneous noise, weakening any apparent correlation. With the refinement of this historical catalogue by primary research, the sizes and the epicentre locations of the regiona!ly felt earthquakes have become more precise!y known, allowing a meaningful comparison with instrumental data, given that a similar refinement in instrumental data quality is made. To achieve the necessary improvement in instrumental data quality, several tasks have been undertaken. First, an attempt has been made to relocate, using modern software, the epicentres of regional earthquakes from the beginning of this century. These old epicentres were calcu!ated rather crudely by hand and should be regarded as provisional. The most important relocation has been for the northern North Sea earthquake of January 24, 1927. Commonly referred to as the Viking Graben event, although Bath (1956) located it east of this tectonic structure, this event has now been located at 59.9°N, 1.8°E on the western

margin of the Viking Graben inside the British sector (Principia, 1985). Another improvement which has been made is to eliminate from the ISC catalogue for the postwar period some of the many errors of dup!ication, omission and mislocation which have sett!ed in the cata!ogue. With its umbre!!a coverage of the world’s seismicity, the ISC cannot be expected to carry out this work using its own limited resources, and it should be the responsibility of all users of this catalogue to assist in its refinement. The most notable advance in the ana!ysis of instrumental data has been the integration of British and Scandinavian seismological data for the northern North Sea covering the 5-year period (1980—1984). Until quite recently, the British and Norwegian stations were run on strictly national lines, but through co!!aboration between the British Geological Survey (BGS), NORSAR and the University of Bergen, the situation now exists whereby the full benefit of seismological stations on both sides of, and in the North Sea itself, can be realized (Bungum et a!., 1986b). A carefu! analysis of the joint seismological EVENTS

ML-2.O Ap(i ABOVE

63.0

,

• ••

-

64.0

-

7~ .‘

~i.•

J-~~7 S~7.0

<

/

:

ss.o•

• 6.0

-2.0



I

2.0

I

6.0

40.0

~Or~ TLJDa (DEG E Fig. 3. Map of earthquakes in the North Sea region from 1980 to 1984, including events with an ML magnitude of 2.0 or above. The regionalization is identical to the one in Fig. 2.

205

data for the northern North Sea leads to the epicentre map shown in Fig. 3. This shows clear evidence for the activity of the Viking Graben as well as offshore Norway, which is a finding previously blurred by the lack of sufficient high quality data. This is consistent with the relocation of the January 24, 1927 earthquake to the Viking Graben, and indeed the entire pattern of instrumental data in the northern North Sea achieves this level of consistency with the macroseismic picture, which has been developed after a much longer time exposure.

4. Geology and tectonics The combination of data on regional seismicity and crustal deformation provide the foundation for the understanding of regional seismotectonics. Using the new historical and instrumental catalogues and through the investigation of stress determinations and strain measurements, including neotectonics, we have constructed a theory of regional seismotectonics. The main features of this theory are explained here. The North Sea region can be seismotectonically partitioned into three main zones: western Fennoscandia, offshore eastern England, and a northwest—southeast oriented aseismic band passing from Scotland to Germany. To the south the seismicity appears to be localized around the margins of regional uplift domes. The central North Sea is subsiding at a rate ten times its 60 Ma-average, but apparently aseismically, and there is very little current activity associated with this zone. The causes of seismicity in the north of the region are probably in part related to crustal deformation associated with lithospheric unloading following ice-cap melting, and to various residual tectonic processes. The Viking Graben is now subjected to extensional tectonics as indicated by an estimated spreading rate of about 0.1 mm year~ (about two orders of magnitude below plate margin values) as well as by a recently obtained focal mechanism with oblique normal faulting and east—west extension. In western Norway the stress field is dominated by horizontal compression in a north-

