You are on page 1of 9

Bulletin of the Seismological Society of America, Vol. 87, No. 3, pp.

701-709, June 1997

Topographic Effects on the Seismic Response of Steep Slopes


by Scott A. Ashford, Nicholas Sitar, John L y s m e r , and N a n D e n g

Abstract

A frequency-domain parametric study using generalized consistent transmitting boundaries has been performed to evaluate the significance of topographic effects on the seismic response of steep slopes. The results show that the peak amplification of motion at the crest of a slope occurs at a normalized frequency 1t/2 = 0.2, where H is the slope height and 2 is the wavelength of the motion. The importance of the natural site frequency is illustrated by the analysis of a stepped layer over a half-space. It was found that the natural frequency of the region behind the crest can dominate the response, relative to the topographic effect, for the conditions studied. Moreover, the effect of topography can be handled separately from the amplification due to the natural frequency of the deposit behind the crest of the slope. This concept of separating the amplification caused by topography from that caused by the natural frequency is advantageous to the development of a simplified method to estimate topographic effects.

Introduction " . . . The effect of the vibration on the hard primary slate, which composes the foundation of the island, was still more curious: the superficial parts of some narrow ridges were as completely shivered as if they had been blasted by gunpowder. This effect, which was rendered conspicuous by the fresh fractures and displaced soil, must be confined to the near surface, for otherwise there would not exist a block of solid rock throughout Chile; nor is this improbable, as it is known that surface of a vibrating body is affected differently from the central part. It is, perhaps, owing to this same reason that earthquakes do not cause such terrific havoc within deep mines as would be expected..." (Barlow, 1933). This quote by Charles Darwin describes the effects of the 20 February 1835 Chilean earthquake and suggests that topographic amplification of seismic motions is a phenomenon that has been well recognized for some time. Certainly, in the recent past, there have been numerous cases of recorded motions and observed earthquake damage pointing toward topographic amplification as an important effect. Examples include observations from the 1971 San Fernando earthquake (Boore, 1972), the 1983 Coalinga earthquake (Celebi, 1991), the 1985 Chile earthquake (Celebi, 1987), the 1987 Superstition Hills earthquake (Celebi, 1991), and the 1994 Northridge earthquake (Ashford and Sitar, 1994). As a result of such observations, a considerable amount of work has been done in an attempt to model, quantify, and predict these effects. One of the first numerical studies of the effect of simple topography on seismic response was carried out by Boore (1972) using the finite-difference method. Subsequent stud701 ies on the effect of topography were conducted using finite elements (e.g., Smith, 1975), boundary methods (e.g., Sanchez-Sesma et aL, 1982), and discrete wavenumber methods (e.g., Bard, 1982). Geli et aL (1988) reviewed these studies and others, and they found that all of them in essence considered the analysis of an isolated two-dimensional ridge on the surface of a homogeneous half-space and that all yielded consistent results: (1) the amplification of acceleration of no more than 2 at the crest, peaking when the wavelength is about equal to the ridge width, and (2) varying amounts of amplification and attenuation along the surface of the slope from the crest to the base. However, these results considerably underestimate amplifications observed in the field, which mostly range from 2 to 10, and up to as much as 30. Geli et al. then analyzed a more detailed model configuration using a layered profile and introduced nearby ridge effects, and they arrived at conclusions similar to those of the previous researchers. In addition, Geli et al. found that neighboring ridges may have greater effect on site response than layering and concluded that future models should be able to analyze SV and surface waves, and three-dimensional geologic configurations. As indicated above, most research on topographic effects has thus far focused on ridges. Some of the procedures and concepts developed for the analysis of the ridges may also be extended to steep slopes. However, there are also significant differences between the response of steep soil slopes and the response of rock ridges simulated as homogeneous half-spaces; foremost are the semi-infinite nature of material in the horizontal direction behind the slope crest and the potential of soil amplification of the motions. With

