OE 2 Seismic hazard and earthquake actions J.J. Bommer and PJ. Stafford 2.1 Introduction Earthquake-resistant design can be considered as the art of balancing the seismic capacity of structures with the expected seismic demand to which they may be subjected. In this sense, earthquake-resistant design is the mitigation of seismic tisk, which may be defined as the possibility of losses (human, social or economic) due to the effects of future earthquakes. Seismic risk is often considered as the convolution of seismic hazard, exposure and vulnerability. Exposure refers to the people, buildings, infrastructure, commercial and industrial facilities located in an area where earthquake effects may be felt; exposure is usually determined by planners and investors, although in some cases avoidance of major geo-hazards may lead to relocation of new infrastructure. Vulnerability is the susceptibility of structures to earthquake effects and is generally defined by the expected degree of damage that would result under different levels of seismic demand; this is the component of the risk equation that can be controlled by engineering design. Seismic hazards are the potentially damaging effects of earthquakes at a particular location, which may include surface rupture, tsunami run-up, liquefaction and landslides, although the most important cause of damage on a global scale is earthquake-induced ground shaking (Bird and Bommer, 2004). The focus in this chapter is exclusively on this particular hazard and the definition of seismic actions in terms of strong ground motions. In the context of probabilistic seismic h award analysis (PSHA), seismic hazard actually refers to the probability of exceeding, a specific level of ground shaking within @ given time. If resources were unlimited simply providing as much ear In practice, it is not feasible to reduce seismic val minimum because the costs would be prohibitive a since they would be for protection against occur during the useful life of the structur, to balance the investment in provision of 4 of damage, loss or distuption that earthqu ; seismic protection would be achieved by thquake resistance as possible to structures. Inerability to an absolute nd certainly not justified a loading case that may not even €. Seismic design therefore seeks cismic resistance against the level ake loading could impose. For thisSeismic hazard and earthquake actions 7 Is of a i cred leve , quantitative assessment and characterisation of the expe expected leven ition Shaking constitute an indispensable first step of een is 1S Process of seismic hazard analysis that is introduced in ee invariant of ground-shaking hazard due 7 fe a for they involves three steps: the development 01° seismicity, moe the location and size (and, if appropriate, the frequent” © 1 fo, the ee in the region; the development of a ground-mot ee Of thea iction of expected levels of shaking at a sive ste a 2 am S€ earthquake scenarios; and the integration of these tw Peer ‘del for the expected levels of shaking at the site of interest (8 Ground-motion model Seismicity model Peak ground acceleration (g) Scenarios Acceleration (cm/s?) Rg Sure & defines cc, Schematic overview of seismic razard anal Ta from thes ‘interest cCatios of earthquakes of magnitude, My a this Md the ground-motion model predicts the shaking Parts of 3 Fesponge COMbination, The results in this ease ate PEE plan SPectray Petra (see Chapter 3 for definition and deta— MM Sh dS lit _,_ememmenmmn naa” 8 Ld. Bommer and py. Stafford The first three sections of this chapter deal with the three steps iluserated in Figure 2.1, that is eismicity models (Section 2.2), pont apa (ection 2.3) and seismic hazard analysis (Section 2.4). The remaining a sections then explore in more detail specific representations of the ae Motion for engineering analysis and design, namely response spectra (Secti = 2.5) and acceleration time-histories (ection 2.6), both with specific estelzit d to the stipulations of EC8. The chapter closes with brief conclusions a parding both the use of EC8 as the basis for defini Tecommendations reg ibe Seismic design loads and possible improvements to the code that cou! ons, made in future revisic 2.2 Earthquake An entire book, let alone a Chapter, could be dedicated to the ae seismicity models, Herein, however, a very brief overview, with ae references, is Presented, with the aim of introducing definitions for the Parameters and the main concepts behind seismicity models, With the exception of son; events, earthquakes are 9, faults, releasing elastic Parameters and seismicity smic waves. ates from the fault rupture in the form of seismic aE The location of the earthquake is specified by the location of the foci hypocentre, whic i ‘asured in Hoan Y seismology, using none at distances of hundreds or thousan ce ‘arthquake, the Source can be approximated as aR ie nphasise that in reality the earthquake source ees Ce is ultimately the part of the crust that Ore Te a result of the fault slip; the dimensions of the eamgu a are controlled by the length of the fault rupture and, to a lesser a ‘lip the amount of slip on the fault during the earthquake. The rupture an a lengths both grow 7 ponentially with the magnitude of the earthuale, | ie shown in Figure 2.2, ‘Two good texts on the geological origin of earthqua and the nature of faulting sre Yeats et al. (1997) and Scholz (2002). ‘The magnitude of an i of energy released in magnitude scales Waves at different urposes of observator Sensitive instruments kilometres from the e; 't 1s important to em very large. The sour relaxation as source ic sured from the amplitude of Ce Magnitude scale proposed et rally denoted by M,, where the subscript stands for lo pee s of body- ‘alogues generally report event size in terms of b ES iw OF Surface-waye magnitude, M, which will often g9 200 12 a g 3 & Rupture slip (m) o Rupture length (km) s “average rea ee Moment magnitude, M, 5 55 6 65 tft 6 5 8 Moment magnitude, M,, Figure 2.2 Medi eqatign 2 Median pred 8 of Wells ar cae vats a ane Iength and slip from the empirical scales mentioned so far different y, ze and are Share a ¢ therefore ortcomi alue bites same earthquake. All of the unable to cena a that they saturate at 2 certain si ; coming does nota aris the sizes of the very largest earthquakes. This Thig eas determined Pa Hes moment magnitude, designated as M, or M, Of the are! based on the the ven long-period part of the seismic radiation. Tigi area of the fault parameter seismic moment, which is the product oat the crust rupture, the average slip on the fault plane and the ismict i future icity model needs to specify th cane to thauakes of differen magnitudes. A wi availatiake aig seismicity models, generally start Rate le since the ae Instrumental recordings of eart eel elle aera Lol the nineteenth century and eve 196 geeues are ae s and low sensitivity of the instrun tigtare The catalogue a incomplete for smaller magni telat ical Scag for a region can be extended throu Onships q its of earthquakes and the inference, thr lerived from twentieth-century earthquakes: Fo, T Soy me To Parts o} ml f the world, historical seismicity can extend the cataloBue 0 ape exrended even further thre, 200 years to s a paleo tt ‘The record ca ; t n aoe field oumolonieal studies (McCalpin, 1996)» which essentially can 128 60-Seismic y of geological fanlts ro assess the deve ancl ample of the §j Obtained eo ute ‘Rdditional constraint on the seismicity model lan, Mee id study of ‘om the tectonic framework and mor’ specifically from Cae. Meas potentially active structures and their signature on the the eS¥ oF se eae ate simgeal deformation USINE traditional tal ee eee chm Siko provide useful input £° estimating ic moment budget (e.g; Jackson» 2001). nl frequency ocation al an de range of data ¢ ing, with regional quakes are only n then th ments meal ides prior to the igh the study of ‘ough empirical of magnitudes. e expected loa fee es a wees 10° JJ. Bommer and Py, Stafford ; The seismicity model needs to first specify the spatial venous future earthquake events, which is achieved by the definition of es pe sources. Where active geological faults are identified and their eae activity can be characterised, the Seismic sources will be lines or Plane its reflect the location of these structures. Since in many cases poate a will not have been identified and also because it is generally not possib 8 unambiguously assign all events in a catalogue to known faults, cue will often be defined. These are general areas in which it is assume: seismicie of mechanism and type of earthquake, a y likely to occur at any location within the source. i Fie ee COese will penerally lie-within aecal sou micity that is not associated with the fault. : ich fixes the Once the boundaries of the Source zones are defined, which fix Spatial distribution of the seismict 4 model for the tem generally referre ¥ IS uniform in term: that events are equall where fault sources that capture the se is that known as the Gutenberg: Richter (G-R) relationship, whic Simple power law relationship bate a and magnitude, The relations a vity (ie, the annual rate of Cae @n or equal to zero or some a the slope of the recurrence ea i mall and large earthquakes; b-valu’ of