Vous êtes sur la page 1sur 17

EARTHQUAKE ENGINEERING AND STRUCTURAL DYNAMICS Earthquake Engng Struct. Dyn.

2006; 35:3955 Published online 30 September 2005 in Wiley InterScience (www.interscience.wiley.com). DOI: 10.1002/eqe.530

Adapting earthquake actions in Eurocode 8 for performance-based seismic design


Julian J. Bommer1;; and Rui Pinho2
1 Department

of Civil and Environmental Engineering; Imperial College London; SW7 2AZ; U.K. 2 ROSE School; c/o EUCENTRE; Via Ferrata 1; 27100 Pavia; Italy

SUMMARY Performance-based seismic design (PBSD) can be considered as the coupling of expected levels of ground motion with desired levels of structural performance, with the objective of achieving greater control over earthquake-induced losses. Eurocode 8 (EC8) already envisages two design levels of motion, for no collapse and damage limitation performance targets, anchored to recommended return periods of 475 and 95 years, respectively. For PBSD the earthquake actions need to be presented in ways that are appropriate to the estimation of inelastic displacements, since these provide an e ective control on damage at di erent limit states. The adequacy of current earthquake actions in EC8 are reviewed from this perspective and areas requiring additional development are identied. The implications of these representations of the seismic loads, in terms of mapping and zonation, are discussed. The current practice of dening the loading levels on the basis of the pre-selected return periods is challenged, and ideas are discussed for calibrating the loading-performance levels for design on the basis of quantitative earthquake loss estimation. Copyright ? 2005 John Wiley & Sons, Ltd.
KEY WORDS:

performance-based seismic design; earthquake actions; return periods; design levels; loss modelling; Eurocode 8

1. INTRODUCTION Performance-based seismic design (PBSD) is essentially the formalization of the often cited objectives of designing structures to withstand minor or frequent earthquake shaking without damage, moderate levels of shaking with only non-structural damage and severe shaking without collapse and a threat to life safety [1]. In the Vision 2000 document [2] this is elegantly stated as the coupling of expected performance level with expected levels of seismic ground motions. Motivation for formalizing performance-based design criteria and objectives has been triggered to a large degree by the fact that in highly developed regions
to: J. J. Bommer, Department of Civil and Environmental Engineering, Imperial College London, SW7 2AZ, U.K. E-mail: j.bommer@imperial.ac.uk
Correspondence

Copyright ? 2005 John Wiley & Sons, Ltd.

Received 1 November 2004 Revised 3 May 2005 Accepted 10 May 2005

40

J. J. BOMMER AND R. PINHO

such as Japan and California, the life-safety objective of earthquake engineering has been achieved to a considerable extent but economic losses in recent earthquakes, most notably Northridge in 1994 and Kobe in 1995, have been enormous. The general principle of PBSD, however, is of relevance to all seismically active regions since the dual objectives of protecting lives and avoiding losses are relevant in all settings: although not comparable to the losses from Kobe and Northridge in absolute terms, the losses caused by earthquakes such as the 1999 Kocaeli earthquake in Turkey have had a devastating impact on developing economies. For the implementation of PBSD to have a major benecial impact on any society, it must be incorporated into the specications and guidelines embodied in regulations or codes for seismic design of buildings. In a very approximate manner, PBSD is embodied in most current design codes through the use of importance factors that increase the design loads for structures required to perform above simple life-safety criteria under the 475-year ground motions. The next generation of seismic design codes is expected to incorporate, to some degree, the principles of PBSD. An issue that will arise is the degree to which this can be achieved through adaptation of current codes formats or if the implementation in PBSD will require a radical change in the way earthquake actions are specied in seismic design codes. This paper addresses the issue of dening suitable actions for PBSD with specic reference to Eurocode 8 (EC8). The pan-European seismic design code will become the exclusive seismic design standard throughout the European Community, an area encompassing regions of low and high seismicity, after 2010. Although many of the issues addressed are equally applicable to other seismic design codes around the world, the paper focuses on EC8 partly because of the importance it will assume in Europe in the near future but also because it could be argued to have moved more in the direction of PBSD of new structures than any other seismic code for new construction. The purpose of this paper is not to prescribe the formulation and values to be used to dene earthquake actions for PBSD in EC8, but rather to identify where additional work is required for such implementation in future revisions of the code. There are two important aspects of EC8 that make it particularly amenable to modications that could make it more compliant with the principles of PBSD. Firstly, each member country of the European Community will adapt the code through a National Annex in which, as is highlighted a number of times in this paper, national code authorities may dene most aspects of the seismic loading as they wish, e ectively using the formulations and recommended values in the main body of the code as guidelines. Secondly, it is envisaged from the outset that EC8 should not be static: the need for updating, revision and completion is strongly recognized so that an improved second generation of EN Eurocodes can evolve [3]. The main body of the paper is made up of two sections, rst, addressing the issue of formats for dening earthquake actions for PBSD and specically exploring the extent to which EC8 currently meets these requirements. Options for mapping suitable parameters to facilitate the denition of suitable earthquake actions in the code are then discussed. The second major section of the paper addresses the critical issue of selecting the design levels to be specied in the code and discusses the options for calibrating these through the use of earthquake loss modelling. The paper concludes with a brief discussion of possibilities and priorities for implementation of the modications identied as necessary or desirable.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

