... aoj ~3uonbozj ptm otu!l snsaos (~;'0 ol I jo sonll~A ol spuodsoaaoo qo~.q~ 'o]!q~ ol ~laglq moaj ~3urpeqs auol£ea~ ,(q patea.~pu!) ~ouza~qoa tllJOjOAeA, ~.
Bulletin of the Seismological Society of America, Vok 86, No. 3, pp. 655-669, June 1996
Seismic Wave-Field Observations at a Dense, Small-Aperture Array Located on a Landslide in the Santa Cruz Mountains, California b y Z h e n g y u Xu, Susan Y. Schwartz, and Thorne L a y
Abstract
A rectangular (4 by 5) array of short-period three-component seismometers with 15-m spacing was deployed to record several U.S. Geological Survey calibration explosions detonated around the Santa Cruz Mountains. The array was located at a site where an earlier station had recorded frequency-dependent polarized site resonances for aftershocks of the 1989 Loma Prieta earthquake. The site is on a hillside believed to be a landslide structure, with the near surface consisting of poorly sorted sediments and weathered rocks with dipping subsurface layers. The primary objective was to explore the site effects in this complex three-dimensional soft-rock environment, characteristic of much of the Loma Prieta source region. The direct P waves from four nearby (15 to 20 kin) explosions at easterly azimuths from the array show counterclockwise arrival azimuth anomalies of 30 ° to 50 °. These deflections are attributed to the presence of more than one dipping velocity contrast beneath the array, with dips of from 10 ° to 50 ° and dip directions generally toward the south. One such boundary may correspond to the landslide slip surface, and the presence of dipping velocity contrasts underlying the site is probably responsible for some of the observed directional site resonance. A slowness vector analysis demonstrates that arrivals early in the P coda have similar azimuthal anomalies, while later scattered arrivals come from many azimuths. Particle motions indicate that the more coherent arrivals in the coda are comprised of scattered P waves and Rayleigh waves, probably associated with scattering from the rough topography in the region. The coda displays greater spatial coherency along the hill strike than down the slope, consistent with a wedge-shaped landslide. The overall wave-field spatial coherence, CCC(f, Ax), decreases with increasing frequency, f, and spatial offset, Ax, and on average can be well represented by CCC(f, Ax) = e -cfAx, with c = 0.6 k m - 1 H z - 1 for the vertical P wave in the first 1-sec window. This behavior is comparable to that found for previously studied hard-rock locations.
Introduction can constrain the large-scale structures and scattering beterogeneities in the crust between the source and the receivers and in the vicinity of the array. Seismic wave coda generally becomes progressively more complex with time, with many interfering arrivals scattered from different directions. Even for small-aperture arrays, there is a transition to an incoherent wave field in which individual arrivals cannot be reliably tracked across the array because of interference from multiple arrivals. Statistical measures of the wave field such as spatial coherence are then used to characterize the overall signal complexity and associated scattering properties of the medium, recognizing that the deterministic substructure of the coda is simply not fully resolvable. The transition from deterministic to incoherent wave field is a fuzzy one, influenced by the signal frequency content, the limitations of the array response, and the local scat-
High-frequency seismic wave propagation in the crust tends to be very complicated, and it is useful to analyze recordings from dense, small-aperture seismic arrays (station spacing from tens to hundreds of meters) to quantify the source and propagation effects that are relevant to earthquake engineering (e.g., Menke et al., 1990; Vernon et at., 1991; Frankel et aL, 1991; Hartzell et al., 1994; Mori et aL, 1994). Array recordings allow the deterministic features of the short-period wave field to be characterized, including the slowness vector and polarization of discrete plane-wave phases that sweep across the array (e.g., Dainty and Toksrz, 1990; Gupta et al., 1990; Bannister et al., 1990; Wagner and Owens, 1993). Localized scattering effects within the array must be treated differently than remote scattering effects, but the scatterers can still be analyzed deterministically (e.g., Hedlin et al., 1991). Identification of the coherent arrivals 655
656 tering properties in the crust. For isolated three-component stations, little can be done other than to monitor the signal polarization as a function of time using the event backazimuth as a reference frame. When an array is located in a homogeneous hard-rock region, the short-period (T < 1 sec) signal coherence drops off significantly over station spacing of just tens of meters (e.g., Menke et al., 1990; Vernon et al., 1991), presumably due to intense scattering from shallow heterogeneity. Even more complex behavior is expected in heterogeneous soft-rock environments. For example, sediment-filled valleys cause frequency-dependent amplifications relative to hard-rock sites (e.g., King and Tucker, 1984; Tucker and King, 1984; Tucker et al., 1984), while roughsurface topography causes short-period signal amplifications (e.g., Geli et al., 1988; Hartzell et al., 1994; Pederson et al., 1994). Many earthquake engineering issues are associated with complex soft-rock sites in tectonically active regions. This article considers three-component recordings of four shallow explosions from a dense, small-aperture array located in the Santa Cruz Mountains above the 1989 Loma Prieta earthquake rupture zone. The objective is to study the short-period wave field in a geologically complex region with rough-surface topography and dipping geological structures. This contrasts with earlier deployments of dense small-aperture arrays in geologically simple environments and provides insight into the possible site effects for isolated sensors deployed for aftershock recordings or earthquake engineering purposes. We initially analyze the deterministic features of the explosion recordings, emphasizing the first arrivals. Many studies of direct P-wave backazimuths have revealed biases of up to tens of degrees (e.g., Niazi, 1966; Buchbinder, 1987, 1990; Magotra et al., 1987; Menke et al., 1990; SuteauHenson, 1990, 1991; Zhang and Langston, 1992). These deflections are commonly attributed to lateral velocity gradients (e.g., Buchbinder, 1987), dipping surface geometry (e.g., Buchbinder and Haddon, 1990), or dipping velocity contrasts beneath the receiver (Niazi, 1966; Cassidy, 1992; Langston, 1977, 1979; Zhang and Langston, 1992). Unusually large backazimuth anomalies are found in our case, which we attribute to dipping subsurface boundaries. We extend the analysis of arrival direction to the P coda using a plane-wave slowness vector procedure (analogous to frequency-wavenumber analysis), along with consideration of the polarization. The local crustal heterogeneity appears to cause significant scattering; longitudinally polarized arrivals impinge on the array from many azimuths. Statistical characteristics of the coda are also considered for comparison with previous work using similar instrumentation and station spacing on hard-rock sites (Menke et al., 1990; Vernon et al., 1991). The influence of the deterministically constrained larger-scale structure on the incoherent signal is examined, along with a consideration of the likely causes of the preferred S-wave polarizations observed in the region. The latter issue will be discussed at greater length in another study (Bonamassa and Vidale, in preparation). The
Z. Xu, S. Y. Schwartz, and T. Lay general conclusion is that shallow structure beneath the site exerts a profound affect on the short-period wave field that would be almost unrecognizable without the advantage of a dense array. This suggests caution in the analysis of isolated station signals for aftershock studies and earthquake engineering applications. The Zayante Dense Array The Santa Cruz Mountains are a tectonically active, geologically complex region with rough-surface topography, contorted sedimentary structures, and relatively low-velocity near-surface rock types. This environment offers many earthquake engineering challenges, including the variability of ground-motion amplification observed for hilltops (Geli et al., 1988; Hartzell et al., 1994) and in sediment-filled valleys (King and Tucker, 1984; Tucker and King, 1984; Tucker et aI., 1984). Among the many interesting site-response characteristics observed in recordings of aftershocks of the 1989 Loma Prieta earthquake, frequency-dependent preferred resonance directions, independent of earthquake location and focal mechanism, were observed in ground shaking at several locations (Bonamassa and Vidale, 1991; Bonamassa et al., 1991). These resonances tend to be spatially highly variable, based on observations from nested tripartite mini-arrays with station spacing of 25 and 300 m. One of the sites that displayed strong frequency-dependent preferred S-wave polarizations was selected as a study area (Fig. 1). The earlier ZAYA mini-array deployed in this region (Bonamassa and Vidale, 1991) exhibited directional resonances at four of six stations, with general consistency in frequency-dependent behavior (Fig. lc), but the other two nearby stations showed no resonances, suggesting very shallow control on the ground motions. To improve our understanding of such site effects, we deployed a small-aperture array, which we call the Zayante array, near the location of the Z5 element of the ZAYA array. The Z5 recordings displayed westerly to northwesterly directions of preferred Swave vibration, loosely paralleling the trend of a nearby ridge. The Zayante array consisted of a 4 by 5 grid of shortperiod Mark Products L-22 three-component geophones with a grid spacing of 15 m (Fig. ld). The signals from the 2-Hz seismometers were sampled at 200 samples per second. The performance of these seismometers is quite good in the 2- to 15-Hz range but deteriorates above this (Menke et al., 1991). The signal-to-noise ratio in our data is highest below 15 Hz, but we do consider the higher-frequency information recognizing that it may have some contamination from instrument nonlinearity. The array was situated on a landslide surface (Gerry Weber, personal comm., 1992) with the hillside dipping about 10° toward the south-southwest. The surrounding structure is the San Lorenzo syncline, a Tertiary structure, with the Lambert shale (Brabb and Dibblee, 1979) underlying the soil in the immediate vicinity of the array. The geometry and depth of the landslide slip surface are not
Seismic Wave-Field Observations at a Dense, Small-Aperture Array Located on a Landslide in the Santa Cruz Mountains
(a)
/
\,
(O)
iI11
2~8" Off
238"
657
15 37" 15'
%..--
\-.
