This document introduces a method for evaluating seismic risk at an engineering project site. The method expresses risk in terms of the expected ground motion parameter (such as peak acceleration) versus average return period. It incorporates the influence of all potential earthquake sources and their estimated activity rates. The derived distributions of maximum annual ground motions take the form of Type I or Type II extreme value distributions if commonly assumed magnitude distributions and attenuation laws are used.
This document introduces a method for evaluating seismic risk at an engineering project site. The method expresses risk in terms of the expected ground motion parameter (such as peak acceleration) versus average return period. It incorporates the influence of all potential earthquake sources and their estimated activity rates. The derived distributions of maximum annual ground motions take the form of Type I or Type II extreme value distributions if commonly assumed magnitude distributions and attenuation laws are used.
Original Description:
This is the first paper talks about seismic risk analysis. Cornell is the first person.
This document introduces a method for evaluating seismic risk at an engineering project site. The method expresses risk in terms of the expected ground motion parameter (such as peak acceleration) versus average return period. It incorporates the influence of all potential earthquake sources and their estimated activity rates. The derived distributions of maximum annual ground motions take the form of Type I or Type II extreme value distributions if commonly assumed magnitude distributions and attenuation laws are used.
This document introduces a method for evaluating seismic risk at an engineering project site. The method expresses risk in terms of the expected ground motion parameter (such as peak acceleration) versus average return period. It incorporates the influence of all potential earthquake sources and their estimated activity rates. The derived distributions of maximum annual ground motions take the form of Type I or Type II extreme value distributions if commonly assumed magnitude distributions and attenuation laws are used.
Bulletin of the Seismological Society of America. V o l . 58, No. 5, pp. 1583- 1606.
Oct ober , 1968
E NGI NE E RI NG S E I S MI C RI S K ANALYSI S BY C. ALLIN CORNELL ABSTRACT This p a p e r i n t r o d u c e s a me t h o d f o r t h e e v a l u a t i o n o f t h e seismic ri sk a t t h e si t e o f a n e n g i n e e r i n g p r o j e c t . T h e results a r e i n t e r ms o f a g r o u n d mot i on p a r a m e t e r (such a s p e a k a c c e l e r a t i o n ) ver sus a v e r a g e r e t u r n p e r i o d . T h e me t h o d i n c o r p o r a t e s t h e i n f l u e n c e o f a l l p o t e n t i a l sour ces o f e a r t h q u a k e s a n d t h e a v e r a g e a c t i v i t y r a t e s a s s i g n e d t o t h e m. A r b i t r a r y g e o g r a p h i c a l r e l a t i o n s h i p s b e t w e e n t h e si t e a n d p o - t e n t i a l p o i n t , l i n e , o r a r e a l sources c a n b e m o d e l e d wi t h c o m p u t a t i o n a l e a s e . In t h e r a n g e o f i n t e r e s t , t h e d e r i v e d di s t r i but i ons o f ma x i mu m a n n u a l g r o u n d mot i ons a r e i n t h e f o r m o f T y p e I o r T y p e I I e x t r e m e v a l u e d i s t r i b u t i o n s , i f t h e mo r e c om- mo n l y assumed m a g n i t u d e d i s t r i b u t i o n a n d a t t e n u a t i o n l a w s a r e u s e d . I N T R O D U C T I O N O w i n g t o t h e u n c e r t a i n t y i n t h e n u m b e r , s i z e s , a n d l o c a t i o n s o f f u t u r e e a r t h q u a k e s i t i s a p p r o p r i a t e t h a t e n g i n e e r s e x p r e s s s e i s m i c r i s k , a s d e s i g n w i n d s o r f l o o d s a r e , i n t e r m s o f r e t u r n p e r i o d s ( B l u m e , 1 9 6 5 ; N e w m a r k , 1 9 6 7 ; B l u m e , N e w m a r k a n d C o r n i n g , 1 9 6 1 ; H o u s n e r , 1 9 5 2 ; M u t o , B a i l e y a n d M i t c h e l l , 1 9 6 3 ; G z o v s k y , 1 9 6 2 ) . T h e e n g i n e e r p r o f e s s i o n a l l y r e s p o n s i b l e f o r t h e a s e i s m i c d e s i g n o f a p r o j e c t m u s t m a k e a f u n d a m e n t a l t r a d e - o f f b e t w e e n c o s t l y h i g h e r r e s i s t a n c e s a n d h i g h e r r i s k s o f e c o n o m i c l o s s ( B l u m e , 1 9 6 5 ) . I t r e q u i r e s a s s e s s m e n t o f t h e v a r i o u s l e v e l s o f p e r f o r m - a n c e a n d e c o n o m i c i m p l i c a t i o n s o f p a r t i c u l a r d e s i g n s s u b i e c t e d t o v a r i o u s l e v e l s o f i n t e n s i t y o f g r o u n d m o t i o n . T h e e n g i n e e r m u s t c o n s i d e r t h e p e r f o r m a n c e o f t h e s y s t e m u n d e r m o d e r a t e a s w e l l a s l a r g e m o t i o n s . S o u n d d e s i g n o f t e n s u g g e s t s s o m e e c o n o m i c l o s s ( e . g . , a r c h i t e c t u r a l d a m a g e i n b u i l d i n g s , a u t o m a t i c s h u t - d o w n c o s t s i n n u c l e a r p o w e r p l a n t s ) u n d e r t h e s e m o d e r a t e , n o t u n e x p e c t e d e a r t h q u a k e e f f e c t s . T h i s e n g i n e e r s h o u l d h a v e a v a i l a b l e a l l t h e p e r t i n e n t d a t a a n d p r o f e s s i o n a l j u d g e - m e n t o f t h o s e t r a i n e d i n s e i s m o l o g y a n d g e o l o g y i n a f o r m m o s t s u i t a b l e f o r m a k i n g t h i s d e c i s i o n w i s e l y . T h i s i n f o r m a t i o n i s f a r m o r e u s e f u l l y a n d c o m p l e t e l y t r a n s m i t t e d t h r o u g h a p l o t o f , s a y , M o d i f i e d M e r c a l l i i n t e n s i t y v e r s u s a v e r a g e r e t u r n p e r i o d t h a n t h r o u g h s u c h i l l - d e f i n e d s i n g l e n u m b e r s a s t h e " p r o b a b l e m a x i m u m " o r t h e " m a x i m u m c r e d i b l e " i n t e n s i t y . E v e n w e l l - d e f i n e d s i n g l e n u m b e r s s u c h a s t h e " e x p e c t e d l i f e t i m e m a x i m u m " o r " 5 0 - y e a r " i n t e n s i t y a r e i n s u f f i c i e n t t o g i v e t h e e n g i n e e r a n u n d e r s t a n d i n g o f h o w q u i c k l y t h e r i s k d e c r e a s e s a s t h e g r o u n d m o t i o n i n t e n s i t y i n c r e a s e s . S u c h i n f o r - m a t i o n i s c r u c i a l t o w e l l - b a l a n c e d e n g i n e e r i n g d e s i g n s , w h e t h e r i t i s u s e d i n f o r m a l l y a n d i n t u i t i v e l y ( N e w m a r k , 1 9 6 7 ) , m o r e s y s t e m a t i c a l l y ( B l u m e , 1 9 6 5 ) , o r d i r e c t l y i n s t a t i s t i c a l l y - b a s e d o p t i m i z a t i o n s t u d i e s ( S a n d i , 1 9 6 6 ; B e n i a m i n , 1 9 6 7 ; B o r g m a n , 1 9 6 3 ) . U n f o r t u n a t e l y i t h a s n o t b e e n a s i m p l e m a t t e r f o r t h e s e i s m o l o g i s t t o a s s e s s a n d e x - p r e s s t h e r i s k a t a s i t e i n t h e s e t e r m s . I - I e m u s t s y n t h e s i z e h i s t o r i c a l d a