Microzonation: techniques and examples

Corinne Lacave1, Pierre-Yves Bard2 and Martin G. Koller1

Affiliation:

1 Rsonance Ingnieurs-Conseils SA, Geneva, Switzerland 2 LGIT, Universit Joseph Fourier, Grenoble, and LCPC, Paris, France Microzonation, site effects, topographic effects, standard spectral ratio, H/V ratio, Nakamura method, numerical simulations, empirical Green's function technique

Key-words:

1. Introduction
It is now well known, and widely accepted amongst the earthquake engineering community, that the effects of surface geology on seismic motion exist and can be large. Two "famous" examples of such effects are San Francisco and Mexico City. In San Francisco, local amplifications over unconsolidated sediments have been shown to be responsible for intensity variations as large as two degrees (MM scale) during both the 1906 "big" San Francisco earthquake and the more recent 1989 Loma Prieta event. In Mexico City, there exist very soft lacustrine clay deposits underneath the downtown area of the city. These led to very large amplifications, which caused a high death toll and large economic losses during the distant Guerrero Michoacan earthquake of 1985. Nearly all recent destructive earthquakes (Spitak, Armenia 1988, Iran 1990, Philippines 1990, Northridge 1994, Kobe 1995, Armenia, Columbia 1999, Turkey 1999...) have brought additional evidence of the dramatic importance of site effects. Accounting for such "site effects" in seismic regulations, land use planning or design of critical facilities thus became one goal of earthquake hazard reduction programs (Bard, 1997). In the first section, the most important site effects are presented, namely amplification effects related with sedimentary sites and strong lateral discontinuities, as well as motion amplifications related with surface topography. Then, experimental and numerical techniques are presented that are available to estimate these site effects, in the framework of microzonation studies. An application is finally presented for the Rhne valley in the Visp area, a narrow alpine valley in the canton Wallis, Switzerland.

2. Site effects
2.1 Effects of soft surface layers It has been recognised for a very long time that earthquake damage is generally larger over soft sediments than on firm bedrock outcrops. This is particularly important because most of urban settlements have occurred along river valleys over such young, soft surface deposits. Many large cities located in earthquake prone areas may be given as examples: Los Angeles, San Francisco, San Salvador, Caracas, Lima, Bogota, Tokyo, Osaka and Kobe, Katmandu, Manila, Lisbon, Thessaloniki, Izmit, and, of course, Mexico City. But one should not forget many other mid-size, recently developed cities, especially in moderate seismicity areas, since they might be the place for future, heavily damaging events, due to the combination of site effects and urban development.

Soft soils may have, and did have indeed, dramatic consequences for the inhabitants of those cities. They have however given engineering seismologists the opportunity to perform numerous macroseismic observations during the last century. These demonstrated very clearly that the damaging effects associated with such soft deposits may lead to local intensity increments as large as 2, in extreme cases even 3 degrees on the MM or MSK/EMS scale. The fundamental phenomenon responsible for the amplification of motion over soft sediments is the trapping of seismic waves due to the impedance contrast between sediments and the underlying bedrock. When the structure is horizontally layered (which will be referred to in the following as 1-D structures), this trapping affects only body waves travelling up and down in the surface layers. When the surface sediments form a 2-D or 3-D structure, i.e., when lateral heterogeneities such as thickness variations are present, this trapping also affects the surface waves which develop on these heterogeneities, and thus reverberate back and forth. The interference between these trapped waves leads to resonance patterns, the shape and the frequency of which are related with the geometrical and mechanical characteristics of the structure. While these resonance patterns are very simple in the case of 1-D media (vertical resonance of body waves), they become more complex in the case of 2-D and a fortiori 3-D structures. An illustration of 1-D and 2-D effects is given in Figure 1.

2.2 Topographic effects It has been often reported after destructive earthquakes that buildings located at hill tops suffered much more intensive damage than those located at the base (see example in Figure 2). There is also very strong instrumental evidence that surface topography considerably affects the amplitude and frequency contents of ground motion: reviews of such instrumental studies and results may be found in Gli et al. (1988), Faccioli (1991) and Finn (1991). However, the number of instrumental studies about topographic effects is extremely low compared to studies dealing with soft soil amplification, so that it remains impossible to derive any statistics from the existing data. Theoretical and numerical models also predict a systematic amplification of seismic motion at ridge crests, and, more generally, over convex topographies such as cliffs. Correspondingly, they also predict de-amplification over concave topographic features, such as valleys and the base of hills. The amount of those effects have been shown (Pedersen et al., 1994) to be rather sensitive to the characteristics of the incident wavefield (wave type, incidence and azimuth angles). Theory also predicts complex amplification and de-amplification patterns on hill slopes, resulting in significant differential motions. Very often, measurements show larger effects than what is predicted by simulations. A possible explanation is that in many cases weathered rock is found at the top of hills that reinforces the topographic effect (Pedersen et al., 1994).

3. Methods for estimating site effects

3.1 Experimental methods 3.1.1 Standard spectral ratio (SSR) The most common procedure (the principle of which is illustrated in Figure 3) consists in comparing recordings at nearby sites (where source and path effects are believed to be identical) through spectral ratios. These spectral ratios constitute a reliable estimate of site response if the