41

2. EARTHQUAKE ACTIONS FOR PERFORMANCE-BASED SEISMIC DESIGN A key element of the implementation of PBSD is the denition of seismic design actions for multiple design levels and in formats that are closely related to the structural and nonstructural damage that the PBSD framework specically aims to control. In this section, the requirements regarding the presentation of earthquake actions for PBSD in seismic design codes are addressed with specic reference to Eurocode 8. Since seismic design and assessment methods based on non-linear time-history analysis are still of relatively limited applicability in routine design o ce practice, discussion on issues such as selection and scaling of natural records and the generation of articial or synthetic accelerograms is not included here. In other words, the subsequent discussion on earthquake actions focuses almost exclusively on its response spectra representation, widely used for the majority of practical applications. The prestandard version of Eurocode 8 [4] has undergone several modications and iterations, and the European standard [5] is signicantly di erent in a number of respects, particularly in the denition of seismic actions for design. In this paper, only the latter edition of the code is considered. 2.1. Displacement vs acceleration seismic input The inadequacy of exclusively force-based approaches to seismic design as a means of controlling damage is widely acknowledged, due both to the poor correlation between transient accelerations and structural damage, and also to the fact that for post-yield response, forces are essentially constant and deformations control the degree of structural degradation and, ultimately, stability. As a consequence, and starting with the work of Moehle [6], a significant number of researchers have developed and proposed over the recent years seismic design=assessment methods that place a focus on controlling displacements=deformations that is at least equal to the emphasis placed on ensuring force=stress capacity. Sullivan et al. [7] provides an extensive, although not exhaustive, list of such proposals classifying them as displacement-based approaches due to the fact that, as previously stated, displacements=deformations become primary design targets, rather than merely secondary nal design checks. Herein, they are instead termed performance-based design methods, re ecting their necessarily hybrid forcedisplacement characteristics. The vast majority of such recently introduced performance-based design methodologies can be categorized as either an (i) equivalent linearization or (ii) displacement modication technique, as thoroughly and clearly described in the recently released FEMA-440 report [8]. In short, equivalent linearization techniques are based on the assumption that the maximum total displacement (elastic plus inelastic) of a SDOF oscillator (which attempts to represent the response of a given structure) can be estimated by the elastic response of an oscillator with an e ective period and an equivalent damping computed as a function of the ductility demand. Alternatively, displacement modication procedures estimate the total maximum displacement of the oscillator by multiplying the elastic response, assuming initial linear properties and damping, by one or more coe cients, the latter being typically derived empirically from series of non-linear response-history analyses of oscillators with varying periods and strengths [8]. Whilst such performance-based design approaches do still present a number of shortcomings, some of which can be readily overcome by the enhancements suggested in FEMA-440, they constitute nonetheless a clear improvement over exclusively force-based approaches, for which
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

42

J. J. BOMMER AND R. PINHO

reason guidance documents have been gradually introducing these to the practicing engineering community over the past few years. Since there is so far no consensus regarding which of the two possible design approaches (equivalent linearization or displacement modication techniques) is to be preferred, there is currently a trend for both to be considered. This leads to the need for the seismic input to be dened in ways that satisfy the requirements of both approaches, as discussed in Section 2.3. 2.2. Acceleration response spectra in EC8 In the majority of current seismic design codes around the world the elastic response spectrum of acceleration is constructed by anchoring a spectral shape dened for each site class to the design peak ground acceleration (PGA). Apart from the lack of geophysical or engineering signicance of PGA, this approach has the signicant drawback that the shape of the response spectrum changes only with the site class, even though it is well established that the spectral shape is strongly in uenced by earthquake magnitude and, to a much lesser extent, by sourceto-site distance. As a result, the spectrum will often not be of uniform hazard [9]. The approximation to the uniform hazard spectrum (UHS) is improved by using two independent parameters to build the spectrum: examples include the use of two hazard maps, one for an acceleration-related parameter and the other a velocity-related parameter, as in the 1984 Colombian code and the 1995 Canadian code [10]. In IBC 2000 [11], the UHS is constructed using the NEHRP approach of maps of spectral acceleration at 0.2 and 1.0 seconds. A less elegantand indeed less e ectiveapproach has been adopted in EC8 [5], prompted by resistance to facing engineers with a second ground-motion parameter: two spectral shapes are presented, one recommended for regions only a ected by earthquakes of magnitude Ms 5:5 and the other for regions a ected by larger events, and each country will adopt the most appropriate shape in its National Annex, which may be either the Type 1 or 2 spectrum or another shape chosen by the National Authority. Figure 1 compares the two spectral shapes with median spectral ordinates for earthquakes of di erent magnitude. The gure illustrates that each spectral shape will approximate that expected for a particular earthquake magnitude but then tend to over- and under-estimate the longer period ordinates for smaller and larger events, respectively. The ability to choose the more appropriate of the two spectral shapes is clearly an improvement on having a single shape for areas of high and low seismicity, but since the practice of anchoring the spectrum to more than one parameter is now well established in other codes, the EC8 solution can only be viewed as a rather unsatisfactory compromise. An important point to note is that EC8 does not only allow each country to select between the Type 1 and Type 2 spectral shapes, but also the values for the three corner periods that dene the spectrum and the soil factors (S) applied to entire period range of the spectrum. In common with most current codes, the site classes dened in EC8 are based on the average shear-wave velocity over the uppermost 30 m, a depth related to the cost of drilling boreholes rather than of any geophysical signicance. Although EC8 denes site classes based on Vs30 values, the code does state that the in uence of local ground conditions may also be done by additionally taking into account the in uence of the deep geology on the seismic actions [5]. However, the code currently provides no indications of how deeper geological structure could be classied or how the spectral ordinates will be in uenced by such classes; this is another area warranting further research.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

43

3 Normalized Spectral Acceleration Type 2 Spectrum 2.5 2 1.5 1 0.5 0 Ms 5.5 Ms 6.0

Ms 6.5

0.5

1 Period (seconds) Type 1 Spectrum

1.5

3 Normalized Spectral Acceleration 2.5 2 Ms 7.0 1.5 1 Ms 6.5 0.5 0 Ms 7.5

0.5

1 Period (seconds)

1.5

Figure 1. Type 1 and 2 spectra from EC8, anchored to the median PGA estimated for rock sites at 10 km from earthquakes of di erent magnitudes obtained from the equations of Ambraseys et al. [12], compared with the median spectra obtained from these equations after smoothing the coe cients.