earthquakes of m threshold level) and and defines the relat, for large areas in much of the world relationship must be trune earthquake that the sei agnitude greater th the b-value, which ig ive proportions of s a small increment to the largest historical event in the earthquake catalogue: The typical form of the G-R relationship is illustrated in Figure 2.3. iv For major faults, it is believed that the G-R Tecurrence relationship eH not hold and that large magnitude earthquakes occur quasi-periodically ea relatively little activity at moderate magnitudes. This leads to alternal is models, also illustrated in Figure 2.3 ifonly large earthquakes occur, then ae maximum magnitude model is adopted, whereas if there is also some activi ; ines in the smaller magnitude ranges then. a model is adopted which com a GR relationship for Jo er h the occurrence ae characteristic earthquakes at higher rates than would be predicted ae extrapolation of the G-R relationship. The recurrence rate of characte! events will generally b ‘ i han ¢ inferred from Palcoseismological studies rather t from the earthquake catg ; : nded infrequent to have ences in catalogues. Highly recomme references on recurrence relationshii nships include Reiter (1990), Utsu (1999) and McGuire (2004),Seismic hazard and earthquake actions i 2 s~ ge 2 4 ee 5 2 ay ge at & ge By = 25 mtou| 3% 2 a 2 z ge ie F Magnitude, m anne a om tom 2 Magnitude, M re d Be 7 z ga a a 3s oF Be 3 Z : gn E £2f 23 3 iad 2 is 5 z sss 3 Magnitude, a i Magnitude, M o Figure 2, 3 ss ypical for rms of earthquake recurrence relationships, shown in non forms, From left to right: cum vulat Gureative ( Sut upper citenbes Richer row) and cumulative (lower row) ake model fel, maximum magnitude model, and characteristic 23 TOund. i = : otion characterisation and prediction The crux the of specifyi Br ying earthquake actions for seismic design lies in estimating inertial Joads that are ulti wd moti ec iidena caused by earthquakes. The aces Upon ae are directly related to the motion of the for the roducing che too ee is built. The present section is concerne! BronnpetPose Bical ols developed, and used, by engineering seismologists eee nd the source of an eartbaake (° the at can be expected at any given site. 2.3, “3.1 A, ccelerog : recording and processing Most of y have d as not obrained until 5 the oe Eee ae in the field of ¢ seismolog He os using pele ener of high-quality time 1233 durin, ees "The first of these ™ ; ating tousands eae ey California, earthauake Bvt ST that eae seismic netw. strong-motion records have been acquired through sing OBTAMS, tec sie across the globe. Prior to the sequisition of the first ora Sebmogeaphs b ings of earthquake ground motions had been made Until gpd aly. ser wut the relatively high sensitivity of these instruments fine bal ‘ong ground motions from being rec lance between creating 4 thest orded. It was not robust yet sensitive instrumentSs aa 12 JJ. Bommer and py, Stafford f ccelerograph, that the field 0} Cnineering seismology was born, in two main forms: analogue and dint ¢ and, while modern instruments are a gue instruments remain in oie and continue to Provide important recordings of. strong ground motions. font n both types of instrument must be processed be ne d applications, Accelerographs simultaneously li TsPect to time in three orthogonal directions nee *wo in the horizontal plane and one vertical) yet, de ite this configuretony it is never Possible to fully capture the true three-dimensional eee the ground as the ins ‘ot ‘see’ all of the ground motion. a acceleration time ser corded may be viewed in the frequen ff domain following i performing this operation = comparing the recorded F; le spectrum with the spectru ae “vant for the instrument, one fin ‘finite bandwidth over which the signal-to-noise Tatto is sufficiently high that one can be confident that the recorled ao wae Benuiinely associated with carthquake-induced ground shaking. Beyon the lower and upper limits if proper filtering is by noise. Boore and Bommer (2005) provide extensive guidance on how one should process ace nsure that the records i Not contaminated, and Bommer (2005) highlight the serie importance of applying an appropriate low-cut filter, particularly whe! ‘ain displacement Spectral ordinates. Poe he the key issue is to identify the maximum Period up to which the filtered dat: can be used reliably. ‘ar and Bommer Processed analog Stiff and soft soil sj rams and concluded that for son ings can be used for determining tl at periods up to 0.65, 0.65 and 0.7 of the long-period filter 08 n tesPectively, whereas for digital recordings these limes acne 9.8, 0.9 and 0.97. This issue is of Sreat relevance as si platemeat pases design methods (Priestley et al., 2007), which rely upon the epecbeuons Cné-Period displacement spectral ordinates, become more widely adopted. An example of the j ce of proper record Processing is shown in wey 2.4 in which both an analogue and a digital record are shown before an after processing ~ this a‘ cement example clearly shows how sensitive the displacem is to the presence of noise. 2.3.2 Ground-motion Parameters Properly processed, many ground motion may be calculated (for aSeismic hazard and earthquake actions 13 +1999 Chi-Chi Earthquake 1940 In pera Valley Earthquake NortrSouth component carded ot TCUPED East-West component recorded at £l Cento 9 a eon nent recorded at El ‘tre th = Maree 200 100 300 ° = ge Time (5) S : ad rime 6) ) and alogue (left ginal records by nd digital the overall me: mean and the pre-event mean for filrered morions are give” Figur digi 24 Accelerati ital (eiingCeletation, velocity and displacement tem, ) vig iB) recordings. G 18 I ings. Grey traces were obtaine’ traces from ai dd from the ori he analogue a otds On the righ eGctvely The dis isplacement axis labels for the un ight-han and-side of the graphs. Modified from Boore and Bommer hese parameters f the recorded y few of these 2 on those eee ne ged during the specification of design Spe states of S pe ay be calculated, peak ground ‘accelerations PGA, 2%), have by rcent damped elastic acceleration response spectra, een used by far the most frequently. parameters that Bur may pre 25 sh Me OWS . s many of the possible ground-motio” cord, Each one of these loser tcula awe tive Sralaanlias an individual earthquake ‘© help unde ters provides some degree ‘of information stand the demands imposed upon @ structure. ‘Although sly specify more Meth od, th, lologi lan Bical fy oO ra i cam oe ee place to simulraneol ion parameter (Bazzurto and Cornell, is (Shome og these 5 al fre) Parameters thi ough to a structura’ des “tiptio POVide nt Of m: ides itiemarte of these, see Kramer, 1996)- Each of tl nn about a different characteristic 0! neerl ned, ver) Broun, id moti lotion. As far as engineering design is CO” |, the additi #8 excagcre tdditi a ee ee that is requi Characters cpiultiee without justifiable benefit in tistics of earthquake engineers will seek fo aecoun ground motions in the future:ae 14 JJ. Bommer and Pj, Stafford os ee, 15. : b Tene 24 ae : dos 5 é a ie °o 10 20 Time (eacond) oe 1900 p——__ Ty*040083 oe E sco 5 2 e dw Al i 2 Ba : B vo aan Oy, ie . Frequency (He) ae ‘7 <= yng 24) 8k, — 5g = 1089 E g 15 é : 9 PCD = 6.5300 cm | Time (seconds) ofad (eon may Figure 2.5 py tation of the types of 8round-motion parameters ha oe be calculated single record. The record in this case is the 020° comp recorded dur Angeles. The three 7 8 5s Si ion in Lo it the Saturn St. station i time-sey \C ration, velocity, and atts J prae saute (rms) values, The pane ee tae bottom, a Husid plot of he build-up of Arias intensity he 5 5~95% of the rotal Arias ene Fourier amplitude spectra along with the mean period and finally the accelera Fesponse spectrum for damping | » 5, and 10% of critical1s Seismic hazard and earthquake actions earth quake fi levelo tom which aah pene ofempini ether came. This type of reasoning is the basis for the pills a relationship i predictive equations for strong ground-motions. cee ee aie suite of observed ground an associated set of independent variables including a Meas ure the 5 of the si sou size of eee wrensure of the distance from ofthe cartidion of te syle of fuking nvOlve8 ar the and some di tecordii lescription : Since ee pea geological and geotechnical conditions eapeion OF these indepen | gander prediction equation is imply some Yale of the groun oe variables that provides 20 estimate of the eapure ofthe disribu motion parameter in consideration 28 well as Cquations hg the developm ion of values about this expected value. around as been Eveeeaet of empirical jground-motion prediction veloc motions, parila ee Pentged upon the prediction of peak response GV), and ordinat PGA and, to a far lesser extents peak ground equation spectra (Douglas, aa of 5 percent damped elastic acceleration si Have also.ibeen ¢ 03; Bommer and ‘Alarcén, 2006)- Predictive fen developed for most of the other parameters 0! previ evio Actioy US secti ns ion, as well a hi well as others not mentioned, but as seismic design the demand for saps ave historic Iy99 tations ae SEE been derived from PGA oF S,( for hgh 00 lesser ae weak. The performanc’ we PGA. (Wald et als indoor woes oe peaiceas S(T) (Priestley, 