"reference site" is free of any site effect, which means that it should fulfil the two following conditions: First, it should be located near enough to the examined station to ensure that differences between each site are only due to site conditions, and not to differences in source radiation or travel path, which is generally warranted when the hypocentral distance is larger than about 10 times the array aperture (in case of necessity a ratio of 5 might still be acceptable). Secondly, it should also be unaffected by any kind of site effect, which is the case when the reference site is located on an unweathered, horizontal bedrock. These two conditions prove to be rather restrictive in practice. The SSR technique, introduced first by Borcherdt (1970), is still widely used. Measurements are usually available only for events weaker much weaker indeed - than the event engineers are interested in. Therefore, due to non-linearity, the site effect estimated by SSR techniques may be considered as an overestimation or upper limit of the actual site effects at high frequencies, and, correlatively, a slight underestimation at frequencies below the "elastic" fundamental frequency. In a nutshell, the principle of this method may be described as follows: for a network of I sites having recorded J events, the amplitude spectrum of the ground motion Rij(f) recorded at site i during the event j can be written as: Rij(f) = Ej(f) . Pij(f) . Si (f) (1)

with Ej(f): source function; Pij(f): path contribution between the source and the local site; and Si (f): local site contribution. Written in logarithmic form, it leads to a simple linear equation: ln (Rij(f) ) = ln (Ej(f)) + ln (Pij(f)) + ln (Si (f)) (2)

The traditional spectral ratio technique corresponds to the fact where the path term Pij(f) is assumed to be site independent, i.e., when the distance to the reference site is small compared to the source-to-site distance. 3.1.2 H/V noise ratio (or "Nogishi-Nakamura" technique) The H/V ratio, i.e. the ratio between the Fourier spectra of the horizontal and vertical components of ambient vibrations (often called "microtremors" or "ambient noise"), was introduced in the early seventies by several Japanese scientists (Nogoshi and Igarashi, 1971; Shiono et al., 1979; Kobayashi, 1980; Nakamura, 1989). Although Nakamura's qualitative explanation looked at least questionable (as indicated in Lachet and Bard, 1994, and again in Kudo, 1995), various sets of experimental data (Ohmachi et al, 1991; Lermo, 1992 ; Field and Jacob, 1993b; Duval et al., 1994; Duval et al., 1995; Field et al., 1995; Seekins et al., 1996; Lachet et al., 1996; Gitterman et al., 1996; Fh et al., 1997; Lebrun, 1997; Riepl et al., 1998) confirmed that these ratios are much more stable than the raw noise spectra. In addition, on soft soil sites, they usually exhibit a clear peak that is well correlated with the fundamental resonant frequency. These observations are supported by several theoretical investigations (Field and Jacob, 1993b; Lachet and Bard, 1994; Lermo and ChavezGarcia, 1994; Cornou, 1998), showing that synthetics obtained with randomly distributed, near surface sources lead to horizontal-to-vertical ratios sharply peaked around the fundamental Swave frequency (Figure 4a), whenever the surface layers exhibit a sharp impedance contrast with the underlying stiffer formations. However, the first three studies mentioned avove also conclude that the amplitude of this peak is not well correlated with the S wave amplification at the site's resonant frequency (Figure 4b). Instead, it is highly sensitive to some parameters such as Poisson's ratio near the surface. Furthermore, Lachet and Bard (1994) proposed that the good match at the fundamental

frequency is due to the horizontal-vertical polarisation of the Rayleigh waves, an interpretation that is in agreement with the early Japanese studies (Kudo, 1995). According to this view, no straightforward relation exists between the H/V peak amplitude and the site amplification. However, this opinion is not shared unanimously. For instance, a one to one average correlation is claimed by Konno and Ohmachi (1998) on the basis of a comparison between observed H/V peaks and numerical estimates of 1-D transfer functions. Furthermore, an empirical relationship between H/V peak amplitude and local intensity increment (MM scale) was argued in Toshinawa et al. (1997) from a strictly experimental viewpoint. A thorough comparison (Bard et al., 1997; Bard, 1999) between observed amplifications derived from earthquake records and observed H/V peak amplitudes at more than 30 sites demonstrates that the latter is almost always smaller than the former (Figure 5). Such an experimental result, although not yet explained by theoretical or numerical work, despite the new computations performed by Cornou (1998), could be very useful indeed. It would mean that the H/V ratio technique could provide a lower-bound estimate to the actual weak motion amplification. This view needs to be confirmed, however, by a larger set of experimental data. Other examples of estimates obtained with this technique are also displayed in Figure 6, for comparison with other estimates. They show that H/V ratio technique allows obtaining, very simply, the fundamental resonant frequency, but fails for higher harmonics, and that peak amplitude is somewhat different from amplification measured on spectral ratios. In practice, the H/V ratios from ambient vibrations are sometimes "non-informative" so that no unequivocal interpretation is possible. This seems to be true particularly in the case of pronounced 2D or 3D structures, i.e. in the presence of significant lateral heterogeneity. Nevertheless, the method proved to be the most inexpensive and convenient technique to estimate fundamental frequencies of soft deposits. It definitely deserves more research efforts so as to elucidate the factors influencing its reliability. 3.1.3 H/V spectral ratio of weak motion Another simple technique consists in taking the spectral ratio between the horizontal and the vertical components of the shear wave part of weak earthquake recordings. This technique is in fact a combination of Langston's (1979) "receiver-function" method for determining the velocity structure of the crust from the horizontal to vertical spectral ratio (HVSR) of teleseismic P waves, and Nakamuras proposal (1989, 1996) to use this ratio with recordings of ambient vibrations (section 3.1.2). This method is obviously interesting, because of its simplicity and economy. It was first applied to the S wave portion of the earthquake recordings obtained at three different sites in Mexico by Lermo and Chavez-Garcia (1993). These recordings exhibit very encouraging similarities between the classical spectral ratios and these HVSR, with a good fit in both, the frequencies and amplitudes of the resonant peaks. The same technique has been also checked on various sets of weak and strong motion data (see Chavez-Garcia et al., 1996; Lachet et al., 1996; Riepl et al., 1998; Theodulidis et al., 1996; Bonilla et al., 1997; Yamazaki and Ansary, 1997; Zar et al., 1999), from which several conclusions appear well established: The HVSR shape exhibits a very good experimental stability. It is also well correlated with surface geology, and much less sensitive to source and path effects (a warning should be issued however for near field recordings of large events, because of the strong directionality of the near-source "fling" or "killer pulse"). However, comparisons with classical spectral ratios (including surface / down-hole recordings), as well as with theoretical 1D computations (see also Lachet and Bard, 1994), also agree on the fact that the absolute level of HVSR depends on the type of incident waves. It

