You are on page 1of 15

Bulletin of the Seismological Society of America, Vol. 86, No. 4, pp.

991-1005, August 1996

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves


b y E d w a r d H. Field

Abstract

various site-response estimates are presented for a linear array deployed in the Coachella Valley, California, during the 1992 Landers/Big Bear aftershock sequence. This systematic comparison is unique in that the response of the site is clearly dominated by basin-edge-induced waves. Average sediment to bedrock spectral ratios for long S-wave windows, which include the basin-edge-induced waves, exhibit amplification factors as high as - 1 8 , below 7 Hz. The deep basin structure, which gives rise to the obvious multi-dimensional effects, produces a fundamental resonant peak that shifts between basin sites (from 0.23 to 0.48 Hz) as the depth to bedrock changes. Above 0.6 Hz, where the largest amplifications occur, the response is remarkably similar between sites and appears to be dominated by a near-surface layer that is relatively uniform across the valley. The apparent fundamental resonant frequency of this layer is between 0.8 and 1 Hz. Sediment to bedrock spectral ratios computed using shorter windows that exclude the basin-edge-induced waves imply that the multi-dimensional effects are significant only below - 4 Hz, where they increase amplifications by an approximate factor of 2. Spectral ratios computed using coda windows, taken at twice the S-wave travel time, exhibit amplifications that are an average factor of 1.7 greater, between 1 and 4 Hz, than those of the S-wave estimates. This discrepancy does not improve by taking coda windows later at four times the S-wave travel time. Horizontal- to vertical-component S-wave spectral ratios do not agree with the sediment to bedrock ratios. However, they do exhibit a clear peak at the fundamental resonant frequency of the deep basin structure. Sediment to bedrock spectral ratios of ambient seismic noise are also inconsistent with the S-wave estimates. However, horizontal to vertical noise ratios exhibit clear peaks near the fundamental resonant frequencies of both the deep basin structure (below 0.6 Hz) and the suspected near-surface layer (between 0.8 and 1 Hz). Therefore, ambient-noise data appear to provide valuable constraints on the basin structure. Ongoing efforts involve multi-dimensional modeling of the observed basin-edgeinduced phases and resonant frequencies.

Introduction The strong influence of near-surface geological conditions, in the form of sediment amplification or site response, has been apparent from the damage distribution of many destructive earthquakes. The long list of examples starts at least as far back as the great Japan earthquake of 1891, for which Milne (1898, p. 82) wrote that "... destruction was marked on every plane between Tokyo and Nagoya, a distance of 200 miles, while those who from the dividing ranges looked down upon the clouds of dust and smoke rising from fallen and burning towns, themselves suffered little if any damage." Modern examples include the 1985 Michoacan earthquake in Mexico (e.g., Celebi et al., 1987); the 1988 Spitak earthquake in Armenia (e.g., Borcherdt et al., 1989); 991 the 1989 Loma Prieta (e.g., Hough et al., 1990) and the 1994 Northridge (e.g., EERI, 1994) earthquakes in California; and the Hyogo-Ken Nanbu (Kobe) Earthquake in Japan (e.g., EERI, 1995). There are many factors that influence the way a site will response to earthquake ground motion [see Aki (1988), Aki and Irikura (1991), and Bard (1994) for seminal reviews]. In addition to the 1D structure, these include (1) the source location, which influences the angle, azimuth, and type of incident waves; (2) the prevalence of energy focused or scattered from lateral heterogeneities, such as surface waves generated at the edge of a basin that propagate as energy trapped within the sediments; and (3) the degree to which

992 sediments behave nonlinearly, which causes the response to depend on the level of input motion. Efforts to make the inclusion of site effects in seismic hazard assessments more sophisticated have been hampered by our limited understanding of these complicating factors. Part of the problem is that the concept of site response is often vague, since its distinction from path effects is somewhat artificial and arbitrary. For example, in the engineering community, it is common to regard the site effect as the influence of the upper 30 m. However, seismologists often think of surface waves generated at the boundary of a sedimentary basin (which may have dimensions of several kilometers) as a site effect. Therefore, it is important that the meaning of site response be clearly defined. For hazard assessment, a practical definition would seem to be the unique behavior of a site, relative to other sites, that persists given all (or most) of the potential sources of earthquake ground motion in the region. For such a definition to be useful, the variability of the response at any particular site, with respect to the different potential earthquake sources, must be less than the differences between the distinguished sites (Aki, 1988). Given this definition, what is the best way to estimate or predict site response? Theoretical approaches are instructive, although conducting the necessary sensitivity tests (with respect to different locations and sizes of possible events) and uncertainty analyses (with respect to our limited knowledge of the Earth's structure) is usually impractical. Ideally, one would like to compare empirical observations from every potential event at every site of interest, which is obviously not possible. Empirical studies typically utilize the aftershocks following a large event to infer the relative response of various sites. For example, Hartzell et al. (1996) recently performed a generalized inversion for the site response of 90 sites using aftershocks of the Northridge earthquake. Similarly, Gao et al. (1996) demonstrated that differing levels of damage between neighboring sites in Santa Monica during the Northridge earthquake can be attributed to basin-edge effects. As stressed by Gao et al. (1996), the results of these studies pertain to sources located within the Northridge aftershock zone, and it is not clear whether they will apply to other source locations as well. In fact, according to the definition of site response given above, it is questionable whether such observed differences between sites should be referred to as site response. The differences could result more from regional path effects specific to the Northridge aftershock zone. Alternatively, one could refer to the differences as site effects, yet acknowledge that the response may be intrinsically variable given other source locations. Regardless of which definition one chooses, the observed differences between sites may apply only to the Northridge earthquake sequence, which we are now less concerned about in terms of hazard assessment. This problem is endemic to many empirical site-response studies. It does not diminish their value, however; to the contrary, it is important that many such studies are conducted using data from dif-

E.H. Field feting source regions so that persistent differences between neighboring sites, and the intrinsic variability at each site, can be ascertained. One widely used technique for estimating site response involves using S-wave coda (e.g., Tsujiura, 1978; Phillips and Aki, 1986; Mayeda et al., 1991; Kato et al., 1995; Su and Aki, 1995). The original motive for this method was to take advantage of the relatively abundant vertical-component network data that is generally clipped on the direct arrivals. However, the most appealing aspect of this approach is that since the coda is thought to be composed of scattered energy coming in from a variety of directions (Aki and Chouet, 1975), the site-response estimate might naturally reflect an average over various source locations. Unfortunately, there is some disagreement as to whether coda siteresponse estimates are consistent with the site response of direct S waves; while some have found agreement between the two (e.g., Tsujiura, 1978; Kato et al., 1995), others have not (e.g., Margheriti et al., 1994; Seekins et al., 1995; Steidl et al., 1995; this study). Clearly, more basic research is needed in order to make our inclusion of site response in hazard assessments more sophisticated. In this article, I present some results of a siteresponse study conducted in the Coachella Valley near Indio, California. This site is of particular interest in that it represents a case where the response is clearly dominated by basin-edge-induced waves (in at least a 2D sense). Furthermore, since the phases are generated at the edge of the basin bounded by a fault, the results should help elucidate the hazard in more densely populated fault-bounded basins (such as those underlying the Los Angeles region). Multi-dimensional site effects have been inferred in several studies, often from the fact that one-dimensional modeling could not predict observed amplitudes or durations (e.g., King and Tucker, 1984; Kawase, 1987), or from arraydata analysis (e.g., Horike, 1988; Frankel et al., 1991; Spudich and Iida, 1993). Direct observations of basin-edgeinduced waves have been less common, and these have usually been of longer-period (2 to 10 sec) surface-wave modes (e.g., Hanks, 1975; Koyama et aL, 1988; Vidale and Helmberger, 1988; Hatayama et aL, 1995). However, shorterperiod (--1 sec) scattered phases have also been observed (e.g., Kinoshita, 1985; Gao et al., 1996; this study), and modeling attempts of these have sometimes been remarkably successful (e.g., Scrivner and Helmberger, 1994). The purpose of this article is to present various empirical estimates of spectral amplification for a site that is clearly dominated by basin-edge-induced waves. A detailed interpretation of these and other observations in terms of the valley structure is presented elsewhere (Field, 1995), and efforts to model the basin-edge-induced waves will be the topic of a future publication. Here, sediment to bedrock spectral ratios computed using long (40 sec) S-wave windows are presented for all of the sites. To investigate the influence of the obvious 2D effects, these long-window estimates are then compared with sediment to bedrock ratios