2003; Akkar and Ozen, 2005), structs tents ihihe ae, structural damage has begun to be questioned. into ee earthquake @ jaboration between engineering seismologists an and th really are th — has prompted the emergence ‘of research the pean spectral pea ey descriptors (such as inelastic spectral ordinates Assess nd motion ee, for damping ratios other than 5 percent) © Sen cll, of damage i a ‘of importance to structural response ‘and to the R a Eee) structures (Bozorgnia et al., 2006; ‘Tothong, and i SS 0 redicnon ae he measur 7 variables, ation can be represen! Tey is the Nes H(M,R,0), and a oe term, ground motion measure: : eration, eric functlo} tion (2-1) ¢ in consid ted as a 8° €0,, as Equat lo By = ' H(M,R,0) + £6, - ‘any ” ASS y dey. He, f Physio. ers_of ground-motion 79). In oe to the terms in d ec ; Dongig, Sed as ieee possible to derive theoretical © select 2002), Althous for selecting appropriate Franeional forms been gan 2PPropriat hough these theoretical considerations interprenducted the 4 functional forms, he regression ae souaaliine! of regression coefficients @ physical meaning as correlations of varying degrees aly, YS ei exist by St Detwee: n the coefficients for different terms of thi ations atcempt £0 scully derived function prediction eat he empiri 1 equations that16 J.J. Bommer and PJ. Stafford For most ground-motion measures the values will increase with increasing magnitude and decrease with increasing distance These two scaling effects form the backbone of prediction equations and many functional forms have been proposed to capture the variation of motions with respect to these two predictors (Douglas, 2003). For modern relationships distinctions are also made between ground motions that come from earthquakes having different styles of faulting, with reverse faulting earthquakes tending to generate larger ground motions than either strike-slip or normal faulting events (Bommer et al., 2003), Historically, account was also taken for site conditions by adding modifying terms similar to those used for the style-of- faulting effects ~ stiff soil sites have larger motions shen rock, and soft soil sites have larger motions still. In Europe thie use of dummy variables for generic site classes remains the adopted approach in the latest generation of Prediction equations (Ambraseys et al., 2005; Akkar and Bommer, 20072, 20076), primarily due to the absence of more denied site information. However, in the US, site response is now modelled using the average shear- Wave velocity over the upper 30m, as introduced by Boore et al. (1997). Furthermore, the influence of non-linear site response, whereby weaker motions tend to be amplified more so thar stronger motions due to the increased damping and reduced strength associaredt with the latter, is also taken into account (Abrahamson and Silva, 1997; Choi and Stewart, 2005). Figure 2.6 demonstrates the form of the non-linear site amplification functions adopted in two recent prediction equations developed as part of the Next Generation of Attenuation relations (NGA) project in the US. The 5 pa mes 3 | zooms | < aooms | a @ 2 some | iii oom |S 215 g 2 2 Ba: 5 & 8 3 ri gu & oor one Sa 2 EQPGA ri) (9) E(PGA yi) (9) Figure 2.6 Comparison of two nonlinear site response models for peak ground (etletation. Both models are from the NGA project with Abrahamson and ae (2007) and Chion and Youngs (2006) on the left and right respectively. The Abrahamson and Silva (2007) model shows amplificesion with respect to the expected value of PGA at a site with Vos9 = 1100 m/s while the Chiou and Youngs (2006) modellgtiows ché sxmplificarion Wiel reqpoct wo ei vedied aaetiods ob aie with V,.. = 1130 m/sSeismic hazard and earthquake actions 17 difference in site amplification relative to rock for sites with differing shear- wave velocities and varying input rock ground motion is striking, with both models predicting de-amplification at strong levels of input rock motion. In addition to the basic scaling of ground motions with magnitude, distance, site conditions, etc., there are additional situations that may result in modified ground motions that are commonly either omitted from developed equations or are later applied as correction factors to the base models. The most common examples include accounting for differences between sites located on the hangingwall or footwall of dip-slip fault sources (Abrahamson and Somerville, 1996; Chang et al., 2004), accounting for rupture directivity effects (Somerville et al., 1997; Abrahamson, 2000), including models for the velocity pulse associated with directivity effects (Bray and Rodriguez Marek, 2004), basin effects (Choi et al., 2005) and topographic modifiers (Toshinawa et al., 2004). The most recent predictor variable to be included in prediction equations for peak ground motions and spectral ordinates is the depth to the top of the rupture (Kagawa et al., 2004; Somerville and Pitarka, 2006). Currently, none of these effects are incorporated into any predictive equations for ground motions in Europe, nor is any account made for non-linearity of site response. Again, this is primarily a result of the lack of well-recorded strong earthquakes in the region. 2.3.4 Ground-motion variability For any particular ground-motion record the total variance term given in Equation (2.1) may be partitioned into two components as in Equation (2.2): (2.2) log y,, =n0m, 7,,0,) + 8, + 8,5 iduals The terms 8,, and 6, represent the inter-event and intra-event res respectively and quantify how far away from the mean estimate of logy, the motions from the ith event and the jth recording from the ith event are respectively (Abrahamson and Youngs, 1992). Alternatively, these terms may be expressed in terms of standard normal variates (z,, and z,,,) and the standard deviations of the inter-event (t) and intra-event (6) components, ie. 6,, =z,,1 and 8,, =z,,0. The total standard deviation for a predictive equation is obtained from the square root of the sum of the inter-event and intra-event variances, i.e. from 6% = 1” +07. Later, in Section 2.4 regarding, PSHA, mention will be made of epsilon, €, representing the number of total standard deviations from the median predicted ground motion. Often ground-motion modellers represent the terms 8,, and 8,, by 1, and &, respectively. Under this convention care must be taken to not confuse the epsilon, €, with the intra-event residual, €,,, term — the two are related via 6, Le. €=(5,, +8, ‘6, using our notation. T anit Jf Or 1B the expression e =(n, +€,)), Each of these components of variability mayibe modelled as functions of other parameters such as the magnitude of the earthquake (Youngs et al.,7 a aan lH 18° JJ. Bommer and PJ. Stafford iva 1995), the average shear-wave velocity of the site Mena how 2007), or the amplitude of the ground motion (Campbell, 1997). ethodology these components are calculated depends upon the repression medio Sd that is used to derive the equations, However, the most common ee groun to adopt random cffects procedures where the correlation nome the samme Motions observed within any Particular event is assumed to be & chow across events and is equal 0 P= /(t? 46%), This concent 1 schematically in Figure 2.7, Many people think of ¢ igtributio™ ” as providing an estimate of the ee le notions for a given set of Predictor variables such Fredo From this perspective, the rea misfit of the model is cae: model’s distribution represenvs the true distribution of 0 not role etather than how lange are variance components, People ia tolike large varig his implies that we cannot pre tive . is perspect certainty. However, this persp interevent residuals — 9, Intra-event residuals > oy 3 No.7") 2 Ib a 4, q0LELE LE ,oa , a Intr-ovent residual Jy : eile zo 1s 3 2 i 3, No.0?) g TZ Metonprtton aeons) g Median pedcton (evan 3 F ona | Median preticion (event 2 08 Indl code (ovent 1 & Indicates fren, 2 pS ef rte an o oor + + io op “oyner-Boore distance, Fy (km) Intra-event residual, xy Figure 2.7 Explanation of motion ‘ance components specified in sronnd or #0 Prediction equations, ‘Tee left panel shows hove the median predi wall events ~ inelvidual event may be hig: or lower than the median predictioe wor a a the inter-event residuals, 8.,, About this median Prediction for each event are Variations in ground megs, the ‘ siieeaneGa the intra-event residuals, 8,,,. The histograi right show how and intra with zero means aie gyent residuals"are normally Sten and variances of 2? and o¢ Tespectively, The median area are generated for an M., 6.5 earthquake with an R,, distance of 10km for faulting and rock ¢ i (2007b)s © eduations'of Akkar and Bommer (201 Figure based on oungs etal. (1995) he vari both the interearthquake actions 19 's closely relate predictable ae io the paradigm that ground motions are ultimately chinplde ate only through a result of inadequate modelling and ‘and, one views ree that the apparent variability arises. If, on the other a cetin views arblind notions as being inherently unpredictable (beyond Of the misfit ale then one: must view the variability mot as.8 INERT Bie fates oistce rather as an additional part of the model that describes Paradigm there eervable ground motions given an event Under this latter mesnre snp isin