west—southeast direction (Havskov and Bungum, 1986). While this is consistent with a glacial rebound model (NORSAR et al., 1984), it is also noteworthy that this stress regime is also consistent with what is observed in other parts of Norway (Kibsgaard, 1985) and Scandinavia (Slunga, 1985), as well as in the rest of northwestem Europe ~Bungum and Fyen, 1980). This shows that stress-generating mechanisms of continental and plate-related origin are important also for the areas in and around western Norway, while the question of where earthquakes occur of course also depends on how this stress field interacts with existing faults, fracture- and weakness zones (Sykes, 1978). The boundaries between these seismotectonic provinces appear in part to be determined by major tectonic fracture zones, which need to be recognized in determining the geometry of seismic source areas used in hazard modelling. Also of importance to hazard modelling is evidence for neotectonics, particularly in the vicinity of an engineering site. Some unsubstantiated claims for neotectonic deformation have been made in a number of ice-eroded regions of England, Scotland, Denmark and western Norway. For sites in offshore areas, it is necessary to have access to and examine data from geophysical sources to determine the presence of local active faulting or other forms of tectonic deformation. From an inspection of multi-channel seismic reflection profiles, coupled with data from shallow seismic traverses and available oil-well logs obtamed in the course of hydrocarbon exploration and reservoir development, it is possible to investigate both regional and local structural movements. This is done by establishing periods of faulting through examination of displacements along faults and the disturbance of geophysical reflectors above faults, and it is also accomplished by establishing periods of regional flexuring through examination of horizons which can reasonably be assumed to have been horizontal at the time of deposition. This method of investigation has been applied to several specific sites in the Norwegian sector of the North Sea with interesting results which confirm the value of such work. In particular, evi-

206

dence has been found for significant tectonic phases of Tertiary faulting. The øygarden Fault off the coast of southwestern Norway has been found to have undergone substantial amounts of post-Pliocene displacement. Data from geological sources on fault length and slip-rate have been used as input parameters for the modelling of active faults within seismic hazard analysis, which is now discussed.

a ooo



DO

10

5. Seismic hazard analysis

I 2

The research undertaken into historical earthquakes, instrumental seismicity, geology and tectonics has been directed towards application to engineenng seismic hazard analysis. Rehance in previous hazard studies on catalogues and other data prepared long before the advent of offshore drilling has obliged past hazard analysts to carry out their computations with a data base often ill-suited to the selected purpose. From the new catalogue of integrated instrumental and historical data, seismicity parameters associated with the various seismic zones can be calculated to fit the Gutenberg—Richter magnitude-frequency relation 1 og N



a



b~M,

“(3\~‘

In particular a regional b-value of 1.3 has been calculated using data included within a spacewindow of Fig. 2, and a time-window of 100 years from 1884 to 1983 (Fig. 4). Similar b-values are found also when smaller areas are analysed. This high b-value, which indicates the rarity of large earthquakes in the region, diminishes the contribution of large events to the regional seismic hazard, which is dominated by moderate magnitude earthquakes. Thus there is little sensitivity to the value of maximum magnitude, if this is taken to be above 6.0. In the computation of the probability of ground motion exceedance, the specification and parameterization of a seismic source model must be accompanied by appropriate relations for ground motion attenuation. Since M, is the type of magnitude used to quantify seismicity, it should also be used as the scaling parameter in the description of

3

4

L I

5



I 6

MA GM I T U CE

Fig. 4. Magnitude—frequency distribution for the data plotted in Fig. 2. The straight line is the result of a linear fit through the cumulative distnbution, with a slope (b-value) of 1.30± 0.04. The bars inside the graph represent the incremental number of events.

attenuation. This alone requires the development of a new set of attenuation relations. A serious limitation here is that no strong-motion data are available from the North Sea area, and also that very few data are available from areas that are similar geologically and tectonically. We have approached this problem then by selecting a data set of hard ground strong-motion records fromregions intraplate southern including Europe and eastern a number U.S.A., of Australia and Germany. After a correction for differences between intensity attenuation relationships for the areas where the records are taken from and the North Sea area, spectral pseudovelocity attenuation relations have been constructed covering the frequency range of engineering interest. For deep-water structures this is less than 1 Hz. The records have been corrected with a recursive elliptic filter method implemented by Shyam Sunder and Connor (1982), which avoids some of the cut-off problems connected with the standard Ormsby filter. In the selection procedure for the strong-motion data set, importance has been attached to the magnitude distribution of events represented, since the magnitude range of interest is lower for the North Sea region than for more active zones. Using this data base and analysis procedure, we have derived the following attenuation rela-

207

tionship for peak ground acceleration PGA (in cm ~_2) as a function of magnitude M, and hypo-

quake source model as for peak ground acceleration.

central distance R (in km) ln PGA

=

2.65

i- 0.92

M,

The computation of seismic hazard for a given



0.97 lnR

+

ln ~

(4)

where c is a log-normally distributed error term with expectance 0 and standard deviation a. The value adopted for a has been of the order of 0.55 (Bender, 1984). For pseudo-velocity FSV (in cm s’), similar relationships have been derived for a large number of frequencies of engineering interest, taking the general form ln PSV(f)

=

c

1(f) + c2(f) . M,



c3 (f) ln R + in e

(5)