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves

993

obtained using shorter S-wave windows that exclude the basin-edge-induced phases. In addition, comparisons are made with sediment to bedrock spectral ratios obtained from the coda and from ambient-noise measurements. Finally, site-response estimates obtained from horizontal- to verticalcomponent spectral ratios of both S waves (Lermo and Chavez-Garcia, 1993) and ambient noise (Nakamura, 1989) are also examined. The results of these comparisons are placed within the context of previous findings in the discussion section. Since the aftershocks examined in this study represent relatively weak motion, the widely debated issue of nonlinear site response is not addressed here. Readers interested in this topic are referred to Beresnev et al. (1995) for a brief review and recent case study, and to Wennerberg (1996) and Chin and Aki (1996) for a lively debate.

for analysis here because they were recorded at both the reference site (ROKW) and at least one of the sediment sites. These 101 events are listed in Table 2, and their epicenters are plotted in Figure la. In addition to the aftershock recordings, ambient seismic noise measurements were obtained with the 5-sec velocity sensors at five of the sites (ROKW, GOLF, FRED, JEAN, and ROKN) in order to address the question of how well ambient seismic noise can be used to predict earthquake site response. Finally, since we did not record any ground motion that might be high enough to excite a significant nonlinear response, SSA-ls were installed on a semi-permanent basis at ROKW, FRED, and FIRE to record any future strong ground motion. The Working Group on Southern California Probabilities (1995) has placed a 22% chance of rupture before 2024 for the Coachella Valley segment of the San Andreas Fault.

Coachella Valley Deployment On the day following the 28 June 1992 Landers (M 7.3) and Big Bear (M 6.4) earthquakes, a site-response study was initiated in the Coachella Valley under support from the National Center for Earthquake Engineering Research [see Field et al. (1992) for a preliminary report]. A regional map showing the aftershock zone and study area is given in Figure la. The deployment was constructed to address four currently unresolved questions with respect to sediment amplification: (1) To what extent are nearby bedrock-site recordings representative of the ground-motion input at the base of the sediments, and can this source of uncertainty be isolated from an intrinsic variability in the site response itself? (2) To what extent do surface waves generated at the valley edges influence ground motion within the sediments (socalled two- or three-dimensional effects)? (3) How well can one predict sediment amplification using ambient seismic noise? (4) To what extent do the sediments behave nonlinearly during strong ground motion? Eight sites were occupied during the 10-day experiment (Fig. lb, Table 1). The overall deployment was in the form of a linear array stretching from an outcrop of Mesozoic granite on the northeast side of the valley (site ROKE near Fargo Canyon), across the Quaternary sediments occupying the valley (sites JEFF, FIRE, JEAN, FRED, and GOLF), and ending on equivalent Mesozoic granite at the southwest side of the valley (site ROKW near La Quinta). An additional rock site (ROKN) was also occupied on the west side of the valley during part of the experiment. Each site was equipped with a three-component set of either Kinemetric 5-sec velocity sensors (SH- 1, SV- 1), Terra Technology force-balanced accelerometers (FBAs), or both. These were paired with EDA PRS4 digital recorders. Except where otherwise noted, all of the data presented in this article were recorded with the 5sec velocity sensors. Although 227 associated events were recorded during the experiment, 101 of these were chosen Coachella Valley Structure The Coachella Valley constitutes the northern extension of the Salton Trough in southern California. Unfortunately, the sedimentary structure beneath the valley surface is not well known (Biehler et al., 1964; Sylvester and Smith, 1976). The main structural constraints come from gravity data. While the pioneering work was conducted by Biehler (1964), a more recent gravity-data compilation and inversion has been conducted by Bob Jackens at the United States Geological Survey (personal comm.). Both studies suggest that the basin is a half-graben tilted toward the northeast. A schematic cross section based on the Jackens data is shown in Figure lc, which implies that the sediments gradually thicken in going from the southwest edge at ROKW and reach a maximum thickness of 5 ___ 1 km just before the steep basin edge at the San Andreas Fault (SAF). The gravity data do not resolve any significant sediment thicknesses northeast of the SAF, even though they are present in the surface-geology map in Figure lb. Damte (1995) recently presented a cross section based on gravity measurements for the northwestern Mecca Hills area ( - 1 5 km southeast of Indio, as shown in Fig. lb) that is consistent with the results of Jackens. Unfortunately, the gravity studies cannot resolve any structure within the sediments. Detailed mapping in the Indio Hills area (e.g., Popenoe, 1959; Rymer et al., 1987) and mapping, dense gravity, and shallow refraction profiles in the Mecca Hills (e.g., Sylvester and Smith, 1976; Potter, 1992; Rymer, 1994) suggest that the region just northeast of the SAF is composed of anastomosing high-angle faults with associated folding. Therefore, while the surface sediments may be relatively thin northeast of the SAF, these studies, and indeed the very existence of the Indio Hills and Mecca Hills faults, suggest that the northeast boundary of the basin may be very complex. The depth extent of this complexity remains uncertain.

994
a 24300` 243" 30' 244' 00`

E.H. Field

34"30'

34" 30'

34" 00'

34'00`

33 ~ 30'

33" 30'

243"00'

243" 30'

244" 00`

243.6E

244,0E 33.8 N

Quaternary,DuneSand Nonmarinesediments (Pleistocene)

U-]

33.6N
Quatem,'u 3, Alluvium & Lake Deposits Mesozoic Granitic Rocks

ll

A
ROKW GOLF FRED JEAN FIRE

s~
JEFF

A t

_ _ &

ROKN

Figure 1. (a) Regional map showing aftershock epicenters (stars) and recording sites (triangles). The Coachella Valley is the fiat region northwest of the Salton Sea. The thin solid lines represent mapped faults. (b) Simplified geological map of the study area (outlined by the rectangular box in Fig. la) based on the California Division of Mines and Geology Santa Ana Sheet. The triangles mark the deployment locations. (c) Schematic sedimentbasement contact cross section for the dashed line A-A' in Figure lb. This is based on an inversion (Bob Jackens, written correspondence) of 2-kin-grid gravity data using a density/depth function based on well information throughout Nevada and from basins along the San Andreas Fault. Jackens believes that while the absolute depths are uncertain (about _ 1 km at the deepest locations), the relative depths are accurate. The triangles mark the approximate points where the station locations project onto the dashed line. Note that the inversion does not resolve any significant sediment thicknesses northeast of the San Andreas Fault.

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves Table 1 Station Locations
Site ROKW FRED FIRE JEFF ROKE GOLF JEAN ROKN Latitude 33.6672 33.6881 33.7081 33.7454 33.7450 33.6812 33.7006 33.7021 Longitude 243.7161 243.7438 243.7856 243.8382 243.9004 243.7288 243.7679 243.6994

995

T i m e Series Observations Figure 2 shows an example time series for a magnitude 4.4 event. The direct P and S waves are amplified at the sediment sites relative to the rock sites. However, the largest amplitudes are produced by a phase that appears to propagate across the valley from the northeast edge (toward ROKW). The S-wave arrival-time delays across the array are consistent with the basin model suggested by the gravity data. That is, the large delay at FIRE relative to JEFF and the shortening of delay times at sites between FIRE and ROKW are consistent with the half-graben model depicted in Figure lc. Note the anomalous ringing behavior at site JEFF, especially on the vertical component. Although outside of the main basin suggested by the gravity data (Fig. lc), this site was actually located on the Indio Hills Fault. Therefore, the anomalous behavior at JEFF may be related to fault-zone wave-guide effects, as observed elsewhere (e.g., Li et al., 1994), or to the general complexity of this region, as discussed above. The relatively large-amplitude arrivals following the direct S waves at the main basin sites were observed consistently among the aftershocks. For example, as in Figure 2, the highest amplitude observed at FRED was often produced by a phase arriving --12 sec after the direct S wave on the north component (this arrival is shaded in Fig. 2). Shown in Figure 3 is the peak amplitude of this phase (determined from velocity seismograms filtered below 2.5 Hz), relative to that of the direct S wave, as a function of epicentral location (see caption for details). This plot shows that the largest excitations are generally produced by Landers aftershocks located just south of the mainshock epicenter but that relative amplitudes greater than unity also occur for other event locations (including some of the Big Bear aftershocks). No significant dependence on hypocentral depth is observed. The average relative amplitude among all of the events plotted in Figure 3 is 1.9, whereas the average relative amplitude measured in an identical manor at ROKW is 0.47. Eighteen of the aftershocks exhibited both strong apparent basin-edge-induced waves and remarkably similar waveforms (these events are plotted with solid stars in Fig. 3 and are underlined in Table 1). All 18 are Landers aftershocks, with 15 of them clustering near 243.6 E and 34.07