"
f
T00,
/
/
/
/
f
(c)
/
/
/
(d)
IHz & 15Hz
/ /
.~
~ t v n z
,~
-
lo~**-~
3Hz ~ Z1 ,~16Hz
z~
\ N
I
51
52
53
54
41
42
43
44
31
32
33
34
21
22
23
24
(-3)
(~
(1)
(2)
12
13
14
11
\
N
15m
\ 7_6['-]
0
100
\
[]z5
Figure 1. Maps of the ZAYA and Zayante arrays. (a) Base map indicating the location of the study area in the Santa Cruz Mountains of California. (b) Local topography near the San Andreas fault and locations of the ZAYA array and four explosion shotpoints for the U.S. Geological Survey calibration shots. (c) Locations of the six elements of the ZAYA array along with frequency-dependent directions of preferred Swave ground shaking (if any) indicated by the arrows (Bonamassa and Vidale, 1991). The small Zayante array was deployed near element Z5 of the ZAYA array. A nearby ridgetop is indicated. (d) Geometry of the Zayante array, with stations names given by row/column numbering with rows trending east-west and columns, north-south. The 20-element array has 15-m sensor spacing. The numbers in parentheses are the relative elevations of the array elements in meters. known, but presumably, there is a southerly dip. A shallow refraction survey was subsequently conducted in the array vicinity (Bonamassa et al., 1993), finding that the near-surface layer varies in thickness from 1 to 20 m with a P velocity of about 550 m/sec, while the velocity below this layer is about 1250 m/sec. The quasi-planar velocity discontinuity dips about 20 ° toward the southwest. The Zayante array was deployed to record several explosions detonated by the U.S. Geological Survey as part of an effort to calibrate shallow velocity structure in the Santa
Cruz Mountains (Eberhart-Phillips and Michael, 1991). Four of the explosions (Table 1; Fig. lb) were recorded with good signal-to-noise ratio. These events are at distances of 15 to 20 km from the array and have easterly backazimuths varying by about 63 ° . While some ambient noise time intervals were also recorded, no events at other azimuths were detected during the short deployment of the array. The array was primarily intended to explore variability of local ground motions in the vicinity of the site with polarized resonances and was not designed to include outlying elements that
658
Z. Xu, S. Y. Schwartz, and T. Lay Table 1 Explosion Source Information
Shot I D
1 2 3 5
L~
(°N)
37.02 37.11 37.17 37.09
Lon CW)
12t.90 121.85 121.79 121.86
Distmce
(km)
16.14 15.55 20.64 15.08
Back~imu~
139.76 ° 100.80 ° 77.29° 110.32 °
Elev~ion
(m)
79 1034 175 530
53
54
32 33 34
22
would have enhanced the resolution of plane waves traversing the array, but the high sample rate and good signal-tonoise quality of the recordings still allow a limited use of the recordings as a dense small-aperture array.
23
0"1"2"3
'4
5
Time
6
7:8"9"
(sec)
Analysis of Direct-Arrival B a c k a z i m u t h Anomalies The signals recorded at the Zayante array provide a high-quality data set for four nearby explosions of precisely known location. Some of the waveforms for one of the explosions (event 2) are shown in Figure 2, with the station labeling defined in Figure ld. Only minor instrument amplification variability is expected among the sensors, but the emplacement sites were all in pits in the sediments no deeper than 1 m, so ground coupling variability is definitely possible. Our analysis uses normalized amplitudes at each sensor, and we focus on relative waveform characteristics rather than absolute amplitudes. As is the case for all four explosions, the seismograms for event 2 (Fig. 2) have good signal-to-noise ratios, especially on the vertical components, and the first P arrival on each trace is readily identifiable, allowing it to serve as a reference phase. The horizontal components tend to exhibit more waveforrn variability, but even the vertical components show some waveform changes over 15-m separations (compare traces 21 and 22). The early portion of the waveforms is enriched in higher frequencies. For this event, there is significant signal amplitude for about 9 sec after the P arrival, with longer-period energy arriving about 5 sec after P. It is worth noting that there is slightly greater visual coherence among the arrivals on a given row of the array than between rows, suggesting that local structure under the array varies down the hillside. We will quantify this tendency later. The bandwidth of the signals is indicated by Figure 3, which shows the average spectra from all vertical-component stations using 9-sec windows for two separate shots, along with the average spectra from an eventless interval of microtremor. The explosion recordings have reasonable signal-to-noise ratios from 3 to 22 Hz, with little signal above 22 Hz, especially later in the recording. The stronger explosions, such as event 2, have good signal-to-noise ratios at frequencies as low as 1 Hz. The array-averaged microtremor spectra are fairly smooth, but individual spectra vary substantially.
Figure 2. Vertical-component seismograms from 12 stations in the Zayante array for event 2. The signals are shown with true relative amplitudes, and the station names follow the labeling in Figure 1.
Average Spectrum . . . .
i
~.i
-8
. . . .
I . . . .
! . . . .
i
~ ~[i~~] ~
. . . .
! . . . .
! . . . .
................ Shot 2 Microtremor
-10
o .1
-12
-14 5
10
15
20
25
30
35
Frequency (I/z)
Figure 3. Amplitude spectra of 9-sec windows of vertical-component seismograms (approximately proportional to ground velocity) averaged across the Zayante array for two explosions and an event-free interval of microseismic noise.
We analyze two attributes of the direct P arrivals from each explosion to characterize any backazimuth anomaly. The particle motions of the horizontal components are used to estimate the backazimuth (Fig. 4b). The P-wave polarity measurement is made at each station using the unfiltered recordings, with the backazimuth estimates being averaged across the array for each event. An independent estimate of the backazimuth is provided by the relative arrival times of the P signals. Cross-correlation of a short time window (Fig. 4a) spanning the direct arrival is used to measure the tiny travel-time differences across the array, with the vertical
Seismic Wave-Field Observations at a Dense, Small-Aperture Array Located on a Landslide in the Santa Cruz Mountains
(a)
(b)
52
0
Effects of the Dipping Surface on Apparent Backazimuth Estimates
z d~
3, I . . . .
1
r
W-E
2 (sec)
(c)
N
N
I
~
r
E
(d) //
/
659
///jJf~ ~.
/./
/e 3' 2'
........................ .~
3 1t~
E
Arra,.
Figure 4. Procedures used to estimate the propagation direction of the initial P wave. (a) Verticalcomponent seismograms for shot 2, illustrating the time window used for frequency-wavenumber (f-k) analysis. A very short, highly coherent interval is used. (b) Horizontal particle motion for station I 1 for shot 2. The nearly linear motion in the first cycle of the P arrival is used to estimate the apparent backazimuth. (c) Broadband f-k analysis of the direct Pwave power spectrum for shot 2 in the window shown in (a). A maximum likelihood method is used, with the spectrum contoured in the wavenumber domain. The peak contour indicates the northeasterly backazimuth. (d) Summary of estimated (open circles) and actual (filled circles) backazimuths for the four shots. The average estimates from the particle motion andfk methods are shown. Biases of 30° to 50° are found for all four events, with systematic counterclockwise rotations.
components providing the more stable differential times due to their waveform coherence. These differential times are fit with a plane wave to estimate the backazimuth, but the angle of incidence of the wavefront cannot be separately resolved due to the planar array geometry. The relative timing also controls the backazimuths inferred from frequency-wavenumber analysis. Results of this analysis (Fig. 4c) confirm the values of backazimuth obtained from fitting a plane wave to the relative arrival times. In each case, the direction that is obtained is an "apparent" backazimuth, uncorrected for dip of the surface under the array. The apparent backazimuths estimated by the particle motion and relative arrival time methods differ by no more than 7 ° for the four sources. For each event, large counterclockwise biases of the apparent backazimuths are observed, as illustrated in Figure 4d. Several factors may contribute to these biases, and we consider simple models to explain the systematic 30 ° to 50 ° backazimuth anomalies.