Figure 5 shows as an example the derived attenuation relationships for four selected frequencies between 0.3 and 40 Hz, for an M, 5.5 earthquake in the case of horizontal motion and with 5% damping. With these parameters pseudo-velocity response spectra can be estimated directly and probabilistically according to the uniform risk method (McGuire, 1977), using the same earth=

engineering site has been performed using EQRISK (McGuire, 1976) for area seismic sources, and PRISK (Principia Risk Program) for individual fault sources. The program PRISK has the capability of modelling first-order non-planar geometries so that fault curvature and listricity can be taken into account. An example of a simple earthquake source model, without specific faults, is given in Figs. 2—3. The seismicity within each of these source areas (or tectonic provinces) is assumed to be uniform, and determined by the parameters a and b in eq. 3. It should be emphasized here that, while this model is an appropriate average model for the North Sea, it would normally not be appropriate for a specific site. The reason for this is that a site-specific analysis normally requires a more detailed delineation of the sources near the site, with the possible use of specific fault modelI

I

I

0

( (

5.5 Ms 2.0Hz

:‘

__________

__________

0.3HZ

:.

20Hz

___________________

~

400

1000 DISTANce (KN)

Fig. 5. Attenuation of pseudo-velocity (in cm s”l) vs. distance (in km) for frequencies of 0.3, 2.0, 20 and 40 Hz, for horizontal motion (5% damping) from an M, = 5.5 earthquake. The 40Hz curve corresponds directly (after converting from PSV to PGA) to the PGA relationship in eq. 4. The relations have been derived for North Sea conditions at base rock level,

C

8

,Bo_~

a

S v ‘S.’

‘6 ~

I

/

‘.~

,\ :‘~

‘..

‘~SSSS’

A

0 10

p

~

a.

U’

______

/

/

/

/

I

--

a, U’

___________________

(

5

-2

1

‘~

0

2

4

if~. /

‘°o~s~_-~ 6

a

to

L ON 01 T U 0 E

Fig. 6. Contour diagram for peak ground acceleration (PGA) in cm ~2 at a i0~ year’ probability level, using the area source model in Figs. 2—3 and the PGA attenuation relationship in eq. 4 together with a ‘best estimate’ model for the rest of the input parameters.

208

ling. Nevertheless, in using this source model and the attenuation relationship defined in eq. 4 above, we obtain peak ground accelerations (PGA) at a iO~ y~ probability level for western Norway and the northern North Sea as shown in Fig. 6. For the same source model, and with the derived PSV attenuation relations from eq. 5, we obtain response spectra for a site at 59.5°N,4.0°E(not close to any known oil field) as shown in Fig. 7. Because the seismic hazard in the North Sea region is dominated by events in the magnitude range 4.5 to 6.0 M~,which have smaller low frequency content than large magnitude earthquakes, these site-specific response spectra are typically lower in the frequency range of offshore platforms than generic response spectra such as API RP 2A produced for regulatory purposes in America, and intended to cover all seismic offa&

..-..‘

..

.•~.•.‘

~

0

_____________

~

~

I IS 425 435 ~

I

~

48519520:2,5

I

225

Fig. 8. Histogram of incremental probability (in %) vs. acceleration (in cm 2) at a io—~year — probability level, using the hazard model defined above for a site located at 59 5°N

‘‘~

4.0°E. API RP-2A 10

j~..

.

-

shore environments from Alaska and California to the Gulf of Mexico and New England The PGA and PSV estimates obtained (for

....

-

illustration purposes) in Figs 7—8 are valid only for average North Sea hard ground conditions, and they are therefore base rock estimates where local soil conditions have not been taken into

-2

10

-

~.

t..

.•‘~

i~..

-

-

10

-

.

I Id’

I

...

~R~QLJENJC’~’

10’

10

consideration For offshore structures, local soil charactenstics can modulate the base rock ground motions to a considerable degree. Hence a full investigation of soil data and analysis of soil amplification are essential for engineenng seismic design cntena (Bungum et al, 1986a) In the site specific studies undertaken recently in the Norwegian sector of the North Sea, the investigation of dynamic soil effects has been carried out using both linear and non-linear methods.

(4-IZ)

Fig. 7. Computed response spectra for North Sea base rock conditions, and horizontal motion and 5% damping, normalized to a PGA value of 1 g. The values are computed using the PSV attenuation relationship in eq. 5, with values as illustrated m Fig. 5. levels The spectra to iO~ 1 probability (top tocorrespond bottom), and the and ~ 10 2 year linear API RP2A response spectrum for rock is shown for comparison (American Petroleum Institute, 1984).