N. Stacks of these events are shown in Figure 4 starting at 1.5 sec before the S-wave arrival (see the caption for processing details). Since the basin sites are dominated by energy propagating from the northeast edge, the horizontal components have been rotated into a coordinate system parallel and perpendicular to the valley axis (N45W). Various individual phases can be seen following the S waves in Figure 4. For example, the valley-axis perpendicular component at JEAN exhibits three distinct arrivals at about 6, 8, and 12 sec after the S wave. The first may represent a wave that has propagated directly from the basin edge, and the second and third arrivals may represent phases that have reflected once and twice, respectively, from the top and bottom of the basin layer. By the time they get to the site GOLF, these phases overlap to such a degree that it is probably more instructive to think of them as a surface wave. As shown in Field (1995), the arrival times of many of these phases can be predicted using a simple model where the waves are scattered from the basin edge implied by the gravity data (at the SAF; Fig. lc). However, as discussed in that article and summarized later here, an interpretation mutually consistent among this and other structural constraints (e.g., the gravity-data depths, the spectral analyses presented in this article) has not yet been obtained. Therefore, a detailed explanation of the various observed phases is not attempted here but postponed until the analysis of ongoing multi-dimensional modeling is complete. In spite of our incomplete understanding of the time series, the observations are clearly dominated by waves scattered from the northeast basin edge (at the SAF). The abundance of aftershock data, and the presence of nearby bedrock sites, provides the unique opportunity to examine various estimates of spectral amplification in a valley that exhibits a strong multi-dimensional response (in at least a 2D sense).

Observed Spectral Amplification Sediment to Bedrock Spectral Ratios Log-average spectral ratios, relative to the bedrock site ROKW, were computed for each site using 40-sec windows starting 1 sec before the S-wave arrival. Again, since the basin sites are dominated by energy propagating from the northeast edge, the horizontal components have been rotated into a coordinate system parallel and perpendicular to the valley axis (N45W). The results are shown in Figure 5. Before taking the ratios, each spectrum was smoothed with a boxcar window of width that increased geometrically with frequency (0.05 Hz at 0.1 Hz, 0.2 Hz at 1 Hz, and 0.34 Hz at 10 Hz). Furthermore, data were excluded if the spectral amplitude of the signal was not three times greater than that of the pre-event noise (scaled to account for possible differences in window lengths). Information on the 101 aftershocks used is given in Table 2, which also shows which stations recorded each event. ROKW was chosen as the reference site since it recorded

996

E.H. Field

Table 2 The Landers/Big Bear aftershocks used in this study. All occurred in July of 1992. The event information is from the Southern California Seismic Network. The asterisks indicate that the event was recorded at that site. The underlined events are those used in the stacks shown in Figure 4.
Event (d:hr:min:sec)
01:07:40:29 01:19:36:28 01:20:53:56 02:16:05:37 02:16:32:46 02:17:01:56 02:17:56:38 02:23:08:50 02:23:57:05 03:01:05:50 03:01:18:35 03:03:51:22 03:04:15:50 03:04:54:14 03:05:28:06 03:05:55:42 03:06:27:32 03:19:46:44 03:19:49:56 03:20:38:06 03:20:52:46 03:21:18:22 03:21:52:20 03:22:52:51 04:00:12:09 04:02:32:00 04:03:29:04 04:04:41:43 04:04:48:50 04:04:54:15 04:05:29:51 04:05:48:42 04:05:52:04 04:06:09:52 04:06:22:02 04:07:11:54 04:08:08:08 04:08:25:31 04:08:56:46 04:09:23:40 04:09:36:01 04:10:33:40 04:14:03:39 04:17:16:24 05:01:58:18 05:03:23:12 05:04:45:44 05:05:49:38 05:11:14:38 05:11:28:22 05:11:34:11 05:11:58:12 05:12:01:54 05:19:31:22 05.:22:33:45 06:00:15:30 06:00:44:47 06:01:11:37 06:02:13:58

Mag
( M L )

Depth
( k m )

Latitude (degrees)
34.332 34.171 34.281 34.326 34.340 34.245 34.261 34.105 34.069 34.008 34.244 34.063 34.209 34.151 34.249 34.018 34.424 34.047 34.122 34.022 34.291 34.619 34.055 34.199 33.993 34.089 34.103 34.078 33.927 34.222 34.419 34.255 34.091 34.089 34.246 34.180 34.070 34.131 34.390 34.565 34.293 34.311 33.978 33.873 34.314 33.929 34.645 33.945 34.068 34.170 34.204 33.957 34.245 34.141 34.583 34.101 34.471 34.222 34.564

Longitude (degrees)
243.538 243.589 243.269 243.540 243.330 243.636 243.239 243.589 243.626 243.651 243.611 243.634 243.230 243.572 243.277 243.652 243.489 243.649 243.588 243.644 243.280 243.357 243.620 243.131 243.627 243.144 243.144 243.622 243.673 243.227 243.524 243.564 243.621 243.144 243.215 243.568 243.634 243.602 243.535 243.479 243.114 243.572 243.721 243.743 243.546 243.669 243.346 243.601 243.636 243.180 243.133 243.661 243.210 243.136 243.696 243.602 243.538 243.137 243.717

R O K W

G O L F

F R E D

J E A N

F I R E

J E F F

R O K E

R O K N

3.6 3.2 3.0 2.6 3.2 2.5 2.9 2.6 2.7 2.5 2.8 2.9 3.0 2.4 2.6 3.3 2.5 2.8 3.7 2.8 2.9 3.6 3.3 2.9 3.0 3.3 2.9 2.6 3.3 3.1 2.8 2.8 3.1 3.1 2.6 2.9 2.7 2.6 2.6 2.8 3.3 3.4 2.8 2.8 3.2 2.4 3.0 3.1 3.7 3.0 2.9 2.6 3.2 2.3 4.3 3.0 2.9 3.5 3.4

9.00 0.35 1.42 1.59 0.14 9.37 0.61 3.59 1.83 4.71 0.01 2.14 10.8 1.19 3.42 2.98 3.04 1.86 2.48 0.72 0.00 4.84 1.98 3.74 1.33 2.77 1.73 3.64 2.83 2.85 8.50 4.30 1.73 2.97 6.00 4.64 2.78 1.12 5.14 1.00 3.41 7.09 3.11 2.43 7.80 3.80 4.23 3.20 2.53 9.44 4.42 0.00 7.46 10.0 0.00 4.77 3.94 2.56 0.01

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

---------. ---. --. . -----. --. --. . . ------. . -----* * * * * * * * ------. . . . . . . . . . .

* --* * * * * * . * * * . * * . . * * * * * . --. --. . . -* * * * * . . * * * * * * * * * * * * * * * * * * -. . . . . . . . .

---. . . . --. . --. . . . . .

-* * . . . . ---

* --. . . . ---

---m

* * *

. * *

. * * * * * * * * -* * --* * *

---

-*

-* * *

-----. --.

* * * * *

* * * * *

* *

* * *

-* * * * * * * * * * * * * * ---

---

* *

* * * * *

-------

* ---* --

* * * * * * * *

-. ---* . * * * . * * . * * . * -. . . . .

* . * -* * . * * * . * -. * * . * * .