702 this in mind, it is important to review studies undertaken specifically for soil slopes. One of the first studies to specifically consider the seismic response of soil slopes was conducted by Idriss and Seed (1967). The researchers were prompted by the extensive landslides generated during the 1964 Alaskan earthquake and conducted a parametric study of the response of 27 clay slopes, and later of 45 slopes (Idriss, 1968), using triangular viscoelastic finite elements to model the slopes on rigid foundations. When considering the N/S component of the 18 May 1940 E1 Centro seismogram, the authors found that magnitude of the peak surface acceleration was in all cases greater at the crest of the slope than at points lower on the slope. However, when comparing the peak surface acceleration at the crest to that at some distance behind the crest, they found that while in some cases the acceleration at the crest was much greater, in other cases there was little difference between the response at the crest and the response at some distance behind the crest. Vertical motions generated by the horizontal component of the base motion were greatest near the crest of the slope; however, the vertical component of the base motion had little effect on the horizontal shear stresses within the embankment. Their results suggest that the natural period of the soil column behind the crest of a slope was responsible for much more amplification of the input motion than the slope geometry itself. Kovacs et al. (1971) performed laboratory shaking table experiments on clay banks, in part to further validate the use of the finite-element method (e.g., Idriss and Seed, 1967) in this type of analysis. The physical model results agreed favorably with the finite-element analyses, and Kovacs et al. concluded that the thickness of the soil deposit was the predominant factor in determining the site response. May (1980) studied the effect of horizontally propagating SH and Love waves on vertical scarps in a half-space and a layer over a half-space using the finite-element method. He found that reflection off the scarp face played a large role in the response and that the effect of the scarp could be related to the ratio of slope height/4 and the wavelength of the motion under consideration. He also performed tests using an instrumented granite block to validate his numerical model and found a good comparison between the two models. Sitar and Clough ( 1983 ) used an equivalent linear, two dimensional finite-element model to analyze the seismic response of steep slopes in weakly cemented sands and found that accelerations tended to be amplified in the vicinity of the slope face. Their results show up to a 70% amplification at the crest of the slope as compared to the free field behind the crest. However, in contrast to Geli et al. (1988), they noted that these topographic effects tended to be small relative to the amplification that occurs in the free field due to the site period. Our study was motivated by an interest in evaluating the response of steep coastal bluffs, such as are found along much of the California coast. We explored a complete range

S.A. Ashford, N. Sitar, J. Lysmer, and N. Deng of slope angles between 30 and 90 to produce a set of generalized results applicable to steep slopes. Both vertically propagating SH and SV waves were considered; in addition, wave splitting due to vertical incidence on an inclined slope was incorporated into the analysis. Finally, the responses of a stepped half-space and a stepped layer over a half-space were compared to determine the effect of the fundamental frequency of the material behind the slope crest. Computational Model The computational model used in the study is the generalized consistent transmitting boundary (GCTB) developed and validated by Deng (1991) for two-dimensional seismic site-response analysis. Kausel and Tassoulas (1981) compared several types of transmitting boundaries and found excellent agreement between the consistent transmitting boundary and the closed-form solution for the problem of a time-harmonic antiplane line load acting on a stratum. The generalized consistent transmitting boundary is an extension of the consistent transmitting boundary, developed by Lysmet and Waas (1972), to allow for a boundary of arbitrary shape. One of the key advance' s by Deng is the formulation of the solution to the equation of motion along an arbitrarily shaped boundary in a layered system, specifically along a rectilinear curve, and this representation is the basis for the formulation of the GCTB. The prefix "generalized" refers to the ability of these elements to conform to arbitrarily shaped boundaries. The frequency-domain model is linear viscoelastic and utilizes the complex response method. Only the key elements of the computational model are presented below. The GCTB is formulated by using the exact analytical solution in the horizontal direction and a discretized first- or second-order displacement shape function along the arbitrarily shaped boundary (Deng, 1991). The boundaries of these elements transmit energy accurately in the horizontal direction and represent the perfect "infinite" boundary condition. Nodal points exist only at the boundaries between the regions, and only the motions at the nodal points need to be solved in the global equations of motion. Once the nodal point motions are obtained, the motions within each region can be recovered through a nodal expansion process. A slope model using the GCTB method is presented in Figure 1. The site is divided into two semi-infinite regions on the left and right sides, respectively. Each region is divided into a group of perfectly horizontal layers, with material properties varying from layer to layer, and with the entire model resting on a simulated viscoelastic half-space. Material Properties The material properties used to model each layer consist of the mass density (p), the S-wave velocity (Vs), the P-wave velocity (Ve), and the respective values of the fraction of critical damping for each wave type. The material properties are assumed to be uniform for each layer. The shear modulus