follows that the determination of the absolute level of amplification from only HVSR is not straightforward. Field and Jacob (1995) also applied this technique in their systematic comparisons, and found that the method reproduces very well the shape of the site response, but underestimates the amplification level (see Figure 6e). They also found very different results when applying this technique to the P-wave part of the recordings. They therefore conclude that HVSR, when applied to the S-wave signals, reveals the overall frequency dependence. Based on our own investigations, we agree only partly with this conclusion, since in some cases we did find a good similarity in "spectral shapes", but in a few other HVSR we were only able to identify the fundamental resonance frequency. Furthermore, it should be pointed out that this technique has been applied and checked for soft soil sites only, and might not be valid for other kinds of site effects.

3.2 Numerical methods When the geotechnical characteristics of the site or of the area are known, site effects can be, in principle, estimated through numerical analysis. The prerequisite of a sufficient geotechnical knowledge generally implies that such ground response analyses be made on a site by site basis, but the density of boreholes and geotechnical information in some large cities may be sufficient to allow a numerically-based zoning. Such an approach, however, requires an in-depth understanding both of the analytical models and of the numerical schemes that are used. When the required expertise is lacking, it may occur that sophisticated numerical analyses lead to less reliable results than simpler and cruder, but more robust, approximations. Thus, the present section will mainly emphasize the latter ones. 3.2.1 One-dimensional response of soil columns There exist a number of simple analytical methods which allow computation of the seismic response of a given site with only the help of a small personal computer. Amongst these, the most widely used makes use of the multiple reflection theory of S waves in horizontally layered deposits, very often referred to as "1-D analysis of soil columns". Such a soil column is excited by an incoming plane S wave, generally considered as vertically incident, and corresponding to a surface bedrock motion representative of what is likely to occur in the area. The specific parameters required for such an analysis are shear-wave velocity, density, damping and thickness of each layer. These parameters may be obtained either through direct in situ measurement, or from drillings and subsequent laboratory measurements, or from known approximate relationships with other, more usual geotechnical parameters such as the SPT number from standard penetration tests. These analyses may be performed considering either a linear or a non-linear behaviour for the soil. The non-linearity is very often approximated by a "linear-equivalent" method that uses an iterative procedure to adapt the soil parameters (i.e., rigidity and damping) to the actual strain it undergoes, according to the curves depicted in Figure 7. The SHAKE program is one of the most widely used for such calculations (Schnabel et al., 1972). Recently developed packages incorporating true non-linear constitutive models are now available, that also allow to account for liquefaction phenomena (example: CyberQuake program, 1998). However , these non-linear analyses require a quantitative knowledge of the actual nonlinear material behaviour, which can only be obtained by means of sophisicated laboratory tests. Some generic average curves have been proposed for different types of material, as sand or

clay, but the actual behaviour of a given soil at a given site may strongly depart from these averages. This was precisely the case in Mexico City where the clays proved indeed to behave almost linearly despite the large strains experienced during the 1985 event, while they were previously believed to be highly non-linear because of their very low rigidity. 3.2.2 Advanced methods Although all numerical methods have the same base i.e. the wave equation many different models have been proposed to investigate several of the various aspects of site effects, which involve complex phenomena. For example, one has to consider various types of incident wave fields (near-field, far-field, body waves, surface waves); the structure geometry may be 1-D, 2D or 3-D; or the mechanical behaviour of earth materials may encompass a very wide range (viscoelasticity, non-linear behaviour, water-saturated media, liquid domains, etc.). Typically, these advanced methods may be classified into four groups: Analytical methods, which can be used only for very simple geometries, are extremely valuable as benchmarks. Ray methods, which are basically high frequency techniques. It is uneasy to use them when wavelengths are comparable to the size of heterogeneity, a situation which is generally the most interesting one. Boundary based techniques (including all kinds of boundary integral techniques, or those based on wave function expansions), which are the most efficient when the site of interest consists of a limited number of homogeneous geological units. Domain based techniques (such as finite-difference or finite-element methods), which allow to account for very complex structures and material behaviour, but are very heavy from a computational viewpoint. As stressed by Aki and Irikura (1991), "() these methods have their advantages and disadvantages, and, in general, those which can deal with more realistic models are less accurate, while those achieving a higher accuracy are more time consuming. Most of these methods are still actively developed, because each has its own merit that effectively applies to a certain class of problems ()". It is not our aim in this section to elaborate on the intrinsic reliability of each numerical technique, but simply to discuss the applicability of numerical methods, considered as a whole, to the estimation of site effects for engineering purposes. Although these methods need heavy computational processes, their main advantage rests in their flexibility and versatility (combined with their cheapness on standard computers), which have lead to significant breakthroughs in the understanding of site effects during the last two decades. Not only they allow to carry out phenomenological and parametric studies, they can also be used to assess the uncertainty in a sites seismic response, given the imperfect knowledge regarding the mechanical and geometrical characteristics of the considered site. Advances could be achieved along two avenues: Proper consideration of diffraction effects in complex surface or subsurface topography. There now exist many numerical techniques to account for 2-D or even 3-D geometries, as the Akil-Larner modelling for example (Aki and Larner, 1970), or the combined mode summation and finite-difference technique proposed by Fh and Suhadolc (1994). However, because of computational limitations, as well as lack - at reasonable expenses - of sufficiently detailed knowledge of the underground structure, analyses with true 3-D models are generally restricted to low frequencies (typically below about 1 Hz). More realistic modelling of strong non-linear behaviour in soft soils, and especially saturated sands, as done by the program CyberQuake for example (CyberQuake, 1998).