-* --. ---. --. --. ---

-* ---

----

---

---

---

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves

997

Table

Continued
Event (d:hr:min:sec) Mag Depth (km) Latitude (degrees) Longitude (degrees)

(ML)
2.6 2.6 3.2 3.1 2.9 3.2 3.2 3.3 2.4 4.4 3.2 3.5 2.9 3.0 3.2 3.6 4.3 2.9 3.3 3.1 3.5 3.4 3.4 3.0 3.1 3.2 3.6 3.4 3.1 4.7 3.3 2.7 3.4 3.5 3.3 3.1 2.9 3.1 2.5 2.6 3.0 3.3

ROKW

GOLF

FRED

JEAN

FIRE

JEFF

ROKE

ROKN

06:02:59:26 06:04:48:34 06:06:00:44 06:11:03:13 06:11:35:33 06:11:38:03 06:11:42:53 06:11:46:34 06:11:59:33 06:12:00:59 06:12:26:12 06:12:48:25 06:13:15:09 06:14:19:45 06:18:06:36 06:18:27:27 06:19:41:37 06:20:56:52 07:01:45:38 07:08:21:03 07:08:38:03 07:13:38:03 07:13:53:28 07:14:51:55 07:16:17:54 07:16:47:35 07:21:01:10 07:22:09:28 07:22:21:45 08:02:23:11 08:08:05:38 09:00:22:29 09:01:43:57 09:02:34:35 09:02:34:36 09:02:56:50 09:02:56:54 09:03:14:18 09:05:37:37 09:22:45:45 10:03:40:15 10:05:48:43

4.41 10.2 2.82 10.7 10.1 9.25 5.67 1.99 2.60 1.79 1.71 4.40 5.00 3.85 0.48 1.11 3.27 3.38 5.51 3.24 1.76 0.01 0.46 0.00 0.90 0.80 7.78 2.53 0.32 6.00 10.5 7.39 0.01 0.66 0.00 0.32 0.00 1.81 10.9 0.89 1.16 3.62

34.174 34.052 34.101 34.165 34.093 34.094 34.127 34.221 34.131 34.092 34.047 34.312 34.238 34.370 34.457 34.152 34.082 34.099 34.232 34.069 34.210 34.230 34.398 34.299 34.247 34.431 33.952 34.341 34.273 34.576 34.605 33.954 34.239 34.224 34.141 34.226 34.227 34.425 34.182 34.091 34.325 34.108

243.568 243.572 243.617 243.172 243.554 243.552 243.084 243.255 243.595 243.631 243.635 243.542 243.143 243.534 243.524 243.594 243.622 243.619 243.106 243.619 243.234 243.168 243.536 243.529 243.285 243.521 243.643 243.533 243.275 243.664 243.649 243.642 243.163 243.156 243.577 243.153 243.149 243.518 243.133 243.629 243.349 243.600

* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *

* * * * -* * --* * -* ------* * * * * * * * * * * * * * * * * * * * . ---

* * * * * * * * * * * * * * * * * * * * * * * * * . * * * * * * * * * * * * * . * * . .

. * * * * . * * . * * * * . * * * * * * * * . . . . * * . * * . . . . . . . . . . *

. * . . * . * * . * * * . . * * * . * * . . . . . . * * . * * . . . . . . . . . . -. . . . . . .

. . . .

. -. . -. --. ---. . . ---. --. . . . . . . . --. --. . . . . . . . .

. -. . -. --. ---. . ---. --. . . . . --. --. * * * * * * -----------------

. . .

. . .

. . --

. * *

--

more sponse may Since higher (which

events

than the other rock cannot be ruled rather bedrock

sites. A significant out, so the spectral absolute

site reratios

pected

from the gravity

model

in Figure

lc. Above

-0.6

Hz,

at ROKW

the spectral ratios look quite similar between sites; in fact, the difference between

the main basin is greater. spectral is sur-

represent

relative on

than

site response. significant at

components

site effects frequencies,

are generally

That is, all of these sites have statistically amplifications, prising given on any particular differences

equivalent which

results are shown range

here only below of engineering

10 Hz

component,

is also the frequency 5 reveals overall up to -7 of

interest).

seen in the time-domain main difference between

observathe two Hz seen

Figure

amplification

at the sediment

tions (Figs. 2 and 4). The horizontal components

sites for frequencies as high main as a factor

Hz, with peak values reaching first prominent peak at the Fault:

is a prominent

peak at -2.4 axis, which

18. The southwest

in the direction

parallel to the valley direction.

is not presthe estimates

basin

sites (those

of the San Andreas occurs from below 0.23 --0.6 + is

ent in the perpendicular for the axis-parallel and sometimes look

Furthermore,

FIRE, JEAN, GOLF, frequency FIRE

and FRED)

Hz. The Hz at

component

have peaks near 0.8, 2.4, 4.0, at JEAN), which frequencies of a should

of this peak _+ 0 . 0 6

increases Hz

0.07

5.6 and 7.2 Hz (especially like the a half-space.

to 0.48

at GOLF,

which

qualitatively
shift ex-

conspicuously over

SH

resonant

consistent

with the fundamental

resonant

frequency

single layer

That

this component

998

E.H. Field (northeast of the SAF). Of particular interest is the anomalously large response on the vertical component, which may be related to its being positioned on the Indio Hills Fault. The bedrock site ROKN, located on the same side of the valley as the reference site ROKW, exhibits no significant relative site response. This gives some confidence that ROKW might represent an adequate reference site, although it remains possible that both of these sites have a significant, yet similar, site response. The other bedrock site, ROKE, located on the northeast side of the valley, exhibits relative amplifications of up to a factor of 3. This small amplification could result for two possible reasons. One is that we were not able to locate this site directly on bedrock, but rather on shallow sediments within 15 m of bedrock. Another is that this site, by virtue of being on the opposite side of the SAF, is on a separate lithospheric plate, so the path effects might differ. The observation in Figure 5 that all of the main basin sites exhibit similar site response above 0.6 Hz, yet differences between components, was observed for the other types of estimates discussed below. Therefore, in what follows, only the results for site FRED will be shown as a representative example. The same geometrically increasing smoothing width has been applied to all spectral estimates, and the earthquake data were always excluded if the amplitudes were not greater than 3 times that of the pre-event noise. Results for site JEFF will not be shown since this site exhibits anomalous behavior presumably related to its location on the Indio Hills Fault (which, although interesting, is not the focus of this article). Sediment to Bedrock Spectral Ratios with Short Windows In Figure 6, sediment to bedrock spectral ratios for the 40-sec windows (reproduced from Fig. 5) are compared with those for 9-sec S-wave windows that just exclude the basinedge-induced waves. While both estimates exhibit some of the same apparent resonances, the basin-edge-induced waves increase the relative amplification by a factor of - 2 on average between 0.2 and 2.5 Hz and by up to a factor of 3 at some individual frequencies. Above --4 Hz, the two estimates are consistent, implying that the basin-edge-induced waves are significant only at lower frequencies. Sediment to Bedrock Spectral Ratios of S-Wave Coda Average spectral ratios computed from 10-sec windows of coda, taken at both two and four times the S-wave propagation time (2Ts and 4Ts, respectively), are shown in Figure 7 along with the direct S-wave spectral ratios (reproduced from Fig. 5). The coda-estimate amplitudes are generally greater than the direct S-wave estimates below - 5 Hz for the earlier (2Ts) windows and below --3 Hz for the later (4Ts) windows. Although the frequency range of disagreement is less for the later coda windows, the amplitude discrepancy at lower frequency is larger. The discrepancy for the earlier coda windows (2Ts) is about a factor of 1.7 on

14.0 10.5 -

s
i. . . . . . . . . '

North
ROKW GOLF FRED

7.03.50,0 -3.5 -7.0 SAF ~

I i

I 0.5 crrds JEFF ROKE

East
ROKW GOLF FRED FIRE

i
,..~

t4"0 t 10.5 7.0 3.5 J

i ..... '~ ,

0.0 -3.51
-7.0 14.0 i0.5 7.03.50.0 .,i -3.5 -7.0

SAF

~ 0.5 cm/s

JEFF

ROKE

l;
t I ' ~ ....