The surface topography at the Zayante array is approximately planar with a 10° dip and a dip direction of S14°W. This dip causes arrival-time variations that bias the backazimuth estimated by f-k or wave-front fitting procedures. By transforming the ray projections from the dipping plane to a horizontal plane, we can correct for the effects of the surface dip, without needing to know the shallow velocity structure. The geometry is such that the predicted bias in backazimuth is clockwise by about 20 ° to 30 ° for all four shots. This is opposite to the bias in the observations, indicating that structure beneath the array is responsible for the backazimuth deviations. Before we can determine subsurface structure consistent with our backazimuth observations, the large effects of the free surface must be corrected for. A dipping surface also affects polarization estimates of backazimuth when the wave front sweeps obliquely across the surface, because the dipping surface defines the boundary controlling the receiver function that partitions amplitude onto the locally defined (i.e., relative to the dipping surface) vertical and horizontal components (Buchbinder and Haddon, 1990). The extent of backazimuth bias varies with the angle of incidence, dip of the surface, and ratio of S velocity to P velocity. We can estimate the angle of incidence from the particle motions, but this is also an apparent angle, influenced by the free-surface interaction. For a horizontal free surface, the relationship of the true incidence angle (relative to the vertical), ip, to the measured incidence angle, ip (ip = tan-1 ph/pv; where Ph is the horizontal-component amplitude of the P wave and Pv is the vertical-component amplitude), is given by COS lp"' = 1 -
2 2 sin 2ip, 2V~/V~
(1)
where Vs and Vp are the S-wave and P-wave velocities near the surface, respectively (Bullen, 1963). The velocity terms enter into the calculation through the receiver function. Using (1), it is possible to infer the true incidence angle based on the surface motions. An additional correction is needed to account for the coordinate rotation when the surface dips. For the two cleanest events, the measured values of ip are 11 ° to 19°. For a Poisson ratio of 0.25 (Vp = 1.732 Vs), this indicates true subsurface incidence angles that are a few degrees smaller. One can compute the effects of the dipping surface on polarization azimuths for a given incidence angle of the P-wave vector by rotating to the coordinate system orthogonal to the dipping surface, calculating the P-wave motions in that coordinate system, and then projecting the motions to the geographic coordinate system. Unlike the large arrival-time backazimuth bias of 20 ° to 30 ° caused by surface topography, the 10° dip of the Zayante array predicts a very small polarization backazimuth bias. For a typical soft-rock ratio of Vp/V~ = 2.0, we find that the surface effect on polarization azimuths actually goes to zero. Uncertainty
660
Z. Xu, S. Y. Schwartz, and T. Lay
in the precise near-surface Vp/V~ ratio will always make it difficult to reliably correct polarizations for surface dip. Effects of Dipping Interfaces beneath the Array A likely cause of large backazimuth anomalies is the presence of a dipping interface with a strong velocity contrast below the array. Niazi (1966) considered the effects of a dipping Moho on array measurements at the surface and provided expressions for calculating the bias in backazimuth as a function of velocity contrast and dip of the interface. We use his results with minor corrections for inconsistencies in sign convention. Aside from plausibility, there are two lines of evidence to suggest that dipping interfaces below the array are present. The first is the geological structure, with the array lying on a southwestward dipping landslide deposit located in the southwestward dipping flank of a synclinal structure. The second is more direct, with the shallow reflection/refraction survey in the area, indicating that there is an interface separating layers with P velocities of 550 m/ sec and 1250 m/sec with a dip angle of 20 ° and a dip direction of $29°W (Bonamassa et al., 1993). A schematic of the shallow structure under the Zayante array is indicated in Figure 5. For specified true backazimuths and incidence angles, the biasing effect of the shallow dipping interface was calo
r-~
_LI_O . . . . . . . . :-::: :~:i:!::::'""
"
culated and found to cause an azimuthally varying counterclockwise bias for sources that lie eastward from the array. This is illustrated in Figure 6 for a generic dipping discontinuity. The near-surface dipping interface produces counterclockwise backazimuth biases on the order of 10° for the travel-time method. This has the correct sign but is much smaller than the observed biases and, at most, compensates the opposing bias induced by the surface dip. It appears that at least one more dipping interface, probably with a larger velocity contrast, exists below the array, as indicated in Figure 5. We seek simple models with an additional deeper interface that yields backazimuth biases for both travel-time and polarization methods consistent with the observed biases. Modeling Results Any modeling effort is complicated by our ignorance of the ray-path effects between the sources and receivers, with the ray paths not penetrating very deeply into the heterogeneous crust. We therefore calculated apparent backazimuths using arrival-time and polarity methods for a wide range of incident ray parameters below the modeled structure, assuming all the bias is due to local structure under the array. The limited class of models considered includes the known surface geometry; the geometry and velocities of the shallowest low-velocity layer; and one additional dipping interface below the array, separating material with a P velocity of 1250 m/sec from deeper structure with P velocity varying up to
E
::ii~: .... 40
20 ,n
0
-20
/ Source
-40
Source
O,
1~
2~
3~
Backazimuth (deg.) Figure 5. Block diagram of the shallow structure under the Zayante array as viewed from the northeast. The array is located on a 10° dipping surface plunging S 14°W and overlies a 20° dipping plane ($29°W) with a P velocity increase from 550 to 1250 m/sec. This interface is from 3 to 10 m below the surface and is constrained by shallow refraction surveying. The deeper structure inferred to cause the array backazimuth bias is a southward-dipping plane with a stronger velocity contrast. The deflection of ray paths from sources east of the array are schematically indicated to suggest the counterclockwise sense of the bias in backazimuth.
Figure 6.
The azimuthally varying pattern of bias in backazimuth resulting from a dipping velocity contrast beneath an array where the shallow layer is lower velocity. The up-dip direction is assigned a reference backazimuth of 0°, and there is no bias. Arrivals from azimuths of 0° to 180° will have negative (counterclockwise) backazimuth biases, while those from 180° to 360° will have positive biases. The pattern of backazimuth perturbation can constrain the dip if the velocity contrast is known, or vice versa. The pattern is not unique when both dip and velocity contrast are varied.
Seismic Wave-Field Observations at a Dense, Small-Aperture Array Located on a Landslide in the Santa Cruz Mountains
around 4 3 0 0 m/sec. W e include corrections for the surface dip and the shallow-dipping interface. W e h a v e nine reliable observations o f angles: four apparent backazimuths f r o m the arrival t i m e method, two apparent i n c i d e n c e angles, and three apparent backazimuths f r o m the polarity method. In order to stabilize the m o d e l i n g , w e i m p o s e two constraints. First, differences b e t w e e n predicted and o b s e r v e d i n c i d e n c e angles and b a c k a z i m u t h s are not a l l o w e d to e x c e e d the estimated m e a s u r e m e n t uncertainties o f 3 ° and 7 ° , respectively. The second constraint is on the m i n i m u m v e l o c i t y at the turning point o f the ray path. This v e l o c i t y m u s t e x c e e d the a v e r a g e v e l o c i t y f r o m source to receiver. T h e average v e l o c i t y is close to 4 k m / s e c for shots 1, 2, and 5 (all with epicentral distances near 15 km) and close to 5.7 k m / s e c for shot 3 at a distance o f 21 km. W e m a k e a first-order a p p r o x i m a t i o n that the turning point velocity, Vt, is directly proportional to the average velocity, Va. The h i g h e r the turning point velocity, the smaller the ray p a r a m e t e r and the steeper the i n c i d e n c e angle. Variations in Va b e t w e e n events are used to scale differences in their incidence angles proportionally. W e p e r f o r m a grid search on four parameters: w a v e inc i d e n c e angles j u s t b e l o w the deep interface for the f o u r shots, v e l o c i t y contrast across the deep interface, dip angle, and dip direction o f the deep interface. F o r the polarization m e a s u r e m e n t s , w e m u s t also c o n s i d e r the near-surface Vp/ V~ ratio to correct for the effects o f the dipping surface. W e calculate the turning-point v e l o c i t y and the contributions to each m e a s u r e m e n t f r o m the deep discontinuity, the shallow discontinuity, and the dipping surface. The subsurface discontinuities contribute identically to the travel-time and polarization m e a s u r e m e n t s in that they simply cause refraction o f the ray paths. Thus, the key to finding a consistent m o d e l for the overall data set is the b e h a v i o r o f the surface contributions. But here w e run into a p r o b l e m ; w h i l e our observations o f b a c k a z i m u t h for the two m e t h o d s agree, the dipp i n g surface causes a large c l o c k w i s e azimuth bias for the travel-time method, while it predicts either a counterclockw i s e or at m o s t a small c l o c k w i s e bias for the polarization method, d e p e n d i n g on the Vp/Vs ratio. F o r hard-rock ratios such as Vp/V~ = 1.732, w e obtain the m a x i m u m c l o c k w i s e v a l u e d bias, but it is only about 3 ° ( c o m p a r e d to 20 ° to 30 ° for arrival times), thus a unique m o d e l c o m p a t i b l e with all o f the observations is not possible. W e found c o m b i n a t i o n s o f m o d e l parameters that satisfy the arrival-time backazimuths and i n c i d e n c e angles within their m e a s u r e m e n t b o u n d s (Table 2), m o d e l s that satisfy the polarization b a c k a z i m u t h s and i n c i d e n c e angles (Table 3), but no m o d e l s fitting the c o m p l e t e set o f observations within their m e a s u r e m e n t bounds. B o t h data sets indicate that a d e e p e r interface dips toward the S or S S E with poorly r e s o l v e d dip ranging f r o m 7 ° to 51 °. A strong v e l o c i t y contrast across a shallow-dipping (7 ° to 20 °) d e e p e r interface is preferred, as this m a k e s the m o s t sense geologically. The sets o f m o d e l s fitting each data set do not overlap, w h i c h m a y reflect our underestimation o f the m e a s u r e m e n t uncer-
661
Table 2 Models Based on Relative Arrival Times (all models satisfying measurement bounds on each datum). Deep InterfaceGeometry Dip Angle
Dip Azimuth
15° 15° 15° 15° 15° 15° 15° 15° 19° 19° 19° 19° 23° 27° 31 ° 35° 39° 43° 43 ° 51 °
140° 140° 150° 140° 150° 140° 150° 140° 140° 140° 150° 150° 140° 150° 140° 140° 140° i40 ° 150° 140°
VelocityContrast
0.29 0.29 0.29 0.29 0.29 0.29 0.29 0.29 0.39 0.39 0.39 0.44 0.49 0.59 0.64 0.69 0.74 0.79 0.79 0.84
Incidence Angle
Error*
65° 70° 70° 75° 75° 80° 80° 85° 45° 50° 50° 35° 35° 30° 25° 25° 20° 20° 20° 20°
0.63 0.62 0.59 0.59 0.57 0.56 0.54 0.56 0.67 0.70 0.55 0.62 0.67 0.54 0.68 0.65 0.76 0.65 0.66 0.60
*Error is an uncertainty weighted variance of the misfit to the relative arrival time backazimuths and angle of incidence measurements.