eecon like or dislike a particular ground-morion isttibution, hi ecause predictive equations for this measure have a broad na ee aT sarjadging the importance of a ground Structural respo iglarleesaiond the ability of this measure tO accurately predict Sag evens mesa OS jess variability in the ground: ate will translate into less variability in the response. Seismic hazard and 2.4 Seismi eismic hazard analysis The prin analyses nae) ebjecuts of engineering seismology i ™ : © essential pcondleciet The two previous sections have provided 8S most basic k ground required to understand seismic hazard analy analysis are ee ‘As will soon be demonstrated, the mechanics of at ae Felatively straightforward, Howetrs ¢ thorough under Prerequisite ae laid out in the sections thus fart, as well as many others, 18 @ for conducting a high-quality hazard analysis. Unfortunately, in Currey ‘nt prac tice this . : ¢ this prerequisite is all too often nor met. smic hazard J most of is at 9 enable se 2.4.1 Probabil Bommer - bilistic vs. deterministic approaches sn (2002) presen a comprehensie discussion of ards peeween. probabilistic chet eterministic approaches 10 seismic ies go ae ile ie pape rermgrerministic methods wow "egarding ated the conception that there is ongoing academic debate Cea ead, the truth of the Mauer is that = eismic hazard analysis (DSHA) is simply a special case of earthquake seenmin hazard analysis (PSHA) in which only @ small number of considered, Ie oe (combinations of magnitude, distant and epsilon) are . In contrast, in PSHA all possible scenarios that are deemed to © of engineeri rahamsnn ane interest are considered (Abrahamson 2006; Bommer and Hise sate 2006). Much of the discussion yegarding PSHA and DSHA has rminolons Parent issues that really stem from miganderstandings of the tome of ree cae often loosely sed in PSHA. Bommer (2003) highlights © most common misundersta 1 relation to eatme arly in nt sei of uncertainty, and urges the p the differences and i i dings, particul Be onents of DSHA 0 ty (0 heir approaches level lop ac a consis istent set of terminology for20° JJ. Bommer and By. Stafford 24.2 Basics of PSHA, haza It is perhaps unto: somewhat intimid are actually very often a particul rd curves and return periods rtunate that the lating for some as simple, lar level site. Now, suppose th Benerates eg the rate is mathematical formulation of rs the mechanics behind the Hes how For example, imagine one wanted to pean i | of some ground-morion ineasufe 1 excels el a ‘Pere is a seismic source near this ste shat rege arthquakes of a particuls, magnitude and further See his rate at which these earthquakes occur may be quantified. Once sround- 's obtained it may be Combined with an estimate of how often the gi eee motion level at the site is exceeded when this earthquake scenario ce af For example, an event of magnitude M may occur once every six mon! roun cach time it does there is 4 50 percent chance of exceeding a target er motion ~ this target level ig then exceeded by this scenario, on ta once every year, [f one then considered another earthquake scenario, is round repeated the above Procedure, one would determine how often the g the iat level in consideration y, se cxceeded for this alternative ela the first scenatio resulted in an exceedance of the ground-motion ae times per year 2 times per year, then for these To hows the ground-me eeded 4.42, dmes per year ae aaa PSHA is conducted: all £ne has to do to complete the earnpneeat the above steps for all of the possible carthquake scenarios that may tions the site, calculate the rates at Which these scenarios result in ground etwas above the target level, and then add them all up. OF course, it is not al cut, mon Bhtforward to ascertain hey often different earthquake scenarios se which, '!*8YS obvious howe, Most appropriately determine the at, which the ground motions are exceeded given these scenarios. How that none of these issues change the Simplicity of the underlying framewor! Constitutes PSHA, Cornell, 1968, 197 1). With this simpie explanation fee Krenn it is now timely 1g relate this to what is more commonly seen in literature on this subject, anutto Formally, bag PSHA may be Fepresented as in Equation (2.3) (Bazz! and Cornell, 1999), eu (gm) = DUS Sitom > gin*lmr, chosen where the capital letters qypresent random variables (GM = a ay 8round-moti magnitude, R = distance and E = aor a i alisations of these rand a s occur having a ee rce 7 is denoted by v; i: ‘ay be taken outside of the cs NY Tepresentations of this equation). ad jtnction of magnitude, distance and epsilon is gi ral and I[GM > gm*|myr,e] is an indicator function eq xceed ]ifrune (mre), dmdrde} 2.3) 4 constant for e, T each source it m; integral, + 8 is commonly joint probability densiey °Y Pasnw (tyr,e),ae Seismic hazard and earthquake actions 21 hal Ce and zero otherwise. Finally, and most importantly, gor vis exceeded zl annual rate at which the target ground-motion value, tera; Tas page ‘often the way that PSHA is presented in the and loreemss the aie of the joint probability density faction it and iis consequent epsilon may be intractable forthe non-cognescenst Equation (25) nlyrwo spending some time to describe this key term reopen 3) Using some bs re ncepts of probability theory we may joint probability density function (pdf) into mor accable Parts 2 rts as in Equation (2.4). Wh ifs Ongre) ne), =, free aaa allay Vag, Oop) fa (tne) fo 2) OM ton pees ee Ot mud Thow many times per ye — nem this event ECE ow icy a Sinn chk fara ire ra es pone nit pa, while already annotated, Each “ach of these ce ese components of the joi planation: dese; TVES 5 on: ne additional comment and ex! centre equal 10 Xiyp> tion within source #- and that was made in Cornell’s f hypocentres are equally likely This assumption requires ire of activity for the + be, (Xp) where = the pdf for an event having # hypo ih tie (longitude, laritude, depth) is any posi original a assumptions that is made, to occur ai entation of PSHA, is that the least follies within a seismic soure seismic oie of information regarding the natu ‘ a (M2 Pe ~ the conditional pdf of magni instead analvers hazard analyses this term is May occur ysts simply take the previous assumptis also assur with equal probability anywhere within deemed a a ese evened mimy have hel fullJEANRE upon the Ne due thiewource: In tisieaetle an not of ingninnte. Goa position and one simply recovers fy (mn), the pdf ie. However, some analysts May wish ro address this problem More thoroug i oughly and make alternative gsumptions Using analyses such 1 Mai et al. (2005). For example, a8 those of § it ae eens er al, (1999) and i that large earthquakes vend to have relatively deep Magnitude a el the pdf may be modified accordingly. ‘The pdf of distribution f often assumed to follow 2 doubly-bounded exponential orm of the fam areal sources (Cornell and Vanmarcke, 1969)5 modified characterj famous G-R equation (Gutenberg and Richter, 1944), and a 1984) ag stic distribution for fault sources (Schwarte ‘and Coppersmith, relate, i emioned in Section’ 2.2, Howeves any distribution that is permis; ee rates of occurrence of earthquakes of different sizes fe (on ible. the’ Rts Xing.) = the conditional Ré-motion prediction equation Bive? aude given the hypocentral hot implicitly considered ‘on that earthquakes ‘aseismic source and of magnitudes conditioned » measure used in pdf of the distane® 1 n the rupture surface of22 JJ. Bommer and PJ. Stafford the earthquake, the size of the including the str depth boundarie: on which the mn ntre, The rupture surface depends upon the hypoce vent and various in 0, 5 in 8; other parameters gener Ne ike and dip of the fault plane (for fault sources pa enic zone, the segment of the fa ¥ 5 slates » ete. This term is important as it tral fees the assumptions regarding the potential locations of earthquakes measures of distance that are pare iction “ppropriate for use in empirical nee m1 Is necessarily different for each dis equations. Note that this tern measure that is considered, + fe(@) ~ the paf of epsilon. tt ig in always simply the pdf of t Teason it is not necess nportant to note that this term is he standard normal distribution, For ad ary to make this a conditional pdf with respect fo anything else. Although standard deviations from eround me Predictive equations may be dependent upon predictor variables such as se magnitude, the pdf of epsilon remains statistically independent of the: other variables (Bazzurro and Cornell, 1999), Given this more also modify the in in ‘ - i st now Complete representation of Equation (2.4) one must tegral to be expressed in terms of the relevant a AWation (2.3). In reality, this is not at all cumbersome as the ane the not evaluated analytically anyway and all that is required is to disererbe the range of possible parameter values and to determine the contribution © the hazard from each Permissible set of these values, The general oe alluded to in the introductory “xample and elaborated upon in the above ‘ «I schematically in Figure 2.8. In this figure, the method via which the probability that the ground motion exceeds the target lev : ms se of the cagbresented two ways: 1) in a continuous manner through the use of! cumulative