Vertical

.ow
GOLF FRED JEAN JEFF

" [ 0.5 crrds

SAF

,t"
I I

10

20

I 30

40

I 50

ROKE

(sec) Figure 2. The time series observed across the linear array for a magnitude 4.4 aftershock (event 06:12:00:59 in Table 2). All of these records were obtained with the 5-sec velocity sensors, except for those at JEFF and ROKN, which are integrated acceleration (FBA)data. The epicentral distance and backazimuth, with respect to ROKW, are --48 km and N9W, respectively. The seismograms are offset by the distance of the site from the San Andreas Fault (SAF). Note that the east-component record at ROKE is only digital noise. The light solid lines connect the P- and S-wave arrival times between the two rock sites (ROKW and ROKE).

exhibit an SH-looking response is not surprising since the valley-parallel direction corresponds to SH motion for energy scattered from the basin edge. What is surprising is that the spectral amplifications look one-dimensional when the time series do not. As discussed more later, one possible explanation for the similarity in response above 0.6 Hz is that there is a shallower near-surface layer that is relatively uniform (i.e., one dimensional) between the main basin sites. The fundamental resonance of this layer could be responsible for the highest amplitude peak observed near 0.8 to 1.0 Hz in the horizontal-component ratios. The effect of the deeper basin structure would then be to channel more energy toward this shallow layer. Figure 5 also reveals a significant response at the sediment site JEFF, located outside of the main part of the basin

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves

999

243*00'

24330 '

244000 '

~r r

34 30'

34* 30'

Big Bear Aftershocks

-Y3
~r

~ . ~ .
r ~r

Landers Aftershocks

3400 '

34 00'

~ - = factor of two r = unit ratio

AA

km
0 33 30' 243" 00' 10 20 33* 30'

243"30'

244"00'

Figure 3. Amplitude of the phase arriving 12 sec after the direct S wave on the north component at FRED (shaded in Fig. 2), relative to the direct S-wave amplitude, as a function of epicentral location. The relative amplitudes were determined by filtering the particle-velocity seismograms below 2.5 Hz (third-order Butterworth filter), measuring the peak amplitudes that occurred in a 5-sec window centered on both the S-wave arrival and 12 sec later, and then dividing the latter amplitude by the former. The filled stars represent those 18 events used to make the stacks shown in Figure 4 (note that most plot on top of each other).

average between 1 and 4 Hz, with differences greater than a factor of 2.5 at some individual frequencies. This amplitude discrepancy was also seen for coda estimates computed for windows starting at a constant time lag of 35 sec (which is --2Ts for the most distant event). As discussed more later, this amplitude discrepancy may be related to prolonged reverberations within the basin. Horizontal- to Vertical-Component Spectral Ratios of S Waves Average horizontal- to vertical-component (h/v) spectral-ratio site-response estimates were also computed using 40-sec S-wave windows. The results for site FRED are shown in Figure 8, along with the sediment to bedrock spectral ratios reproduced from Figure 5. In contrast to the results of some previous studies (Lermo and Chavez-Garcia, 1993; Field and Jacob, 1995), the h/v ratios look nothing like the sediment to bedrock ratios. There is, however, a clear peak at 0.24 _+ 0.04 Hz that is near the fundamental resonant frequency apparent in the sediment to bedrock ratios. The results for the other main basin sites look similar, except the low-frequency peak shifts systematically from 0.16 + 0.02 Hz at FIRE to 0.42 + 0.05 at GOLF. This shift is consistent with that seen in the sediment to bedrock spectral ratios (Fig. 5) and with the notion that it reflects a shift in the funda-

mental resonant frequency as the deep-basin sediments thin toward the southwest edge. Ambient-Noise Spectral Ratios Sediment to bedrock spectral ratios of ambient seismic noise are compared in Figure 9 with the S-wave sediment to bedrock ratios (reproduced from Fig. 5). Also shown are h/v spectral ratios of the noise. These noise estimates were computed from 16 consecutive 40-sec windows obtained simultaneously at both sites just after midnight. Although there is some agreement between the horizontal-component sediment to bedrock ratios below - 0 . 9 Hz, above 1 Hz the amplitudes of the noise ratios are up to five times greater than the S-wave estimates. The agreement is worse for the vertical-component sediment to bedrock ratios. The h/v noise ratios in Figure 9 look nothing like the Swave results. However, there is a prominent peak near 0.25 _+ 0.04 Hz that exhibits a shift in frequency at the other main basin sites (not shown). These peak frequencies are statistically equivalent to those observed in the S-wave h/v ratios. Again, these probably reflect the fundamental resonant frequency of the deeper basin structure. There is also a secondary peak near 1 Hz in the h/v noise ratios that is present at the same frequency at the other main basin sites (not shown). Therefore, it may well correspond to the fundamen-

1000

E.H. Field

12"

Axis

Parallel

Parallel
~D

Perpendicular GOLF

Vertical

9-

15

~,

llllllll

I +IIIIIII

FRED

1"
r.,l

1296-

Axis Perpendicular
FRED

15 I0 '

~" " A

<

FIRE 3-

o
-3 12-

San Andre,as

Fault

JEAN
JEFF

15 ~10 ~ "

+~+: '~ ' '

Vertical
; ~ ~ GOLF FRED IgAN
~ F I R E

9- ~ 63-

i i i illlll

i i

iiii1~1

FIRE
15 <+

S a n Andreas Fault 0-3 0


i i i i i I |

0 ~ I "?,/,;;'"+;"T
0.1
1.0

(sec)

10 O.l

1.0

10

0,1

10

I0

Figure 4. Stacks of shear-wave traces at the sediment sites for 18 events that exhibit similar basinedge-induced waves (solid stars in Fig. 3 and underlined in Table 2), beginning 1.5 sec before the S-wave arrival. The horizontal components have been rotated to valley-axis parallel (N45W) and perpendicular (N45E) directions Before stacking, each trace was filtered below 2.5 Hz with a third-order Butterworth filter and normalized by the peak amplitude on the valley-parallel component at FRED (therefore, relative amplitudes have been preserved, but absolute amplitudes have not). The + one standard deviations are shown by the shaded regions The slight differences in S-wave arrival times seen among sites in Figure 2 have not been preserved, and the stacks have been offset by the distance from the San Andreas Fault.

5o]

JEFF

,Jr

i iiiiiii

+ i lllllll

I ltllllll

I I lllllll

ROKB

+t
i i Iiiiiii + i lllllll 10 I |IIIIIII I I lllllll 10

ROKN

tal resonant frequency of the shallow uniform layer suggested earlier as the source of the highest amplification seen in the sediment to bedrock S-wave ratios.

01

1.0

frequency (I-'Iz)

10 0.1 t.0 to o.1 10 ,o frequency (Hz) frequency (Hz)

Discussion
The time series observed in the Coachella Valley are dominated by waves scattered from the steep basin edge at the SAF. These waves cause amplification factors, as measured by sediment to bedrock spectral ratios of 40-sec Swave windows, to reach values as high as 18 at some frequencies. The main basin sites exhibit a prominent peak

Figure 5. Average sediment to bedrock spectral ratios, with respect to ROKW, for each site computed using 40-sec windows starting 1 sec before the S-wave arrival. The horizontal components have been rotated to the valley-axis parallel and perpendicular coordinate system The shaded regions represent the 95% confidence limits (___ two standard deviations of the means), and the dotted line marks the unit amplitude level.

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves

1001

Parallel
18 15-

Perpendicular

Vertical
Figure 6. Comparison of average sediment to bedrock spectral ratios at FRED for 9-sec Swave windows (solid line), which do not include the basin-edge-induced waves, with those of the 40-sec windows (dashed line, reproduced from Fig. 5). In both cases, the windows started 1 sec before the S-wave arrival. The shaded regions represent the 95% confidence limits.

.~
~

129 6

i illtlll

I lllllll

I I |lllllt

I Ill|Ill

I I I Itlll

I I I Itlll

0.1

1.0

l0 O.1

1.0

frequency (Hz)

frequency (Hz)

lO 0.1

frequency (Hz)
Vertical

1.0

10

Parallel

Perpendicular

15

! ~ i i I I

I I lllll|l

t lill|ll

1 I I I IIIII

I [ | IJlll

0.1

frequency (Hz) Parallel

1.0

10 0.1

frequency (Hz)

1.0

10 0.1

frequency (Hz)

1.0

I0

Figure 7. Average sediment to bedrock spectral ratios for site FRED computed from 1 0 - s e c coda windows taken at twice (solid line) and four times (dot-dashed line) the S-wave propagation time. Shown for comparison are the 40-sec S-wave ratios (dashed line) reproduced from Figure 5. The shaded regions represent the 95 % confidence limits (to avoid clutter, no uncertainties are shown for the ratios of coda windows taken at four times the S-wave propagation time).