t a , g e o l o g i c a l i n f o r m a t i o n , a n d o t h e r f a c t o r s i n t h i s a s s e s s m e n t . T h e l o c a t i o n s a n d a c t i v i t i e s o f p o - t e n t i a l s o u r c e s o f t e c t o n i c e a r t h q u a k e s m a y b e m a n y a n d d i f f e r e n t i n k i n d ; t h e y m a y n o t e v e n b e w e l l k n o w n . I n s o m e r e g i o n s , f o r e x a m p l e , i t i s n o t p o s s i b l e t o c o r r e l a t e p a s t a c t i v i t y w i t h k n o w n g e o l o g i c a l s t r u c t u r e . I n s u c h c i r c u m s t a n c e s t h e s e i s m o l o g i s t u n d e r s t a n d a b l y h a s b e e n l e d t o e x p r e s s h i s p r o f e s s i o n a l o p i n i o n i n t e r m s o f o n e o r t w o s i n g l e n u m b e r s , s e l d o m q u a n t i t a t i v e l y d e f i n e d . I t i s u n d o u b t e d l y d i f f i c u l t , i n t h i s s i t u - 1 5 8 3 1584 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA ation, for the seismologist to avoid engineering influences; the seismologist's estimates will probably be more conservative for more consequential projects. But these de- eisions are more appropriately those of the design engineer who has at hand more complete information (such as construction costs, system performance characteristics, etc.) upon which to determine the optimal balance of cost, performance, and risk. Seismologists have long recognized this need to provide engineers with their best estimates of the seismic risk. Numerous regional seismic zoning maps have been de- velopcd. Familiar examples appear in the Uniform Building Code (1967) and Richter (1959). Despite reference to probabilities they are seldom clear as to how the (single) intensity level for each location is to be interpreted. More recently these values have been associated with specific average return periods (~JIuto, Bailey and Mitchell, 1963; Kawasumi, 1951; Ipek et al, 1965). In any case, more information is needed to define a relationship between a continuous range of average return period and intensities. Other attempts have been made to provide this more complete in- formation at regional levels (Ipek, 1965; Milne and Davenport, 1965). These ap- proaches, which are usually large scale numerical studies based directly on historical data, have difficulty giving proper weight to the known correlation between geological structure and most seismic activity. They also are not successful at a fine or local scale. Lacer (1965) has presented a numerical, Monte Carlo technique designed to estimate the distribution of the intensity of motion at a particular site given the occur- rence of an earthquake somewhere in the surrounding region. He is able to account for geological features, such as faults, but he assumes all the assigned "point" sources are equal likely to give rise to this earthquake. In this paper a method is developed to produce for the engineer the desired rela- tionships between such ground-motion parameters as Modified Mercalli Intensity, peak-ground velocity, peak-ground acceleration, etc., and their average return period for his site. The minimum data needed are only the seismologist's best estimates of the average activity levels of the various potential sources of earthquakes (e.g., a particular fault' s average annual number of earthquakes in excess of some minimum magnitude of interest, say 4). If, in addition, the seismologist has reason to use other than average or typical values of the parameters in the function used to describe the relative frequency of earthquake magnitudes or in the functions of intensity, say, versus magnitude and distance, he may also supply these parameter values. The tech- nique to be developed provides the method for integrating the individual influences of potential earthquake sources, near and far, more active or less, into the probability distribution of maximum annual intensity (or peak-ground acceleration, etc.). The average return period follows directly. The results of the development appear in closed analytical form, requiring no lengthy computation and permitting direct observation of the sensitivity of the final results to the estimates made. Unlike the analogous flood or wind problem, in the determination of the distribu- tion of the maximum annual earthquake intensity at a site, one must consider not only the distribution of the size (magnitude) of an event, but also its uncertain dis- tanee from the site and the uncertain number of events in any time period. The presen- tation here will show the mathematical development of a simple ease. Results of other eases of interest will be displayed without complete derivations. An illustration will demonstrate the application of the method. Finally, the assumptions and limitations will be discussed more critically. Extensions and advantages of the method will con- elude the presentation. ENGINEERING SEISMIC RISK ANALYSIS 1585 LINE SOURCE DERIVATION For illustration of the development of the method of solution, the determination of the distribution of the annual maximum Modified Mercalli intensity at a site due to potential earthc uakes along a neighboring fault will be considered. As illustrated in C A a) Perspective Si t e ~/ 2 =- X - - ~ ~/2 Site b) ABD Plane FIG. 1. Line source. Figure la, the site is assumed to lie a perpendicular distance, A, from a line on the surface vertically above the fault at the focal depth, h, along which future earthquake loci are expected to lie. The length of this fault is l, and the site is located symmetri- cally with respect to this length. Concern ~i t h focal distances restricts attention to the ABD plane; Figure lb. The perpendicular slant distance to the source is d = %/ ~ --F A 2 (1) 1586 B UL L E T I N OF T HE S E I S MOL OGI C AL S OC I E T Y OF AME R I C A The focal distance, R, to any fut ure focus located a di st ance X from t he poi nt B is R = v / d ~ + X ~ (2) Since - l / 2 <= X <-_ 1/2, t he distance t o any ear t hquake focus is rest ri ct ed to 0 =< R < r0 i n which r0 = %/d 2 -t- 12/4. I n general t he size and location of a fut ure ear t hquake are uncert ai n. They shall be t r eat ed therefore as r andom variables. ( Random variables are denot ed by capi t al letters. ) We first seek t he condi t i onal di st ri but i on of t he Modified Mercalli I nt ensi t y, I , at t he site given t hat an ear t hquake occurs at a focal di st ance R = r from t he site. For i l l ust rat i on we us t he common assumpt i on (Ipek, 1965; Est eva and Rosenbl uet h, 1964; Wiggins, 1964; Kanai , 1961 ) t hat in t he range of i nt erest t he i nt ensi t y has t he following dependence on magni t ude, M, and focal distance, R: I = cl + c~M - c 3 1 n R (3) in which In denotes nat ur al l ogari t hm and c~, i = 1, 2, 3, are semiempirical const ant s on t he order of 8, 1.5, and 2.5, respect i vel y for fi rm ground in sout her n California ( Est eva and Rosenbl uet h, 1964). Given t hat an ear t hquake occurs at focal distance R = r, t he probabi l i t y t hat I , t he i nt ensi t y at t he site, is great er t han a