Nevertheless, the routine application of numerical models for the evaluation of site effects raises some concerns: Until recently, numerical models had been only very rarely objectively tested for their ability to predict actual effects. Indeed, comparisons between observations and theoretical computations had almost always been carried out a posteriori : the predictors knew what to find. Numerical methods are not panacea, but can be applied only to some limited cases. Unfortunately, individuals who lack an understanding of their limits may use the software implementing these methods outside of their validity domain, which may consequently lead to wrong estimations. Even when the methods are properly used, their actual cost may be high, perhaps well in excess of the cost using instrumental means. The reason is that they require detailed geotechnical or geophysical investigations for the site to provide the constitutive properties needed as input parameters. This issue may sometimes be overcome through parametric studies, but this is useful only when the results do not exhibit too much sensitivity (which is rarely the case). 3.2.3 Empirical and semi-empirical methods Empirical attenuation laws Many empirical attenuation laws have been derived on the basis of available strong motion recordings. They all relate a given ground motion parameter (pga, pgv, Sa , duration, Arias intensity, etc.) to the magnitude and distance of the seismic event, and they also very often take into account a site parameter. Very often this site parameter is simply a binary descriptor, such as "rock" and "non-rock". Only rarely is the site geology characterized in a more refined manner, for instance with distinction between thin and thick deposits, or with S-wave velocity values : the reason is that detailed information on strong motion recording sites is generally missing. Significant efforts are made throughout the world, however, to fill this gap : the most striking example is the K-NET network installed in Japan after the Kobe event, for which a 20 m deep borehole has been drilled at each of the 1000 sites, and the S and P wave velocity profile has been obtained. It is thus possible to modify the ground motion parameters according to the site geology. However, as these modifications are based on a very crude classification of soils, and on statistical studies which, in essence, smooth out the extreme values, such an approach may lead to a dangerous underestimation of amplifications at sensitive sites. Conversely, there is a significant probability of overestimating the motion at common sites. Empirical Green's functions technique (EGF) The empirical Green's functions (EGF) technique is known essentially in the seismological community, as a tool for studying the source process of past large earthquakes using records from both mainshock and aftershocks (e.g. Mueller, 1985 ; Courboulex et al., 1998). It is far less known in the engineering community, where its potential to predict the expected strong ground motion during future large events has not yet been sufficiently exploited. A key feature of the EGF technique is its capability to synthesise physically realistic, site specific acceleration time histories. This is of particular interest for the seismic analysis of critical facilities. Since critical facilities should withstand strong earthquakes with long return periods, typically of the order of 10 000 years or more, it is common that no recording of such an event is available for the facility's site. This lack of appropriate time histories can elegantly be overcome with the aid of the EGF technique. Its basic idea is to interpret recordings of small seismic

events at the site of interest as reasonable approximations of Green's functions and to convolute them suitably, using earthquake scaling laws, in order to simulate time histories that correspond to larger earthquakes. The EGF technique was first put forward by Hartzell (1978) and has since been further developed by numerous scientists. Its main interest is that the true propagation and site effects are automatically accounted for ; its main disadvantage is that it cannot, on its own, account for non-linear soil behaviour. If non-linear soil behaviour cannot be neglected, the EGF technique should be combined with geotechnical methods, as outlined by Heuze et al. (1995). There are, roughly speaking, two different "families" of EGF techniques. In both cases, the EGF is taken at several times and added up so that a larger earthquake, referred to as the "target" event, of the same focal mechanism is synthesized (see Figure 8). The difference lies in the way how the summing up of the EGF is performed : with or without kinematic modelling of the target event's rupture process. Irikura (1983, 1986), Hutchings (1994) and Irikura and Kamae (1994) are all representatives of the family of kinematic modelling techniques. The other family uses essentially statistical tools that allow to sum up the EGFs in a way that the relevant earthquake scaling laws will be respected. An overview on this family is given by Tumarkin and Archuleta (1994).

4. Example: Microzonation of the Visp area (Wallis, Switzerland)

Alpine valleys are often characterised by significant site effects, due to the presence of deep young alluvial deposits in narrow glacial valleys, overlaying a bedrock formation. The town of Visp is located in the Rhone valley, Wallis, Switzerland. The aim of the following work was to establish a microzonation directly applicable in engineering practice, by means of proposed design response spectra for different zones in the considered area (Rsonance et al., 1999).

4.1 Geological context and soil parameters The dimensions of the valley are approximately 220 m in depth (down to the substratum) for a width of 1500 m. Geological information were gathered from a systematic bibliographic study. Soil parameters were deduced from SPT (Standard Penetration Test) values with the aid of empirical relations. The Rhone valley in the region of Visp can schematically be described as follows : In the centre of the valley, recent alluvial deposits are found down to a depth of 170 m, with S-wave velocities estimated to range from 250 m/s at the surface to 900 m/s at depth. These deposits are characterised by silts and fine sands in alternation with gravel and sand layers. Then a morainic layer of 50 m is found, characterized by velocities estimated around 1000 to 1200 m/s. Finally, the rock substratum is characterized, in this study, by an assumed S-wave velocity of 2500 m/s.

