This study is focused on three steps: (1) detailed mapping of site response functions using microtremors recordings, (2) use geological information and borehole data with empirically obtained response functions to derive subsurface models for different sites across the study area and (3) estimate the seismic hazard in terms of Uniform Hazard Site-Specific Acceleration Spectra.

 The Nakamura technique is used to empirically estimate the site response functions. 362 sites in Lod and in Ramla were instrumented for varying length of time. The site response of soil sites exhibit peaks of amplification factors ranging from 4 to 7 in the frequency range 0.5 to 2.5 Hz . These findings suggest that there are significant differences in the uppermost sedimentary layers. The shape of the empirical transfer functions determined in this study, as well as good agreement between both horizontal components, strongly suggest that the site response functions in this region are of 1-D nature, i.e. can be fairly inferred analytically from 1-D subsurface models.

The extensive database of microtremors recordings and the availability of boreholes data concerning thickness of the sediments enabled a reliable comparison between the empirical and the analytical response functions. The shear-wave structures for different sediments were deduced by trial-and-error fitting of the calculated functions to the empirical transfer functions. Consequently we obtained the following values of share-waves velocities: in sand – 350 m/s, in sandstone – 600 m/s, in clay – 500 m/s, in marl – 700m/s, in chalk – 800 m/s and in dolomite –2100m/s. In turn, fixing the S-wave velocities in the different materials, it became possible to adjust the thickness of the sedimentary layers to obtain a reasonable fit between the calculated and observed site response functions.

Once the site response functions are determined, it became possible to apply the SvE procedure to compute uniform hazard site-specific acceleration spectra that meets the criterion of accepted hazard in the Israeli Standard 413, i.e., a 10% probability of exceedence during an exposure time of 50 years and a damping ratio of 5%. The shape of the spectra obtained is significantly different from the one prescribed by Israel Building Code (IS-413). We thus conclude that the current requirements in the  Israel standard 413 to design buildings to sustain earthquakes are inappropriate for Ramla and Lod and we provide an improved design spectrum to be applicable for building in Lod and in Ramla.

Local site response may be evaluated by empirical and theoretical methods. The theoretical method allow detailed analysis of the parameters used in the evaluation; however, they require detailed geo-technical information about the materials through which the seismic waves propagate to the surface. The analytical response of plane SV waves impinging on a single layer overlying a half-space is well known and widely used (Burridge, 1980; Shearer, 1987; Lermo and Chávez-García, 1993).  Empirical methods are based on seismic records on sites; thus, dominant frequency and amplification are determined directly.  Empirical methods can be separated in two categories: those that use two sites and those that use only one site. Borcherdt (1970) introduced the sediment-to-bedrock ratio (the most common approach) that consists of dividing the spectrum of the measured earthquake motions at a site by that of a nearby reference site (rock site).  If the two sites have similar source and path effects, and if the reference site has a negligible site response, then the resulting spectral ratio constitutes an estimate of the site response. This approach identifies, in most cases, the fundamental resonant frequency and is considered to be the most reliable (Rogers et al., 1984, Singh et al., 1988, Jarpe et al., Darragh and Shakal 1991, Borcherdt and Glassmoyer 1992, Gutierrez and Singh, S. K., 1992, Satoh, et al., 1995; Aguirre and Irikura, 1997; Su, et al., 1998; Beresnev, et al., 1998; Hartzell, 1998; Reinoso and Ordaz, 1999). Many investigators, among them Tucker and King (1984), McGatt et al. (1991), Field et al. 1992, Jongmans and Campillo (1993), Liu et al. (1992), Gagnepain-Beyneix et al. (1995), Zaslavsky et al., (1995), Carver and Hartzell (1996), Hartzell et al. (1996), Zaslavsky et al. (2000) evaluated site response function from moderate to weak motions of earthquake. However, this technique requires a number of earthquake records. In regions with relatively low seismicity, it would be necessary to wait for a significant period of time to obtain a usable data set.

Several studies (e.g., Otha et. Al., Yamanaka et al., 1994) applied the Borcherdt’s approach, using ambient seismic noise instead of earthquake. For frequencies smaller than 0.5 Hz, seismic noise is categorized as microseisms and, for higher frequencies, as microtremors. The main advantage given by this approach is the fact that the spectral characteristics of microtremors have been recognized to be associated with the site conditions (Kanai and Tanaka, 1961; Katz, 1976; Katz and Bellon, 1978, Kagami et al., 1986; Zaslavsky, 1987; Gutierrez and Singh, 1992). It has been shown that with microtremors it is possible to identify the fundamental resonance frequency of nears surface soil deposits (see among others Otha et. Al., 1978; Lermo et al.,1988;  Hough et al., 1991,1992).