Table 3 Models Based on P-Wave Polarizations (all models satisfying measurement bounds on each datum). The Vp/V~is 2.1. Deep InterfaceGeometry Dip Angle
Dip Azimuth
7° 7° 7° 7° 7° 7° 7° 11° 11° 11° 15°
180° 190° 150° 160° 170° 180° 190° 170° 180° 190° 190°
VelocityContrast
0.69 0.69 0.84 0.84 0.84 0.84 0.84 0.84 0.84 0.84 0.84
Incidence Angle
Error*
25° 25° 20° 20° 20° 20° 20° 20° 20° 20° 20°
0.73 0.70 0.75 0.73 0.69 0.66 0.65 0.75 0.72 0.70 0.72
*Error measure is an uncertainty weighted variance of the misfit to the polarization backazimuths and angle of incidence measurements.
tainty or possibly an intrinsic failure o f the ray m e t h o d to account for finite-frequency effects. The fact that the raw b a c k a z i m u t h m e a s u r e m e n t s f r o m polarizations and arrival times are similar is fortuitous, as they actually are e x p e c t e d to differ due to the g e o m e t r y o f the problem. O n e possible factor is that the free-surface interaction controlling the polarizations i n v o l v e s finite-frequency arrivals that m a y effectively sense average near-surface properties rather than the v e r y shallowest structure. W e h a v e h i g h e r confidence in the results based on relative arrival times, as these are insensitive
662
Z. Xu, S. Y. Schwartz, and T. Lay
to the precise near-surface Poisson ratio. For Vp/V~ ratios greater than 2.1, no model could be reconciled with the polarization-based observations, because the azimuthal pattern of backazimuth biases could not be matched. Results of the refraction experiment (Bonamassa et al., 1993) indicate Vp/ Vs ratios of up to 2.1, and we conclude that the polarization observations may have been contaminated in some manner. The models fitting the arrival-time data attribute large backazimuth anomalies on the order of 50 ° to 60 ° to a deeper interface, which may correspond to the slip surface of the landslide that underlies the Zayante array. Of course, several interfaces could actually be involved. While some ambiguity remains, and only simple models have been considered, the robust aspect of the analysis is that the very shallow subsurface effects produce large backazimuth biases that are difficult to predict a priori. Array Analysis of P Coda Analysis of secondary arrivals in the seismograms can further reveal the very near receiver effects of the site. First, we attempt to identify coherent secondary arrivals and their trajectories across the array. We do this using a slowness vector analysis (Xie and Lay, 1994) that is basically similar to f-k analysis. The procedure calculates the signal power distribution in the horizontal slowness plane in discrete time intervals. Figure 7 shows the summed power distributions for all three components of motion for shot 2, in 1-sec time windows (see Fig. 2 for time scale). The first window contains the onset of the P arrival, and the peak energy has a slowness vector pointing at an azimuth of 240 ° (i.e., incident at an azimuth of 60 °, which is 40 ° counterclockwise of the true backazimuth of 100°), which is consistent with the corresponding f-k analysis shown in Figure 4c. The strongest coherent plane-wave energy is in the window from 1 to 2 sec (Fig. 7a) with a northeasterly arrival azimuth; this is the main P-wave motion. For later windows, the peak energy has highly variable propagation directions, although there is some tendency for energy to arrive from the northeast, clearly seen in the individually normalized plots (Fig. 7b). As the lag time increases to about 4 to 5 sec, the energy spreads more uniformly in the slowness plane, indicating increased levels of scattering from various directions. After 4 to 5 sec, there is some stability in the energy distribution over the slowness plane, with relatively well-defined peak arrivals that come from different directions. This corresponds to a transition to somewhat longer-period arrivals in the waveforms (Fig. 2), which presumably involve a larger percentage of scattered surface-wave energy. The pattems for the other events are similar but differ in detail. We can analyze the polarization of the coda to help identify the coda constituents. The power distributions in the horizontal slowness plane for different particle motion directions for shot 2 are shown in Figure 8. For this analysis, the two horizontal components were projected onto the indicated particle motion azimuths,
and the resulting signals were processed using the slowness method. By considering all polarization directions and the associated array slowness analysis, we can consider the correspondence between sense of ground motion and direction of wave propagation, which can help to identify the wave type. Of course, the presence of overlapping phases scattered from different directions can confuse this simple polarization analysis, but our impression is that much of the coda involves scattered, but relatively coherent, surface waves. In the first two time windows, the strongest energy has a polarization of N60°E, arriving from the same azimuth, and there is almost no signal in the orthogonal direction of N150°E. This reflects the nearly linear polarization of the direct P wave, along the biased backazimuth. The next few seconds of coda have little coherent polarized energy, perhaps due to strong near-receiver scattering of the direct P phase that does not produce coherent plane waves across the array. The 4- to 5-sec window shows an energy peak for east-west polarized motion propagating toward the west across the array. This is consistent with a Rayleigh-wave arrival, arriving about 20 ° off the true backazimuth. Polarization analysis of the full ground motion supports this interpretation. This is probably the onset of the direct surface wave from the source, arriving with a velocity near 2 km/ sec, but may also involve P to Rayleigh scattered energy from the nearby topography. Later coda intervals, as well as similar polarization analysis for the other three shots, reveal similar behavior, azimuthally deflected direct P waves with coda that is dominated by P-SV-wave-type arrivals. More sophisticated methods to decompose the coda into distinct particle motion contributions exist (e.g., Wagner and Owens, 1993, 1995), but we do not pursue this given strong variability observed in the event-to-event behavior. This indicates that near-receiver scattering is very sensitive to the source azimuth, and our limited attempts to do any further deterministic imaging of the scattering structure were not encouraging. We turn our attention to the statistical aspects of the data. Waveform Coherence Spatial coherence of waveforms can be quantified using the cross-correlation coefficient (CCC) for pairs of seismograms as a function of the receiver spatial offset. The CCC can be computed in either the time or frequency domains. Similar to previous studies (Menke et al., 1990, Vernon et al., 1991), we use a conventional squared coherence measure. In the time domain, the CCC between two waveforms s(xl, t) and s(x2, t) at location x 1 and x 2 is
ccc
(t, at, ax) = (s(xl, t)-s(x:, O)2/(s(xl, 0 • S(Xl, t)). (s(x2, t). s(x> 0),
where ( ) denotes the average value of the enclosed function over a time interval At centered on t; the offset Ax = Ix1 x21.
Seismic Wave-Field Observations at a Dense, Smaii-Aperture Array Located on a Landslide in the Santa Cruz Mountains
sec
663
b
Fraction of max, power" o,eo o.~,~ o.eo o.r~ ~.~
Figure 7. Slowness vector analysis for shot 2. The summed power distributions in the horizontal slowness plane (up is north, right is east) for three components of ground motion are shown in 1-sec time windows. (a) Relative power (indicated by shading, with darker shades indicating relatively higher power) normalized to the largest value in the 10-sec window. (b) The same results normalized separately in each 1-sec window. Crosses indicate the peak power. Note that the slowness vector indicates the direction the plane wave is traveling as it traverses the array, not the backazimuth, which is opposite to the convention off-k plots such as Figure 4c. The larger and smaller circles correspond to horizontal velocities of 1 and 2 km/sec, respectively. The true backazimuth is 101 °, but few arrivals have slowness vectors near 281 °.
Before calculating the CCC for a pair of waveforms, one must contend with the difficult issue of time shifts. With overlapping signals arriving from many possible azimuths, one cannot assume an appropriate move-out velocity for alignment, yet failing to shift the traces should bias the coherence measure low. For our alignments, we used relative time shifts among the array channels based on broadband cross-correlations of the traces. W e select one of the stations
near the center of the array as a reference trace. For each array element, a cross-correlation function is computed using a time window of 0.25 sec with sliding steps of 0.2 sec. The time lags for each window for all stations relative to the reference are applied to shift the data to common alignment when interstation correlations are computed. This procedure emphasizes the dominant frequencies in the signals and effectively shifts the signals to optimal coherence for the larger
664
Z. Xu, S. Y. Schwartz, and T. Lay
Azimuth of Particle Motion N-S
30N
60N
E-W
120N
150N
0-1
1-2 (D v
O3
2-3 C
(1)
E 3-4
4-5
FraCtion of m a x .