distribution function of the standard normal distribution, and ina discrete manner whereby the range Of epsilon values is discretised ane the contribution to the total hazard is determined for each increment. a of these approaches will Bive very similar answers but the latter SRS offers advantages in terms of later representing the total hazard and als for the selection of ac leration time-histories to be used in seismic design (McGuire, 1995; Bazzurro and Cornell, 1999; Baker and Cornell, 2006). i Thus far we have only been Concerned with calculating the rate at which a single target ground motion x exceeded. If we now select a series a ig oat Sround-motion levels and calculate the total rate at which each level we acceded we may obtain a hazard curve, which is the standard onsput of a PSHA, ice. a plot of Ru (gm) against gm*, Examples of the fort re |-motion of typical hazard curves are given in Figure 2.9 where the ground-mot measure in this case is PGA, The curves shown in Figure 2.9 the aleatory variability in the ground results of a seismic hazard recently discussed this issu further represente demonstrate the strong influence ibe “motion prediction equation has on i analysis, Bommer and Abrahamson (2006) ee te in detail, reviewing the historical developmen!soypeoadde ruayeamba ow Suysn umoys st oueuass stip uaatd poposoxa Sutog (,u3) uonour punos’ ro81e2 ayp jo Ayigegord ous AYBE BM UO “oxeNbuBIEO oup Jo apmyuseut aup st [[2M se aoanos atp Jo soustiaiseseys snorzea uodn spuadap aoeyins ommadnu sty Jo Arras JY ~seIs ayp Aq WAT VOMISOd [eNUAI0d sy dup moge somadna (4918 aySqp) somos yNey v Jo (Xox8 sep) WOIIOA & Yay ap UE “Ss2d0d YHSd AU Jo WoRRudsosdax gHEMTOYDS g°T 24"3H| uor22[9s proses 495 saTeueape sey Yeosdde aaIDs1p aU). + saisty aup st ypwordde snonumuod auf + sourpuaas aamdna stp 104 * tu “uonou DAOgE att} JO [IP OATS OUS SUP OF POUEISIP 2 + =pumo.3 193121 ap Suipaaoxa jo Arjiquqoud ayp arwjnoye> yyney dup Jo Anawoad ayy pue apmyuseur (a3 (a) aup ‘anussodsy aty uaats advgans aamdns paumnsse a) + anuacody ay jo wornisod ayy « vo 0 20 vO o. Fo eo zo 10 0 289 pauyap oureuaas amdny, | ‘o =3 <—— [orm if WS Soy 9 oyns i w(gud)s<'sf'e=alag [eu faus)o the associat pies the difference between being in one zone oF another with Sich a cine cre r-trivial changes in ground motions. Under boundaries sa had regulatory authorities must take care in defining the imits to ined ¢ relevant sources; common practice is t0 adjust the zone See political boundaries in order to prevent ambiguity. SE the star with the introduction of EC8 looming, 4 Comparative analysis as under ip ational hazard maps within sixteen Europe countries the pctaken (Gareia-Mayordomo P5004). ‘The study highlighes developed fos methodological differences chat exist berween hazard maps Ga cide wae countries across Europe. Many of the differences lo so as a result of the differing degre¢s © xist Toughout th f seismicity that ¢ a rem geeut the region, but some of these differences are exacerbated as *[ processes not b witha Of Barochialism despite geolosie cing concerned of efforts aul or political boundaries. There ar, however, other examples Stee exe ee erate “Ievelop consistent seismic hazard maps GSHAP (Gi led regions. The two primary examples of such efforts 2% the at integr iardini et al, 1999) and SESAME (Jiméne? & wl 2001) projects °F global ate national hazard information in order to develop continental be viewed cale hazard maps. These examples of regional hazard maps may Gsuapy , at the following URLs: the GSHAP map at ‘wrwweseismo.cthZ-cb VAP/ and the SESAME map at te paca ene oth approaches Mor trul Tg he gust azar! mapsito be developed the best of b c evel Tw" upon. For example, ground-motion predicrion equations mech as those of Ambrasey® et “fvelope al, 2008) from large regional datasets ‘ Tobust wi or Akkar and Bommer (20072 2007b), are likely to be more & more wren applied within individual countries than those developed from Motion m, ited national dataset (Bommeh '2006b). Furthermore, ground °f Euro iodellers working in low-seismicity regions, such as in most parts With ype offen make inferences regarding the sealing of ground motions {© themnitude on the basis of the Bare magic cata tha available &xist When doing so, researchers find egional differences that egional en making, comparisons between 1 predictions of F larger eoantmotion models derived predomi ecordings shown shay enue earthquakes ve pera g0ey eae work has pare ney te unfounded and that particular care ms range op heh extrapolating empirical ground Is beyond On the th patent which they were der ymer ct al ee inelidedh ee vletailed assessments of seismoBe led for national hazard map and zonati28 J.J. Bommer and PJ. Stafford not fully incor Porated into region relatively poor. ion is al studies where the spatial resolutior 2.5 Elasti Most seismic design is based on n the €Presenting the earthquake actions in form of an equivalent static force applied to the structure. These forces See determined from the maximum acceleration response of the structure oe the expected earthquake-induced ground shaking, which is represente nah the acceleration response g ing point is an elastic ine pectrum, which is subsequently reduced by factors thar account for a CRAY of the structure to dissipate the seismic energy through inelastic deformations, The definition of the elastic response spectrum oe Fonversion to an inelastic spectrum are Presented in Chapter 3; this ane focuses on how the elastic design response spectra are presented in seist design codes, with particular reference to BC8, The purpose of representing earthquake action: such as EC8 is to circumvent the ne Seismic hazard analysis for eve regions. For non-crit Provide a zonation m, throughout the regio Parameters rey the near-sur spectrum at in a seismic design code Cessity of carrying out a ae TY engineering project in seismically active ures it is generally considered sufficient a ‘ap indicating the levels of expected ground cae n of applicability of the code and then to use tl f Presented in these zonations, together with a classification © * - sponse ace geology, in order to construct the elastic design respo! any given site. ‘al struct 2.5.1 Uniform hazard spectra and code spectra The primary output from a PSHA is a suite of hazard curves for ee spectral ordinates for different response Periods. A design return perio . then selected - often rather arbitrarily as noted previously (e.g. ee 2006a) and then the respons Parameter at this return period is determine’ at each response period and used to construct the elastic response spectrum. X spectrum produced in thie way, for which it is known that the a Period associated with several response periods is the same, is known . a uniform hazard spectrum (UHS) and it is considered an STE Probabilistic representation of the basic earthquake actions at a eae location. The US will often be an envelope of the spectra associated Fc different sources of Sessmicity, with short-period ordinates controlled fe nearby moderate-magnitude carthquakes and the longer-period part of o Spectrum dominated by larger and more distant events. As a Someeacr the motion represented by the UHS may not be particularly realistic, ue interpreted as being associated with some design scenario, and this become an issue when the motions need to be represented in the form of acceleration time-histories, as discussed in Section 2.6. If the only parameter of interest tOSeismic hazard and earthquake actions 29 the engi, te engine i er 1S. * . in its fandamer 7 maximum acceleration that the structur® will experience ntal mode of vibration, regardless of the origin of this motion oF any of a Sats Ss features (euthiss “faation), then the UHS is @ perfectly ollowing ae for the representation of the earthquake actions. Whe gates ach is assumed that the UHS is a desirable objective: Dia ee seismic design codes invariably presented a single ‘eee usually for a return period of 475 yeaus showing values of Spectral shay that in essence was the PGA. This value was used tO anchor a Renae ‘ed for the type of site, uswally Wlefined by the nature of are eae erin the elastic design spect” In many Wout a4 G aloo be multiplied by an imporanes factor, which Period) ae the spectral ordinates (and thereby the effective recumn ee required co perform (0 2 higher level of dama one earthquake actions, either becatst of the consequences Would pene sree company Ot ie materials) oF because the facility Station or a operational in a pos arthquake situation (e-8- fire Glen spital). ‘ve spectrum constructed in thi ' ated oe where the spectra ne hazard co; rn period would often not be the target value 0 : ‘ Mia ee Simplified into zones with * single representative technique ae the entire area. More importantly, thi al construction Seth a ete specification of seismic Toads ro accour aS Well iat shape of response spectrum ¥: hh earthquak the PGA with site classification (Figure 2-11)> wri the result hat e¥ER | igs cio male was ‘associated with the exact design return Perio’, Would Paemikely indeed that ‘he spectral ordinates 4 eriods ave the same return period (McGuire, | 7). Conseque « way would almos » PGA, the Even | acceleratt £475 years since ASSOC}, aries Wit g 2 