Nakamura (1989) proposed a method that requires only one recording station. Nakamura hypothesized that site response could be estimated from the horizontal-to-vertical ratio of microtremors. This technique was tested, experimentally and theoretically, at many sites all over the world, by different authors (Ochamachi et al., 1991; Lermo and Chavez-Garcia, 1993, 1994; Lachet and Bard, 1994; Field and Jacob, 1995; Zaslavsky et al., 1995, 1998, 2000; Malagnini et al., 1996; Seekins et al., 1996; Gitterman et al., 1996, Teves-Costa et al., 1996; Theodulidis et al., 1996; Konno and Ohamachi, 1998; Mucciarelli, 1998). Results obtained by implementing Nakamura’s technique support such use of microtremors measurements for estimating the site response of surface deposits.

Figure 1 presents the geological map of the Lod-Ramla area.

This map is a part of geological map of Israel to a scale of 1:200,000 (Sneh et al.). The surface comprises Quaternary sediment rocks and some remnants chalks of Eocene age.

Borehole data are provided by Gvirtzman (1969- PhD Thesis). The borehole data of the Lod-Ramla area are presented in the Table 1. Complementary geological information of the area under investigation, including boreholes data and well drillings is obtained mainly from the Geophysical Institute of Israel and the Hydrogeology Division of the Geological Survey of Israel. 

Table 2 presents the basic lithological units of the rocks overlaying the carbonates of the Bina formation in the investigated area.

         Bina formation, Shadmon, 1959. Bedrock. (by Fleisher L., Gelberman et al.)

Structure of the Top Judea Group in presented in Figure 2B (Fleischer, Gelberman, and Wolf, 1993; Fleischer and Gafson, 2000). The Bina Fm. consists of white to gray limestone and dolomite, containing rudist and coral fragments. The upper contact of the Bina Fm. is unconformably overlaid almost everywhere by the ‘En Zetim Fm. The structural pattern of the Lod-Ramla area is marked by primary synclinal structure, crossed by normal fault along the axis. The depth of the top Bina Fm. increases from 60 m in the northeast part of the area to 350 m in the west part.

        Mount Scopus group   (Flexer , 1968)

‘En Zetim Fm. is composed of marl-chalky facies. The formation is presented as a syncline in the eastern part of the area with 110 m maximum thickness and wedging out in the western part of the area.

There are few local lenses of Taqye and Ghareb fms. (3-15 m) that consist of shale and chalky limestone.            Avedat group (Braun, 1967).

The area of Lod and Ramla also contains the lower part of the Avedat Group, composed of silicified chalks (0-23 m). It has a local spreading in the central-east part of the area and its unconformably overlaing almost everywhere the ‘En Zetim Fm

Saqiye Group ( Gvirtzman and Reiss, 1965).

The Saqiye Group is associated with four major sedimentary cycles, each starting during a marine transgression and terminating with a sharp westward regression, accompanied by deeply cutting erosional unconformities caused by the tectonic and sea level movements (Gvirtzman, 1969). The group overlies, uncomfortably, the Avedat and Mt.Scopus group. The first cycle of Oligocene age consist of deep marine chalky marls – Bet Guvrin Fm. and limestone of Lakhish Fm, deposited in a synclinal position from south to north. Thickness of marl and limestone vary from 0 to 80 meters and from 0 to 15 meters, respectively.

The Lakhish, Beit Guvrin, Taqye, Ghareb, and ‘En Zetim formations are represented by marl facies. Therefore, for the purpose of site response modeling, they will be given as united lithological unit as displayed in Table 3. The isopach map of marl unit is presented in Figure 2A.

The fourth sedimentary cycle of Pliocene age is represented by the transgressional Yafo Fm. (Gvirtzman, 1969). The Yafo Fm. consists of homogenous clay and clayed marl with some coquina beds and a few calcareous sand intercalations, in the upper part of the section. Clay fills the main syncline in the central part of the area. Thickness of clay varies from 0 m in the western part to 150 m in the eastern part of the area, as illustrated in Fig. 3B. The Yafo Fm. is overlaid by the diachronous Kurkar Group.