~r:
o,o~
~25
o, so
o.'~5 1.oo
Figure 8. Horizontal slowness vector power distributions for shot 2 (as in Fig. 7) using the horizontal components along different particle motion directions (obtained by rotating the north-south and east-west recordings onto various directions ranging from 0 ° (N-S) to 150° with 30° intervals. The relative power in all o f the panels is normalized to the largest amplitude arrivals. The stronger (darker) arrivals have propagation directions roughly aligned with the particle motion, indicating that these are longitudinally polarized P- or Rayleigh-wave arrivals with backazimuth of 60° to 80°. The true backazimuth is 101°.
arrivals that could be traversing the array in each time window. Each of our seismograms is filtered into a set of traces using Butterworth zero phase filters with a 5-Hz passband, moving steps of 3 Hz, and a total bandwidth from 3 to 33 Hz. The alignments found for the broadband signals are held fixed for correlation of the narrowband filtered traces, rather than optimizing the correlation in each passband, which could involve many cycle slips relative to the stronger arri-
vals. For each frequency band, we consider 0.5-sec portions of the waveforms with moving steps of 0.2 sec, aligned by the cross-correlation shifts. We compute the CCC between all record pairs for variable time lengths, T w. Tw depends on the center frequency of the band, Fc:
Tw =
0.5 ×
5.5/Fc (sec),
where 5.5 is the Fc of the lowest passband of 3 to 8 Hz and
665
Seismic Wave-Field Observations at a Dense, Small-Aperture Array Located on a Landslide in the Santa Cruz Mountains
0.5 is the time segment width for alignment. For example, T w (F c = 5.5 Hz) = 0.5 sec; Tw(Fc = 30.5 Hz) = 0.09 sec. Thus, for a center frequency of 5.5 Hz, there is one window for calculating the CCC, while for a center frequency of 30.5 Hz, we calculate the average CCC from five windows. Menke et al. (1990) have considered the possible bias due to frequency-dependent alignment decisions and find that it is relatively small for time windows like those we used, which we confirmed by following their exact procedure for some of our data. Our procedure estimates the fiequency-dependent decrease in coherence relative to the stronger broadband arrivals in each time window, which is right at the transition from the deterministic array analysis to the statistical behavior. This approach is valid given the weak frequency-dependent behavior we found in anf-k analysis of the data, which did not reveal the presence of strong plane-wave arrivals with distinct frequency content and diffeting arrival directions in each time window. Variations of Coherence Down the Hillside We noted earlier that Figure 2 suggests stronger waveform similarity along rows of the array than down the columns. To test this quantitatively, we calculated the average CCC between elements on rows, CCC .... and between elements on columns, CCCcov The mean value for 2-sec time windows for the vertical components from all four shots was computed for the 3- to 9-Hz frequency band as a function of the receiver spatial offset. We then computed the fractional difference in coherence:
CCCfc d
=
2,
(CCfro
w
-
CCCcol)/(CCfro
w
-}- C f f c o l )
,
which is shown in Figure 9. This effectively measures the anisotropy in waveform coherence, with positive values indicating stronger coherence along rows. For spatial offsets of 15 and 30 m, the CCCfcd is positive, confirming the statistical tendency for receivers at relatively uniform elevations (Fig. 2d) to record more similar waveforms. This is consistent with the presence of the southward-dipping surface and subsurface interfaces discussed in the previous section. A wedge-shaped structure will cause more similar reverberations at stations overlying comparable depths to interfaces. The fact that the array is not perfectly aligned with the structure, and the irregularity of the structure itself, causes the 45-m offsets to be less coherent and to have less spatial pattern. This spatial pattern in the statistical measure indicates that we are again measuring right at the level of transition from deterministic control on the array signals to a stochastic wave field. While supporting the general expectations of the near-surface structural effects, it is difficult to further constrain the structure with such statistical measures.
*~
0.2
. . . .
I
. . . .
I
. . . .
(
. . . .
I
. . . .
I
. . . .
I
se¢ O
~= 0.1
~ ~= "~ ~
. ~
"~-~.
-
-
--
--
0-2
2-4 4-6 -- 6-8
0
O
-o.1
. . . .
15
i
,
r
i
i
I
,
,
i
,
I
,
,
,
i
I
. . . .
I
. . . .
30
i
45
Spatial Offset (m)
Figure 9.
The difference in signal coherence on vertical components between east-west and northsouth sensor offsets averaged for four shots. The difference is normalized by the mean coherence at each offset. Results are shown for four time windows in the signals, and the frequency band used is 3 to 9 Hz.
Variation of Coherence with Frequency, Spatial Offset, and Time Lag Generally, the spatial coherence across the Zayante array reduces with increasing frequency, sensor offset, and time lag for recordings of all four shots. Average coherence versus spatial offset in various frequency bands and time windows for vertical-component seismograms for the four shots, for microtremor intervals, and for synthetic Gaussian noise are shown in Figure 10. The Gaussian noise has an average coherence value of 0.1 and the coherence is never more than 0.2. Therefore, coherence values for the data of less than 0.2 are not considered significant. The microtremor coherence has an average value of about 0.3, and there is less variance in the 3-sec microtremor window than in the 1-sec window. The coherence is not very dependent on the frequency band or the spatial offset for microtremor and Gaussian noise. The microtremor spectrum is relatively fiat (Fig. 3), and the weak coherence suggests that the wave field involves much localized scattering of the signals traversing the array. For the high-frequency energy (>15 Hz) in the four shots, especially in the 3 to 9-sec window, the coherence approaches the level of the microtremor. This probably reflects the low energy level (Fig. 3), and the scattered nature of the coda (Fig. 7), which is comparable to the microtremor signal. We infer that coherence greater than about 0.3 reflects the presence of more coherent signals traversing the array than is typical of microseism. Following Menke et aL (1990), the array coherence can be expressed by a simple empirical equation, e -cfAx, where c is a constant determined from the data. For the first 1-sec window for the vertical-component signals of the four shots c =- 0.6 kin-1 Hz-1. This value is similar to the values of
s!q~ u! guotuouoqd pIOtJ-OAgA~£ouonboaj-q$!q oql ffu!u!umxo ao3 lxo:moo ~umaodtu! ue op!Aoad o~l~nbq~.reo ~o.u d etuoq 686I ottl £q pos!ea sonss! ~u!:~oou!guo O~lenbqla-eo poleo!Id -moo jo lsoq oql pug 'sm.~uno]AI zru;D r;1ugS Oq:l u~. ~uotuuoJ qAUO leO!~OIOOgoq~ JO leo!d£1 ~OA s! £g~.re s!ql j o ol!s oq,L •pio~I OAgA~OItUS!OSpogod-uoqs oql JoJ sloojjo o~!s j o OlOJ luel.~odtu! oql o~eslsuomop s~u!paooo.~ £eJae oqu~£trZ oqz suo!snlaUO D ptm u o ! s s n o s ! o "S]UOmUOJ!AUOJoqlo UI. punoj lgtI1 ol £l!xolduloo lgUgt.s o l q ~ d m o o soonpu! £l!ouogoao -loq oogjans-.mou OleOS-ilems oql leql sJeodde 1! 'lUOUlUO:qA -uo leO!goIoo~ xoldmoo oaom e u! pole3o I st ,(e.ue olu~£eZ oq~ ol!qAX "(I66D 7 v ~a uou~oA ptm (066I) "l~ ~ O~lUOIAI £q pozgi~ue sol!s ~tooJ-paeq Otll l~z osoql ol xeI!m!s oJe sloqs oql Joj ~osjjo pue '(os!ou o1 leU~!S poo~ seq q~!qA~ 'zH IZ o1 pueq oq~ u!) £ouonboaj 'otup tla!A~ spuo-u pue SOnleA o~uo -.IOq03 Oq£ q0sJJO pue 'Louonbo~j 'gel otup tp!A~ oouozoqoo
j o oseoJoop IeJOuog aei!tu!s e £eldst.p SleU~!S uo!soldxo oql mq 'luouodtuoo Z oql ugql moJOtlOO sso[ l~qaxotuos s£eaxie oae sluouodulo3 S - N oq& "ZH LZ moqe j o so!ouonboJj ol dn 'splmq £ouonbo~j lie u! oouozoqoo lsoq~!q oql OA13qS|13U~I.S :loqs oq:l JO oos E o1 [ ]s3[J oq~L "IOAO[[eU*l.S punoJgsloe q oql sou!jop pug £3uonbo~j pug ouI!l tll!ax stuo~ted aeI!tU!S sgq sluouodtuoo S - N pug IeOgJOA UO OOUOaOqooaotuo~loao -!m oqZ "SA~OpU!maotuoatozo!tu oql ptm 'sloqs lie Jo o~eaOAe oql 'E loqs aoj uot3otu jo stuouodmo3 (S-N) qlnos-ql~ou oql pue sluouodtuoo (Z) l~39aOA oql qtoq aoJ UA~OqSO.m slInso~I "(I [ "~.I~) ~{I~3.I13~]1113~{1~z OLI1 .loJ ~losJJo Iet.leds pue '£3uonbo~g 'otu!l jo uol3oung ~ se oouoaoqoo pz!leds o~eaoAe oql m.elqo OA~ 'losJJo UOAtg e JoJ (xV 'iV 'l)DDD IIe gU~.gt3JOAV "S]I.I~U.IUO.II.A[IO 2{00.I
-paeq pue -lJos qloq Joj ooupns oql A~O[Oq SO!l!OUO~'O30~,oq Jei!tu!s ~q POiloaluoo oq ~etu sao]otu jo spoJpunq ol suo~ moJj sqlguOIOAeA~ ~ OOUOJOqo0Otll 'ZH 0E O~ E JO offueJ £ouonb -o~J oql u! ~ q l ~Idm! £mu s!q& "(066I '7 ° la O~lUOl/q) sol!s 51oo.t-p~q le S[13A.I./XI313oJ!p ~oJ punoj t z H t -tt°l L'0 ol 17"0
•osmu o(t ol poaop!suoo oae DAOI s.vql A~OIO q SOrI~ApOA.IOsqopug 'E'0 ueq1 SSOI St. OS!OUue!ssneo oql Jo oouoaoqoo oq,L •0q~.u luo~oq) ~st.ou ueissne D m.loq~ugsaoj se IIOA~Se '(lq~.u oipp!m pug ~oddn) os!ou aotuoJlo:tortupug 0jOl) s:loqsznoj jo sm~.~goms!os~u~uodtuoo-l'eO.rlaOAzoj SA~Opm.taotuyl pug spueq £ouortboaj sno.ueA u.~ losJJO le!leds snszoAOouo.loqo0OgeZOAV "0[ °'that.t:[ (m) losjjo le!l~ds OL
~ 09
. . . .