For regions a ected by signicantly di erent sources of seismicity, such as major subduction zones and moderate magnitude crustal earthquakes, the variation of the bedrock spectral shape across a country may be pronounced. In the 1994 Spanish code [10], the variation is included via a contribution coe cient, Kmapped as contours on top of the basic zonation map that represents the degree of in uence of large o shore earthquakes and raises the longperiod spectral ordinates accordingly. Neighboring Portugal adopts a more radical approach, dening two separate spectraone for local events and one for large, distant earthquakes each of which must be considered separately in design [10]. The 1989 Chinese code [10] also denes separate spectral shapes for near- and far-eld earthquakes. EC8 also envisages such a possibility: When the earthquakes a ecting a site are generated by widely di ering sources, the possibility of using more than one shape of spectra should be contemplated to adequately represent the design seismic action. In such circumstances, di erent values of PGA will normally be required for each type of spectrum and earthquake [5]. An important aspect
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

44

J. J. BOMMER AND R. PINHO

of these approaches of dening separate response spectra for di erent sources of seismicity is that they represent a departure from the principle of the uniform hazard spectrum, and this challenges the primordial importance that is often assigned to the selected return periods for which the UHS is derived. 2.3. Computation of displacement demand in EC8 As described above, the two main performance-based design=assessment approaches used in practice require two distinct seismic input denitions. Methods based on the equivalent linearization technique make use of a response spectrum that must feature spectral values dened over a wide range of periods (since the e ective, rather than initial, period is employed) and dened for a wide range of damping levels (since e ective damping is employed to represent the energy dissipation capacity of the structure at di erent levels of ductility demand). Methodologies based on the displacement modication approach, on the other hand, make use of an elastic response spectrum, the ordinates of which are then multiplied by a number of coe cients so that the expected inelastic displacement demand can be obtained from its elastic counterpart. In Eurocode 8, these two di erent approaches for determining displacement demand are already introduced, albeit in informative annexes rather than in the body of the code. Annex B gives rules for computing inelastic displacement demand from elastic response ordinates, as required by the N2 method [13], a displacement modication approach that has been adopted in Eurocode 8. Annex A, on the other hand, provides a displacement demand spectrum that was derived with application to equivalent linearization methods in mind, although currently there is no guidance at all on such design and assessment methods in the code; the only reference made to Annex A is a note regarding the prediction of spectral ordinates for periods greater than 4 s. In order to emulate the approach of FEMA 440, in which both design methods are presented as options for the engineer, future revision of the code will have to present an equivalent linearization approach (e.g. Reference [14]). For both approaches, the starting point for dening the earthquake actions is the 5%-damped elastic displacement response spectrum and a key issue is the control period TD that marks the start of the constant displacement plateau (Figure 2). The value of TD in Eurocode 8 was xed at 2 s for the Type 1 spectrum applicable in areas of high seismicity, based mainly on the work of Tolis and Faccioli [15] that made use of digital accelerograms from the 1995 Kobe earthquake, ltered at 20 s in most cases. The average spectral shape from these recordings did imply that the constant displacement plateau begins at a period of around 2:0 s for the recordings obtained at short distances, although the plateau began at periods between 4 and 5 s for recordings from intermediate distances, which were mainly on soft soils. In more recent work, Faccioli et al. [16] again present displacement response spectra with the corner period at 2 s although these authors note that the dependence on magnitude of the TD control period dening the beginning of the plateau of the elastic displacement spectrum should be carefully considered, especially for large magnitude earthquakes. This statement has become all the more pertinent following the publication of the 2003 NEHRP guidelines [17] in which a constant displacement plateau has been introduced at a control period designated TL , which is equivalent to TD in Eurocode 8. Values of TL are mapped for the U.S., ranging from 4 to 16 s depending on the dominant magnitude of earthquake in each region; these values are based on seismological theory and digital recordings of large earthquakes. Seismological theory dictates
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

45

= 5% 10% Spectral Displacement 20% 30% PGD

TBTC

TD

TE Period (s)

TF

Figure 2. Displacement response spectra derived from Eurocode 8.

that the period at which the constant displacement plateau begins, TD , increases exponentially with earthquake magnitude (e.g. Reference [18]). The derivation of displacement spectra for seismic design by Bommer and Elnashai [19] included a variation of this control period with magnitude but the dependence was severely underestimated as a result of their analysis being based entirely on analogue accelerograms from which reliable spectral ordinates at periods even as long as 4 or 5 s cannot generally be obtained (e.g. Reference [20]). This is clearly an area in which additional research is urgently needed and in writing National Annexes for EC8, national authorities might consider using larger values of TD than are currently specied in the code. A key issue in predicting long-period displacement spectra is the role of the sub-surface geology. It has been pointed out that for long-period ground motions, the wavelengths are such that the shaking is not strongly in uenced by the properties of the uppermost 30 m (e.g. Reference [21]). Particularly for the specication of long-period displacement spectral ordinates, Vs30 is likely to be found wanting as the basic parameter for site classication. This has led to some ground-motion prediction equations including the depth to basement rock, which boosts long period motions [22], as an explanatory variable. As noted in the previous section, EC8 provides for the deeper geological structure to be included in the denition of the site classes and thereby o ers the possibility to address this requirement. For the specic case of equivalent linearization methods, guidance needs to be provided on calculating equivalent damping values. However, the issue of equivalent viscous damping to model the e ects of hysteretic energy dissipation is complex. Relationships between ductility and equivalent damping have commonly been derived using di erent hysteretic models; work on this topic is ongoing, exploring the variation of ductilitydamping relationships with e ective period (e.g. References [2325]). In Eurocode 8, spectral displacements for e ective damping values higher than 5% of critical are obtained by applying scaling factors that are dened in terms of the damping ratio and which vary with period, being constant from the
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