oe al = a= é & asf ampbell (1997) & g Sq) gz S~ Hf \ 3 BE | Ss 3 aS] —enns| | = S zB Be — se z 3 ie a g os a | § e ; § 3 8 ° — eee ee ae 3 fee a aeeea 74 o> 05 1 Period (seconds) period (soc0nd) hy igure 2 210 kn ge tiged 10 PGA, for a rock site an a perenne NOt nea iAy O km from a Ciempbell reer etaakes a different ma 997) and Boore et al. (1997)ee —xeEeEeEeEeEeEeEE 30° JJ. Bommer and PJ. Stafford objective of a UHS is not met by anchoring spectral shapes to the zero period acceleration, Various different approaches h; better approximation to the UHS than one ave been introduced in order to sehisi® 8 in design codes, generally by as Parameter to construct the spectrum. The 1984 Colombes ea 1985 Canadian codes second zonation map for PG' an in effect used PG short-period part of the spectrum ra PGV for the intermediate spectral ordinates. Since the zonation maps "i ic design the two parameters were different, the shape of the resulting elastic Spectrum varied from place to pl lace, reflecting the influence of sarthgual of different magnitude in controlling the hazard, The 1997 edition o! the Uniform Building Code (UBC) used two Parameters, C, and C,, for short- and intermediate. Period portions of the spectra (with the subset Fe indicating relations with ¢leration and velocity) but curiously the Hs the two parameters was the same in each zone with the result that the shaP of the spectrum did not vary except with site classification, i In the Luso-Iberian Peninsula, seismic hazard is the result of modera ‘s magnitude local earthquakes and large-magnitude earthquakes offshore iy the Atlantic. ‘The Spanish seismic code handles their relative influence By anchoring the response Spectrum to PGA but then introducing a erat Set of contours, of a factor called the ‘contribution coefficient’, K, ih controls the relative amplitude of the longer-period spectral exes ee values of K occur to reflecting the stronger influence of the larg offshore events. The simply presents and distant events. ‘The Portuguese code is an intere effectively abandons the ULIS conce return period of the individual spe OF 475 years associated with the r design codes, Within the drafting coms about how the elastic desi decision being an inelegane and almost with spectral shapes anchore from the target UFIS, how the west, i the Pt, although it is noteworthy chat the ctra is 975 years, in effect twice the a sponse spectra in most European sei mittee for EC8 there F ns were extensive discussio! 8M spectra should be constructed, with the ne anachronistic compromise oe dlonly to PGA. In order torednee the divengen ; “ver, the code introduced two different sets é Spectral shapes (For different site classes), one for the higher seismicity = : of southern Europe (ype 1) and the other for adoption in the less aa areas of northern Europe (Type 2). The ‘Type 1 spectrum is in effect anchored apncrthawakes of magnitude close to M, ~7 whereas the Type 2 s pectrum i aPPropriate to events of M55 (c.g, Rey et al., 2002). (See Figure 2. sey = any location where the dominant earthquake event underlying the ree to different from one or other ny these magnitudes, the spectrum will rem diverge from the target 475-year UMS, especially at longer periods. e ‘he importance of the vertical component of shaking in terms of t ‘I demand on structures is a subject of some debate (e.g, Papazoglou an31 Seismic hazard and earthquake actions Response Acceleration (g) 1 Period (seconds) 18 05 ordinates from the Europea ground-motion Fe 10 kam from small al. (1996) for rock sites a ey the ECB Type 1 and 2 rock spectra Fi igure 2 ure 2.12 Medi Preicon - Median predicted spectral Tre equations of Ambraseys et anchor a magnitude events, compare the median predicted PGA of structures and structural ding could be at all and » with Elna shai, 19 996) but there are certain tyP& fh the vertical loa ements, inporare Ny as cantilever beams, for whic! the ce that a Sealy codes Yo not provide a vertical spectrum ie inates redlaced by specify it as simply the horizontal spectrum or ort period ara y one-third. Near-source recordings have shown that of eatal motion, a ao the vertical direction ca" actualy exceed the con®, vertical mane it has also been clearly ¢s “plished that the shape enPonents of motion fe spectrum is very different from the horizonta rath some eee li (e.g. Bozorgnia and Campbell, 2004). In this respec ase chat through in specifying the vertical response spectrum separately or 8 the work sae of the hor SPeeam; this approach Wa prove close to f Elnashai and Papazoglou 1997). As a result at least agile ty cae ca ceeaeireresai ‘of an earthquake, the ECS vertical spectrum ¢ realistic estimation of the vertical motion the nm nany Seismic de: design codes (Figure 2-13)- ver zontal sP* nis achieved 25.2 "9:2 The i ¢ influ ence of near-surface geology o” response spectra ally experience Th ef ie fact Stro, that loca Nger cae underlain by soil deposits general 7 than rock sites during earthquakes has bech recognised 5 of earthquake effects and from logy on ground equations. The nature © road site lasses usually «by the explicit value of 7.14 shows the influence Or man Teegpai Years: ordings of wy both from field studie ground motions. The influen Mot; tions j ns is ie now : defn at Surface routinely included in predictive leposits is characterised either by (v), or el ined the nea by x. le ‘anges 2 Over dene of shear-wave velocities uppermost 30 m at the site. Figure ce of surface32° J.J. Bommer and PJ. Stafford 18 05 = Bezerania & Campden (2004) | Suvcode a cre ol foes 0.01 0.02 01 02 Period (seconds) Vertical / Horizontal Spectral Ratio \ = Rock | = Stiff soil 3 Soft soi & £ 8 8 £ = 5 & = M=65 : SS ne = 65 0 05 1 15 2 Period (seconds) . er et erche dicted spectral ordinates from the equations of Bomm al. (2003) for different sits classes 55.5 es at 10 km from strike-slip earthquakes of Ms and Ms 6.5 as indicated ally al shapes for different site classes generally sig stlect of softer soil layers, resulting in ineeed ae Sites, and the effect on the frequency content, whic! - fi fi a jiate "acceleration plateau and higher ordinates at intermedi and long response periods. The EC8 Type 1 spectra for different site classes are illustrated in Figure 2 1 toa wider constanand earthquake actions 33 Seismic hazard Site class A Site class BL Site lass C | —— Site class D [eorz_Steoaase | | | S 06 04 Spectral Acceleration (g) 02 cee ee : i: G2 2 Period (seconds) he response of k lay f the non-lin the soil, the il, whence haking. AS woked by n recent fer depends s to », Previous! sh 7 a 0 ly mentioned in Section 2.3.3, ear oy notions , mite strength eae upwards from an underlying OF higher’? Of soil ( the incoming rock motions as @ result of We ee elder Figure 2.6). The greater the shear strain in inp Ping and the lower ‘ies shear modulus of the $° roa OFthumby ne tends to be amplified far more than stronger § ‘on Bc hy soil response can be expected © be i , Bround-mo beyond 0.1-0.2 g (Beresnev and Wen, 199°) Dae es led the ae prediction equations developed for Cali eat a eating oe ee swith greater 131108 OA her (Bigg, Motions th ‘or magnitude-distance combinations resulting in wea’ re 7g than those for which stom $ 5 7 tempts to find evidence of non-linearity in the deriva y Small the lack a using European data have pot been ¢ ive PelativelY ateg (amber of Bood-quality data on site characteristics and the emt Mose kar and Beans of genuinely strong motion in the codes Don nably th ommer, 2007a: Bommer er ale, 2008). Some design oot The eeatity in eh 1997 alition of UBC, have included che effects ofS Cua Plied Fee pec ation of amplification f@ ra t org OPS and desi ication factors for rock motion’ from a few @ ates lesign regulations for snvermediate-period 5 S are compared in Figure 2.16. tical, dy tical e34° JJ. Bonner and PJ. Stafford 4 — BC 1997 Zone 1 }} UBC 1997 Zone 4 | ‘Ambraseys Simpson Bommer (1996) | ‘Akkar & Bommer (2007) | + ECB Type 1 ECB Type 2 | * Bore Joyner Fumal (1997) Amplification Factor 0 200 400 = 600800 ~—«t000~~«1200 30m Shear-Wave Velocity (m/s) ’ ifferent Figure 2.16 Amplification factors for 1.0-second spectral acceleration for oe site shear-wave velocity values relative to rock motions; for Boore er al. (1997), has been assigned a shear-wave velocity of 800 wie A number of interesting observations can be made regarding the curves in Figure 2.16, the first being the wide range of proposed amplification factors lor different sites, especially those overlain by soft soil layers. The second observation that can be made is that amplification factors assigned to bes site classes will often be rather crude approximations to those obtained for specific sites where the V_ profile is known, The UBC spectra for Zone | (low hazard) and Zone 4 (high hazard) have quite differene amplification Hacer with non-linear soil response leading to much lower soil amplification in t y high hazard zone. A similar feature seems to be captured by the ‘Type 1 an ‘Type 2 spectra from ECR. 2.5.3 Displacement response spectra In recent years, exclusively force-based approaches to seismic design pe been questioned, both because of the poor correlation bebween transi i accelerations and structural damage, and also because for Pose ta