•
Of
~
01~ 0g 0~ ..............
0t
0 0"0 ZT0
(m) 1osjjo i e g g d s 0L
o9
o~
o~
0~
....... . . . . . . . . . . . .
o~
0t
0
if0
~E0
I,'0 ~
puooos [
8"0 ~
8"0 ~
9"0 N
9"0 ~
0'~ o Z'I t,'I
0'I ¢¢
puooos 6-£
z'l
. . . . . . . . . . . . . . . . . .
t"l
~opuIAk as.toN ue!ssne D
09
0£
09
0L
0g
fit,
0g
0Z
0L
01
O9
Of
0~
0£
0Z
0I
0 Z'0
~'0
i o'o
0"0
t"0 9 9"0 ~" 8"0 ~
1,'0 9 9"0 ~ 8"0 ~
0"I
spuooas
puooos ~-0
O
iS'I
~,'!
~opu!A~ JotuoJaoJo!IAI 0~
0t,
0£
0Z
09
0I
Of
0¢
0£
0Z
0I
0 E'O
~'0
i00
0"0
~,'o ~
I,'0 ~ 9'0 8'0 ~
Lg~lZ T~~Sl
0'1, ~
puooas I
~L-6
0"!
puooos I-0
6-£ ZH
1"I
Z'I
o~moAV
A~OpU.IAtt JOIllOalO.tOl.lt ~
~eq "Z pue 'ZU~A~qos"A "S 'nx "Z
999
.qaql soougqua qa!qA~ 'SaOATOaOaoql aopun luop!ou! £Idzzls 0113 SOAeA~ OlUlSIOS lOoa!p 'S~UOtU!pOs ooejans-a~ou ,{l!oo 1 -OA ~O 1 ~dOA qa!A~ 'su!elunolAI ztu D rltreS oql o:,I!I uo~goa e uI "suogoalJOp Ieqmtu!ze ogxeI Otll .toj olq!suodsaa ore slse.rl m o o ,(I!OOIOA gUOZS ~ia.rej q~!t~ saaael gu!dd!p aaepnsqns
~gql s.modde 1! 'Iopotu o|dtu!s ~ ql!,~ suo!leAaasqo Jo slos qloq oIDuoa~a ,(IInJ lou pInoa a ~ 91!qM "o0g m o0~ tuoaj j o s~s~!q qmtu!zw4~q o~a~I mea!pu! satu!l FApa~ OA!l~pa ptm uo!lez -!x~Iod o a ~ - d ~ o q mq 'sluoa3 aA~m ~U!tUO0U! JO uo!mlo -soa q~!q op!Aoad ol lI~ms ool s! oanlaode 3~a.m oq,L "uo!~oa
-(moa aa~Ol) ~opu!~ aotuoaloao.nu ~ql aoj ptm '(~oa alpp.nu) sloqs anoj jo ~ a 0 A ~ ~qt aog '(A~Oaaoddn) loqs lsn.faoj UA~OqSoa~ (sutunloo OA~I~ . U ) s~uouodmoa qmos-qlaou ptm (sutunloo OA~t ~Jal) sluouodtuoo F3.na~A ao~ sllnsoa aqx "t~p aql jo slasqns sno.m~Aao~ tu ~17ptm gl jo slosj3o aoj ~3uonbozj ptm otu!l snsaos (~;'0 ol I jo sonll~A ol spuodsoaaoo qo~.q~ 'o]!q~ ol ~laglq moaj ~3urpeqs auol£ea~ ,(q patea.~pu!) ~ouza~qoa tllJOjOAeA,~ "[ [ 9jtl~.l d 0"¢ [ZH]/~uenbeJ=I
e'o
9'0
~r0
[zN] A~uenbeaq
~:'e:e0UeJeq00 [ZH]
6~
I~
~ouaaba~9 6t
'PL
[ZH]/¢menbsJ4
6
i
.....
......
~,~
,i
.......
.o
,
-
o
L
......
.lowen,oJo!N (O)
{
-4 ®
i!i:i;ii~
W
{1D
s~oqS mO=l ;o e6eJeAV (q) 6E
l'&:
6t
-t
3
g oqs
LU £17= ,esJ O
uJ cj L= ,esj o
ueuod oo S-N L99
lueuodLuo0 Z
su~.v~unol4t znaD muvg a~tl u~. op~.Fpuzrl v uo Falvao'-I ffoaJ V aJnlJadv-iIvut S 'asua(1 v 1D suot.lz~aaagqo plal.d-OaDRl a~.uts?~
gu!gem! UO~laaljaJtuoJj aJnlamls ~Ol[~qs aql jo £pms ~oq .,IoSllnSaJpap!A -rod ~ss~tu~uo8 ~liau~O "Re:u~mu~£~Z aql :Io luatu£o[dap aql u! pal~d!a!l -.red sa!uolaa,Ljo a:tm.~lsuI aql u! s.Iaqa.masaJpu~ sluapms al~anpe.~$,~U~IA I
sluatugpaIA~ou:,DV "Jog palaa~oa '{Iqe!la.t aq ueo laa~,la m!s aql SSalUn 'slaajja aoanos jo smaal ut. [tu~!s '{ouanba.tj-q~!q go uo!ltlaadaalu! u! uo!lnea .toj~ aauap!Aa.taqlanj sap!aoad s!q£ "saalam go sum 3o OleOS aql uo '{[p!d~a 3,1o sdmp aaua.taqoo '{ouanba.t~-q~!q leql st. a.tnleag umm aql 'palaalap a.ta~, a.tnloruls ssmg aql '{q poouantlu! sa!lemals'{s it!ltds plum amos al!qAk "sm!s >Iom-.tapamq le pamasqo l~ql ol alqt.mdmoa SaAmd qa!q~ 'aaua.taqoa ttuog -aAe~ £ouanba.tj-q~!q go '{eaap It!rods p!d~a aql S! t l t p alut -'{eZ aql U! laagJ_a m!s go uo!ltlSag!utm .teo|o .laqlo aqJ~ "aouaaojJolu! I~UO!lOaa!p aauequa leql stualled uo!le.taqaaaa.t alnq.nluoo oslt '{tin laa~Ia ol!s .tNno!l.md s!ql go od~qs a~I!I -a~poA~ aEL "~!auagmolaq aleaS-ilgmS posodm.uadns ql!m 'slmluoa l-eamalms ssmg aql ap!Amd '{oql ltql '{Ia~I!I s! 1! ' ( [ 6 6 I ) 'W go g s s t m t u o ~ ' { q paMasqo Saoutuosa.t ituo!loo~!p aql go lie ol asu aAtg 1OU '{tin saJmamls gu!dd!p aldm!s ai!qA~ "alSUt aauap!au! pue 5alamoa~ jo uo!launj xaldmoa t se 'paq.uasap se stxt '{allta aql 8UOlt dolaaap op 8u.r:,Itqs punoJ~ go SUO!laaa!p pa~uagaJd ltql A~oqs ( [ 6 6 D "W Ia .t.mtl~ go suo!ltlnm!s iea.uamnu aql '(uo!le.tedaad u! 'a|ep! A put tss~umuoa '{q paDnpuoa ~u!aq s! pu~) lOaJga Ie.tmomls aql '{j!luenb ol papaau s! uo!le$~dmd aAet6 [euo!suam!p-aa.tql
",{aIll~Aaql JO S!Xl~aql p.l'et~O1pmasaJaa s!