46

J. J. BOMMER AND R. PINHO

beginning of the constant acceleration plateau to the end of the constant displacement plateau (Figure 2). Recent work by Bommer and Mendis [26] showed that for e ective damping values of 20% and higher, the scaling factors demonstrate a strong dependence on the duration of motion, which is manifest as a decrease of the scaling factors with increasing magnitude and distance. These ndings suggest that the scaling factor currently presented in EC8 may represent an excessive simplication. With regards to the specic case of the displacement modication design approach prescribed in EC8, a number of developments are also needed since the in uence that factors such as pinched hysteretic shape, sti ness degradation, strength deterioration and dynamic Pe ects have on the actual inelastic response of structures is currently not accounted for in the computation of seismic demand, described in Annex B of EC8. To this end, it is believed that the pertinent improvements suggested in FEMA-440 [8], where the e ects listed above, together with others (e.g. inelastic response of short period structures), have been examined and=or revisited, need to be somehow taken onboard and introduced in EC8, with obvious adaptation=revision where necessary. 2.4. Hazard mapping for PBSD For site-specic seismic hazard analysis, it can be expected that the ground motions for design may be specied in considerable detail and re ect the location of the site with respect to active geological faults in the immediate region as well as the shallow geological prole at the site. Furthermore, the design ground motions may be entirely consistent with the selected basis for the design as specied by an annual frequency of exceedance (the reciprocal of which yields the return period). For code applications, simplications are inevitable since detailed site-specic information cannot be included in the denition of the earthquake motions and the in uence of many continuous variables such as bedrock accelerations and site amplication will be represented by a series of discrete zones or classes. Therefore, some degree of approximation of either the return period or the corresponding loadsand in most cases both of these quantitiesis inevitable. Therefore, di erent options for presenting the earthquake actions in design codes should be judged relative to one another rather than with respect to what might be attainable for the site-specic hazard assessment. Somerville [27] points out that seismic hazard curves are ideally suited to PBSD because they specify the ground motions that are expected to occur for a range of di erent annual probabilities (or return periods) that correspond to di erent performance objectives. If it were found that the hazard curves for a given ground-motion parameter were of the same shape at all sites in a country, then it would be su cient to provide 475-year maps of the selected ground-motion parameters and an equation to scale these amplitudes to other return periods. Since the shape of the hazard curves will often vary throughout a country, advantage can be taken of the approximately linear relationship (over the range of return periods of usual engineering interest) between the logarithms of exceedance frequency and of the ground-motion parameter to determine two coe cients that relate them; from maps of these two coe cients the value of the ground-motion parameter can be determined for any return period [28]. This elegant approach would allow current code approaches such as that in IBC 2000 to cover all return periods of interest with just four hazard maps: two for the short-period spectral acceleration and two for the 1.0-s ordinate.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

47

The premise of Eurocode 8 that hazard should only be mapped in terms of a single groundmotion parameter is clearly anachronistic, although one possible defense of this decision is that in the current seismic design codes that Eurocode 8 will replace, more than half of the seismic hazard zonation maps are based on macroseismic intensity [29]. However, seismic hazard mapping by seismologists in nearly every country is far more advanced and it may safely be assumed that the rst revision of Eurocode 8, due a few years after it comes into e ect, will move beyond the insistence on mapping only PGA and in the process abandon the unsatisfactory compromise of the Type 1 and 2 spectral shapes. For PBSD, assuming that it will involve some form of design requiring displacements as input, the denition of long-period spectral ordinates will require at least a third parameter to be mapped: one option is the corner period that denes the constant displacement plateau as mapped in the 2003 NEHRP guidelines; another is that proposed by Bommer et al. [30] of mapping PGA, PGV and PGD and then dening the corner periods from the ratios of these parameters. Using the method of Grases et al. [28] this leads to six maps, whereas if four return periods are to be consideredas recommended in Vision 2000and mapped individually, then the code would include 12 separate maps, which starts to be cumbersome (although this could be overcome to an extent by embedding the maps within a GIS provided on CD-ROM with the code). The situation becomes more complicated if maps for other parameters, such as independently dened vertical motions and strong-motion duration, are also needed. 2.5. Seismic actions and earthquake scenarios Disaggregation of seismic hazard (e.g. References [31, 32]) allows the contributions made to the hazard at a site by di erent magnitudedistance pairs to be identied. In the U.S.A. there is the unique situation wherein the U.S. Geological Survey operates a web site that allows users to enter the co-ordinates of the site and obtain disaggregations of the hazard at selected response and return periods; since the hazard maps and UHS in the seismic design codes are very closely related to the national hazard maps of the USGS, this e ectively allows users of the code to identify hazard-consistent scenarios. The controlling magnitudedistance pairs identied for horizontal spectral ordinates at the structural period of interest and for the selected return period can be used to estimate other parameters such as the strong-motion duration. This approach is preferable to producing separate maps for these other parameters since the horizontal spectral accelerations and the strong-motion duration, for a given return period, will often be controlled by di erent earthquake scenarios. This would imply that the motions were not compatible hence the use of multiple maps could result in the elastic spectral ordinates controlled by one source of seismicity being reduced by a duration-related ductility or damping factor controlled by another. From the preceding discussion it is clear that the basic information required for all representations of the earthquake ground-motions to be considered in design is the magnitude and distance of the dominant earthquake scenario. For most current design codes, however, there is no facility for performing disaggregation and the controlling earthquake scenarios are e ectively invisible to the design engineer. Providing insight into dominant earthquake scenarios not only enables comprehensive denition of the ground motions but can also facilitate communication regarding seismic risk [33]. An obstacle to adopting alternative formats for presenting the seismic actions on the basis of earthquake scenarios is the insistence on adhering to the total probability theorem as embodied in the UHS, i.e. that the most important
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