ny number i is, using equat i on 3, P [ I >= i I R = r] = P[cl -Jr" c2M - c31n r => i l R = r] (4) in which P[ A I B] is read t he probabi l i t y of A given B. Assumi ng probabilistic inde- pendence of M and R, P [ I >= i , R = r] = P [ M >= i + c31n + c l l = I - - F M[ i + c31nrc2 + cl l (5) in which F M( m ) is t he cumul at i ve di st ri but i on funct i on of ear t hquake magni t udes. For example, Ri cht er' s wi del y verified (19, 20) rel at i onshi p bet ween number, n, ~, and magni t ude, m loglo nm = a - - bm implies 1 - - F ~( m) = e - ~(' ~- m) m => m0 (6) i n which ~ = b In 10 and mo is some magni t ude small enough, say 4, t hat event s of lesser magni t ude ma y be ignored by engineers. Thi s rest ri ct i on to larger event s implies t ha t t he probabilities above are condi t i onal on t he occurrence of an event of i nt erest , t ha t is, one where M > m0. The par amet er b is t ypi cal l y (Isacks and Oliver, 1964) such t hat f~ is about 1.5 t o 2.3. E NGI NE E R I NG S E I S MI C : RI SK ANAL YS I S Combi ni ng equat i ons 5 and 6, t he resul t is 1587 The l i mi t on t he definition of FM( m) , namel y m ~ m0, implies t hat equat i on 7 holds for i + c~ lit r + 0 -~ L = i , ' 0 C2 o i l i ~ c2mo- - 0 -- cs hl ~. f R ( r ) ~ d r o r FIG. 2. Probability density function of focal distance, R. At smaller val ues of t he ar gument , i, t he pr obabi l i t y ( equat i on 7) is uni t y t hat I exceeds i (gi ven t he occurrence of an event of magni t ude gr eat er t ha n mo at di st ance r ) . I n or der t o consider t he influence of all possible val ues of t he focal di st ance and t hei r r el at i ve likelihoods, we must i nt egr at e. We seek t he cumul at i ve di st r i but i on of I , F~(i ), gi ven an occurrence of M >= m0, f d r O 1 -- F, ( i ) = P[ I >= i] = P[ I >= i l R = r ] f ~( r ) dr ( 9 ) in whi ch f R(r) is t he pr obabi l i t y densi t y f unct i on of R, t he uncer t ai n focal di st ance. For t he i l l ust rat i on here, i t is assumed t hat , gi ven an occurrence of an event of i nt er est al ong t he fault, i t is equal l y l i kel y t o occur anywher e along t he faul t . For mal l y, t he l ocat i on vari abl e X is assumed t o be uni f or ml y di st r i but ed on t he i nt er val ( - 1 / 2 , +1/ 2) . Thus I X I, t he absol ut e magni t ude of X, is mf i f or ml y di st r i but ed on t he i nt er val (0, 1/2). The cumul at i ve pr obabi l i t y di st ri but i on, FR(r), of R follows 1588 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA i mmedi at el y: F R ( r ) = P [ R ~ r] = P [ R 2 ~ r 2] = P [ X 2 -4- d 2 ~ r e] = P[IxI ~ ~ 7 - d 21 2 d ~ r - - 1 / 2 d- < r - < r0. (10) Therefore, t he probabi l i t y densi t y funct i on of R i s d F . ( r ) d ( 2 ~ / ~ ) f R ( r ) - d r - d r 2 r ' = d _-< r _-< r0. (11) - ~ / ~ - d2 Thi s densi t y funct i on is pl ot t ed i n Fi gure 2. Subst i t ut i ng equat i on 11 i nt o equat i on 9 and i nt egrat i ng is complicated by t he awkward limits of definition of t he functions, but in t he region of great est interest, namel y larger values of t he i nt ensi t y t he resul t is 1 - - F i ( i ) = P [ >= i ] [ = ll C G e x p - ~ i ~ i ' (12) i n which i ' is the lower l i mi t of val i di t y of t hi s form of t he result and equals i t = c l -4- c2mo Ca In d (13) and in which C and G are const ant s. The first const ant is rel at ed to paramet ers in t he vari ous relationships used above: [ ( C l ) ] C = exp ~ ~-4- mo . (14) The second const ant is rel at ed to t he geomet r y of i l l ust rat i on: f ro d r G = 2 r ~ / ~ _ d2 "~1 se-l[rO/d] (COS U) 5'-1 d u ( 15) 2 d'Y ,-1o in which = ~ c_~_ 1. ( 1 6 ) C2 E N G I N E E R I N G S E I S M I C R I S K A N A L Y S I S 1 5 8 9 The integral in equation 15 must be evaluated numerically. Results appear in Figure 3. For typical paramet er values and sufficiently long faults it is conservative and reason- able to replace r0 by infinity. I n this case G is given by 2~ r(~) G- in which r (V) is the complete gamma function and v is restricted to positive values. The results above yield t he probability t hat the site intensity, I, will exceed a 5 - \ \ 2 1.0. 0. 5 ro/d =CO / s e c - I r o / d Q = / (cos u)Y-Jdu 0 \ - - . < - ~ " - ~ 1.10 ro/d = 1.05 I I I I I -I 0 +1 +2 +8 FIG. 3. Numerical values of integral in equation (15). Y 1590 BULLETI N OF THE SEI SMOLOGI CAL SOCI ETY OF AMERI CA certain value, i, given t hat an event of interest ( M _-_ m0) occurs somewhere along the fault. Next we must consider the question of the random number of occurrences in any time period. For illustration, it is assumed t hat the occurrences of these maj or events follow a Poisson arrival process (Parzen, 1962; Cornell, 1964) with average occurrence rate (along the entire fault) of v per year. Then, N, the number of events of interest along the fault in a time interval of length t years is known to be Poisson distributed p~( n) = P [ N = n] = e- ~e( ~t) " n = 0 , 1 , 2 , . . . . ( 1 8 ) n! I t is easily established that, if certain events are Poisson arrivals with average ar- rival rate v and if each of these events is independently, with probability p, a "special event , " then these special events are Poisson arrivals with average rate pv. (This is said to be a Poisson process with ("random selection.") I n our case the special events are those which cause an intensity at the site in excess of some value i. The probability, p ~ , t hat any event of interest ( M _-> m0) will be a special event is given byequat i on 12. p~ = P [ I >= i] = ~ C G e x p [ - ~ i ] . (19) Thus the number of times N t hat the intensity at the site will exceed i in an interval of length t is p ~ ( n ) P I N n] e - ' ~ t ( P ~ vt)'~ = = - n = 0, 1, 2, . - . . (20) n! Such probabilities are useful in studying losses due to a succession of moderate inten- sities or cumulative damage due to two or more major ground motions. Of particular interest is the probability distribution of I(n[~)x the maxi mum intensity over an interval of time t (often one year). Observe t hat ~( t ) P [ I .... =< i] = P[exactly zero special events in excess of i occur in the time interval 0 to t] which from equation (20) is v(t) - " (21) P [ . . . . <= i] = P [ N = 0] = e ~t . If we let I .... (1) equal I ...... the a n n u a l maximum intensity, t = 1, and Fx~(~)~ = e -vi~ = exp - ~CGe xp - ~ i i > i ' in which now the ratio ~ = v / 1 appears. This ratio is the average number of occur- rences per unit length per year. The conclusion is t hat for the larger intensities of engineering interest, the annual maxi mum intensity has a distribution of the double exponential or Gumbel type. This distribution is widely used in engineering studies of extreme events. I t is i mport ant to ENGI NEERI NG SEI SMI C RI SK ANALYSI S 1591 realize that, here, this conclusion is n o t based on the intuitive appeal to the familiar asymptotic extreme value argument (Gumbel, 1958), which has caused other investi- gators to seek and find empirical verification of the distribution for maximum magni- tudes or intensities in a given region (Milne