,8]
3 0.1

Perpendicular

I 1111111

I 1t|tt11

frequency (Hz)

1.0

I0 0.I

frequency (Hz)

1.0

I0

Figure 8. Comparison of average horizontal- to vertical-component spectral ratios (solid line) with the sediment to bedrock spectral ratios (dashed line, reproduced from Fig. 5) for 40-sec S-wave windows (starting 1 sec before the arrival) at site FRED. The shaded regions represent the 95% confidence limits.

below 0.6 Hz, which shifts toward higher frequency in going from FIRE (0.23 + 0.07 Hz) toward GOLF (0.48 + 0.06 Hz). Since both gravity data and S-wave arrival-time differences seen between these sites (Figs. lc and 2, respectively) suggest that the sediments thin in this direction, this lowfrequency peak likely represents fundamental reverberations of the basin. Above 0.6 Hz, where the largest amplifications occur, the response is remarkably similar between the main basin sites (although different between components). Furthermore, the component of motion parallel to the valley axis has prominent peaks near odd multiples of 0.8 Hz, which

suggests the one-dimensional S H response of a single layer. This similarity in response between basin sites is surprising given the obvious multi-dimensional effects observed in the time domain. The present interpretation involves a relatively shallow near-surface layer that is relatively uniform (i.e., one dimensional) across the basin. The fundamental resonant frequency of this layer, which gives rise to the largest amplification factors, appears to be between 0.8 and 1.0 Hz. By comparing the spectral ratios of long S-wave windows (40 sec) with those of relatively short windows (9 sec) that exclude the basin-edge-induced waves, the multi-dimensional effects are found to be influential at frequencies below - 4 Hz. Above 4 Hz, the scattered energy appears to be completely attenuated by the valley sediments. At lower frequencies, the basin-edge-induced waves increase the amplification levels by an approximate factor of 2 (on average between 0.2 and 2.5 Hz). Knowing the threshold frequency where multi-dimensional effects are no longer significant will be useful in theoretical modeling since computational constraints usually govern the highest frequency that can be considered. Sediment to bedrock spectral ratio estimates obtained using coda waves show significantly greater amplification factors than those obtained from the 40-sec S-wave segments. This is especially true between 1 and 4 Hz, where the discrepancy is a factor of 1.7 on average and up to a factor of 2.5 at some frequencies. Such an inconsistency between direct S wave and coda estimates at sediment sites has been observed in at least three other studies (Margheriti et al., 1994; Seekins et aI., 1995; Steidl et al., 1995), which

1002

E.H. Field

Parallel
50_

Perpendicular

Vertical
. ~

e~

0.1

I I Iltlll

I IIIIlll

I I ~tlltl

I IIIIl~l

.......... i"

i IIIIHI

f tllllll

0.1

1.0

10 0.1

1.0

I0 0.1

1.0

10

frequency (Hz)

frequency (Hz)

frequency (Hz)

Figure9. Average spectral ratios at FRED computed using sixteen 40-sec windows of ambient seismic noise. The dot-dashed line represents those taken with respect to the bedrock site (ROKW), and the solid line represents those taken with respect to the vertical component (Nakamura's estimate). For comparison, the average S-wave sediment to bedrock ratios (reproduced from Fig. 5) are shown with the dashed line. The shaded regions represent the 95% confidence limits.

calls into question the reliability of applying coda-amplification factors. However, there are at least two studies showing consistency between the two at sediment sites (Tsujiura, 1978; Kato et al., 1995), leaving the issue unresolved. As first suggested by Phillips and Aki (1986), the discrepancy may be related to the prolonged reverberation of energy at the sediment sites. There is, however, an alternative explanation. Coda estimates are believed to reflect the average response for energy incident at a variety of angles and azimuths, whereas the direct S-wave estimates in this study represent the response to waves emanating from the Landers/ Big Bear aftershock sequence. Therefore, it is possible that the coda estimates accurately reflect the average response expected given a variety of source locations, which could be larger than the estimates obtained for the limited source locations available in this study. Unfortunately, neither one of these explanations can be ruled out here. It is interesting to note, however, that the coda-amplification factor of --9.0 for the vertical component, averaged between 1 and 4 Hz, agrees with that of the direct S-wave estimate for the horizontal component. Since most of the previous coda-amplification studies were based on vertical-component data (e.g., Phillips and Aki, 1986; Mayeda et al., 1991; Su and Aki, 1995), this agreement implies that these previous results may be applicable to horizontal-component S-wave amplification factors. For several sediment sites in Mexico, Lermo and Chavez-Garcia (1993) found that h/v spectral ratios of S waves revealed the frequency and amplification factor of the fundamental resonant frequency observed in sediment to bedrock spectral ratios. Field and Jacob (1995) found that such h/v ratios revealed the overall frequency-dependent character of site response, including several higher-mode peaks, but that the amplitudes were off by a frequency-independent scaling factor of - 1.6. Other more recent studies have found generally mixed, yet encouraging, results (Lachet et al., 1995; Seekins et al., 1995; Malagnini et al., 1996). These estimates are sometimes referred to as receiverfunction-type estimates (Field and Jacob, 1995) because of the analogy with the so-called studies that utilize teleseismic P waves to infer crustal and upper mantle structure (Langston, 1979; Owens et al., 1984). The success of this method appears to be based on the fact that sediments have relatively little influence on the vertical ground motion (predominantly

P waves converted from S waves). Therefore, the vertical component may provide a convenient estimate of the source and path effects (at least for the frequencies where the sediments are significantly amplifying). In this study, the h/v spectral ratios of S waves look nothing like the sediment to bedrock ratios. This may be due to the vertical component being dominated by basin-edgeinduced waves rather than S-to-P conversions at the bedrock contact below the site. There is, however, a prominent peak at low frequency with a shift between sites as the depth to bedrock changes. This suggests the peak is related to the fundamental resonant frequency of the basin. The frequency of this peak is generally about 0.8 times that observed in the sediment to bedrock ratios, although given the uncertainties, this difference may not be statistically significant. Therefore, while the h/v spectral ratios of S waves do not agree with the sediment to bedrock ratios, they do appear to reveal the fundamental resonant frequency of the deep basin structure. Although sediment to bedrock ratios of ambient seismic noise do not compare favorably with those obtained from S waves, the h/v noise ratios appear to reveal resonant frequencies. Many recent studies, both observational (Nakamura, 1989; Omachi et al., 1991; Field and Jacob, 1993; Lermo and Chavez-Garcia, 1994; Yamanaka et al., 1994; Field et al., 1995; Field and Jacob, 1995; Seekins et aL, 1995) and theoretical (Field and Jacob, 1993; Lermo and Chavez-Garcia, 1994; Lachet and Bard, 1995), have shown that h/v ratios of ambient seismic noise can reveal the fundamental resonant frequency of sediment deposits. This is often referred to as Nakamura's estimate, named after the first person to demonstrate the method empirically (Nakamura, 1989). However, since the peak at the fundamental resonant frequency is thought to be a manifestation of Rayleigh-wave particle motion (Lermo and Chavez-Garcia, 1994; Lachet and Bard, 1994), h/v noise ratios are also referred to as "ellipticity" estimates in the literature (Yamanaka et al., 1994). The h/v noise ratios observed in this study have two peaks. The first occurs below 0.6 Hz, which is statistically equivalent to that observed in the h/v S-waves ratios. This peak also has the associated shift in frequency between the main basin sites, implying that it too reflects the fundamental resonant frequency of the deep basin structure. There is also