6. Seismic hazard uncertainties

Probabilistic earthquake hazard estimates are in general subjected to two kinds of uncertainties, namely, inherent uncertainties, tied especially to the prediction of earthquake recurrence, and stat.

209

istical uncertainties, tied to the estimation of parameters in the earthquake hazard model (Kulkarni et al., 1984). The first kind of uncertainties cannot be reduced, while the second kind can be reduced as more data are accumulated. This can be illustrated by the a-value describing the variability in strong-motion attenuation, which is easily demonstrated to be one of the largest sources of uncertainty in most seismic hazard analyses (Bender, 1984). There are two main contributing factors to this variability, namely, intra-earthquake effects tied to variations in stress release along faults, and site effects, especially those tied to variations in local geology. The first of these cannot be reduced, while the second can be reduced through a proper site response analysis. It has been found (Joyner and Boore 1981~McCann and Boore, 1983) that each of these factors contribute to a variability of about 1.35 which corresponds to a a-value of 0.30. The total a-values used by us have been m the range 0.500.65. The input parameters for seismic hazard cornputations are in our approach considered to be variables with statistical distributions that can only be fully accounted for within the scope of Bayesian analysis. This means that alternative parameter values are described with associated likelihoods that are estimated partly on the basis of subjective scientific (expert) judgement. For each seismic source modelled, the parameters (such as activity rate, b-value, focal depth, maximum magnitude) are subject to statistical variations describable empirically by probability weights. Furthermore, the sigma parameter in attenuation is a global variable subject to variations as described above. This is a Logic Tree approach which also has the advantage of allowing the definition of dependencies among model parameters. When all the input parameters have been defined, with their range of variation, likelihoods and possible interdependence, there are essentially two ways of computing the associated hazard estimates. The first approach is to loop through all possible combinations of input variables (Kulkarni et al., 1984), and the second one, which is the one used by us, is to apply a Monte Carlo technique (Bernreuter et al., 1984). In both cases the results will be hazard estimates with associated probability density functions.

~.

t ~,

g

.

~

~

0

.. .~

i~ ~

0

tOO

120

I 440

160

I ISO

200

220

Acce Lerot ~.on (cm/ss) 2) Fig. (in %) vs. acceleration cm s the at a 9.10Excedent year probability probability level. TIus figure (in shows cumulative distribution of the data in Fig. 8, and the smooth line indicates a log-normal distribution.

Within the Monte Carlo approach the probability distribution of seismic hazard at a given excedent level is constructed out of the results of several hundred individual runs of the hazard program in which each input variable with an assigned discrete distribution has been sampled using a random number generator. The error introduced by this method is inversely proportional to the square root of the number of runs, and the stability of the results can be tested through a simple sensitivity analysis using running averages. In using this technique with the source model defined above and with input parameters defined as described above, we get a peak ground acceleration hazard distribution as shown in Fig. 8. The cumulative distribution shown in Fig. 9 is approximately log-normal, with a mean (expected) value slightly above the 50% level. The percentile level to choose in engineering applications is of course dependent upon acceptable risk as well as upon the level of conservatism used in deriving the estimates. In addition to this procedure, which gives a possibility of computing confidence intervals at

210

given probability levels, we have also in our sitespecific analyses regularly performed sensitivity tests by varying the input values for particular parameters while keeping the others fixed. The basis for this is a best-estimate input model, where each parameter is given a value that normally is the one with the highest likelihood in the Monte Carlo approach. The result is usually a hazard estimate very close to the mean value of the distnbution in Fig. 8. Acknowledgements This research has been supported by Norsk Hydro a.s., Oslo, Norway, and by Statoil a.s., Stavanger, Norway, through several contracts on earthquake hazard analysis. We thank E. Smith (Norsk Hydro) and O.T. Gudmestad (Statoil) for encouragement and support and R. Muir Wood (Principia), T. Løken and F. Nadim (both Norwegian Geotechnical Institute) for active participation in much of the work reported on in this paper. References Adams, R., 1985. The ISC catalogue of British earthquakes. Proc. Conf. Earthquakes and Earthquake Engineering in Britain, Univ. of East Anglia, April 1985. Thomas Telford Ltd., London. Ambraseys, N.N., 1985. The seismicity of western Scandinavia. J. Earthq. Eng. Struct. Dyn., 13: 361—399. American Petroleum Institute, 1984. API recommended practice for planning, designing, and constructing fixed offshore platforms. API RP2A, Fifteenth Edition, October 22, 1984. BMh, M., 1956. An earthquake catalogue for Fennoscandia for the years 1891—1950. Sver. Geol. Unders., Ser. C, No. 545, Arsbok 50. Bender, B., 1984. Incorporating acceleration variability into seismic hazard analysis. BulL Seismol. Soc. Am., 74: 1451—1462. Bernreuter, DL., Savy, J.B., Mensing, R.W. and Cherry, D.H., 1984. Seismic hazard characterization of the eastern United States: methodology and interim results for ten sites. Prepared for U.S. Nuclear Regulatory Commission, NUREG/ CR 3756. Bungum, H. and Fyen, J., 1980. Hypocenter distribution, focal mechanism, and tectonic implications of Fennoscandian earthquakes, 1954—1978. Geol. Fören. Stockholms Förhl., 101: 261—273. Bungum, H., Gudmestad, O.T., Løken, T., Muir Wood, R., Nadim, F., Swearingen, PH. and Woo, G., 1986a. Seismic hazard in the northern North Sea. Proceedings from the 8th European Conference on Earthquake Engineering, Lisbon, September 8—12, 1986, in press.