  Kurkar Group (Gvirtzman, 1969).

Lower part of the Kurkar Group of Pleistocene age are characterized by marine and eolian calcareous sandstone (“kurkar”) of Pleshet Fm. (0-15 meters) and conglomerates of the Ahuzam Fm. (0-10 meters). The upper part is characterized by eolian sands of the Rehovot Fm. Holocene rocks, composed of reddish silty-clayey sand (“Hamra”) with thickness 0-10 m are present in the east and alluvium sediments (sand, soil, gravel, clay and loess with thickness of 0-10 m) are in the western part of the area, as shown in Table 2 and on the Figure 3A.

On the basis of available geological data we selected lithologically homogeneous units (presented in Table 3) for geological modeling of the subsurface to be incorporated with microtremors measurements.

The share-wave velocity ranges were determined directly by seismic refraction surveys in different sites across Israel. 

1.     Dominant frequency of the Fourier spectrum of microtremors.

When applying the assumptions that microtremors consist of vertically propagating S waves and that the source spectrum of microtremors is mainly white noise then the dominant frequency of the Fourier spectrum of the microtremors is a close approximation of the resonance frequency of the soils column. Good results were achieved by Lermo et al. (1993,1994) who applied this technique to determine dominant periods in Mexico City.

2.     Sediment-to-bedrock spectral ratio.

This technique is based on comparing the spectrum of a seismic event recorded at the investigated site by the spectrum of the same event recorded in a nearby reference site (hard rock). This is also termed the classical technique. The classical technique requires that the propagation paths to the two sites are almost identical, i.e. the investigated site and the reference site are very close to one another with respect to the epicentral distance

3.     The Nakamura technique.

An alternative method for removing the source effect was proposed by Nakamura (1984) and gained much interest because of its low cost, rapid field operations and simplicity of the analysis.   

Figure 4 shows the simple model used by Nakamura. The Nakamura technique is based on the assumption that:

(1)             Microtremors are composed mainly of Rayleigh waves, propagating in soft surface layers overlaying an half-space;

(2)             The vertical motions are not affected by the soft soils.

(3)             The microtremors are originated by local surface sources (traffic and industrial noise) and they have no contribution from deep sources;

(4)             The amplification of the vertical component is exclusively associated with the depth dependence of the surface (Rayleigh) waves motion.  

Figure 4. Simple model assumed by Nakamura (1989) to interpret microtremor measurements.

According to Nakamura, the transfer function (i.e., the site response function) for Rayleigh waves, and compensated for the source spectrum is: 

                                     (1)

  Where Zs= Hs/Vs and Zb=Hb/Vb. Under the prescribed assumptions, the vertical component is not amplified by the surface layer, i.e.,

                                   (2)

we obtain

                                   (3)

  i.e. the vertical component of microtremors on the surface retain the characteristics of horizontal component of the hard rock.

     Ground motions were recorded by GII-SDA system, which is a multi-channel, PC-based, digital seismic data acquisition system designed for site response field investigations (Shapira and Avirav, 1995). The seismometers used are sensitive velocity transducers with a natural frequency of 1.0 Hz. Digital records were obtained in the 0.2-25 Hz band-pass filter with a sampling rate of 100 samples per second. Each of the stations is equipped by one vertical and two horizontal seismometers (oriented north-south and east-west). Table 4 shows the time of measurements at the various sites. The length of recording for each measurement is an important parameter. Too short a period will result in an unreliable average spectral ratios, a too long period increases project efforts and costs.

     At each site, the microtremors were recorded continuously, for two hours, creating data files of 3 minutes of microtremors data. Based on previous experience, we selected from each data file several segments of only 30 sec for spectral calculations. This time window is proven to be sufficiently long to provide stable results. The selected time windows were Fourier transformed using cosine tapering before transformation. The spectra were then smoothed with a triangular moving Hanning window (0.4 Hz) and re-sampled every 0.024 Hz in the 0.5-10 Hz frequency range. After data smoothing, and in order to obtain spectral ratios, the spectra of an EW and NS channel at a site were divided by the spectra of the vertical channel (Nakamura estimate). The arithmetical average of each individual ratios were also computed. We observed that there is practically no difference between arithmetical and geometrical averaging. The ratios are computed using the program of N. Perelman (Perelman, 2001).  If the averaged ratios of the NS and the EW component are similar, than the average of the two component ratios will be the site amplification function:

     Prior to the recording of microtremors, we checked and determined the transfer functions of the instrumentation. This information is essential to facilitate transformation of the recorded signals into ground motion data, i.e., particle velocity. The individual seismometer constants (free-period, damping and motor constant) were determined from sinus and step calibration signals. The instrument characteristics of the stations are given in Table 5.