and Davenport, 1965; Nordquist, 1945; Dick, 1965). The form of the distribution is dependen~ on the functional form of the various relationships assumed above. Others, too, have found (Dick, 1965; Epstein and Lomnitz, 1966; Epstein and Brooks, 1948) that the combination of Poisson oc- currences of events and exponentially distributed "sizes" of events will invariably lead to the conclusion that the largest event has a GumbeLlike distribution (the true Gumbel distribution is non-zero for negative as well as positive values of the argu- ment). Any combination of assumptions which leads to the exponential form of the distribution of I will, in combination with Poisson assumption of event occurrences, yield this Gumbel distribution. The exponential form of F i ( i ) does not require the exponential form of F ~ ( m ) . If the logarithmic dependence of I on R (equation 3 ) is retained, for example, even polynomial distributions (Housner, 1952) of magnitude will lead to the exponential distribution of I. If the annual probabilities of exceedance are small enough (say _-<0.05), the dis- tribution of I ..... can be approximated by 1 - - F ~ ( ~ ) ~ = 1 - e - p ~ ~___~'~ 1 - ( 1 - p i ~ ) pl y ~ ) i ' ~ - - ~ C G e x p - ~i i ~ . (23) The average return period, T~, of an intensity equal to or greater than i is defined as the reciprocal of 1 - F~[:) or T i __-- ~ exp i i >= i' (24) or, the " T - y e a r " intensity is i ~ C2 .t _--- ~ In ( ~CGT~) i >-_ ~. (25) Consider the following typical numerical values of the parameters and site constants, applicable to a particular site in Turkey, where in one region in 1953 years it was found (Ipek et a l , 1965) that log10 n m = a - - b m = 5.51 -- 0.644m in which n~ is the number of earthquakes greater than m in magnitude. Assuming these earthquakes all occur along the 650 km of the maior fault system in the region, the average number of earthquakes in excess of magnitude 5 (i.e., m0 --- 5) per year per 1592 B UL L E T I N OF T HE S E I S MOL OGI C AL S OC I E T Y OF AME R I C A uni t l engt h of faul t is n 5 - ( 1953) ( 650) - Also 1.5 X 10 -4 (year) -1 (ki l omet er) -1. = b In 10 = 0.644(2.30) = 1.48. c2 = 1.45 c3 -- 2.46 t he following numeri cal results are obt ai ned for a site located a mi ni mum surface distance, A, of 40 km from a line source of eart hquakes at dept h h = 20 km: d = %/'h 2 + A s = 44. 6km /~ C3 -- - -- 1 = 1.52 C2 = exp ? + 2~ r(~) G----~ (2d)~ [ r (3' _+2 1 ) ] 2 = _ _ 7"04 X 10-~" (Numeri cal i nt egrat i on gives G = 6.58 X 10-8). Thus, t he i nt ensi t y at this site with r et ur n period T~ is 2 i ~ ~- In (~CGT~) 0.98 In (6.9T~). Not e t he l ogari t hmi c relationship bet ween i and Ti . The risk t hat a design i nt ensi t y will be exceeded can be hal ved ( T doubl ed) by increasing t he design i nt ensi t y by about 0.7. Thi s equat i on is pl ot t ed in Fi gure 4 for t he range of val i di t y i > i ' where .t = c l + c 2 mo - c3 In d = 6.08. I f i nt erest extends t o smaller intensities, i t necessitates more cumbersome i nt egrat i ons not shown here. PEAK GROUND MOTION RESULTS The previous section developed t he desired di st ri but i on results for t he Modified Mercal l i i nt ensi t y, I , and a uni form line source, wi t h a part i cul ar set of assumptions on magni t ude di st ri but i on and t he i nt ensi t y versus M and R relationship. Engineers are generally more di rect l y concerned wi t h such ground mot i on paramet ers as peak- cl = 8.16 Using at t enuat i on const ant s found empi ri cal l y ( Est eva and Rosenbl uet h, 1964) for California ground acceleration, A , peak-ground velocity, V, or peak-ground displacement, D, t han with intensity itself. An argument parallel to t hat in the preceding section can be carried out wi t h any functional relationship bet ween the site ground-motion variable, Y, and M and R. For example, the particular form Y = b l e b ~ R - b 3 (26) I 0 has been recommended by Kanai (1961) and by Est eva and Rosenbl uet h (1964)* for peak-ground acceleration ( Y = A ), peak-ground velocity ( Y = V), and peak-ground displacement (Y = D) . The l at t er authors ( Est eva and Rosenblueth, 1964; Est eva, o 8 7 " o 0) 0 6 I I / / I / / / ~ I I / I / I I E NGI NE E R I NG S E I S MI C R I S K ANAL YS I S 1593 I I I I I I I 00 2 0 0 500 I 000 2 0 0 0 5000 Ti , y e a r s 0 . 0 1 0 . 0 0 1 I - F I m Q x ( i ) FIG. 4. Numerical example: Intensity Versus return period. 1967 ) (on theoretical and empirical grounds) suggest t hat the constants { bl, b2, b3 } be {2000, 0.8, 2}, {16, 1.0, 1.7}, and {7, 1.2, 1.6} f or A, V, a ndD respectively in southern California, with A, V, and D in units of centimeters and seconds and R in kilometers. For the general relationship in equation 26, an argument like t hat in the previous section yields for the annual maximum value of Y from a uniform line source Fr~.a) = exp [--~CGy -~lb2] y >= y ' 1 - F r ( . )~ ~ ~ CGy -~/b~ y >= y ' 1 y~ib2 T y ~ nCG (27) (28) *More recently, Esteva (1967), it has been suggested that the focal depth, h, in kilometers, be replaced by an empirically adjusted value, ~/ ~ -4- 203, which increases the formula's accuracy shorter focal distances. (29) 1594 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA in whi ch C = e ~ m b l ~/b2 (30) and G is as gi ven in equat i on 15 ( or equat i on 17) wi t h 53 v = ~ - -- 1, (31) O2 The lower l i mi t of t he val i di t y of t hese forms of Fr(g2x is y ' = b l e b~mo d - b 3 . (32) For dur at i ons, t, ot her t ha n one year , ~ shoul d be repl aced by ~t in equat i ons 27 and 28. Not i ce t ha t equat i on 27 is of t he general f or m of t he Type I I asympt ot i c ext r eme val ue di st r i but i on of l argest val ues ( Gumbel , 1958). Thi s di st ri but i on, too, is com- monl y used in t he descri pt i on of nat ur al loadings on engi neeri ng st ruct ures, t he most fami l i ar bei ng maxi mum annual wi nd velocities ( Ta s k Commi ssi on on Wi nd Forces, h Site Point Source FIG. 5. Point source, cross section. 1961; Thom, 1967). The j ust i fi cat i on t her e is based on asympt ot i c (l arge N) argu- ment s while t ha t here is not . The resul t s here are a consequence of t he forms of t he rel at i onshi ps assumed. Using resul t s such as t hese t he desi gner can comput e for his site t he peak- gr ound vel oci t y, v, and peak- gr ound accel erat i on, a, associ at ed wi t h t he same, say t he 200-year, r et ur n period. For t he numer i cal exampl e in t he previ ous sect i on and t he val ues of t he par amet er s referred t o in this section, t hese val ues are appr oxi mat el y v = 7.5 cm/ sec = 3 i n/ sec a = 80 cm/ sec 2 = 0.08g. GENERAL SOURCE RESULTS I n or der t o faci l i t at e represent i ng t he geomet r y and pot ent i al source condi t i ons at ar bi t r ar y sites, i t is desi rabl e t o have addi t i onal resul t s for poi nt and area sources. I t will be shown t hat t hese resul t s can be used t o r epr esent qui t e general conditions. I f a pot ent i al source of ear t hquakes is closely concent r at ed in space rel at i ve t o i t s