4.2 Methodology 4.2.1 Regional seismic hazard The regional seismic hazard to be taken into account in the area of Visp, for ordinary buildings, is characterised by a peak ground acceleration ag = 0.16 g, following the seismic zonation of Switzerland (SIA 160, 1989). This zonation is based on damage estimation from historical

catalogues and empirical correlation with intensity. The corresponding acceleration values are thus representative of a soil type which is neither a "good" rock site, neither an alluvial site (as villages were rarely built in alluvial valleys, due to the flow hazard). Following the work done by Rttener (1995), and Smit and Rttener (1998), we chose to consider a reduced value of ag = 0.12 g, for an ideal rock site in the Visp area. 4.2.2 H/V measurements Measurements of the ground natural frequency were conducted at 80 points in the considered area, using the H/V ratio experimental technique. The a priori estimated S-wave velocities of the numerical model were then adjusted, keeping the layer thickness fixed, so as to find the measured natural frequencies for weak motion. In the present case, velocity adjustments between 10 and 25 % were necessary. An example is given in Table 1. 4.2.3 Numerical simulations The geological profile determined as described above was used to perform numerical simulations of the local amplification due to site effects. A pragmatic methodology was chosen for these calculations, using SHAKE (1-D, pseudo non-linear approach) and Aki-Larner (2-D, accounting for geometrical effects) programs with the following procedure: 2-D calculation by means of the Aki-Larner technique for weak motion, adjustment of the S-wave velocities according to the measurements of the H/V ratio, in order to obtain the correct fundamental natural frequency in the transfer function; 1-D calculation for strong motion, accounting for the approximate influence of non-linear behaviour of the soil, followed by an adjustment of the S-wave velocity and the damping (Q values), as a function of the mean soil deformation, for each layer (Table 1 gives an example of the pseudo non-linear velocity adjustments and Q values in the centre of the Rhne valley);

2-D calculation for the strong motion with the new values for S-wave velocities and damping.

4.3 Results The numerical simulations made it possible to characterize site effects in the Visp area. The results obtained are of the same order as site effects already observed or calculated in other similar areas. This study proposed to distinguish three zones : zone "Rhne valley", zone "alluvial cone", zone "rock". The whole Rhne valley has been defined as a single "zone" in this microzonation, accounting for a "mean" 2-D effect of the valley. This approach is in agreement with the recent conclusion of the work performed by Chavez-Garcia and Faccioli (2000), indicating that "the 2-D basin amplification factor can in first approximation be assumed constant throughout the surface of the valley". Following the results of the microzonation study of Visp, an elastic response spectrum has been defined, for each zone, for the design of standard buildings (see Figure 9).

5. Conclusions
One of the main points of this overview is that the instrumental approach based on analyses of earthquake recordings is principally a reliable technique. However, the existence of non-linear soil effects a firmly established reality compromises the validity of amplification factors obtained from weak motion measurements to a certain extent. Whenever non-linear effects must be taken into account in numerical approaches, however, the computations are significantly impaired by the uncertainties in the measurement or estimation of the non-linear constitutive characteristics of the soil. These uncertainties are at least as large as those that affect the measurement or estimation of the soil characteristics at small deformations. Numerical approaches remain of primary importance to help understand the physics of site effects. Their advantages for practical estimations of amplification factors at specific sites must not be overlooked, since the instrumental approach may not always be applicable, for instance in urban areas with weak seismicity (gap areas, or intraplate seismicity zones with large return periods). However, sensitivity studies (see Field and Jacob, 1993a) draw attention to the need for multiple, redundant geotechnical measurements (which increases the actual cost of numerical estimations). Ambient vibration methods are clearly important because of their low expense, as the H/V ratio technique based on horizontal to vertical spectral ratios, for example. Additional research must be carried out to establish the reliability and actual limits of these techniques. For instance, can the H/V ratio method be used to provide quantitative estimations of site amplification factors ? The main lessons learned on the physics of site effects are i) the growing reconciliation of seismologists' and engineers' viewpoints on non-linear soil effects and ii) the accumulating experimental and numerical evidence on the engineering importance of 2-D or 3-D effects (wave diffraction by surface or subsurface topography). Much work, both in research and of regulatory character, remains to be done in order to transfer the accumulated knowledge about site effects to the engineering practice. For instance, regulations are still needed on how to anchor peak ground accelerations for soil and rock sites in design-response spectra so as to account for amplification effects caused by surface and subsurface topography. In summary, although significant advances have been achieved in recent years, some issues regarding the physics of site effects as well as the manner in which to consider them in engineering practice remain unresolved (Bard, 1997) : Basic research is needed, with both theoretical and experimental approaches, in order to better understand some particular aspects of site effects: surface topography effects, effects of strong lateral discontinuities, actual importance of non-linearity in soil response, actual level and effects of differential motion on structures, site-city interaction effects in densely urbanized areas. Methodological work is required to better assess and compare the reliability, cost and usefulness of the various methods available for the prediction of site effects. Last, but not least, some regulatory work is needed to better account for site effects in seismic codes. The extreme importance of site effects in recent damaging earthquakes calls for special efforts to apply right now what we already know regarding site effects without waiting for results of further research. The state of the art is such that it is now possible to perform "present day"

10

microzonation studies. The task at hand is certainly not simple, and it will not only encounter just technical problems. It must be remembered that the results of such microzonation studies are intended for use by local authorities, city planners, land-use specialists and civil engineers, whose technical background and preparation differ greatly. Thus, the recommendations must be very clear and based on sound judgement. Programs recently launched in several countries will eventually clarify the methodologies required for this important goal.