     In addition, all seismometers were placed at the same location and in the same orientation to record the same waves (Figure 6). These measurements provide relative calibrations between the different channels of the entire monitoring system. Figure 7 presents, as an example, the seismograms and the corresponding spectra of vibration of 8 horizontal and 4 vertical seismometers. The amplitude spectra show that the measured motions and thus the transfer functions of the seismic channels, are practically identical.

While the details of spectra of different time windows vary, their shapes are similar. These spectra are relatively flat between 0.2 and 1.0 Hz, but increase in amplitude near 3 Hz and decay with an approximately constant decay rate to 10 Hz. Figure 9B presents individual and average horizontal-to- vertical spectral ratio for this point No. 176 (i.e. the empirical transfer function as estimated with the Nakamura technique). As shown, the transfer function has a dominant peak near 1.0 Hz. The observed amplification factor is about 5. Examples of different time windows of the horizontal and the vertical components of microtremors at Point No. 88 are shown in Figure 10A. The solid lines present the average. All of the spectra show a well-defined peak at about 3.0 Hz, while spectral ratios (Figure 10B) clearly exhibits the resonant peak at about 1.9 Hz with amplification up to 7. Figure 11A shows individual and average spectra of microtremors at Point No. 208 . Two spectral peaks are visible in all horizontal spectra: a peak centered around the low frequency of 0.6 Hz and a peak centered near the higher frequency of 3.0 Hz. The peak at frequency 0.6 Hz is probably associated with the site response characteristics. Figure 11B presents individual and average horizontal to vertical spectral ratios. These curves show prominent peak at about 0.6 Hz with amplification of about factor 4 in a fair agreement with the spectra of the horizontal components. However, in most cases, due to the influence of sources from the dense population, high traffic and various industries the resonance frequency cannot be directly identified from microtremors spectra (Zaslavsky, 1988, Zaslavsky et al., 1998).

The individual and average horizontal-to-vertical spectral ratios obtained from microtremors for Points 91 and 261 are shown in Figure 12. These figures also demonstrate a similarity among the individual functions not only in terms of the peak position and magnitude, but also in the whole shape. The dominant feature of all spectral ratios for Point 91 is the high spectral ratio level at a frequency of about 1.8 Hz with amplification up to a factor 7.0. In the case of Point No. 261 we can see prominent peak at about 0.6 Hz with amplification up to factor 4.5. The shape of the empirical site response functions, as well as good agreement between both horizontal components strongly suggest that site response functions in the Lod-Ramla region are of 1-D nature.

Examples of the average experimental site response functions obtained by Nakamura technique for the northeastern, central and southwestern parts of the region under study are displayed in Figures 13, 14, 15. We noticed that parameters of site effects are remarkably robust: Comparing two neighboring points we find that the differences in the location of the fundamental frequencies and amplification level are small and the general shapes of the two horizontal components are similar. These findings significantly increase the reliability of the obtained information and emphasize the importance of a dense grid of observation points in microzoning studies.

Dominant frequencies and amplifications at all measurement sites across the Lod-Ramla region are summarized in Table 6.

The microtremors measurements yield a series of parameters that can be used to estimate the expected ground motion during earthquake. The key elements for seismic hazard scenarios (Shapira et al., 2001) are maps of the predominant frequency and maximum relative amplification of ground motion. The map in Figure 16, created on the basis of measurement data (see Table 6) shows the distribution of the fundamental frequency across the Lod-Ramla region. This figure integrates all the experimental data obtained in this study. The site response functions of the soil sites exhibit peaks at dominant frequencies between 0.6 to 2.5 Hz, which, in turn, are correlated with the total sediment thickness. The lower resonance frequencies (range from 0.5 to 0.9 Hz) are attained at those sites in West-central part where the depth bedding of the top Bina Fm. is 350 meters, at most. The higher resonance frequencies (range from 2.0-2.7 Hz) are attained at those sites in Northeast part of the area, where the depth of the top Bina Fm. is only 60-80 meters from the surface.