di st ance, d, f r om t he si t e, i t sat i sfact ori l y ma y be assumed t o be a poi nt source, Fi gur e 5. ( Exampl es mi ght be sites one or t wo hundr ed ki l omet ers f r om New Madr i d, Mo. or Charl est on, S. C. ) I n t hi s case t her e is no uncer t ai nt y in t he focal di st ance, d, and t he E NGI NE E R I NG S E I S MI C R I S K ANAL YS I S 1595 pr evi ous resul t s (e.g., equat i ons 22, 25, 27, 29) hol d wi t h ~ equal t o t he average number of ear t hquakes of i nt erest ( M ->_ m0) per year ori gi nat i ng at this poi nt and wi t h a geomet r y t er m (i n place of equat i on 15) equal t o G = d -(~+1) ( 33) For i nt ensi t i es 7 is gi ven by equat i on 16 and for vari abl es wi t h rel at i onshi ps of t he t ype shown i n equat i on 26, 7 is gi ven by equat i on 31. For a poi nt source, for val ues of t he ar gument less t ha n i ' or y ' , t he cumul at i ve di st r i but i on f unct i on ( equat i on 22 or 27 ) is si mpl y zero. I n some si t uat i ons, owing t o an appar ent l ack of correl at i on bet ween geologic st ruc- t ur e and seismic act i vi t y or owing t o an i nabi l i t y t o observe this st r uct ur e due t o deep over bur dens, i t ma y be necessary for engi neeri ng purposes t o t r eat an area sur r oundi ng t he site as if ear t hquakes were equal l y likely t o occur anywher e over t he area. I t can F I G, 6. Annular sources, perspective. be shown t ha t for an annul ar areal source sur r oundi ng t he site, as pi ct ur ed in Fi gur e 6, t he di st ri but i ons above ( equat i ons 22 and 27 ) hol d wi t h a geomet r y t er m equal t o G - ( ~ : ~) d~_ 1 wi t h 7 gi ven by equat i on 16 or 31. The val ue of p shoul d now be t he aver age number of ear t hquakes of i nt erest ( M >= m0) per year per uni t area. I n t er ms of ~, t he average number per year over t he ent i re annul ar region, p is P = (35) ~ ( l 2 _ ~ 2 ) " For val ues of t he ar gument less t han i ' or y ' , t he cumul at i ve di st ri but i on f unct i on ( equat i on 22 or 27) is zero. Not e t ha t d will never be less t ha n h. Thus t he geomet r y f act or remai ns finite even when t he site is " i mmer s ed" i n t he areal source, i.e., when A = 0, and nil ear t hquake di rect l y below t he site is an ( i mpr obabl e) possibility. 1596 B U L L E T I N O F T H E S E I S MO L O G I C A L S O C I E T Y O F A ME R I C A When more compl ex source confi gurat i ons exist, t he di st r i but i on f unct i on for t he maxi mum val ue of some gr ound mot i on var i abl e can be f ound by combi ni ng t he re- sul t s above. For exampl e, if t her e exist i ndependent sources (1, 2, n) of t he vari ous t ypes discussed above, t he pr obabi l i t y t hat t he maxi mum val ue of Y, t he peak- gr ound accel erat i on, for exampl e, is less t ha n y is t he pr obabi l i t y t ha t t he maxi mum val ues f r om sources 1 t hr ough n are all less t ha n y, or ~I I l UX IIIUX. l l l a X 2 ITt ~Xn = lYI Fr(~> j = l m a x j in whi ch Fr(~) is t he di st ri but i on of t he maxi mum Y ( say peak accel erat i on) f r om m a , x j source j , as gi ven by equat i on 27 wi t h t he appr opr i at e val ues of t he par amet er s pj., C j , G~.. Not e t ha t t he di fferent possible focal dept hs on t he same f aul t can be account ed for in this manner . For t he exponent i al f or m of t he ,. ~r~ (s) funct i ons ( equat i on 27) I _Sj/b22 7 Fr(m~2~ = exp - - ~ ~jCjG~y j y > y ' (36) ! ! where y' is t he l argest of t he Yi For y less t han y , t he di st r i but i on can be f ound wi t h ease (unless a line source is i nvol ved). I f t he const ant s/ ~, bl , b~, ba are t he same for all t he sources in t he regi on ar ound t he site, equat i on 36 becomes si mpl y in whi ch Fr(~) = exp [ - C~ Gy -~/b2] y > y ' max (37) pG : ~ pjGj (38) A si mi l ar conclusion hol ds for Modi fi ed Mercal l i intensities, equat i on 22. I n shor t t he di st ri but i ons r et ai n t he same forms wi t h t he pr oduct , ~G, equal t o t he s um of t he correspondi ng pr oduct s over t he vari ous sources. Wi t h respect t o t hese pr oduct s, t hen, l i near superposi t i on applies. Thi s concl usi on is a refl ect i on of t he fact t ha t t he s um of i ndependent Poi sson process is a Poi sson process wi t h an average ar- ri val r at e equal t o t he sum of i ndi vi dual rat es. Thi s conclusion can be used t o det er mi ne geomet r y fact ors for unsymmet r i cal source geomet ri es. For exampl e, for t he condi t i on in Fi gur e 7a, t he geomet r y fact or, G, must equal one-hal f of t ha t for t he symmet r i cal si t uat i on. The geomet r y f act or for t he situa- t i on in Fi gur e 7b must equal one-hal f of t ha t for a symmet r i cal source l engt h 2b minus one-hal f of t ha t for a symmet r i cal source of l engt h 2a, or G -- [ G' - G"] (39) i n whi ch G' and G" are cal cul at ed f r om equat i on 15 wi t h val ues ro t and r0" respect i vel y. An exampl e will follow. Thi s resul t also per mi t s easy t r eat ment of a f aul t wi t h a (spa- t i al l y) non- const ant aver age occurrence rat e, each di fferent por t i on of t he f aul t being t r eat ed i ndependent l y. %, d S i t e r o a ) Case I / J I - - b ~. , / I I . , i d I r o r o I Si t e b ) C a s e 2 Si t e ~Of h c ) Case 5 , P e r s p e c t i v e F I e . 7. Un s y mme t r i c a l s o u r c e s . 1597 1 5 9 8 BULLETI N OF THE SEISMOLOGICAL SOCI ETY OF AMERI CA i n the same manner the geometry factor for an area such as that shown in Figure 7c is found to be _ Ol G. ~G~ W E FI(~. 8. Numerical examples: plan in which G~ is the result for the complete annulus, equation 34. As will be shown in a numerical example, an areal source of arbitrary shape can be modeled with ease by approximating it by a number of such shapes. Note that the approximation to equation 37 for smaller values of the probability 1 - - F y(v) becomes ma : ~ 1 - - F~(I~)~,: _~ Cy --~/b~ ~ f,~Gi (41) ENGI NEERI NG SEISMIC RI SK ANALYSIS 1599 suggesting t hat the (small) probability t hat the anrmal maximum, Y . . . . exceeds y, in any year is made up of the sum of the probabilities coat ri but ed by each of the sources. Also, for larger values the ret urn period is approximately 1 y~/b2 ( 4 2 ) T~ ~-- CX~jG~ " TABLE 1, PART i NUMEI~ICAL EXAMPL'm. Source d ro G~ A V D Li ne 1 Ri ght por t i on 104 115.3 5. 12 X 10 -7 1. 73 X 10 -4 3. 66 X 10 -a Le f t por t i on 104 241 1. 06 X 10 -6 3. 17 X 10 -~ 5. 99 X 10- 9 Li ne 2 Tot a l 49 206 3. 18 X 10 -5 1, 98 X 10"- 3 1. 55 X 10 -~ Por t i on al ( - ) 49 57. 5 - 7. 28 >( 10-0 - - 0. 94 X 10- 3 - 1. 082 X 10- a Areal Annul us 1, a = 2~ 28. 3 45 24. 9 X 10 -4 2. 44 >( 10 - i 1. 76 2, ~ = 4.38 45 75. 5 7. 0 X 10- 4 1. 14 X 10 -1 1. 28 3, ~x = 3.78 75. 5 123. 5 2. 11 X 10 -4 0. 65 X 10 -1 0. 98 4, ~ = 3.44 123. 5 252 0. 80 X 10 -4 0. 61 X 10 - i 1. 16 5, a --- ~ 252 ~o 0. 23 X 10 -4 0. 60 >( 10 -1 10. 49 "Poi nt " 216 4. 7 X 10 -19 4. 4 X 10- 7 1. 1 X 10 -6 As s umpt i ons : h = X/ 20 Y + 202 = 28.3 km; ~ = 1.6; m0 = 4 Pe a k a c c e l e r a t i on: bl = 2000; b~ = 0. 8; b~ = 2 ; C - 2. 4) < 109 Pe a k ve l oc i t y: bl = 16 ; b2 = 1. 