(sAmoa
-~ papeoq-ot~l iietus) Su.~qs aAe~A-S Jo asuas oql lsql alan "tlmtu!zs palasajaa aql salsa!pu! aaepns aql uo too.tin p!Ios u!ql aq£ 'sap~.s $u!dd!p aql £q polasag -aa lou J! aaeq plno~ s£ea asoql tDntu!z'e atll alea.q3 -u! aaspns agl uo s~o.u¢ paqs~p u!ql aqz "sau!i ploq aql ql!m 'sqmtu!z~ l~aaAaS ao3 ut~oqs a~ £alIeA aql ,~OIaq tuo.tJ luop!au.~ sqled aea OAeA~-S '£alleA aql Jo spxe aql ~uoI~ poluapo £IW!luo.toJoadoq [I!A~~'u.'~eqs aA~A~-Saql 'u~oqs S~ £alIeA pad¢qs-A s aod "uo!laaa.rp d!p-dn atD tuoJj mop!an! SaAe& aoj oa~g.talu! gu!dd!p aql Jo a:,t.UlS aql p~t~ol uo.rlotu ala!l~¢d aa~pns aql oleloJ uot.l'glsoql ql~ouoq soagpolu! Su!dd!o 'oJmaruls aaCpnsqns £q pasn~a ~u.r,4eqs Imuoz.uoq aA~A~5 30 uo!laaa!p pa.uajaad jo uo!l~D.tdaaluI "~[ mn~!d
£~"I "Z ptm 'zl.mtAqaS "X "S ' n x 'Z
ql!A~ uooelnm!s al!qA~ "a~glnmn3o~ ol suo!laa;!p o~g!oads p~eAxo1 uo.tlo~Ij~ Jo loajja aql sA~oIIe qg!q~ 'sqlnm!z~ £ u e m mo.tJ SI~A.LI.te SaAlOAm. epoo aql l t q l 10tJ aql ol anp ieu~!s 10a.t!p Oql u! u~ql ~po0 0Itl U.t paaunouoad aaom aq ol '{Ia~!I s! ~tt.D[tqs Jo Ho!ltlua!Jo pa~ajaad jo luamdoIaAa p aq,L "lualxa qldap .ta~el e .taAO slaaJJa I¢atUomls atll a~l~laAe satauanb -a~.I.taA~Olal!qA~ 'aamatuls ~ o w q s £aaA O1 aAD!SUaS a.tom aq ip.t6sa!ouanba~j .taq~!H "pu~qss~d qa~a aol uoD~aH!ldttm ptre LIO!lOt,~.tJO.t~0 ][Ia].XO at[] U! ~m..tagZ~p slo~jga '{ouanba.tj-al!u -H ql!A~ 's~moru~s pa.ta£~[ xa[dmo3 ~oj paloadxa s! sl.lo!loaJ -!p ~ u . ~ q s p a ~ a ~ a d aql jo aouapuadap £auanba.td "~lamo - ~ |t.tnlonlls o~!oads aql pue s ~ m a m ~ d aA~A~ luap!0u.t aql jo uo!loun3 e s! paonput tualled as~aaad aqJ~ "uoDa~1!p d!p aql ~uoI~ st ~u.p4eqs IelUOZ.Uoq ~eql os 13e~Ja.t [I!aX ap.ts d!p -UA~Opaq~ mo.tj auap!au! s a a e ~ oI!qA~ 'paDeaJoa '{IssI!m!s aq II.tax ~a|IeA aq~ 30 ap!s d!pdn aql mo.t~ luap!ou.~ SaAeaX 's)IuelJ ~a[[eA aid Jo auo ~U!'{[JOAO SOI!S .I0d ''{a[ll!A aI.ll J.O stx'g
aql ql!A~ uff![e O1 spual ~U.DI~qsJo luauodmo3 Ieluoz.uoq aql ltql os palaezJa.t aq H!A~sqlnm!zt lua.taJ~!p m m j sqled '{eJ aAtA~-S atl130 UO!Da.t!p aql 'Z[ a.tn~!~[ u! UA~OtlSSt 'It.umem £1!aOlaA-A~OI ~u!£I:~apun aanDmls £aI[eA padtqs-A e ~to s!xt aql ~uoIv "£alamoa~ aldm!s ~ aap!suoa o1 Ingasn s! *I "suo9 -ms oql .tapun saaepolu! gu!dd!p 30 aauasaad aql ol anp 1.rod u! aq £ e m (I66[ "I v ,a t s s t m t u o f l ) £t.at VAVZ aql m gut -)[tqs aatAat-~ gO uo!laa.t!p pa=aga.td go uo!leaaasqo Oil& •uo!lom go sluauodmoa amos aauanl~u! £iitguaaagaad u~a l~ql aatjamu! aoe~nsqns aql £q loaj~o uo!lot.tjaa lau t oq lI!~ aJaqz "sqlnm -tzt luaaajj!p mo.tj a.mlotuls ames aql no luap!ou! SIeA.U.m aAt~-S pue -d '{uem ql!A~ sp|atl aAtA~ xaldmoo .tog ploq II!t~ s!q£ "suo!loa.t!p amos u! UO!lOm jo uo!l~o0!Idum tll!A~ 'pla0 aAtA~-S aql gO uo!mz.uelod a!ltmals'{s amos asnea ol aq snql tmo aatjaalu! ~u!dd!p e 3o loagga lau aq~L "ItaUOZ.Uoq a.totu aq m palat.tja.t st. uo!ltZptlod AS aql ptm 'luauodmoa AS aql .tog paaunoumd lsom s! uo!lat.tga.t mmg loagga uo!lta0!Idum aqt 'uo!lom :Io sluauodmoo AS pue HS atll au~Iap ol aaejJalu! gm.dd!p aql asn aA~ ~I "(9 "~!d) aatpalu! aql 3o uo!laa~.p dip aql p~tao1 put 'aotj~alu! aql Ol ~ttnou aql p.t~a~m UO!Da.t!p aAtA~ aql 1Ot.tJa.t O1 S! slse.tluoa ~!aOIaA aA.tl!sod RI!A~saaej -.talu! ~u!dd!p aasj~nsqns go laajja aql 'SaAe~A S put d qloq .tod "epoo oql olu! am!l ~e I Jo uoDaun J t se sast!q aql agtmqa plnoqs SItA.U.m Empuooas go saI~ue luap!au! aql u! SUO!m! -SeA qSnoqll~ 'paauan[tu! aq ol pla0 OAeA~a.t!lua atI1 slaadxa auo ltql qans s! a.tmamls ~aAtaOa.t aq£ "SltA.U.m ~ p u o a a s uo Slaa~ja aql .tap!suoa m ~u!lsa.talu! s! 1! 'aAtA~ d laaat.p atll uo saat~alu! ~u!dd!p-~oI~qs go laagja ~uo~ls aql UaA!D •am!~a~ laa.uoa aql u! aq ol .madde a.tmatuls aql go slaa~a pme.t~alu! aql lnq 'xa|dmoa a.tom '{Ipa~qnopun s! a.tm -alms [treat aq£ ''{atans uo!latajaa atms~as A~olgqs aql '{q paltA!lOtU Iapom '{l!aOlaA |tuo!suam!p-aa~ttl aldm!s t RI!A~ suo.ntA.tasqo aql .tog lunoaae '{IaA!lel!Itnb aA~ "slaaJja al!s '{q pastt.q aq O1 sluama.tnstam '{~=t a.tnlaadt-l[~ms Uaaa ptm .tosuas-ai~u!s aoj i~!lualod q~!q s! a.taql 'Ja'{eI pa.taqlea~ aql u! sa!l!aua~o.talaq put saap[noq '{q paanpmd ' s l u a ! p t ~ ,(1! -aOlaA-l~.tmtI aatauns-~au ~umts RI!A~pau!qmoD "saamamls ~u!dd!p '{[mtaapom '{q uo!laalgap paqlnm!zt ol '{I!A!I!StIOS 899