the forces effectively remain constant and damage control zeguperlimneeten of the ensuing displacements. Most of the recently introduced per ortnat based design methodologies can be classified as being Se Leni displacement modification techniques or else equivalent a aeaie FEMA-440 (ATC, 2005) presents both approaches, alieea ae pare to select the one felt to be more appropriate, acknowledging, in effect, thaSeismic hazard and earthquake actions 35 opinions are currently divided as to which is the preferred approach. EC8 also envisages the potential application of these two general approaches to the computation of displacement demand, and provides guidelines on the appropriate seismic actions in informative annexes A and B. The equivalent linearisation approach to displacement-based seismic design requires the characterisation of the design motions in the form of elastic displacement response spectra, The inelastic deformation of the structure is reflected in the longer effective period of vibration, which requires the spectral ordinates to be specified for a wider range of periods than has normally been the case in design codes. The dissipation of energy through hysteresis is modelled through an increased equivalent damping. Based on a proposal by Bommer et al. (2000), the EC8 acceleration spectrum can be transformed to a displacement spectrum by multiplying, the ordinates by T?/4n’, where T is the natural period of vibration. The critical question is at which period should the constant displacement plateau begin, which, as can be discerned in Figure 2.17, was set at 2 seconds for the Type 1 spectrum in EC8. This value has since been recognised to be excessively small; the corner period of the spectrum increases with earthquake magnitude, and for the larger events expected in Europe (M ~7) the period could be expected to be in the order of 10 seconds (e.g. Bommer and Pinho, 2006). The inadequacy of the corner period, T,,, being set at 2 seconds has recently been demonstrated by new European equations for the prediction of response spectral ordinates up to 4 seconds (Akkar and Bommer, 2007b). Figure 2.17 compares the displacement spectra from EC8 with those from Akkar and Bommer (2007) 30 = Me | = oi | | 10 Spectral Displacement (cm) Period (seconds) Figure 2.17 Comparison of $%-damped displacement response spectra for a stiff soil site at 10 km from earthquakes of different magnitudes from Akkar and Bommer (2007b) with the EC8 Type 1 spectra for the same conditions, anchored to the PGA value predicted by the equation presented in the same study36 I} Bommer and py, Stafford ' each for stiff soil sites at 10 km from, earthquakes of different a 7 case, the EC8 spectra have been anchored to the predicted me cevestindl from the equation of Akkar and Bommer (2007b). A number of i hace observations can be made, the first being that the fixed spectra the result EC8 is un fad tt influence of varying magnitude, seal fr te that the short-period jhectral ordinates are severely ore Peo sinaller magnitudes. The secon | observation is that the fixed COE OF 2 seconds is clearly inadequate and the dependence of this Pee 6, the magnitude is very clears for carthquakes of greater than Tne eee a Gomer period is longer than 2 Seconds, and for the larger events great 4 seconds, inal The spectral ordinates with damping. ratios higher than the oes af 5 percent of critical are obtained by multiplying the spectral oe is intermediate periods by factor, detived by Bonner cy at oa rocctl 8 function only of the targes damping level. These factors replaced ¢ ed in an early draft of EC8, and many other factors have since been prep005) the literature and in ocha sign codes. Bommer and Mendis ae explored the difference, aOnest the various factors and found # mata amount of reduction of the $ Percent-damped ordinates required to are the ordinates amping levels increases with the duration of the Bround motion. Since the Type 2 spectrum in ECR pomesponds 16 Yat small magnitude earthquakes, which will Senerate motions of short dura ’ in spectral * Was proposed that the existing sealing equation in EC8 to obtain sp! displacements, SD, and different damping values, e: SDE) should be retai this should be rept; Bommer (2006): SD(E) 2.6) SD(S%) Ev ( SHE ai ct “trum, whereas for the Type 2 ee Wing expression derived by Meni cm 2.7) SD(S%). 304E ( 2.6 Acceleration time-histories : ich Although seismic de ign invariably begins wich methods of analysis in _ the earthquake Actions are represented in the form of response nies 2 Some situations require fully dynamic analyses to be performed and in es Cases the earthquake actions must be represented in the form of accclera Hop time-histories, Such Situations include the design of safety-critical ee highly irregular buildings, bas “isolated structures, and structures designe fora high degree of ductility. For such Projects, the simulation of eee response using a scaled elastic response spectrum isnot considered ie and suites of celerograms are required for the dynamic analyses. nél guidance given in the Malority of seismic design coder on the selection a y ofrd and earthquake actions 37 such purposes is either it practically impossible Seismic haza sealin 5 of suites Si time-histories for ate or else so prescriptive as to make to ide ntify realisti and Ruggeri eye accelerograms that meet the specified criteria (Bommer » 2002). A point that cannot be emphasised too strongly is that - aniform hazard spectrum, but time-historj rather Fie should never be matched 10 the case of codes thi corresponding to a particular earchauale scenario. In ‘loin viifficalt since the code generally provides an Bene de ee the UHS and offers no possiblity £0 ste aoe event-specific spectrum. : ore histories f umber of options for obtaining sul i Spectrum com dynamic analysis of seructuress including the generation Ss now wid ee accelerograms from white noise, 4 method that ii inappropriace because the resulting signals are teal cee oaks aroma maptiony Te ermost popular option is 10 use nse cee which can be selected either on the basis of having Ise matching ay mmilar, at least in shape, to the elastic design spectrum, OF tistance a Ae an earthquake scenario in terms of magnitude, soure o-site latter appro, Possibly also site geology (Bommer and ‘Acevedo, 2004). The Sign core neweven is enerally not feasible in the context of seismic can cod applications, because informanen garding the wnderiing cin cathe he oe Traually not available 0 the seh Selecting records es hiloet an oF appropriate magnitude is ryan ise fhe duration of Seismic d considered an important parameter in determining the degree ebate in the see ne ve pords impose, which i" Fer ‘of ongoing | Once he technical literature (Hancock and Bommer 2008). ey Pectral shane, of records has been selected, whether on the basis & the “Ngineer iene or an earthquake scenario, the next question for the design tienen eee vecords are needed: Most of the seismic SES recorae that address this issue, including ECS, cpecify that a minima vine should be used, and that if less than 7 records are used ceo the ‘ r more tructural response must be used as the basis for design, whereas” CAN be weg pre histories are employed then the averae® ‘ructural response namie Sts The use of the maxim en ee response obese from Will in co may never be appropriate since the input accelerograms Spectr me sense have been adjusted to approxima’ to the elastic design already 1 which, if determined from 2 probabilistic hazard assessment Wl Geet include the influence of the ground-morion viability. The largest bly correspond £0 @ cord that is 80 nse the of acceleration time. ai aic ore eee will proba Netefore hee spectrum, and in a se! agate key ing taken into account twice. Able es ‘question then becomes how many records are requires | i pane of the mean inelastic resP Il depend © thos, Is a : . those of the es adjusted so that their spe atch the t, elastic design spectrum: the more closely # arget elastic design spectu™ the fewel38° J. Bommer and PJ. Stafford —— Linearly Seated Target Spectum — spectrally Mate Spectral Displacement (em) Oc cenee. on 02 0304 2 Petiog (seconds) Period (seconds) rH? 2.18 Comparison of the differe Modified from Hancock er 4 (2006) spectra ‘nee between scaled and matched sp Options include natural Period of the structu ordinates ove e records to match the design spectrum ie re or scaling to match or exceed the ee ‘'nge around this value, the extended ae accounting for | ntributions to the response from higher ae i and also for the elongation of "esponse period due to inelastic ae ae Scaling the records in amplitude is legitimate given that whilst the ampli few of the motion is highly dependent on distance especially within oe fens of kilometres from the soumen the shape of the response spectrut za actually rather insensitive to distance over the range of distances of aie engineering interest (Bommer and Acevedo, 2004). Although scaling li he of a factor of 2 were Propoted at one time, and became enbedded i ‘folklore’ of engineering practice, much larger scaling factors can be app} ing the (Watson-1 amprey and Abrahamson, 2006). Adjusting records by scaling time axis, however, is to be avoided, An alternative to linear