Bungum, H., Havskov, J., Hokland, B.Kr. and Newmark, R., 1986b. Contemporary seismicity of northwest Europe. Ann. Geophysicae, in press. Hainar, G.P., 1979. Tectonic map, North Sea. Statoil, Stavanger. Havskov, J. and Bungum, H., 1986. Source parameters for earthquakes in the northern North Sea, in preparation. 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. Bull. Seismol. Soc. Am., 71: 2011—2038. Kibsgaard, A., 1985. Focal Mechanism Studies of/by the Sv to P Amplitude Ratio. Thesis, Univ. of Oslo, 115 pp. Kulkarni, R.B., Youngs, R.R. and Coppersmith, K.J., 1984. Assessment of confidence intervals for results of seismic hazard analysis. Proc., Eighth World Conf. on Earthquake Engineering, Prentice-Hall Inc., NJ, Vol. I, pp. 263—270. McCann, Jr., MW. and Boore, D.M., 1983. Variability in ground motions: root mean square acceleration and peak acceleration for the 1971 San Fernando, California, earthquake. Bull. Seismol. Soc. Am., 73: 615—632. McGuire, R.K., 1976. FORTRAN computer programs for seismic risk analysis. U.S. Geol. Surv. Open File Report no. 76-67. McGuire, R.K., 1977. Seismic design spectra and mapping procedures using hazard analysis directly on oscillator response. Earthq. Eng. Struct. Dyn., 5: 211—234. Muir Wood, R., Woo, G. and Bungum, H., 1985. The history of earthquakes in the northern North Sea. In: W.H.K. Lee, H. Meyers and K. Shimazaki (Editors), Preliminary Proceedings from the IASPEI Symposium on Historical Seismograms and Earthquakes, Tokyo, August 27—28, 1985, pp. 463—472. NGI, NORSAR and Principia, H. Hansteen (Editor), 1985. Seismic Hazard Evaluation for the Sleipner Field. Report for Statoil, Stavanger, 172 pp. NORSAR, NGI and Principia, H. Bungum (Editor), 1984. Earthquake Loads for the Troll field. Report for Norsk Hydro a.s., Oslo, 250 pp. NORSAR, NGI and Principia, H. Bungum (Editor), 1985. Gullfaks C—Seismic hazard evaluation. Report for Statoil, Stavanger, 181 pp. Norwegian Petroleum Directorate, 1983. Forskrift for utforming, beregning og dimensjonering av baerende konstruksjoner til bruk for utvinning eller utnyttelse av petroleumsforekomster pa norsk kontinentalsokkel. NPD, Stavanger, draft 1983. Principia Mechanica Ltd., 1985. North Sea seismicity. Report prepared for U.K. Dept. of Energy, London, HMSO publication. Slunga, R., 1985. The seismicity of southern Sweden, 1979— 1984, Final Report. Swedish National Defence Research Institute Report No. C20572-T1, 45 pp. Shyam Sunder, S. and J.J. Connor, 1982. A new procedure for processing of strong-motion earthquake signals. Bull. Seismol. Soc. Am., 72: 643—661. Sykes, L.R., 1978. Intraplate seismicity, reactivization of preexisting zones of weakness, alkaline magmatism, and other tectonism postdaling continental fragmentation. Rev. Geophys. Space Phys., 16: 621—688.