'g66I ~sngnv 6E pOA!OOo.llduosnuglAI 1~90~6 g!tuoj!leD 'znaD g~ugS znaD glue S 'g!uaoj!it D jo ~I!S~OA!Ufl Molgaoqt~l [go!~olotus!os ~Ioo)I "IA["M put so!uolooL go olnl!lsuI "LE '£9 "Ill7 "salt "zus~.aS 'suo!lounj aOAt.Oooa ~ql uo sloojjo o~moruls G-E pug (I-E "(E66[) uols~ugq "V "D put "f '~ugqz 'Et~E-I~E~ '1~8 "mY "los "ms!a~ "ling 'uo!lgff!lSOAU!oouo~ojj!p -oltu9 e :~q uoIsoldxo JO uo!lgl!oxo OEL "(t~66I) £g~I "£ put "9-'X 'niX "sso~d u! "sa~/"s£tldoao T 'glgp £ta.m orms!os ~uouodmoo-ooJql ~u!sn uo!l~zpo~ogyeqo pug uo!l!sodtuooop pI~OAgA'~IgUOtfloH"(866I) su°"aO "f ".L pug "S "D "z°ugtM "9EgI-EEg[ '0Z "Ill7 "so~I "S£ZldOaD 'gpoo o:,IgnbqLreoItoo I jo £pms oql o3 uo!tgo!lddg a!oql pug t l t p ,~g.tm o.mts!os ]uouodmoo-oo.rq~ Jo spaoooJ otu!l-~u.treoq pu~qpgo.I9 "(E66I) SUOA~O'f ',L Pug "S "D '.IougtAk "966I I - I 86II '96 "sa~I "s£ttd°aD T 'Aga.reoam -.~odt-lIttus g Aq poJnsgom st S'IUgAg[800[ LIIO.IJSOAtA~£poq o!ms!os jo oouo~oqoD "(I66I) oAgqD "V pug 'IIOa.mD "q 'aoqolold "f "~I "~I 'uou~oA '9~I-IE[ 'I~L "uW "a°S "ms~.a$ "llng 'sloojjo ol!s ~lOo~-pagttjo SUOggA --zosqo "(I~86I) AOSOS-tON"~I "I pug 'p|otz~g H "G 'gtt!)t "~I "f "H '9 '.~o3[on,L $ 9 1 - ~ I 'lq, "mV •~oS "ms!aS "ling 'so!Ltodo.Id £OlItA pug opm!ldtug mdu! uo osuodso,t gO!leA poIlg-~uom!pos Jo oouopuodo o '(1~86[) gu.r)I "q 'f pug "H "9 'ao:,IonI "01r1~E-6IVE 'I8 'mV "ooS "vas.~aS "ling 'uo!ltotJ!luop! osgqd pug uo!lgz!.mlod :SS~tDHV pug SS~dON 18 sasgqd IgUO!~a~JO s!s,~lgUgluouodmoa-ooaq, l "(I66I) "g 'uosuoH-ngoans "866I-L86I '0g "utV "los "tuslas "llng 'suo!lgls £gaxe pug luouodtuoo -ooaql tuoJj ssouaxois pug qlntu!zg ~u!lttugs H "(066I) "V 'uosuoH-ngolnS "008[-98L1 'l~g "mV "los "uts!aS 7lng 'so~p!a ssoaog oPm!idtug uo!lotu-punoaD "@66I) pyefl A "d put 'Oli!dmgD "IA['P[OJZltH ( I 'urtlqoq "fl "H 'uosaopod '608-16# '9~ "mY "°°S "u~s~.aS 71n8 '~(l!nu!luoos!p ;3!O!A!.IOqOI/~~u!dd!p t aoj sluo!pg.t~ omt3IOhgalpug sqlntu!zg ~uoagddg oa suo9oo:uo D "(996I) IAI '!Ze!N '960[-680[ '178 "uw "oos "tag!aS 71ng 'e!uotu.iV '!uae D ~g £g.ue ituoisuom!p-oo.rq~ osuop oql tuoaj sluoaj OAgm S pug d jo s~uotuoansto~ "(1~66I) ug!P:tOAq°tH "~I Pug 'uttuo -agqv "A 'ug.r'doqa!tuv "H '*Paoqoao9 "~t '21o!msugaD "~ 'uos[!d "f "f '.UOIAI "EIzE-~EE '18 "utv "°°S "uts~.~S "ling '~g.Lre'TVD -SSVd/SIHI zq~ jo souoqdoo~ popod-~aoqs oql jo oougtmo3aod '(I66I) tut'I-ZOtuzq "V pug 'oD 'H '~uoqsguoH "D 'ploffuoqs "I "V 'M 'z)tuoIA[ "6tdz--0¢*r '08 "mV "°°S "ms!as "ling 'asmo oql u! sog!ouogmoloq/{I!3O[OA o:1 O0UgAOIO.I..I!Oqlpug m!S ~IOOa -paeq t lg SplOgOAtA~omls!os zI-I 0E o1 ~ go oouo~oqoo pug tlogtz!Jg[ -Od "(066I) oooqogd "f put 'jjaopuoqnG 'fl 'tugq-.~otuo'-I "q "V "M 'O~lUOlA[ •[ L6-8~6 'LL "mV "oos "ms~.aS 7lng 'glgp 0nouodtuoo-oo.ntl) uoIlmS-OlgU!S gu!sn uoggooI ooanos pug uo9oo~op IUOAOO!tUSIOS "(L86I) IotqD "H Pug 'pomqv 'N "N 't.Uogt~ 'E9LIr-6~eL1~'178 "salt "s'ftldoaD T 'SOAtA~£poq o.mls!OsololmoJ2 po.uoj -u! 'uo~guiqstM 'aoiu.reH lunolA! aopun oamon~ls (6L6I) "V D 'uolsgugq "0g0I-6EOI '9L "mV •ooS "tus?aS 7lng 'aolomtaed £ta ~uglsuoo aoj sosuodsoa ~OAmOOapug ooanos uo oamotuls ~u!dd!p aeugld jo ~ooJ3o oq,L "(LL6I) "V "D 'uo~s~ug'l 'IgI -L~I '¢L "tuv "aoS "ms!as 7Ing '£oIIgA POlIg-luom.rpos t sso.tot uo!~otu o:h'-enbq~mojo SUO!lg!aeApoAIosqo "(]786[) J~Ion,.L "H '9 put "I "f '~u.r~l "LgE-m9"~E 'I8 "mY "oos "ms~.aS •ling 'SOAtA~ q~!OlggH pug HS m°P~.°uI "II "odgqs [tuo!loos-ssoao Lleallqae jo S£OIItA ltlAnllg l-eOIaPU!I,~0£q SOAt~A0n.UStOSluoPIOU! ,~Ionb!IqoJO uogto!J!IdmV "(1661 ) ttgqs "H "V pug 'ellt(I ")I "S "H "~ 'a!gtl)I "lzLLI-ILLI 'gI Wa"I "s~t "sfzldoaD '£g:ut o~n~Iodt iletUs e ffm.sn gut.grin! ~IOtls-tugo9 "(166I) uno~o "f Pug 'Jolsu!lAI "fl "f "H "V "IAI'uIIPOH
699
"61rEI-9E~[ '178 "mV "°os "ms~.aS 71ng 'gt.uaoJ!IgD 'o~p!~I pooA~u!qo~tlg slooJJO o!qdga~odol pug ol!s jo uogg~llSOAU! Ie!~!uI "(1r66I) ~u!)I "M ")I pug '.IOAag~)"7[ "~ "H "S "IloZLmH "[~'EE-L~EE '808 "mY "°°S "tusk.aS 7lng "~u!ao~tos ooatos-~ou put aOA!OOoaqeou jo soo~nos £j!~uop! o3 elgp uo!lgls oI~U!s put £taxe SSHHON J o s!sAIgUg)l-J pugqpgoa9 •(066I) azuffgAk "V "8 Pug 'qszaJIHOl4[ "Ak "& 'szuu,(q "S "D "N "I 'mdnD "~9-Z*' 'g/, mV "noS "ms!aS "ling 'sllnsoa ~ou pug A~O!hOag :uo.qotu punoa~ o~Ignbql.mo uo Aqdga~odo~ jo loojjo oq3L "(886l) u°!Ilnf "9 pug 'pae 9 'X-'d "q 't.[°D "EE61-006[ '[8 "mV "°°S "u~s~.aS 7lng 'g!tuoJ!Ig D 'OlgAXUUnS u! £galg osuop t mo.~j s~Iooqsaoljg g:lo.tJd gtuoq jo uogtmosqO "(I66[) ~qsn9 "H pug '~oq.u~t "d 'q~noH "S "V '[o:'Iugad "I I ~ '~/, SOft 'ozmomls lingj put ~l!o.rtus!os A~OlIgqs:uoigoa g!uaoj!ig D 'tlo!Jd gtuoq oql u! luom!aodxo o.rtusios OAI.lOt I66I oq,L •(I66I) I°tq°!lA[ 'f "V pug "O 'sd!II!q&lagqa°qH "09EE-EIrEi~ '08 "mY "oos "tus~as "ling '~u!aol -]gos o!tus!os jo s!s£igug ,(g.uV "(066I) z9s~to£ 'N "IAIpug "IAI"V '~lu.reo "~LI~I-Eg171 '~8 "mV "~°S "tuslaS "ling 's!s£Igug uo!l -ounj azA!ooo~pugqptoaq u! s~ttom!aodxo igo!aztunN '(E66I) "d "f 'Ap!sseD "dd I8£ '~IaOAaxzN 'ssoad &!S~OA!Ulqo~p!~q -mr D '£gOlOU~S/aS fo £toazl£ mtl oj uo~.lonpo.ouI u V "(E96[) "H "?,I 'UOlInfl "EgEI-ELE[ '08 "mY "°°S •ms!as "ling ',~qdgaffodol oogJans jo s13ojjo pug 'gpgugD 'so.uo~iJZOL lsoA~q~oN 's:,IooqsaolJg !uugqg N tuo.tJ s[gA!a.rgOAg~-d po!.~od-~oqs .to so