11

6. References
Aki K. and K.L. Larner (1970), Surface motion of a layered medium having irregular interface due to incident plane SH waves, J. Geophys. Res., 75, pp. 933-954. Aki, K. and K. Irikura, 1991. Characterization and mapping of earthquake shaking for seismic zonation, Proceedings of the Fourth International Conference on Seismic Zonation, August 25-29, Stanford, California, E.E.R.I. (editor), Oakland CA, 1, 61-110. Bard, P.-Y. (1997), Local effects on strong ground motion: Basic physical phenomena and estimation methods for microzoning studies, Advanced study course on seismic risk (SERINA), Thessaloniki, Greece, pp. 229-299. Bard, P.-Y., A.-M. Duval, B. Lebrun, C. Lachet, J. Riepl and D. Hatzfeld, 1997. Reliability of the H/V technique for site effects measurement: an experimental assessment. In: Seventh International Conference on Soil Dynamics and Eathquake Engineering, July 19-24, Istanbul, Turkey. Bard, P.-Y., 1999. Microtremor measurements: a tool for site effect estimation ?, State-of-theart paper, Second International Symposium on the Effects of Surface Geology on seismic motion, Yokohama, December 1-3, 1998, Irikura, Kudo, Okada & Sasatani, (eds), Balkema 1999, 3, 1251-1279. Bonilla, L.F., J.H. Steidl, G.T. Lindley, A.G. Tumarkin and R.J. Archuleta, 1997. Site amplification in the San Fernando valley, California: variability of site-effect estimation using the S-wave, coda, and H/V methods, Bull. Seism. Soc. Am. 87-3, 710-730. Borcherdt, R.D. 1970. Effects of local geology on ground motion near San Francisco Bay. Bull. Seism. Soc. Am. 60, 29-61. Bour, M., 1993. Simulation de forts mouvements du sol partir de petits sismes utiliss comme fonctions de Green empiriques, PhD thesis, Universit Louis Pastuer, Strasbourg Chavez-Garcia, F.J., L.R. Sanchez and D. Hatzfeld, 1996. Topographic site effects and HVSR. A comparison between observations and theory. Bull. Seism. Soc. Am., 86-5, 1559-1573. Chavez-Garcia F.J. and E. Faccioli (2000), Complex site effects and building codes: Making the leap, Journal of Sesimology, 4, pp. 23-40. Cornou, C. 1998. Etudes thoriques et numriques sur la mthode de Nakamura-Nogoshi. Mmoire de Diplme d'Ingnieur, EOST Strasbourg-LGIT Grenoble, 132 pp. (in French). Courboulex, F., A. Deschamps, M. Cattaneo, F. Cosli, J. Deverchere, J. Virieux, P. Augliera, V. Lanza and D. Spallorossa, 1998. Source study and tectonic implications of the 1995 Ventimiglia (border of Italy and France) earthquake (Ml= 4.7), Tectonophysics, 290, 245257. CyberQuake (1998), User's Guide, Version 1.1 (Version 2 in Arbeit), BRGM, Orlans. Dobry, R. and M. Vucetic, 1987. Dynamic properties and seismic response of soft clay deposits, Proc. Int. Symp. on Geotech. Eng. of Soft Soils, Mexico city, Vol. 2, pp. 51-87. Duval, A.-M. 1994. Dtermination de la rponse d'un site aux sismes l'aide du bruit de fond: Evaluation exprimentale. Thse de Doctorat, Paris VI, 265 pp. (in French). Duval, A.-M., P.-Y. Bard, J.-P. Mneroud and S. Vidal, 1994. Usefulness of microtre-mor measurements for site effect studies. In: Proceedings of the Tenth European Conference on Earthquake Engineering, Vienna, Austria, Balkema, Duma Ed., I, 521-528. Duval, A.-M., P.-Y. Bard, J.-P. Mneroud and S. Vidal, 1995. Mapping site effects with microtremors. In: Proceedings of the Fifth International Conference on Seismic Zonation, October 17-19, Nice, France, Ouest Editions Nantes, II, 1522-1529.

12

Faccioli, E. 1991. Seismic amplification in the presence of geological and topographic irregularities. In: Proceedings of the Second International Conference on Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, March 11-15, St. Louis, Missouri, S. Prakash (editor), Univ. of Missouri-Rolle, 2, 1779-1797. Fh D. and P. Suhadolc (1994), Application of numerical wave-propagation techniques to study local soil effects: the case of Benevento (Italy), PAGEOPH , Vol. 134, No. 4, pp. 513536. Fh D., Rttener E., Noack Th. and Kruspan P. (1997), Microzonation of the city of Basel, Journal of Seismology, 1, pp. 87-102. Field, E.H. and K. Jacob, 1993a. Monte Carlo simulation of the theoretical site response variability at Turkey Flat, California, given the uncertainty in the geotechnically derived input parameters, Earthquake Spectra, 9-4, 669-702. Field, E.H. and K. Jacob, 1993b. The theoretical response of sedimentary layers to ambient seismic noise. Geophysical Res. Lett. 20-24, 2925-2928. Field, E.H., A.C. Clement, K.H. Jacob, V. Aharonian, S.E. Hough, P.A. Friberg, T.O. Babaian, S.S. Kara-petian, S.M. Hovanessian and H.A. Abramian, 1995. Earthquake site response in Giumri (formerly Leninakan), Armenia using ambient noise observations, Bull. seism. Soc. Am., 85, 349-353. Field, E.H., and K. Jacob, 1995. A comparison and test of various site response estimation techniques, including three that are non reference-site dependent, Bull. seism. Soc. Am., 85, 1127-1143. Finn, W.D., 1991. Geotechnical engineering aspects of seismic microzonation. In: Proceedings of the Fourth International Conference on Seismic Zonation, August 25-29, Stanford, California, E.E.R.I. (editor), Oakland CA, I, 199-250. Gli L., Bard P.-Y. and Jullien B. (1988), The effect of topography on earthquake ground motion: a review and new results, Bull. seism. Soc. Am., 78, pp. 42-63. Gitterman, Y., Y. Zaslavsky, A. Shapira and V. Shtivelman, 1996. Empirical site response evaluations: case studies in Israel. Soil Dyn. Earthq. Eng. 15, 447-463. Hartzell, S.H., 1978. Earthquake aftershocks as Greens functions. Geophys. Res. Letters, 5, 104-107. Heuze, F.E., T.S. Ueng, L.J. Hutchings, S.P. Jarpe, P.W. Kasemeyer, 1995. A New SeismicGeotechnical Strong Motion Approach, Proc. 3rd Int. Conf. on Recent Advantages in Geotechnical Earthquake Engineering and Soil Dynamics, Vol. III, Paper No 8.11, St. Louis, Missouri: 1311-1318 Hutchings, L. , 1994. Kinematic earthquake models and synthesized ground motion using empirical Green's functions. BSSA, 84, 1028-1050. Irikura, K., 1983. Semi-empirical estimation of strong ground motions during large earthquakes. Bull. Disas. Prev. Inst., Kyoto Univ., 33, 63-104. Irikura, K., 1986. Prediction of strong acceleration motions using empirical Green's functions, Proc. 7th Jap. Earth. Eng. Symp., 151-156 Irikura, K. and K. Kamae, 1994. Estimation of strong ground motion in broad-frequency band based on a seismic source scaling model and an empirical Green's function technique, Annali di Geofisica, XXXVII-6, 1721-1743. Kobayashi, K., 1980. A method for presuming deep ground soil structures by means of longer period microtremors, Proc. of the 7th WCEE, Sept. 8-13, Istanbul, Turkey, 1, 237-240.