The map of maximum amplification (Figure 17) reflects the variation of the impedance between the bedrock (carbonates of Bina Fm.) and the overlying sediments. The highest amplifications (range from 5-7) are attained at areas where deposits (clay Yafo Fm.) are lying directly over the rock basement. The lower amplification (range from 3-4) corresponds to areas where marl deposits are placed over the bedrock.

Measurements of ambient noise have been carried out very close to or directly at drilling sites where detailed information about the subsurface structure, namely the thickness of the sediments, is available. At that point, we could combine the borehole information with the observed site response functions to develop a 1-D model of the subsurface and also suggest S-wave velocity estimations for the soils that characterize the investigated region.

Initial parameters were chosen in accordance with geological information about the area investigated. Analytical response functions, computed for multi-layers, 1-D models, were compared with the empirical determinations (horizontal-to-vertical spectral ratios of microtremors). By means of trial-and-error we found the S-wave velocities that yield the best similarity between the analytical and the observed fundamental resonant frequency and maximum amplification.

The sediment rocks of the investigated area and the some characteristic geotechnical parameters are presented in the Table 3. As a first step, data from Table 3 (including velocities) were used to compute the analytical functions. This trial yield unsatisfying results, as demonstrated in Figure 18 and in Table 7  (Model 1). 

In order to obtain more detailed and reliable information about the subsurface structures, we incorporated available logging data. Considering both the logging analysis and the detailed description of the wells, the upper layer (characterized in the Table 3 as sand and calcareous sandstone) were separated into three layers: hamra, sand and sandstone. Figure 19 and Figure 20 show the inferred depth maps of hamra, sand and sandstone. It should be emphasized that the inferred isopach maps in Figures 19 and 20 are based primarily on the information from the boreholes logs. The corresponding S- velocities in these layers are: 200-250 m/s, 300 – 550 m/s and 600 – 1100 m/s.

Figure 18 illustrates the selection of the optimal model for borehole L-11.

B. Thickness of sedimentary layers

Shear waves velocities, derived from modeling at the borehole sites, were summarized and fixed for five selected sediment layers (see Table 9). These fixed velocity values and the microtremors measurements are used for estimating the thickness of the sediments at every point of measurement.

The procedure of 1-D modeling at borehole site is repeated for sites where no borehole information is available. In the trial-and-error process of adjusting the analytical response function to the empirical one, the soil thickness varied until an acceptable match is observed.  Results for 51 selected sites are presented in Table 10 (layer thickness) and in Table 11 (main response parameters).

The thickness of hamra, sand and sandstone were taken from isopach maps, presented on the Figures 19 and 20. We used the 1-D models of near-by borehole sites (where the information is highly reliable) to provide the initial values of thickness of the clay and marl. 

In the trial-and-error process we altered the thickness of clay and marl (and consequently also of the depth of top Judea group). Figure 22 presents isopach maps of clay, created on the basis of boreholes data (Figure 22B) and on the basis of our measurements (Figure 22A). The new subsurface models yield a thickness of the synclinal area to be 40 m more than is given in Figure 22B and reaches a thickness of 270 m.  In the southwest part of the area, the thickness of clay is 140 m, i.e. 40 m less than inferred from contouring the boreholes data.

Figure 23A displays the isopach map of the marl, which is the sum of marly facies of Lakhish, Bet. Guvrin, Adulam and ‘En.Zetim formations. In the central-eastern part of the area, the estimated thickness of marl is 25-30 m more than initially estimated. Also, at syncline B we estimate a thickness of 95 m, which is 45 m thicker than estimated from the initial isopach maps.

Comparing the structural data given in the map of Top Judea Group (Fleischer & Gafson, 2000) (depth taken from the surface) with our estimates,  we can note, that structural elements generally coincide but the depth of occurrence of the Top Judea Group in the central part of the area is 50 m deeper than that on the Fleisher map. Two anticlinal crests A and B were well distinguished in our map, as shown on the Fig. 24A.

The concept of a standard response function raises some seismological difficulties. The most significant among them are:

1. The probability of exceeding PGA is not necessarily the same as the probability of exceeding the frequency dependent accelerations obtained from the re-normalized response spectrum.

2. The response spectrum must depend on the spectra of ground motions from earthquakes of different magnitudes, mechanisms and distances from the investigated site.