Topographic Effects on the Seismic Response of Steep Slopes

703 wavelengths of the motion in the half-space for the frequency under consideration, as recommended by Chen et al. (1981). The choice of this thickness is based on the observation that fundamental-mode Rayleigh waves in a halfspace decay exponentially with depth, have very small amplitudes at a depth corresponding to 1.5 wavelengths, and become insignificant at a depth of two wavelengths (Ostadan, 1983). In our analysis, the total thickness of the layers is adjusted according to the frequency under consideration, and all layers in the extended region are of uniform thickness. The second technique, also recommended by Chen et al. (1981), is to attach Lysmer-Kuhlemeyer (1969) viscous dashpots at the base of the simulated half-space, thus the base becomes a viscous boundary instead of a rigid boundary. Since the dashpot representation of the half-space is only exact for the vertically propagating P and S waves, and since the directions of the scattered motions are usually unknown, this technique is approximate in the sense that some of the scattered energy may still be able to bounce back into the system. However, the use of both techniques gives very satisfactory results in most practical problems (Deng, 1991). Analysis o f a Stepped Half-space The problem of a steep slope in a uniform viscoelastic material can be simplified to that of a stepped homogeneous, isotropic half-space. The analysis of this problem is very useful for the development of an understanding of the parameters necessary to quantify the effect of topography on seismic response, because the only variables are the slope height and the wavelength. This allows the analysis to focus on the relationship between these two parameters without having to incorporate the natural frequency of the site. Once this relationship is examined, then the influence of other variables such as slope angle and wave inclination can be assessed. The model used in the analyses is similar to that shown in Figure 1, where the left and right GCTB and the viscoelastic half-space have uniform properties: a shear-wave velocity of 300 m/sec and a Poisson's ratio of 0.3, with a slope height H of 30 m. The control point for the input motion is located 90 m directly below the base of the slope. The fraction of critical damping (fl) was varied from 1 to 20%, and it was found that damping had very little effect on amplification at the crest of the slope. Thus, the results for 1% damping are presented for the parametric study, though results for 5% are presented for the first case for comparison. The results of the analyses are presented as a function of 1-1/2, i.e., the ratio of the slope height and the wavelength of the motion under consideration. This definition of the normalized wavelength differs from earlier studies of ridge effects (e.g., Boore, 1972; Geli et aL, 1988) and dams (e.g., Gazetas and Dakoulas, 1992), in which the correlation was made between the wavelength and the width of the topographic feature but is similar to the "dimensionless fre-

Z---Left GTE RightGTE ~_


io,-,nout i ot on

i i i i ! i i i i i ! i i i ! i i i ! i i -]Contl;,-ddt

Figure 1. Generalized consistent transmitting boundary representation of a steep slope.

(G) is obtained from the shear-wave velocity and mass density using the relationship G = (Vs)2p, and the constrained modulus is obtained in a similar manner. Material damping is accounted for through the use of a complex modulus. In our analyses, the complex shear and constrained moduli are defined as G' = G(1 - 2fl2 + i2fls 1,/~s) M" = Mc(1 - 2fl2 + i2fleSfl 2)

(1) (2)

where fls and fie are the fractions of critical damping for S and P waves, respectively. Though it is possible to use different values for fls and tip, both were assumed to be the same in the analyses presented herein. As noted above, the model is linear viscoelastic. If the effect of soil nonlinearity is to be taken into account, then strain-compatible soil properties must be developed elsewhere. Simulation of a Semi-Infinite Base In order to better model realistic conditions, a simulated viscoelastic half-space is used as the model base underlying the generalized consistent transmitting boundaries. Two techniques are used in order to minimize the energy reflected back into the mesh to account for radiation damping (Deng, 1991). This is accomplished by adding additional layers to the model and by the inclusion of viscous dashpots at the model base. The use of the combination of both techniques was originally validated by Chen et al. (1981) against exact solutions for a uniform half-space, a single layer over a halfspace (Mooney and Bolt, 1966), and two layers over a halfspace (Stoneley, 1957). Excellent agreement was found in all cases. One or a combination of both techniques are commonly used in finite-element analyses of seismic site response (e.g., Hudson et al., 1994; Lysmer et al., 1981). The first technique is to add some additional layers to the original model of the site. The total thickness of the additional layers varies with frequency and is set to 1.5