scaling of the records j Fast Fourier Transform or wavelet transformati shape that approximates to thy Acevedo, 2004). The transform, a period to make adjustments, asin site achieved specs, get design spectrum (Bommer pe ‘ay to achieve this is using the wave! m1 ses the alteration of the original a noe but ar the same time can achieve a very good Spectral match (Hancock et al d 2006). An example of the difference between linearly scaling a record an matching spectra through wavelet transformations is given in Figure 2.18. Most elegant w; ation, which minimi 2.7 Conclusions and recommendations For most engineering be represented by an for inelastic deform ji i ing can Projects in seismic zones, the earthquake loading ¢: 2 accour acceleration respon: ectrum, modified to accou il ‘ation of the structure, The elastic design spectrum w:39 rthquake actions nich provides ; istic seismic hazard analysis, reais ies the most Tate framework for handling the large Pelion. Most seismic neh te gels for seismicity and ground-mowm™ associat lerived probabilisicaly pods present zonation maps and response Smosited witha rerun per even though these design Toads are often cothe resulting iset et whose origin is 2 fairly aebitrary selection, ; a if a uniform ioe eee is generally a poor approximation to the ain ard spectrum. lading is alo ee femic codes offer the engineer to bypass the very considerable effort, exp ense speense and ti nuld ime re : é eee gure for a full site-specific hazard assessment. This be interpreted to mean that the engineer should not be jvation and preseneation of the aware of earthquake assumptions underlying the der Sis ‘ae as well as their limitations. gine de rather fee seismic design codes in tha produits design. ma complet se ve jefinitions of earebauake Map sho member state of the European Union wi re Showing PGA values ae yocument, including @ se specifi Pe 2 spectrum jor the 475-year returt period, select either the Se otation Ml die ulees and, if considered appropriat®> adapt decals of the cre td Purpose of asses and spectral parametets- Interestingly, although Could well be j EC8 is harmonisation of seismic design across Europes jumps in the level of seismic desi8 Joads acro re ers : zona ee ss national ati rentl j Nee : ion ly there is no official project fora ‘community-wide hazard nm Althouen to ough t the he re are a number of innovativ' definy «. Pecificati davition of ee of design earthquake actions cemetit-baied doc response spectrum and the prov inno: ttontal elas design approaches, the basic mechani the yauons in oe design spectrum is ‘outdated and sigh CUS Its to be ant Sali tn Js es of the worlds MEY notably Contry out 5 yea pet that the first major revision of EC8, which should Pilon tectini ain after its initial introduction, will modify the spectral nique, incorporating at least one mo! g parameter ‘rable, including in ada to th, uo” to PG, t belong priod orton of he snodiications ate tthe’ Fea eee srtjcement specu a th Testa BR seist higher than 5 percent of critical. ‘ani mic codes provide useful guidance for the there are cases Where the code following: Specific, mnples may include the Seismic hazard and ea’ Most fre, which quently be obtained through probabil in terms of earthquake ly a cemplate e actions for II have t rit is actual C its oF i wn National Applicatio 8 with regards gs the separate of input for sm for defining vficantly behind e features 0 such jsion fo t desi an ign of many structures, ation 'S Wi ill not be sufficient. Exa” for which nearsourey Proj Tojects | Hects located i sa ‘ nrectivity ae in proximity to active faults fects associated with the fault rupcure need to be considered jdered in the 97 edition of UBC nth e desi sign (5 Ut Not in aia effects are consid40 JJ. Bommer and PJ. Stafford Projects in areas where active faults are and for which surface displ, for the performance of the Projects on sites with deep effects of the near-surf, be well captured by th shapes in the code; Projects for which return periods si 475 years are considered any project for which ful guidelines on Preparing in many respe ent, known or suspected to be eer ‘acements would be a critical considerat structure; : : the and/or very soft soil deposits, for vn i ‘ace geology on the ground motions are unlil el ¢ simplified site classes and corresponding spe inal ignificantly longer than the nom appropriate; 2 i : EC8 ly dynamic analysis is required (since the time-history input for such analyses is lac If it is judged that then this needs to be considered budgeted a site-specific seismic haz planned careful 4s an integral part of the cordingly. If investigation Part of the assessment, then the increase very significantly, ! Seismic hazard analysis is a highly specialised discipline that is constantly evolving and advancing, and in which 4 great deal of expert pudgernee 'S required. Nowadays it is fairly straightforward to obtain Beslene maps, satellite imagery, earthquake catalogues, published sound mone Prediction equations and software fox Performing hazard calculations, i 4 many cases from the Internet and free of charge. The art of seismic hazard analysis, however, lies not primarily in accessing and analysing. thes resources but rather in judging their completeness and quality, and ass me the uncertainties associated with the data and the applicability of models t the specific region and site under consideration, ‘ard assessment is required, ly and in good time — it should d site investigation, and scheduled oe S of active geological faults are to ag time and budget requirements are likely t References Abrahamson, NLA, hazard analy: Zonation, Pal (2000) “Effects of rupture » Proceedings of the Im Springs, CA, Abrahamson, N.A. (2006) Seismic hay and future developments’, and Seismology, Geneva, Switzerland, Abrahamson, N.A. and Bommen, J.J. (2005) ‘Probability hazard analysis’, Earthquake Spectra, 21.60, 3-607. tral atrenuation Abrahamson, N.A. and Silva, W] (1997) ‘Empirical response spectral attenuat relations for shallow crust, A Al earthquakes’, Seismological Research Letters, 68: 127. Abrahamson, N.A. and Silva, W, relations for the c seismic directivity on probabilistic sciamic Seisr 6th International Conference on Seis current practice ard assessment: problems with current pra ingineering First European Conference on Earthquake Engineering ee and uncertainty in seismi n 2.007) ‘Abrahamson & Silva NGA ground 7 Beometric mean horizontal component of peak and specearthquake actions 41 t Generation Attenuation +, Richmond, CA. anging wall and Bulletin Seismic hazard and 8tound ca rahams Sy 5 acs Motion. models, aa Interim Reports of Nex! ny NA cn Earthquake Engineering Research Cente re ground pees PG. (1996) ‘Bifects of the h a 'eismological ae recorded during the Northridge eartl Bie ae of America, 86:893-899- Ime Using the ra ae TUR, (1992) ‘A stable algorith by rcs $2:505-510 ont eifects model’, Bulletin of the Seismo TnsoRs NAA., Birkha ee, ad Ron B0t B, Koller, M- Mayer-Rosa, D., Smit, P Sprecher, C» i “onferesssessment fae 2) _PEGASOS: a comprehensive probabilistic seismic ka, § ee On Earth ninclear power plants in ec ertand’, 12th Europea spect Bomer, quake Engineering, London, U-K- Bsa diguen (2006) ‘Influence of long-Per ‘Akar, i 6s. Earthquake Engineering . and Bown mer, J.J. (2007a) ‘New empirical predict equations for peak Hh from Europe and the Middle quake’, m for regression logical Society OF jod filrer cut-off on elastic nd Structural ‘Dynamics, Bround velo} reutd velocity deri Ag» Bulierin eran from strong-motion recor ¢ Seismological Society of America 97:511-930- mf elastic displacement response nd Structural Spectgg nd Bomme: if ca Be ‘Prediction Way ts 36:1275-1 the Middle Easy Sarthquatke Es NS 8: and Ore Fr re mgt’ for oe cpffect of peak ground tbraseyys aos systems’, Earthquake Engineering and 5! i LN., Si impson, K.A. and Bommer, Ja. (1996) The prediction of ring anc Struct ering an yelocity 07 deformation ruclueral Dynanics vtizontal Ynarntal res ‘bees NIN a a in Europe’, Earthquake Bnsi eS, NIN. . imation of sone Ja, Sarma, $.K, and Smit pt, (2005) “Eawations for aqhPeetal Europe and ti ground motions from shallow vceustal earthquakes using, C ptt acceleration’ the Middle East: Horizontal peak ground saecelerarion a ecw 5) Improve * Bulletin of Earthquake Engineers 3:1-53- ker, ype CAE Peon rontinear static seismic analyse procedures pEMA-4405 Joye CA? Applied Technology Counc silon and record selection’, PIM. and Bane tithe Pe ne seen spectral shapes &P ites »P and ae Structural Dynamics 1077-1095: vara logical Soci GAA, (1999) ‘Disaggregation Of seismic hazards » Band Cor iety of America, 89:901 5 ao Aegon) ‘vector-valued Bro .S. Bulletin of seismic hazard abilistic Conference om National anal nalysis (Vy PSHAY, rt fl Dereon Make Ey AY’, Proceedings of the Seven b the LA. ahd Wee oe (ONCE), Bosto”, a “yer, ismological Sects (1996) ‘Nonlinear soi response-@ reality? Bulletin of alee ee iy of America, 86:1964-1978- vg Otion? ruben ‘Relationships betwee” smedian values berween ig. NON Balin of J 2006) "Relations of the Hows csamponet Ol 4 in of the Seismological Society °! ica, 95 2-1522. ye ground air \ and Bo. mmer, J.J. (2004) “Earthquake los lag, ‘since ing Geology, 75:147-179.