0; b~ = 1. 7; C = 4.98 X 104 Pe a k di s pl a c e me nt : bl = 7 ; b~ = 1.2; b3 = 1. 6; C = 7.8 X 10 s TABLE 1, P A R T 2 NUMERICAL EXAMPLE ~iGi Source A V D Li ne 1 Ri ght por t i on 5. 12 >( 10 - l i 1. 73 X 10- g 3. 66 )< 10 -7 Lef t por t i on 10. 6 )< 10 - l l 3. 17 X 10 -8 5. 99 X 10 -7 Li ne 2 Tot a l 318 X 10 - i ' 19. 8 )< 10 -8 15. 5 X 10 -7 Por t i on ai ( - ) - 72. 8 X 10- n - 9. 4 X 10 -8 - - 10. 82 X 10- 7 Areal Annul us 1, a = 2~ 249 >( 10 - n 24. 4 X 10- s 17. 6 X 10 -7 2, a = 4.38 70 >( 10 -11 11. 4 X 10- 8 12. 8 )< 10- 7 3, a = 3.78 21. 1 X 10 -11 6. 5 X 10- 8 9. 8 X 10- 7 4, a = 3.44 8. 0 X 10 -11 6. 1 X 10 - s 11. 6 X 10 -7 5, a = 7r 2. 3 X 10 -11 6. 0 X 10 -.8 104. 9 X 10 . 7 "Poi nt " 4. 3 X 10- ii 4. 1 X 10 -8 9. 8 X 10 -7 Sum 616 )< 10 - l l 73. 7 X 10 -8 182 )< 10 -7 ' I 600 BULLETIN OF TH:E S~ISMOLOGICAL' SOCIETY OF' AMERICA ',, :i : ~ ' : NUMERICAL EXAMPLE "~ For:illustration We treat th~ hypothetical situation shown in Figure 8: The site is located on a deep alluvial plane (shaded) such t hat the geological structure below the site and to the south and east is not known in detail. Historically, earthquakes have occurred throughout this plane, but not often enough to determine fault patterns. The engineer chooses to treat the region as if t he next ' earthquake were equally likely to occur in any unit area. The average rate, P = 1.0 X 10 -~ per km 2, was estimated by dividing the region's total number of earthquakes (with magnitudes in excess of 4) by its total area. The exception, historically, is a Small area, some 200 km southeast. It is also below the alluvial plane. The frequency of all sizes of earthquakes there has been relatively high, including several of larger magnitudes. Although the engineer can easily account for any suspected local difference in the parameter ~ (smaller values imply higher relative frequencies of larger magnitudes), he chooses to use the same ~ value, 1.6, for th e entire region. In other words, he chooses to at t ri but e the Small area' s observed larger magnitudes to the same population f , ( m) . The justification is t hat the larger the :average arrival rate , the larger is the number of observations and the more likely it is t hat larger magnitudes will be inclL~ded among the observations of a given period of time. Exactly what area (here shown as 30 by 30 kin) is used to estimate the areal occurrence rate, ~ = 1.0 10 -4 per km 2, is not critical in this cuse since the area is small enough and far enough from t he site t hat the entire source will be treated as a point with rate ~ = 0:09. Finally, to the northwest where the geological structure is exposed two faults have been located. Neither can be assumed inactive. Past activity on the first (and other geologically similar faults) suggests an average occurrence rate of P = 1.0 X 10 -4 per kin. No earthquakes on the second, closer fault have been recorded, but its geo- logical similarity to the first suggests t hat it be given a similar activity level. The sectors of annuli used to represent the areal region are shown in Figure 8. The geometry factors, G~, for the various sources are shown in Table 1 along with the products ~G~, and their sums for peak-ground acceleration, A, peak-ground velocity V, and peak-ground displacement, D. The conclusion is t hat the maximum ground ac- celeration, velocity, and displacement during an interval of t years have distributions F~(?~)~ = exp [ - 14.7ta -2] Fv~(~2x = exp [-0.0367tv -1'6] F~,~i ~ = exp [-0.142td-1'33]. The annual maxima have the approximate distributions 1 -- FA[~2x ~ 14.7a -2 1 -- F. ~2 X ~ 0.0367v -16 1 -- F.(~) __--~ 0.142d -1'33. ma x In terms of return periods T~ ~ 0.0681a 2, Tv --~ 27.3v 1"6, T~ _~ 7.05d 1"~3, or a ~ 3.83Ta '5, v --~ 0.126Tv '62~, d _--~ 0.231Td '75. ENGI NEERI NG SEI SMI C RI SK ANALYSI S 1601 If a design response spectrum based on a 200-year return period were desired, it should be based on design ground-motion values of a200 = 55 cm/sec 2= 0.054g, v200 = 3.5 cm/sec, d2o0 = 12.5 cm. Using the method for constructing a spectrum sug- gested for design by Newmark (1967), the dynamic response spectrum in Figure 9 is obtained. Z ( / ) E 0 ~10 t - O ( / ) 0 n ~ 5.C 0 0 Q) > i o 2. 0 "0 0 u I.O a . 0.1 0 . 2 0 . 5 1.0 2 . 0 5. 0 I 0 Nat ur al P e r i 0 d , sec Design Response Spectrum; 2 % of Critical Damping FIG. 9. Dynami c response spectrum. Notice that, for a proportional increase in all ground-motion factors, the risk ( as measured by 1 - F or 1/T) decreases more rapidly for peak acceleration than for peak velocity, and more rapidly for velocity than for displacement. The implication is t hat shorter-period structures can achieve greater risk reductions for the same per- centage increase in design level than longer-period structures. 1 6 0 2 B U L L E T I N O F T H E S E I S M O L O G I C A L S O C I E T Y O F A M E R I C A Inspection of the individual "contributions" to Ev~G~ (or, approximately, to the risks 1 - F) in Table 1 reveals t hat the closer faults have the predominant influence on peak acceleration risks, since even relatively small, more frequent magnitudes can give rise to high accelerations locally.* More distant potential sources contribute sig- nificantly to the risk of longer-period structures, as evidenced by their contributions to the E,~G~ factor for velocity and displacement. The slow decay with distance of peak displacement causes a large contribution even from long distances. This explains why the displacement of 12.5 cm is considerably larger than those associated with specific earthquakes records with peak accelerations of the order of 0.05g. ASSUMPTIONS AND EXTENSIONS While the assumptions made in the method are considered reasonable for most purposes of engineering design, a number of them can be relaxed without significantly altering the basic method. In particular, the distribution of magnitudes and the rela- tionships used to relate site ground-motion characteristics to the magnitude and focal distance can be replaced with ease, only the results of certain integrations will change. In the derivations above the distribution of magnitudes has been assumed to be the unlimited exponential distribution. For the larger, rarer magnitudes there are in- sufficient data to substantiate with confidence this or any other assumption (Rosen- blueth, 1964). The shape as well as the parameters may in fact vary among different regions for these larger values. The magnitudes of earthquakes may be bounded. Relatively clean analytical results can be obtained for distribution functions of poly- nomial form and for the limited exponential distribution. Their influence, which may be significant for larger return periods, are under investigation. For different focal distance relationships, the existing results can be used with piece- wise fits to the other functions. For example, if it is assumed that there is no attenua- tion of Y with distance for a certain distance, r', from a source (Housner, 1965; Ipek e t a l , 1965) an annular source canbebroken into two