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-Induced Waves a secondary peak in the h/v noise ratios near 1 Hz that does not shift in frequency between the main basin sites. This may reflect the fundamental resonant frequency of the inferred shallow near-surface layer (that is relatively uniform between sites). Therefore, while the noise data do not look promising in terms of directly predicting site response, the h/v noise ratios appear to reveal the fundamental resonant frequencies of two widely different size scales, that of the deeper basin and that of the inferred shallow uniform nearsurface layer. This information provides valuable constraints on the basin structure. The fundamental resonant frequency of a sediment site is generally equal to the average S-wave velocity divided by four times the sediment thickness (the so-called "quarterwavelength" rule). Therefore, if constraints can be placed on either the sediment thickness or the S-wave velocity, then the other can be determined from the observed resonant frequency. Efforts to determine the basin structure from such considerations are detailed elsewhere (Field, 1995). Unfortunately, a mutually consistent picture has not yet emerged. For example, the gravity data (Fig. lc) suggests a depth to basement of 5 _ 1 km at FIRE. With the observed resonant frequency of --0.23 Hz and the quarter wavelength rule, this translates into an unrealistically high average S-wave velocity of 3.7 to 5.5 km/sec. Therefore, either the gravity-data depth estimates have a gross systematic error or the impedance contrast giving rise to the observed resonance is somewhere above the basement contact within the sediments. By making various assumptions, one can try to estimate shear-wave velocities from the arrival time delays seen in Figure 2 or from the various phases seen in the shear-wave stacks in Figure 4 (Field, 1995). Given observational uncertainties and the range of assumptions one can make, the average shear-wave velocity estimates vary between --0.4 and --0.9 km/sec. With the quarter-wavelength rule, this translates into an impedance-contrast depth between 0.5 and 1.1 km at FIRE (Field, 1995). Constraining the average-velocity/ depth trade-off for the relatively shallow uniform layer inferred between the sites is even more problematic. At this point, there is no independent evidence for its existence, just the persistent peaks at odd multiples of 0.8 Hz at the main basin sites. Although the absolute depth and S-wave velocities in the basin are not yet known, the relative change in depth to basement is well constrained from the observed shift in resonant frequencies, arrival time delays in the direct S-wave, the gravity data, and P- to S-wave conversions at the basement contact (Field, 1995). Clearly, borehole and/or refraction/reflection studies would provide additional constraints and, therefore, may be pursued in the future. In the meantime, it is hoped that ongoing multi-dimensional modeling of the basin response will simultaneously provide additional constraints on the structure and make sense of resonant peaks, arrival times, particle motion, and Hilbert-transform effects seen in the observations. If a satisfactory fit can be obtained, then the important question of how variable the

1003

response is with respect to other source locations can be explored with additional simulations.

Conclusions Several conclusions can be made regarding this site that exhibits a clear multi-dimensional response (in at least a 2D sense). First, by comparing sediment to bedrock spectral ratios of long S-wave windows with relatively short windows, the basin-edge-induced waves are found to be significant only below --4 Hz, where they boost spectral amplifications (on average between 0.2 and 2.5 Hz) by an approximate factor of 2. Therefore, multi-dimensional effects might be approximately accounted for by applying a factor of 2 to 1D-based predictions at those frequencies. Second, siteresponse estimates based on coda waves significantly overestimate amplitudes below 3 to 4 Hz (by up to a factor of 2.5). Third, sediment to bedrock ratios of ambient noise show little agreement with those based on S waves. Similarly, h/v ratios of both S-wave data and ambient noise do not compare favorably with sediment to bedrock S-wave ratios, except in identifying fundamental resonant frequencies that can provide valuable structural constraints. These conclusions apply to a specific set of sites and sources (weakmotion Landers/Big Bear aftershocks), so additional studies will be needed before generalizing to other sites, to other source locations, and to strong ground motion.

Acknowledgments
I would like to thank Klans Jacob for obtaining the initial financial support for this project. Paul Friberg, Noel Barstow, and Doug Johnson were all vital participants in the field deployment, and Amy Clement helped with data processing and preliminary analysis. I am also grateful to Kei Aki, Susan Hough, Yutaka Nakamura, and Paul Spudich for reviewing this manuscript, especially Paul, whose comments led to a significant revision. Finally, I would like to thank Geoffrey Ely, whose assistance saved me hours in preparing Figure la (made with GMT), and Bob Jackens for sharing his gravity-inversion results. The data collection and processing was supported by the National Center for Earthquake Engineering Research Grant Number NCEER-92-1002, and the research was supported by the Southern California Earthquake Center (SCEC), which is funded by the National Science Foundation under cooperative agreement EAR-8920136 and United States Geological Survey cooperative agreement 14-08-0001-A0899. This is SCEC publication number 227.

References
Aki, K. and B. Chouet (1975). Origin of coda waves: source, attenuation, and scattering effects, J. Geophys. Res. 80, 3322-3342. Aid, K. (1988). Local site effects on strong ground motion, Proc. Earthquake Eng. Soil Dyn. II, 103-155. Aki, K. and I. Irikura (1991). Characterization and mapping of earthquake shaking for seismic zonation, Proc. of the 4th International Conference on Seismic Zonation, 1, Stanford, California, pp. 61-110. Bard, P.-Y. (1994). Effects of surface geology on ground motion: recent results and remaining issues, Proc. of the lOth European Conference on Earthquake Eng., t, 305-325. Beresnev, I. A., K. L. Wen, and Y. T. Yeh (1995). Seismological evidence

1004

E . H . Field

for nonlinear elastic ground behavior during large earthquakes, Soil Dyn. Earthquake Eng. 14, 104-114. Biehler, S. (1964). Geophysical study of the Salton trough of southern California, Ph.D. Thesis, California Institute of Technology, Pasadena, 139 pp. Borcherdt, R. D., G. Glassmoyer, A. Der Kiureghian, and E. Cranswick (1989). Results and data from seismologic and geologic studies following earthquakes of December 7, 1988 near Spitak, Armenia, S.S.R., U.S. Geol. Surv. Open-File Rept. 89-163A. Celebi, M., C. Dietel, J. Prince, M. Onate, and G. Chavez (1987). Site amplification in Mexico City (determined from 19 September 1985 strong-motion records and from recordings of weak motions), in Ground Motion and Engineering Seismology, A. S. Cakmak (Editor), Elsevier, Amsterdam, 141-152. Chin, B.-H. and K. AM (1996). Reply to "Comment on Simultaneous study of the source, path, and site effects on strong ground motion during the Loma Prieta earthquake: a preliminary result on pervasive nonlinear effects" by L. Wennerberg, Bull. Seism. Soc. Am. 86, 268-273. Damte, A. B. (1995). Gravity modeling of the basement surface in the Mecca Hills, southern San Andreas Fault (abstract), EOS 76, 597. EERI (1994). Northridge Earthquake of January 17, 1994, Preliminary Reconnaissance Report, Earthquake Engineering Research Institute, Publication Number 94-01, John F. Hall (Technical Editor). EERI (1995). Hyogo-Ken Nanbu Earthquake of January 17, 1995, Preliminary Reconnaissance Report, Earthquake Engineering Research Institute, Publication Number 95-04, C. D. Comartin, M. Greene, and S. K. Tubbesing (Technical Editors). Field, E. H., K. H. Jacob, N. Barstow, and P. A. Friberg (1992). Coachella Valley site-response experiment using Landers earthquake aftershocks in southern California, Bull. Natl. Center Earthquake Eng. Res. 6, 8-11. Field, E. H. and K. H. Jacob (1993). The theoretical response of sedimentary layers to ambient seismic noise, Geophys. Res. Lett. 20, 29252928. Field, E. H., A. C. Clement, V. Aharonian, P. A. Friberg, L. Carroll, T. O. Babaian, S. S. Karapetian, S. M. Hovanessian, and H. A. Abramian (1995). Earthquake site response study in Giumri (formerly Leninakan), Armenia, using ambient noise observations, Bull. Seism. Soc. Am. 85, 349-353. Field, E. H. (1995). Structural constraints for modeling basin response in the Coachella Valley, California, Southern California Earthquake Center Report Number 297 (avail. from the author). Field, E. H. and K. H. Jacob (1995). A comparison and test of various site response estimation techniques, including three that are not reference site dependent, Bull. Seism. Soc. Am. 85, 1127-1143. Frankel, A., S. Hough, P. Friberg, and R. Busby (1991). Observation of Loma Prieta aftershocks from a dense array in Sunnyvale, California, Bull. Seism. Soc. Am. 81, 1900-1922. Gao, S., H. Liu, P. M. Davis, and L. Knopoff (1996). Localized amplification of seismic waves and correlation with damage due to the Northridge earthquake, Bull. Seism. Soc. Am. 86, $209-$230. Hanks, T. C. (1975). Strong ground motion of the San Fernando, California, earthquake: ground displacements, Bull. Seism. Soc. Am. 65, 193-225. Hartzell, S., A. Leeds, A. Frankel, and J. Michael (1996). Site response for urban Los Angeles using aftershocks of the Northridge Earthquake, Bull. Seism. Soc. Am. 86, S168-S192. Hatayama, K., K. Matsunami, T. Iwata, and K. Irikura (1995). Basin-induced Love waves in the eastern part of the Osaka Basin, J. Phys. Earth 43, 131-155. Horike, M. (1988). Analysis and simulation of seismic ground motions observed by an array in a sedimentary basin, J. Phys. Earth 36, 135154. Hough, S. E., R. D. Borcherdt, P. A. Friberg, R. Busby, E. Field, and K. H. Jacob (1990). The role of sediment-induced amplification in the collapse of the Nimitz freeway during the October 17, 1989 Loma Prieta earthquake, Nature 344, 853-855. Kato, K., K. Aki, and M. Takemura (1995). Site amplification from coda