48

J. J. BOMMER AND R. PINHO

factor is the frequency of exceedance of individual ground-motion parameters, regardless of the source of seismicity that produces them and regardless of the overall nature of the shaking with which they are associated. To move beyond the UHS as the unique representation of the design earthquake actions does not, however, imply abandonment of probability: adjustments to the design return periods on each map can take account of the reduction of exceedance frequencies caused by separating seismic sources. Given the rather arbitrary basis of the return periods currently used in design codes (see Section 3.1), and the large uncertainties associated with calculating the ground motions at the selected exceedance frequencies (and the approximation that inevitably results from presenting the continuous variations of hazard in discrete intervals of ground motion), it does not seem reasonable to oppose a change to the format of seismic codes on the basis of preserving nominal adherence to the total probability theorem. Clearly there is much to be gained by providing some information regarding the disaggregated hazard in addition to the hazard maps in a code. There is, however, another possibility, which is to substitute the maps of ground-motion parameters for a given return period with maps of the magnitude and distance pairs that dominate the hazard at a given return period for each ground-motion parameter (e.g. Reference [34]). There is also the possibility to go one step further and present maps of M-D pairs corresponding to the dominant source for each site, or if the hazard is a ected by two di erent types of source, by pairs of maps for each source [35]. The third element of the disaggregated hazard, the number ( ) of standard deviations away from the logarithmic mean, needs to be accommodated in the calculation of the resulting ground motions. There is no way to prescribe a procedure for drafting such maps, since they would need to be adapted to the characteristics of the seismicity of each region. Such maps di er from the scenario ground motion maps proposed by Anderson [36] in that rather than presenting a single ground-motion parameter on each map, two maps (one of magnitudes, one of distances) provide the information required to determine all the required features of the design ground motions. At this point it is important to make a clarication: to represent the hazard from a given seismic source by a single M-D pair, coupled with a constant value of , is an approximation, since the hazard at di erent response periods will be dominated by di erent M-D combinations and di erent degrees of aleatory variability. The acceptability of this approximation, however, needs to be assessed not in absolute terms but in comparison with the much cruder approximations made in current code formats for expressing earthquake actions for design. One clear advantage of identifying earthquake scenarios in the specication of seismic design loads is the facility that this o ers with regards to the selection of acceleration timehistories for dynamic analysis.

3. CALIBRATION OF DESIGN LEVELS IN SEISMIC CODES As stated in the Introduction, the essence of performance-based seismic design is coupling performance targets with expected levels of ground motion. All formulations for PBSD presented to date assume that the di erent design levels considered will be determined from probabilistic seismic hazard assessment (PSHA) and a series of selected return periods (e.g. Reference [37]). Although PBSD may ultimately aim to provide a check on performance over the full range of hazard, for code applications it is inevitable that the design will focus on a few selected performance targets, usually coupled with particular loading levels. The current
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

49

practice of dening design loads on the basis of selected return periods is critically examined, before discussing the calibration of the selected loading-performance couples using earthquake loss modelling. 3.1. Pre-selected return periods as the basis for design loads The rst seismic design regulations to be based on a probabilistic seismic hazard map appeared in ATC 3-06 [1]. The hazard map for the U.S.A., showing PGA values with a return period of 475 years, was taken from Algermissen and Perkins [38]. The return period of 475 years was the result of selecting 50 years as the exposure period, although it was acknowledged that the use of a 50-year interval to characterize the probability is a rather arbitrary convenience, and does not imply that all buildings are thought to have a design life of 50 years (ATC, 1978). Algermissen and Perkins [38] stated that for structures which should remain operable after large, damaging earthquakes, the 10% exceedance probability in 50 years seems reasonable, although the choice of 10% was adopted on the rather arbitrary basis of being a signicance level often taken by statisticians to be meaningful [39]. A very interesting observation made in ATC 3-06 is that it was not decided a priori to base the design seismic actions on the selected 475-year return period: a map of e ective peak acceleration was drafted for ATC 3-06literally having been drawn by a committee [1], or, in other words, based entirely on expert judgmentand was found to be consistent with the 475-year PGA map of Algermissen and Perkins [38]. On this basis the 10% in 50 year map was adopted in ATC 3-06 and subsequently in the 1988 edition of UBC. The extensive commentary in ATC 3-06 provides a rational and honest examination of the risk implications of basing structural design on the ground motions with a return period of 475 years, although the estimates of losses are based entirely on expert judgment rather than modelling, and the tone is very much one of assessing and judging the chosen design basis as being reasonable and at least as stringent as the design basis in use at the time. The formulation and arguments presented in ATC 3-06 represented an important landmark and are laudable when viewed in a historical context. With time, however, the issues have been re-visited and examined in the light of improved understanding of seismic hazard and the relationships between ground-motion intensities and structural damage. These considerations have led to the adoption of 2% in 50 yearsa return period of 2475 years, considered to be more closely related with the probability of structural collapseas the design level of hazard in the 1997 NEHRP guidelines and in IBC 2000, although the actual design motions are obtained by dividing the 2475-year spectral ordinates by 1.5 [40]. A return period of 2475 years has also been adopted in the 2005 edition of the Canadian seismic code [41]. However, the seismic design codes of nearly every other country in the world, regardless of di erences with the U.S.A. in terms of seismicity, construction practices and economic prosperity, have adoptedgenerally without any clear risk-based rationalethe 475-year return period as the basis for the ground motions considered in design. A notable exception to this is the 1986 Costa Rican code, which provides maps of PGA for return periods of 50, 100, 500 and 1000 years and allows the designer to calculate the appropriate return period explicitly considering the importance, the design life and the ductility of the structure. Although most codes are based on a single hazard map, some performance-based considerations are actually present in most codes through the use of importance factors that increase the spectral ordinates for structures required to perform above simple life-safety criteria under
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