regions, one ford -< r -< r' and the other for r' -- r -- r0. In the first region b3 should be set equal to zero, and the values of bl and b2 appropriate for near-source conditions adopted. Coupled with a limited magni- tude distribution, this process facilitates incorporation of any suspected upper bounds on maximum ground motions (Housner, 1965). These functions are, in any case, no better than the parameter estimates used in them. One primary advantage of an analytical method, as opposed to a numerical one (Ipek et a l , 1965; Lacer, 1965) is that the sensitivity ot final conclusions to the ac- curacy of these parameter estimates can be assessed. Other of the more basic assumptions in the method can also be relaxed with relative ease. Specifically, these include the two assumptions (a) that the radiation of effects can be treated as if the earthquake generating mechanism were concentrated at a point and (b) that isoseismals are circular. These assumptions are commonly made in design studies. This is done not so much because it is thought to be true, but because alternative methods and information are seldom available. The vast majority of sta- tistical data on attenuation and scaling laws, for example, are available in forms (averages, etc. ) based on these two assumptions. At the expense of added mathematical complexity alternative assumptions (e.g., finite mechanism length, elliptical isoseis- mals) can be incorporated into the method above if sufficient data are available to justify their inclusion. *Also, of course, the durations are correspondingly short, a factor not explicitly appearing in the method for construction of response spectra proposed by Newmark. ENGINEERING SEISM IC R I S K ANALYSIS 1603 The more fundamental assumptions are those of (a) equal likelihood of occurrence along a line or over an areal source, (b) constant-in-time average occurrence rate of earthquakes, and (c) Poisson (or "memory-less") behavior of occurrences. If data or judgement rule against the equM-likelihood assumption and in favor of other relative values they can be included by simply treating each portion of the source over which the equally likely assumption is reasonable as an individual source using the superposition method described above. If the engineer and the seismologist are prepared to make an assumption about the time dependence of the average occurrence rate, other than t hat of constant in time, a minor modification in the method suffices to account for this non-homogeneity in time (Parzen, 1962; Cornell, 1964). The influence appears, for example, in Equation 21 as I f0 t 1 PL[I u)m~ = < i] = exp -- p~ ~(r) dr (59) in which v ( r ) is the average occurrence rate at time T. The assumption that the occurrences of earthquakes follow the behavior of the Poisson process model can be removed only at a grea~er penalty, however. The Poisson assumption does not reflect earthquake swarms or aftershocks, nor is it physically consistent with the elastic rebound theory, which implies that a zone of recent past activity is less likely to be the source of the next earthquake than a previously active zone which has been relatively quiet for some time. These limitations can, in principle, be removed by adopting more general renewal process or Markov process models (Aki, 1956; Vere-Jones, 1966). For engineering purposes the Poisson results are con- sidered adequate for numerous reasons (Rosenblueth, 1966; Lomnitz, 1966). When swarms and aftershocks are excluded, data does not clearly reject the Poisson assump- tion (Lomnitz, 1966; Wanner, 1937; Knopoff, 1964; Niazi, 1964)for the rarer, major events of engineering interest. Even when more accurate theoretical models become available, it is not evident that sufficient statistical data and other information will be available in many regions to permit the seismologist to adopt a non-Poisson assumption or to estimate any more parameters than the average occurrence rates. The structural engineer is concerned more directly with a design response spectrum. For random forcing functions such as earthquake ground motions, the duration of motion also influences the pe~k-response values (Rosenblueth, 1964; Crandall and Mark, 1963). Given a relationship between duration and M and R, and given a func- tion relating (expected) peak response to duration and to expected peak-ground ac- celeration or velocity, a simple application of the same method will produce such response spectra. In addition, inclusion of the randomness of peak response to random motions with given parameters (Ilosenblueth, 1964; Crandall and Mark, 1963) will permit the construction of response spectra based on prescribed probabilities of re- sponses not to be exceeded in a given lifetime. There is strong reason to believe that this latter influence is negligible (Rosenblueth, 1964; Borges, 1956). Although developed specifically for the seismic risk analysis of individual sites, the method systematically applied to a grid of points would yield regional seismic proba- bility maps. These might take a form similar to those used in determining design winds (Thorn, 1967), namely contours of maximum ground motion of equal return period. Consistent maps could be produced to as fine a scale as desired. Perhaps the greatest advantage of this method for this purpose is that it would insure t hat consistent assumptions were being used for all portions of the region and among different regions. 1604 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA All assumpt i ons made b y t he sei smol ogi st s i nvol ved woul d be explicit and quant i t at i ve, open t o r evi ew and t o up- dat i ng wi t h new evi dence. Ma j o r difficulties woul d r emai n, however , in t he j udgement of act i ve sources, in t he es t i mat i on of t hei r aver age a c t i vi t y r at es, and in det er mi nat i on of local soil influence. CONCLUSION A quant i t at i ve me t hod of eval uat i ng t he seismic ri sk at a par t i cul ar site has t he a dva nt a ge t ha t consi st ent es t i mat es of t hese ri sks can be pr epar ed for var i ous pot en- t i al sites, all per haps i n t he same general regi on but i n si gni fi cant l y di fferent geomet r i - cal r el at i onshi ps wi t h r espect t o pot ent i al sources of ear t hquakes. Such a me t hod is necessar y t o det er mi ne how r api dl y t he ri sk decays as t he resi st - ance of t he s ys t em' s desi gn is i ncreased. Reas onabl e economi c t rade-offs, be t he y wi t h r espect t o oper at i ng regul at i ons, bel ow- s t andar d per f or mance, or s ys t em mal f unct i on, cannot be made wi t hout such qua nt i t a t i ve rel at i onshi ps. The me t hod pr oposed offers t he means b y whi ch t o ma ke t hese engi neeri ng anal yses consi st ent wi t h t he sei smi ci t y i nf or mat i on avai l abl e. Thi s i nf or mat i on is t r ans f er r ed f r om t he sei smol ogi st i n t he f or m of his bes t es t i mat es of t he aver age r at e of seismic act i vi t y of pot ent i al sources of ear t hquakes , t he r el at i ve l i kel i hoods of var i ous