3. The “corrections” applied in the standard response spectrum for different site conditions are often found to be non-realistic.

In an attempt to overcome these significant difficulties, Shapira and van Eck (1993) developed the SvE approach. The SvE flowchart is shown in Figure 25. As described in this flowchart, SvE computations are based on regional parameters such as distribution of seismogenic zones, frequency-magnitude relationships, stress drop, Q-values, seismic moment - magnitude relationship, etc. which have been evaluated and routinely updated by the Seismology Div. of GII from local and regional earthquakes (see, for example van Eck and Hofstetter, 1989, Shapira and Hofstetter, 1993, Shapira and Shamir, 1994, Hofstetter et al., 1996, Shamir et al., 2001 ).  The uncertainties associated with those parameters are incorporated in the SvE by applying Monte-Carlo statistics, i.e., simulating several possible lists of earthquakes over a very long time and synthesizing many accelerograms (an accelerogram is the acceleration time history on the free surface of the investigated site). The regional information is used to synthesize accelerograms for the surface of the underlying bedrock and then they are convolved with the response function of the site under investigation, to yield the expected accelerations on the free surface of that site. Following our practice and that of many investigators world wide, the site response functions must be evaluated empirically.

As stated above, the SvE method was used to evaluate the seismic hazard for the area of Lod and Ramla.

Three hundred sixty measurements of microtremors were registered to produce the distribution map of the predominant frequency of site response. Based on our observations we divided the study area into four zones, each characterized by a fundamental resonance frequency of the soil column (see Figure 26).

The four zones are:

Zone 1. Characterized by a resonance frequency between 0.5 and 0.8 Hz.

Zone2. Characterized by a resonance frequency between 0.8 and 1.4 Hz.

Zone 3. Characterized by a resonance frequency between 1.4 and 1.8 Hz.

 Zone 4. Characterized by a resonance frequency between 1.8 and 2.5 Hz.

The Uniform hazard site-specific acceleration spectra for the 4 zones was computed for a probability of exceedence of 10% during an exposure time of 50 years and damping ratio of 5%. For each zone we used 10-15 analytical response functions. The results of our computations are shown on the Figures. 27-30. We also plotted the acceleration response spectrum that is required in the same area by the current IS-413 for ground types S2 and S3 with the required design horizontal Peak Ground Acceleration (PGA) value of 0.1g. The shape of the spectra obtained foe for all zones are significantly different from the ones prescribed by IS-413, in that IS-413 underestimates theacross the accelerations in the period range from 2 sec. to 0.1 sec.

The Lod-Ramla area is characterized by moderate irregularities of soils yielding an overall amplification of earthquake ground motions with no remarkable differences from one zone to another (see Figures 27 to 30) . In other words, although the study area can be divided into 4 zones with respect to the resonance frequency, the seismic hazard functions do not show a distinctive difference from one zone to the other. Consequently, it is suggested to adopt an overall response spectrum, which accounts for amplification effects throughout the whole investigated area. From the engineering standard point, the area of Lod and Ramla can be characterized by the uniform hazard site-specific acceleration spectra shown in Figure 31.

We strongly recommend that estimations of earthquake loss scenarios should be based on our results. The other functions are presented for reference and information purposes only.

A methodology for using microtremors spectral ratios as an aid to perform earthquake hazard microzoning in intensely populated areas of Israel has been developed. This method requires the accumulation of microtremors data at a grid of sites over the region of interest. The horizontal-to-vertical spectral ratios obtained from microtremors (Nakamura technique) proved to be valuable tool determine frequencies of deep and shallow soft soils with multi layer distribution and linear behaviors. We obtained a good correlation between the empirical and the analytical evaluations of the fundamental frequencies as well as for the amplification levels. The site response spectra exhibit peaks between 0.5 and 2.0 Hz with typical amplification factors of 4-6. We propose two maps that reflect the fundamental characteristics of possible site effects in the area: dominant frequency and maximum amplification.

The acquired data demonstrated that while the spectra of microtremors varied with time due to varying noise conditions, their shapes, for all the sites, are similar and characterized by relatively flat levels between 0.4 and 1.5 Hz. The amplitudes increase at about 3 Hz. Consequently; we conclude that the identification of the fundamental frequency of soil from the observed maximum Fourier amplitude of microtremors is misleading and probably incorrect.