50

J. J. BOMMER AND R. PINHO

the 475-year ground motions. The factors essentially result in safety-important structures being designed for longer return-period groundmotions. In EC8, recommended importance factors for large occupancy and essential buildings are 1.2 and 1.4, respectively. The code is quite specic about the purpose of the importance factor: Wherever feasible this factor should be derived so as to correspond to a higher or lower value of the return period of the seismic event (with regard to the reference return period) as appropriate for the design of the specic category of structures [5]. The shortcomings of this approach are obvious: rstly, it assumes that the variation of ground-motion amplitude with return period is constant throughout the country covered by the code, and secondly, it assumes that designing for life-safety under ground shaking with a longer return period will ensure that the structure remains operational under the 475-year return period. The di erence between the explicit PBSD approach and the approach implied by the use of importance factors applied to increase the loading (rather than being applied to reduce the allowable drifts or displacements) is illustrated in Figure 3. In the now famous matrix of design levels and performance levels (Figure 3) presented in Vision 2000 [2]which renders the importance factors dened in current codes redundant the return periods specied (without explanation or commentary) for design ground motions are 43, 72, 475 and 970 years, which correspond to exceedance probabilities of 69, 50, 10 and 5% during an exposure period of 50 years. Consensus on the appropriate return periods has not yet been achieved and indeed, on the basis of decisions made in the absence of quantitative loss assessments such consensus may never be reached. 3.2. Calibration of seismic codes through earthquake loss modelling The current version of EC8 does to a limited extent already conform to the concept of PBSD [42], with design for no collapse based on ground motions with a recommended return period of 475 years (the design seismic action) and damage limitation controlled for ground motions with a recommended return period of 95 years (the damage limitation ground motion). Here it is important to emphasize that EC8 does not prescribe the values of 95 and 475 years as the return periods to be used as the basis for design, since these are suggested values and each national authority is given the freedom to select values considered appropriate; the current Portuguese seismic design code, for example, denes seismic loading based on a return period of 975 years [43]. This exibility within EC8, which as noted previously extends to almost every aspect of the seismic loading, presents a unique opportunity but in the absence of a clear framework for the denition of the loads, there is the danger that the application of the EC8 template to national design codes throughout Europe will simply involve other arbitrary selections. As noted in the previous section, the commentary of ATC 3-06 discusses possible losses that could be in icted on an exposed building stock designed according to the proposed requirements anchored to a 475-year return period. The proposal of the authors of this paperwhich is evidently not originalis that such checking of the specied design requirements should be carried out for all seismic design codes, taking advantage of the signicant developments in recent years in the eld of earthquake loss modelling. In the case of EC8, such models could be used by national authorities as a procedure for checking that the return periods selected as the basis for damage limitation and no-collapse result in levels of risk that are considered tolerable. For EC8, extensive analytical and experimental investigations were carried out (e.g.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

51

Earthquake Performance Level


Fully Operational Operational Life Safe Near Collapse

Frequent Earthquake Design Level

Occasional

Es se

Ba sic
nt ial /H a za r

Unacceptable
Ob je

Performance
cti ve

Rare

Sa f

et

yC

do

riti

us O

ca l

Very Rare

Ob jec ti

b je cti

ve

ve

Earthquake Performance Level


Fully Operational Operational Life Safe Near Collapse

Frequent Earthquake Design Level

Occasional

Es se

Ba sic

Unacceptable
Ob je

Performance
cti ve

nt ial

Rare

Sa f

/H a

et

za r

yC

do

riti

us O

ca l

Very Rare

Ob jec ti

b je cti v

ve

Figure 3. Upper: the PBSD framework of moving to higher performance targets under each level of loading for structures of higher importance. Lower: the approximation of PBSD through the application of importance factors to the design loads.

References [44, 45]) to calibrate and verify elements of the design specications such as the behaviour factors, q. To the knowledge of the authors, no national or regional loss modelling has been performed to identify the quantitative risk implied by the suggested design levels in di erent regions. This is perhaps understandable, given that EC8 is applicable to the entire European Community, which encompasses a very wide range of seismic hazard levels. However, within the context of application of EC8 to individual countries, it should be feasible
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

52

J. J. BOMMER AND R. PINHO

to develop models to estimate the levels of loss that future earthquakes could be expected to cause to a building stock designed and constructed according to the code requirements. If the expected losses were judged to be excessive, the design return period for one or other limit state could be extended, to impose more severe design loads. The loss modelling could be performed for a representative urban area or a selected region, since the compilation of a detailed exposure and vulnerability model for an entire country could be excessively demanding in terms of time and e ort. Clearly, for consistency it would be necessary to model the ground motion in terms of parameters that are consistent with those forming the basis for design. Furthermore, the loading level considered in the loss model should be consistent with that specied for design, which means that the ground motions considered should correspond to the return period of 475 years or whichever alternative value is selected. Although this can be approximated by simply using ground-motion parameters from hazard maps corresponding to this return period, it is more meaningful to model the hazard through Monte Carlo simulations constrained for compatibility with the hazard at the specied return period. To embrace PBSD entirely, EC8 probably needs to introduce a third and maybe even a fourth design level, which will require the selectionand subsequent checkingof additional return periods to dene the seismic actions for design. The same concept of calibrating the selected design levels by quantitative assessment of the possible ensuing earthquake losses could be applied to any number of loading-performance couples. The task is not, however, trivial, since it involves representing hypothetical future building stocks and accurately modelling their earthquake capacity. With the increasing development and renement of earthquake loss models, however, this becomes increasingly feasible. Providing that incremental changes to the seismic capacity of di erent classes of building can be accurately re ected in the loss calculations, it is possible to estimate lossesin both human and nancial termsfor a large number of di erent levels of seismic resistance. If planners and politicians, informed by the output from such iterative loss modelling, can make decisions regarding minimum levels of protection for life safety, and regarding overall investment in seismic resistance, this process could lead to the selection of appropriate levels of seismic design in codes. Bommer et al. [46] present the framework for such an approach, which does not take the selection of design return periods as the starting point.