waves: validation and application to S-wave site response, Bull Seism. Soc. Am. 85, 467-477. Kawase, H. (1987). Irregular ground analysis to interpret time-characteristics of strong motion recorded in Mexico City during the 1985 Mexico earthquake, in Ground Motion and Engineering Seismology, A. S. Cakmak (Editor), Elsevier, Amsterdam, 467-476. King, J. L. and B. E. Tucker (1984). Observed variations of earthquake motion across a sediment-filled valley, Bull Seism. Soc. Am. 74, 137151. Kinoshita, S. (1985). Propagation of total reflected plane SH pulses in a dipping layer, Zisin 38, 597-608 (in Japanese). Koyama, S., K. Seo, and H. Yamanaka (1988). On the significant later phase on seismograms at Kumagaya, Japan, Proc. 9th World Conf. Earthquake Eng. II, 385-590. Lachet, C. and P.-Y. Bard (1994). Numerical and theoretical investigations on the possibilities and limitations of Nakamura's technique, J. Phys. Earth 42, 377-397. Lachet, C., D. Hatzfeld, P.-Y. Bard, N. Theodulidis, C. Papaioannou, and A. Savvaidis (1995). An experimental study of the microzonation in the city of Thessaloniki (Greece), submitted to Bull. Seism. Soc. Am. Langston, C. A. (1979). Structure under Mount Rainier, Washington, inferred from teleseismic body waves, J. Geophys. Res. 84, 4749-4762. 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. Li, Y.-G., K. Aki, D. Adams, and A. Hasemi (1994). Seismic guided waves trapped in the fault zone of the Landers, California, earthquake of 1992, J. Geophys. Res. 99, 11705-11722. Margheriti, L., L. Wennerberg, and J. Boatwright (1994). A comparison of coda and S-wave spectral ratios as estimates of site response in the southern San Francisco Bay area, Bull Seism. Soc. Am. 84, 18151830. Malagnini, L., P. Tricarico, A. Rovelli, R. B. Herrmann, S. Opice, G. Biella, and R. de Franco (1996). Explosion, earthquake, and ambient noise recordings in a Pliocene sediment-filled valley: inferences on seismic response properties by reference- and non-reference-site techniques, Bull. Seism. Soc. Am. 86, 670-682. Mayeda, K., S. Koyanagi, and K. Aki (1991). Site amplification from Swave coda in the Long Valley Caldera region, California, Bull. Seism. Soc. Am. 81, 2194-2213. Milne, J. (1898). Seismology, 1st ed., Kegan Paul, Trench, Truber, London. Nakamura, Y. (1989). A method for dynamic characteristics estimation of subsurface using microtremor on the ground surface, QR Railway Tech. Res. Inst. 30, 1. Omachi, T., Y. Nakamura, and T. Toshinawa (1991). Ground motion characteristics in the San Francisco Bay area detected by microtremor measurements, Proc. of the 2nd International Conference on Recent Advances in Geotechnical Earth Engineering & Soil Dynamics, 1115 March, St. Louis, Missourri, 1643-1648. Owens, T. J., G. Zandt, and S. R. Taylor (1984). Seismic evidence for an ancient rift beneath the Cumberland Plateau, Tennessee: a detailed analysis of broadband teleseismic P waveforms, J. Geophys. Res. 89, 7783-7795. Phillips, W. S. and K. AM (1986). Site amplification of coda waves from local earthquakes in central California, Bull. Seism. Soc. Am. 76, 627648. Popenoe, F. W. (1959). Geology of the southeastern portion of the Indio Hills, Riverside County, California, Masters Thesis, University of California, Los Angeles, 153 pp. Potter, T. T. (1992). Gravity and seismic refraction investigation of San Andreas system fault strands in the Mecca Hills area, Riverside County, California, Masters Thesis, California State University, Long Beach, 146 pp. Rymer, M. J., J. Boley, and R. J. Weldon II (1987). Nonuniform rotation

Spectral Amplification in a Sediment-Filled Valley Exhibiting Clear Basin-Edge-lnduced Waves

1005

(Pleistocene) along the San Andreas fault in the Indio Hills, southern California (abstract), EOS 68, 1507. Rymer, M. J. (1994). Quaternary fault-normal thrusting in the northwestern Mecca Hills, southern California, in Geol. Investigations of an Active Margin: Guidebook, 1994 Geol Soc. Am. Cordilleran Section Meeting, Redlands, California, 325-329. Scrivner, C. W. and D. V. Helmberger (1994). Seismic waveform modeling in the Los Angeles basin, Bull. Seism. Soc. Am. 84, 1310-1326. Seekins, L. C., L. Wennerberg, L. Margheriti, and H.-P. Liu (1996). Site amplification at five locations in San Francisco: a comparison of Swaves, codas, and microtremors, submitted to Bull Seism. Soc. Am. 86, 627-635. Spudich, P. and M. Iida (1993). The seismic coda, site effects, and scattering in alluvial basins studied using aftershocks of the 1986 North Palm Springs, California, earthquake as a source array, Bull. Seism. Soc. Am. 83, 1721-1743. Steidl, J. H., F. Bonilla, and A. G. Tumarkin (1995). Seismic Hazard in the San Fernando Basin, Los Angeles, CA: a site effects study using weak-motion and strong-motion data, Proc. of the 5th International Conference on Seismic Zonation, Nice, France, 2, 1149-1156. Su, F. and K. Aki (1995). Site amplification factors in central and southern California determined from coda waves, Bull Seism. Soc. Am. 85, 452-466. Sylvester, A. G. and R. R. Smith (1976). Tectonic transpression and base-

ment-controlled deformation in San Andreas fault zone, Salton Trough, California, Am. Assoc. Petrol GeoL Bull. 60, 2081-2102. Tsujiura, M. (1978). Spectral analysis of the coda waves from local earthquakes, Bull Earthquake Res. Inst. Tokyo Univ. 53, 1-48. Vidale, J. E. and D. V. Helmberger (1988). Elastic finite-difference modeling of the 1971 San Fernando, California, Earthquake, Bull. Seism. Soc. Am. 78, 122-141. Wennerberg, L. (1996). Comment on "Simultaneous study of the source, path, and site effects on strong ground motion during the Loma Prieta earthquake: a preliminary result on pervasive nonlinear effects" by B.-H. Chin and K. Aki, Bull Seism. Soc. Am. 86, 259-267. Working Group on Southern California Probabilities (1995). Seismic hazards in Southern California: probable earthquakes, 1994 to 2024, Bull Seism. Soc. Am. 85, 379-439. Yamanaka, H., M. Takemura, H. Ishida, and M. Niwa (1994). Characteristics of long-period microtremors and their applicability in exploration of deep sedimentary layers, Bull Seism. Soc. Am. 84, 1831-1841. Southern California Earthquake Center University of Southern California Los Angeles, California 90089-0740 Manuscript received 3 October 1995.

You might also like