13

Konno, K. and T. Ohmachi, 1998. Ground-motion characteristics estimated from spectral ratio between horizontal and vertical components of microtremor. Bull. Seism. Soc. Am. 88-1, 228-241. Kudo, K. 1995. Practical estimates of site response, State-of-the-Art report. In: Proceedings of the Fifth International Conference on Seismic Zonation, October 17-19, Nice, France, Ouest Editions Nantes, 3, 1878-1907. Lachet C. and Bard P.-Y. (1994), "Numerical and Theoretical Investigations on the Possibilities and Limitations of Nakamura's Technique", J. Phys. Earth, 42, pp. 377-397. Lachet, C., D. Hatzfeld, P.-Y. Bard, N. Theodulidis, C. Papaioannou and A. Savvaidis, 1996. Site effects and microzonation in the city of Thessaloniki (Greece): comparison of different aproaches. Bull. Seism. Soc. Am. 86, 1692-1703. Langston, C.A., 1979. Structure under Mount Rainier, Washington, inferred from teleseismic body waves, J. Geophys. Res., 84, 4749-4762. Lebrun, B. 1997. Les effets de site: tude exprimentale et simulation de trois configurations. Thse de Doctorat de l'Universit Joseph Fourier - Grenoble I, November 27, 208 pp. (in French). Lermo, J. 1992. Observaciones de microtremores en Mexico y su aplicacion en la ingenieria sismica. MS Thesis, UNAM, 75 pp. Lermo, J. and F.J. Chavez-Garcia, 1993. Site effect evaluation using spectral ratios with only one station, Bull. Seism. Soc. Am. 83, 1574-1594. Lermo, J. and F.J. Chavez-Garcia, 1994. Are microtremors useful in site response evaluation ? Bull. Seism. Soc. Am. 84, 1350-1364. Mueller C., 1985, Source pulse enhancement by deconvolution of an empirical Green's function, Geophys. Res. Lett., 12, 33-36. Nakamura Y. (1989), "A Method for Dynamic Characteristics Estimations of Subsurface Using Microtremors on the ground Surface", QR RTRI, 30, pp. 25-33. Nakamura, Y, 1996. Real-time information systems for seismic hazard mitigation UrEDAS, HERAS and PIC. Q.R. of R.T.R.I., 37-3, 112-127. Nechtschein, S., P.-Y. Bard, J.-C. Gariel, J.-P. Mneroud, P. Dervin, M. Cushing, B. Gaubert, S. Vidal and A.-M. Duval, 1995. A topographic effect study in the Nice region. In: Proceedings of the Fifth International Conference on Seismic Zonation, October 17-19, Nice, France, Ouest Edition Nantes, II, 10671074. Nogoshi, M. and T. Igarashi, 1971. On the amplitude characteristics of microtremor (Part 2), Jour. seism. Soc. Japan, 24, 26-40 (in Japanese with English abstract). Ohmachi, T., Y. Nakamura and T. Toshinawa, 1991. Ground motion characteristics in the San Francisco Bay area detected by microtremor measurements. In: Proceedings of the Second International Conference on Recent Advances in Geotechnical Earthquake Engineering and Soil Dynamics, March 11-15, St. Louis, Missouri, S. Prakash (editor), Univ. of Missouri-Rolla, 1643-1648. Pedersen, H., B. LeBrun, D. Hatzfeld, M. Campillo and P.-Y. Bard, 1994. Ground motion amplitude across ridges, Bull. seism. Soc. Am., 84, 1786-1800. Rsonance, Rovina and Tissires (1999), Mikrozonierung der Region Visp, Zwischenbericht, CREALP, Sion. Riepl, J., P.-Y. Bard, D. Hatzfeld, C. Papaioannou and S. Nechtschein, 1998. Detailed evaluation of site response estimation methods across and along the sedimentary valley of Volvi (EURO-SEISTEST). Bull. Seism. Soc. Am. 88-2, 488-502.