4. DISCUSSION AND CONCLUSIONS This paper has presented ideas for modications to future editions of EC8 as steps towards fuller adoption of the PBSD framework. The paper has also highlighted the fact that EC8 has already taken signicant steps towards adopting a performance-based approach to seismic design by providing two performance levels and by allowing each country that adopts the code to dene their own hazard levels for design of new construction. Another important aspect in which EC8 has adopted a PBSD approach is in allowing the design of new buildings on the basis of pushover analysis with direct verications of deformations without a force reduction or response modication factor. One of the most urgent tasks in fully adapting Eurocode 8 to a performance-based seismic design framework is precisely to quantitatively check that the specied design levels do, in fact, result in tolerable levels of risk in terms of deaths, injuries, economic losses and societal
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

53

disruption. Each nation adopting EC8 could perform such checks, through earthquake loss modelling, as part of the process of national adaptation of the code, and in particular in selecting the return periods to form the basis for the damage limitation and no-collapse limit states. The incorporation of additional design limit states may or may not be judged necessary but this is clearly not a matter of urgency. Adapting Eurocode 8 to be a fully performance-based code, however, will also require representing the seismic actions in formats more closely related to damage potential than the seismic forces that are currently used in the code. To the credit of its drafters, Eurocode 8 has already envisaged this development through the inclusion of elastic and inelastic displacement response spectra, albeit in informative annexes rather than in the body of the code, as a move towards the employment of performance-based design=assessment methods (in both their equivalent linearization or displacement modication versions). However, the acceleration spectrum in EC8 is currently anchored only to PGA, which leads to a poor approximation of the UHS and is totally inadequate as the basis for constructing long-period displacement spectra. Therefore, in future editions of EC8, for each return period (or design level), at least three parameters will need to be mapped in order to construct the design spectrum. However, if those responsible for the revision of the code are prepared to look beyond the limits of current practice and embrace the opportunity to set new standards, more elegant and more physical representations of the earthquake loading could be dened, particularly if the outdated concept of the uniform hazard spectrum can be left to one side and if the earthquakes that drive seismic hazard can somehow be made visible in the code.
ACKNOWLEDGEMENTS

We are grateful to a number of people who have contributed in di erent ways to the development of some of the ideas presented in this paper, most notably Edmund Booth, Damian Grant and Juliet Bird. A very special mention is due to Helen Crowley for her invaluable input through many discussions and reviews. We have also beneted from many discussions of these issues, particularly with participants at the Bled Workshop, especially Robert E. Bachman, Allin C. Cornell, Ronald Hamburger and Eduardo Miranda. Previous versions of the manuscript received very thorough and challenging reviews from two anonymous referees, whose critical comments and constructive suggestionsaddressing specics of EC8 and general concepts of PBSD, respectivelyhave led to a major modication and, we believe, signicant improvement of the paper.
REFERENCES 1. ATC. Tentative provisions for the development of seismic regulations for buildings. Report ATC 3-06, Applied Technology Council, 1978. 2. SEAOC. Vision 2000: Performance Based Seismic Engineering of Buildings. Structural Engineers Association of California, Sacramento CA, 1995. 3. CEN. Application and Use of Eurocodes. CONSTRUCT 01/483. Direction G n ral Enterprises, Commission e e Europ enne, Brussels, 2002. e 4. CEN. Eurocode 8, Design Provisions for Earthquake Resistance of Structures, European Prestandard ENV 1998. Comit Europ en de Normalisation, Brussels, 1994. e e 5. CEN. Eurocode 8, Design of Structures for Earthquake ResistancePart 1: general rules, seismic actions and rules for buildings. EN 1998-1: 2004. Comit Europ en de Normalisation, Brussels, 2004. e e 6. Moehle JP. Displacement-based design of RC structures subjected to earthquakes. Earthquake Spectra 1992; 8(3):403 428. 7. Sullivan TJ, Calvi GM, Priestley MJN, Kowalsky MJ. The limitations and performances of di erent displacement based design methods. Journal of Earthquake Engineering 2003; 7(Special Issue 1):201241.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