mag- ni t udes of event s on t hose sources, and t he rel at i onshi ps bet ween site charact eri st i cs, di st ance, and magni t ude appl i cabl e for t he region. The concl usi ons appear in an easi l y appl i ed, easi l y i nt er pr et ed form, sui t abl e f or r evi ew for consi st ency and sensi t i vi t y t o assumpt i ons. For t he mos t commonl y assumed f unct i onal f or ms of t he rel at i onshi ps used, t he upper t ai l s of t he pr obabi l i t y di st r i but i ons of t he desi gn gr ound mot i on pa r a me t e r s are f ound t heor et i cal l y t o be of Ty p e I or Ty p e I I ext r eme val ue t ype. ACKNOWLEDGMENTS This work was supported by a T. W. Lambe and Associates consulting contract with the govern- ment of Turkey and by the Inter-American Program of the Civil Engineering Department of the Massachusetts Inst i t ut e of Technology. This latter program sponsored, in part, the author' s visiting professorship at the University of Mexico, where discussions with Luis Est eva and Dr. Emilio Rosenblueth initiated the author' s interest in this subject. Subsequently I nt . Est eva in- dependently developed a number of the results presented here (Esteva, 1967). The author wishes to thank two co-workers, Octavio Rascon and Erik Vanmarcke, who contributed to this study. REFERENCES Aki, K. (1956). Some problems in statistical seismology, Zisin, 8, 205-228. Allen, C. R., P. St. Amand, C. F. Richter and J. M. Nordquist (1965). Relationship between seismicity and geologic structure in the southern California region, Bull. Seism. Soc. Am. 55, 753-797. Benjamin, J. R. (1967). Probabilistic models for seismic force design, ASCE National Convention, Seattle. Blume, J. A. (1965). Earthquake ground motion and engineering procedures for important instal- lations near active faults. Proc. Third World Conf. on Eq. Engr. (New Zealand), IV-53. Blume, J. A., N. M. Newmark and L. H. Corning (1961). Design of Multistory Reinforced Concrete Buildings for Earthquake Motions, Portland Cement Assoc., Chicago. Borges, J. F. (1956). Statistical estimate of seismic loadings, Preliminary Publ. V. Congress IABSE, Lisbon. Borgman, L. E. (1963). Risk criteria, Proc. ASCE, WW3, 1-35. Cornell, C. A. (1964). Stochastic Process Models in Structural Engineering, Dept. of C. E. Tech. Report 34, Stanford University, Calif. Crandall, S. H. and W. D. Mark (1963). Random Vibration in Mechanical Systems, Academic Press, New York. E N G I N E E R I N G S E I S M I C R I S K A N A L Y S I S 1 6 0 5 Dick, I. D. (1965). Extreme value theory and earthquakes, Proc. Third World Conf. on Eq. Engr. New Zealand. I I I 45-55. Epstein, B. and tI. Brooks (1948). The theory of extreme values and its implications in the study of the dielectric strength of paper capacitors, J. Appl. Phys. 19, 544-550. Epstein, B. and C. Lomnitz (1966). A model for the occurrence of large earthquakes, Nature, 221, 954-956. Esteva, L. (1967). Criteria for the construction of spectra for seismic design, Third Panamerican Symposium on Structures, Caracas, Venezuela. Esteva, L. and E. Rosenblueth (1964). Spectra of earthquakes at moderate and large distances, Soc. Mex. de Ing. Sismica, Mexico II, 1-18. Gumbel, E. J. (1958). Extreme Value Statistics, Columbia University Press, New York. Gzovsky, M. V. (1962). Tectonophysics and earthquake forecasting, Bull. Seism. Soc. Am. 52, 485-505. Housner, G. W. (1952). Spectrum intensities of strong motion earthquakes, Proc. Sym. on Earth- quakes and Blast Effects on Structures, Los Angeles. Housner, G. W. (1965). Intensity of earthquake ground shaking near the causative fault, Proc. Third World Conf. on Eq. Engr. 94-111. Ipek, M. et al (1965). Earthquake zones of Turkey according to seismological data, Prof. Conf. Earthquake Resistant Construction Regulations, (in Turkish) Ankara, Turkey. Isacks, B. and J. Oliver (1964). Seismic waves with frequencies from 1 to 100 cycles per second recorded in a deep mine in northern New Jersey, Bull. Seism. Soc. Am. 54, 1941-1979. Kanai, K. (1961). An empirical formula for the spectrum of strong earthquake motions, Bull. Eq. Res. Inst. 39, 85-95. Kawasumi, H. (1951). Measures of earthquake danger and expectancy of maximum intensity throughout Japan as inferred from the seismic activity in historical times, Bull Earthquake Res. Inst. 29. Knopoff, L. (1964). The statistics of earthquakes in southern California, Bull. Seism. Soc. Am. 54, 1871-1873. Lacer, D. A. (1965). Simulation of earthquake amplification spectra for southren California sites, Proc. Third World Conf. on Eq. Engr. New Zealand III, 151-167. Lomnitz, C. (1966). Statistical prediction of earthquakes, Reviews of Geophys. 4, 337-393. Milne, W. G. and A. G. Davenport (1965). Statistical parameters applied to seismic regionaliza- tion, PToc. Third World Conf. on Eq. Engr., New Zealand, III, 181-194. Muto, K., R. W. Bailey and K. J. Mitchell (1963). Special requirements for the design of nuclear power stations to withstand earthquakes, Proc. Instn. Mech. Engr. 117, 155-203. Newmark, N. M. (1965). Current trends in the seismic analysis and design of high rise structures, Proc. Syrup. on Earthquake Engr. Vancouver, B.C. Newmark, N. M. (1967). Design Criteria for Nuclear Reactors Subjected to Earthquake Hazards, Urbana, Ill. Niazi, M. (1964). Seismicity of northern California and western Nevada, Bull. Seism. Soc. Am. 54, 845-850. Nordquist, J. M. (1945). Theory of Largest Values Applied to Earthquake Magnitudes, T~ans. Am. Geo. Union, 26, 29-31. Parzen, E. (1962). Stochastic Processes, Holden-Day, San Francisco. Richter, C. F. (1958). Elementary Seismology, W. H. Freeman and Co., San Francisco. Richter, C. F. (1959). Seismic regionalization, Bull. Seism. Soc. Am. 49, 123-162. Rosenblueth, E. (1964). Probabilistic design to resist earthquakes, Proc. ASCE, 90, 189-219. Rosenblueth, E. (1966). On seismicity, Seminar in the Application of Statistics to Structural Me- chanics, Dept. of Civil Engr., Univ. of Penn. Sandi, H. (1966). Earthquake simulation for the estimate of structural safety, Proc. Ing. Symp. R.[.L.E.M., Mexico City. Task Comm. on Wind Forces (1961). Wind forces on structures, Trans. ASCE, 126, 1124-1198. Thorn, H. C S. (1967). New distributions of extreme winds in the United States, Conf. Preprint 431, ASCE. Envir. Engr. Conf. Dallas. Uniform Building Code, 1967 Edition, (1967). Inter. Conf. of Building Officials, Los Angeles. Veletsos, A. S., N. M. Newmark and C. V. Chelapati (1965). Deformation spectra for elastic and elasto-plastic systems subjected to ground shock and earthquake motions, Proc. Third World Conf. on Eq. Engr. New Zealand, II, 663-682. 1606 BULLETIN OF THE SEISMOLOGICAL SOCIETY OF AMERICA Vere-Jones, D. (1966). A Markov model for aftershock occurrences, Pure and Applied Geophys. 54, 31-42. Wanner, E. (1937). Statistics of earthquakes I, and II, Ger. Geitr. Geophys. 50, 85-99 and 223-228. Wiggins, J. H. (1964). Effect of site conditions on earthquake intensity, Proc. ASCE, 90, 279-312. DEPARTMENT OF CIVIL ENGINEERING M.I.T. CAMBRIDGE~ MASSACHUSETTS Manuscript received January 2, 1968.