14

Rttener E. (1995), Earthquake hazard evaluation for Switzerland, Gophysique, no. 29, Commission Suisse de Gophysique. Schnabel B., Lysmer J. and Seed H.B. (1972), SHAKE a Computer Program for Earthquake Response Analysis of Horizontally Layered Sites, Report EERC, pp. 72-12. Seekins, L.C., L. Wennerberg, L. Margheriti and H.-P. Liu, 1996. Site amplification at five locations in San Francisco, California: a comparison of S waves, codas and microtremors. Bull. Seism. Soc. Am. 86-2, 627-635. Shiono, K., Y. Ohta and K. Kudo, 1979. Observation of 1 to 5 sec microtremors and their applications to earthquake engineering, Part VI: existence of Rayleigh wave components, Jour. seism. Soc. Japan, 32, 115-124 (in Japanese with English abstract). SIA 160 Norm (1989), Einwirkungen auf Tragwerke, Schweizerischer Ingenieur- und Architektenverein, Zrch. Smit P. and Rttener E. (1998), Spektrale Beschleunigungsabminderung in der Schweiz: Eine erste Anwendung zur Berechnung der seismischen Gefhrdung im Wallis, Sonderdruck aus Bautechnik, 75. Jahrgang, pp. 34-46. Theodulidis, N., P.-Y. Bard, R.J Archuleta and M. Bouchon, 1996. Horizontal to vertical spectral ratio and geological conditions: the case of Garner Valley downhole array in Southern California, Bull. Seism. Soc. Am. 86, 306-319. Toshinawa, T., J.J. Taber and J.J. Berrill, 1997. Distribution of ground motion intensity inferred from questionnaire survey, earthquake recordings, and microtremor measurements - a case study in Christchurch, New Zealand, during the 1994 Arthurs Pass earthquake. Bull. Seism. Soc. Am. 87-2, 356-369. Tumarkin A.G., R.J. Archuleta, 1994. Empirical ground motion prediction, Annali di Geofisica, XXXVII-6, 1691-1720. Yamazaki, F. and M. A. Ansary, 1997. Horizontal-to-vertical spectrum ratio of earthquake ground motion for site characterization, Earthq. Engng. and Struct. Dyn., 26, 671-689. Zar, M., P.-Y. Bard and M. Ghafory-Ashtiany, 1999. Site categorization for the Iranian strong motion network, Soil Dyn. and Earthq. Engng, 18, 101-123.

15

Table 1 :

Estimated and adjusted S-wave velocities and Q factors for the soil layers in the centre of the Rhne valley at Visp.
A priori estimated S-wave velocity for weak motion [m/s] 243 286 412 444 539 641 800 900 1000 1100 1200 S-wave velocity adjusted according to "noise" measurements [m/s] 290 340 490 530 650 770 960 1080 1200 1320 1440 Pseudo non-linear S-wave velocity for strong motion [m/s] 262 282 420 444 555 741 930 1050 1170 1290 1410 Pseudo nonlinear Q factor for strong motion [] 10 6.0 7.4 6.3 7.2 33 38 38 42 42 45

Depth [m] 0-4 4-9 9-18 18-22 22-45 45-70 70-100 100-130 130-170 170-210 210-250

16

Figure 1 :

Differences between 1-D and 2-D behaviors for a perfectly elastic (no damping) sediment-filled valley. These diagrams represent the spatial (x) and temporal (t) evolution of the surface motion of a sediment-filled valley impinged by a SH signal with a characteristic frequency fp=_1/4h (_1 is the S wave velocity in the sediments, h is the valley thickness). a) Results obtained with a 1D approximation ( considering only, for each site x, the local sediment thickness) b) Results obtained a 2D modeling, in the case of a shallow valley: h/w = 0.06 c) Results obtained a 2D modeling, in the case of a thick valley: h/w = 0.70

17

Figure 2 :

An example of a large topographic effect at Pine, France. The 5 top diagrams display average spectral ratios obtained at 5 surface sites ( one strandard deviation), for the EW (in-plane) component. The reference station is not the base station ("Roya"), but the westernmost, mid-slope station ("Bramafan"). A crosssection of the investigated topography is shown on bottom. (Reproduced from Nechtschein et al., 1995).

18

Figure 3 :

An example of use of microtremors through H/V ratio technique. Horizontal to vertical spectral ratios of ambient noise are shown for various locations around an alluvial coastal plain in the city of Nice. The H/V ratios (thick continuous line) are compared with classical spectral ratios with respect to a rock reference site (thin continuous line), and with 1D numerical estimates of the transfer function (thick dashed line). After Duval, 1994.

19

a)

Figure 4 :

Theoretical checks of the H/V ratio technique. Numerical horizontal to vertical spectral ratios were computed from noise models in various (about 15) soil profiles. a) Comparison between the S-wave resonance frequency (fs, computed for vertically incident S-waves) and the peak frequency "observed" in theoretical H/V ratios (fn): the agreement is very satisfactory. b) Same thing for the spectral amplitude at the resonant frequency (As is the amplification for vertically incident S waves, An is the amplitude of the H/V peak obtained from noise modeling): the agreement is very poor... (Reproduced from Lachet and Bard, 1994).

Figure 5 :

Experimental checks on the H/V ratio technique, based on instrumental recordings at 33 different soft sites in different areas from France and Greece. Relationship between the amplitude of the H/V peak and the peak amplification measured on earthquake spectral ratios with respect to a reference site. There is no correlation but an apparent inequality : the former is almost systematically lower that the latter.

20

Figure 6 :

Comparison between various techniques for the estimation of site response transfer function, for two sites in Oakland (California). Adapted from Field and Jacob, 1995. a) Traditional spectral ratios; b) Generalized Inversion spectral ratios whatever the noise ratio); c) Generalized Inversion spectral ratios obtained when only data with signal to noise ratio larger than 3 are kept (and then given an equal weight). d) Parameterized inversion estimates; e) Average Horizontal-to-vertical component spectral ratios for S wave part of earthquake records. f) Nakamura's estimates (average horizontal-to-vertical spectral ratio of ambient vibrations) Curves a) to c) correspond to site-reference techniques, while curves d) to f) correspond to non-reference site techniques. The dashed-lines represent 95 % confidence limits of the mean.

21

Figure 7 :

An example of strain dependency of normalized shear modulus (top) and damping (bottom), for soft soils with varying plasticity index PI (after Dobry and Vucetic, 1987).

22

Figure 8 :

Schematic representation of the different sources of time delays in the summation of the EGF : path length, finite rupture velocity and dislocation rise time (after Bour, 1993).

Figure 9 :

Chosen elastic response spectra for the three zones "Fels (Rock)", "Schuttkegel (alluvial cone)" and "Rotten (Rhone valley)".

23