54

J. J. BOMMER AND R. PINHO

8. ATC. Improvement of nonlinear static seismic analysis procedures. FEMA-440, ATC, California, U.S.A., 2005. 9. McGuire RK. Seismic design spectra and mapping procedures using hazard analysis based directly on oscillator response. Earthquake Engineering and Structural Dynamics 1977; 5:211234. 10. IAEE. Regulations for Seismic Design: A World List1996. International Association for Earthquake Engineering, Tokyo, 1996. 11. ICC. International Building Code. International Code Council, 2000. 12. Ambraseys NN, Simpson KA, Bommer JJ. The prediction of horizontal response spectra in Europe. Earthquake Engineering and Structural Dynamics 1996; 25:371 400. 13. Fajfar P. Capacity spectrum method based on inelastic demand spectra. Earthquake Engineering and Structural Dynamics 1999; 28:979 993. 14. Kowalsky MJ, Priestley MJN, MacRae GA. Displacement-based design of RC bridge columns in seismic regions. Earthquake Engineering and Structural Dynamics 1995; 24(12):1623 1643. 15. Tolis SV, Faccioli E. Displacement design spectra. Journal of Earthquake Engineering 1999; 3(1):107125. 16. Faccioli E, Paolucci R, Rey J. Displacement spectra for long periods. Earthquake Spectra 2004; 20(2): 347 376. 17. BSSC. NEHRP recommended provisions for seismic regulations for new buildings and other structures, 2003 Edition, Part 1Provisions, Part 2Commentary. Report FEMA 450, Building Seismic Safety Council, Washington, DC, 2004. 18. Joyner WB, Boore DM. Measurement, characterization and prediction of strong ground motion. Proceedings of Earthquake Engineering and Soil Dynamics II, GT Div/ASCE, Park City, Utah, 2730 June 1988; 43 102. 19. Bommer JJ, Elnashai AS. Displacement spectra for seismic design. Journal of Earthquake Engineering 1999; 3(1):132. 20. Boore DM, Bommer JJ. Processing of strong-motion accelerograms: needs, options and consequences. Soil Dynamics and Earthquake Engineering 2005; 25(2):93 115. 21. Gregor NJ, Bolt BA. Peak strong motion attenuation relations for horizontal and vertical ground displacements. Journal of Earthquake Engineering 1997; 1(2):275 292. 22. Spudich P. What seismology may be able to bring to future codes (a seismologists perspective). Report ATC 35-3, Applied Technology Council, 1999; 2133. 23. Dwairi H, Kowalsky MJ. Examination of the equivalent viscous damping approach. International Workshop on Performance-Based Seismic Design (PBSD)Concepts and Implementation, Bled, 2004. 24. Iwan WD. The use of equivalent linearization in performance based engineering. Proceedings of the International Conference on Advances and New Challenges in Earthquake Engineering Research, Harbin, China, 2002. 25. Ruiz-Garcia J, Miranda E. Inelastic displacement ratios for evaluation of existing structures. Earthquake Engineering and Structural Dynamics 2003; 32(8):12371250. 26. Bommer JJ, Mendis R. Scaling of spectral displacement ordinates with damping ratios. Earthquake Engineering and Structural Dynamics 2005; 33(2):145 165. 27. Somerville PG. Ground motion prediction for performance based seismic engineering. Proceedings of the 65th SEAOC Convention, Maui, Hawaii, 1996. 28. Grases J, Cascante G, Gajardo E. Earthquake hazard mapping for seismic design: a new approach. Proceedings of the 10th World Conference on Earthquake Engineering, vol. 1, Madrid, 1992; 57775782. 29. Garc a-Mayordomo J, Faccioli E, Paolucci R. Comparative study of the seismic hazard assessments in European national seismic codes. Bulletin of Earthquake Engineering 2004; 2:5173. 30. Bommer JJ, Elnashai AS, Weir AG. Compatible acceleration and displacement spectra for seismic design codes. Proceedings of the 12th World Conference on Earthquake Engineering, Auckland, 2000; Paper no. 207. 31. McGuire RK. Probabilistic seismic hazard analysis and design earthquakes: closing the loop. Bulletin of the Seismological Society of America 1995; 85:1275 1284. 32. Bazzurro P, Cornell CA. Disaggregation of seismic hazard. Bulletin of the Seismological Society of America 1999; 89:501520. 33. Merovich AT. Future characterizations of ground motion for design (an engineers perspective). Report ATC 35-3, Applied Technology Council, 1999; 9 20. 34. Harmsen SD, Perkins D, Frankel A. Deaggregation of probabilistic ground motions in Central and Eastern United States. Bulletin of the Seismological Society of America 1999; 89:113. 35. Bommer JJ. Seismic zonation for comprehensive denition of earthquake actions. Proceedings of 6th International Conference on Seismic Zonation, Palm Springs, California, 2000. 36. Anderson JG. Benets of scenario ground motion maps. Engineering Geology 1997; 48:43 57. 37. Stewart JP, Chiou S-J, Bray JD, Graves RW, Somerville PG, Abrahamson NA. Ground motion evaluation procedures for performance-based design. Soil Dynamics and Earthquake Engineering 2002; 22:765 772. 38. Algermissen ST, Perkins DM. A probabilistic estimate of maximum accelerations in rock in the contiguous United States. US Geological Survey Open-File Report 76-416, 1976. 39. Perkins DM. Written communication (email) to Dr Dario Slejko (Trieste), 2004.
Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955

ADAPTING EARTHQUAKE ACTIONS IN EUROCODE 8

55

40. Leyendecker EV, Hunt RJ, Frankel AD, Rukstales KS. Development of maximum considered earthquake ground motion maps. Earthquake Spectra 2000; 16:21 40. 41. Heidebrecht AC. Seismic design implications of revisions to the national building code of Canada. Proceedings of the Pacic Conference on Earthquake Engineering, 2003; Paper no. 148. 42. Fardis MN. A European perspective to performance-based seismic design, assessment and retrotting. Proceedings of an International Workshop on Performance-based Seismic Design: Concepts and Implementation, PEER 2005=05, Pacic Earthquake Engineering Research Center, Berkeley, California, 2005; 113. 43. Oliveira CS, Campos-Costa A, Sousa ML. Denition of seismic action in the context of EC-8: topics for discussion. Proceedings of the Twelfth World Conference on Earthquake Engineering, Auckland, 2000; Paper no. 2552. 44. Fardis MN (ed.). Experimental and numerical investigations on the seismic response of R.C. inlled frames and recommendations for code provisions. ECOEST-PREC8 Report No. 6, LNEC, Lisbon, 1996. 45. Severn RT, Calvi GM (eds). European activities for the development of Eurocode 8summary report. ECOEST-PREC8 Report No. 9, LNEC, Lisbon, 1996. 46. Bommer JJ, Pinho R, Crowley H. Using displacement-based earthquake loss assessment in the selection of seismic code design levels. Proceedings of International Conference on Structural Safety and Reliability (ICOSSAR05), Rome, 2005; 35673574.

Copyright ? 2005 John Wiley & Sons, Ltd.

Earthquake Engng Struct. Dyn. 2006; 35:3955