The Nakamura technique proved to be valuable tool to determine the dominant resonance frequency of shallow and deep soils with a multi layer distribution. Results from simple numerical simulations indicate that there is a good agreement between the results of Nakamura’s technique and analytical assessments obtained from 1-D model computations. These findings strongly suggests that the technique of Nakamura effectively compensates for source effects in microtremors measurements, which eliminates major limitations to their applications in earthquake engineering.

The relatively large database of microtremors recordings and the availability of logs of borehole information facilitated the comparison between the empirical and the analytical assessments. The shear-wave structures, for different sediments, were deduced by trial-and error fitting of observed and theoretical transfer functions. This in turn, helped to adjust the thickness of the sedimentary layers at locations where no borehole information is available. The practical relevance of the investigation is illustrated by means of map of depth to the bottom of the sediments and maps of thickness of different layers of sediments.

The fundamental frequencies determined in the present study were found to correlate well with the thickness of the sediments in Lod-Ramla area. The sediments are deeper on the southeastern part of Lod-Ramla (site response spectra exhibit peaks at 0.5-0.8 Hz), while they are shallower on the northwestern part (predominant frequency of site response at 0.9-2.0 Hz). The sites in Lod and Ramla show significant amplification. Thus, we divided the study area into four zones, each characterized by a fundamental resonance frequency of the soil column.

Modern practice in designing buildings to withstand earthquakes is to apply a modal analysis, while taking into account the expected ground accelerations that will shake the building. This practice is also applied in many modern buildings codes. The SvE method (Shapira van Eck, 1933) was deployed to comply with such needs under situations when no real acceleration data are available. The computations in the SvE are based on the stochastic method (Boore, 1983, Boore and Atkinson, 1987) and then convolve the synthetic accelerations with the response function of the site. The analytical response functions of the sites are computed by using the computer code of Joiner (1977).

 The Uniform Hazard Site-Specific Acceleration Spectra were computed for a probability of exceedence of 10% during an exposure time of 50 years and damping ratio of 5%. The shapes of the spectra are significantly different from the one prescribed IS-413, in that IS-413 significantly underestimates the requirements (accelerations) across the period range 0.1 sec to 2.0 sec. From engineering standard point it is therefore proposed,  that the whole area of Lod and Ramla be characterized by the uniform hazard acceleration spectrum shown in Figure 31.

Apart from the comments directly related to the Ramla-Lod area, we would like to add the following comments:

1. Microtremors measurements, in combination with Nakamura’s technique, can be a powerful tool to map sedimentary cover layers. In regions of unknown basement morphology, such a procedure may be a way to quickly obtain a general idea of the subsurface structure. The dense grid of site measurements facilitated the development of isopach maps for the marl and top Judea group.
2. The joint application of analytical and empirical techniques for assessing soil response functions can provide useful feedback to improve the reliability of the obtained results. A detailed comparison of the analytical and empirical values constitutes a low-cost, efficient and fast procedure in order to establish the spatial dependence of both suitability and reliability of the method, improvement of models assumed and delimitation of those areas for which in-depth surveys are needed for proper assessment of soil response.
3. As demonstrated in this study, the site response variations are significant over very short distances, thus we strongly recommend that estimation of earthquake loss scenarios should be based on the site response functions obtained over a relatively dense grid of measurement points.
4. It must be realized that the applicability of the Nakamura technique is heavily depended on the amount of data. Analysis of ambient seismic noise encounters significant limitations primarily associated with the great variability of interfering sources of seismic energy. It is therefore essential to monitor the microtremors for a number of hours, sometimes preferably during different hours of the day, to increase the chances of selecting the appropriate time windows (of only 30 sec) that are used in the analysis. Here, quantity is a necessity to achieve quality.
5. An important issue that was raised before and during the investigations is the question of how dense should the grid of measured points be ?

The Lod-Ramla project was the first in Israel where site response functions are determined over a dense grid. In retrospective we may state that we gained reliability to the obtained results ONLY because we had a dense grid of measured sites. Again, it has to do with the application of the Nakamura Technique. Further more, in most of Israel t here is very limited availability of densely distributed geotechnical information such as S-wave velocities and densities of the materials, especially at depth. We could compensate for the need of a dense grid of measured points of microtremors by drilling new borehole, conduct many geophysical surveys and monitor strong enough earthquakes at points across the area. These alternatives are by far more expensive, time consuming and may not always provide the necessary information.