704 quency" proposed by Dakoulas (1993) for the study of S H waves in earth dams. Effect of S H Waves on a Vertical Slope The effect of vertically propagating S H waves (out-ofplane waves) on the seismic response of a vertically stepped half-space is evaluated in the frequency domain over the range of 0.5 to 10 Hz, which includes the typical frequency range of engineering interest and spans the range of dominant frequencies most often observed in large earthquakes. The results are presented as transfer functions, normalized by the free-field motion behind the crest, as a function of the normalized frequencies of motion. The transfer function is the multiple required to transfer the input motion, at a given frequency, from the control point in Figure 1, to the output motion at the point of interest. The actual transfer function is a complex number that accounts for the phase difference between the motions; however, only the magnitudes of the transfer functions are needed to compare the amplification of motion. The results of the analysis are presented in Figures 2 and 3 for damping values of 1 and 5%, respectively, for locations varying from the slope crest to a distance of 4H

S.A. Ashford, N. Sitar, J. Lysmer, and N. Deng behind the slope crest. The transfer functions plotted in Figures 2a and 3a show that increased damping significantly reduces the response of the free field and of the slope, particularly at higher frequencies, as would be expected. These results also show that the effect of the slope is more pronounced at the lower level of damping. However, damping has little effect on the amplification of the motion at the crest as compared to the free field behind the crest, as can be seen from comparison of Figures 2b and 3b, which are nearly identical. Therefore, the remainder of the results for the stepped half-space is presented assuming the case of 1% damping. Considering only the amplification at the crest, a peak amplification of nearly 30% occurs at approximately H/2 = 0.2, a secondary peak of 15% at about H/2 = 0.7, and a null at H/2 = 0.45. The peaks correspond somewhat to the natural frequency of the soil column behind the crest for the height of the slope, which would occur at H/2 = 0.25 and H/2 = 0.75 for the first and second modes, respectively. This implies that the relationship between the slope height and the shear-wave velocity of the soil behind the slope is very important in quantifying the effect of topography. The peak amplification seems to decrease and occur at

-!
1.5-

Cr_~t
...2~... Free~ield

(a)

~1. ~
Z
LL '~"~'-~'~,. " -~10.

it

1.u_ Cr_~t ~
Free.F!eld

(a)

,,=

1I

...... ~.~.~.
i

0.5

~ 0.5

0.01

0.03

0.3 SLOPE HEIGHT/WAVELENGTH


O, 1

0 0.01

0.03 .... 0.1 0.3 SLOPE HEIGHT/WAVELENGTH Cr_~t (b)

,,-4, m ,,,
LL
LI.

Cr_~t _IH_ 1.5


..2~1...

(b)

q
t,i IxI uJ ovId.

_lH_

LU

1..5

...2~...

4H 1 ,;.~-~.~r-~-~-~-

z o F< o
(3.

0.5 <
0

o I--< _ _ 9 _ IJ. <

.4,~f~- - - ~ . ~

0.5

0.01

0.03 0.1 0.3 SLOPE HEIGHT/WAVELENGTH

0.01

0.03 0.1 0.3 SLOPE HEIGHT/WAVELENGTH

'

'

'

'

Figure 2. Horizontal (a) transfer functions and (b) amplifications for vertically incident SH wave on a stepped half-space for various distances behind crest, f l = 1%.

Figure 3. Horizontal (a) transfer functions and (b) amplifications for vertically incident SH wave on a stepped half-space for various distances behind crest,

Topographic Effects on the Seismic Response of Steep Slopes

705 50% of the free-field motion and decreases with increased damping. Finally, the amplitude of the vertical response at the crest tends to increase with increasing frequency and seems to be independent of the horizontal response at frequencies above H/2 > 0.2. Effect of Slope Angle The effect of slope angle on topographic amplification was evaluated by varying the slope angle, m, as shown in Figure 1. Since steep slopes are the subject of this study, only slopes between 45 and 90 are considered (30 and 90 for S V waves). The slope-crest amplification of the SH-wave free-field motion is shown in Figure 6. With decreasing slope angle, the magnitude of the amplification at the first peak decreases from about 25% to about 15%, while the response at higher frequencies tends to increase to about 50%, with no apparent second peak. The horizontal response due to S V waves is shown in Figure 7. In general, the magnitude of the amplification decreases with decreasing slope angle, from about 55% to about 15%, for H/2 < 0.4. Results at higher frequencies, above H/2 = 0.4, indicate no clear trend. The vertical response due to S V waves is presented in Figure 8. As with the horizontal response, the vertical response decreases with decreasing slope angle; however, there is no relationship between the amplification and the natural frequency of the layer behind the slope crest.

a lower frequency with increasing distance from the crest, though in any case, the amplification is on the order of 15 to 20% of the free-field motion. In addition, attenuation occurs at certain frequencies with increasing distance from the crest. At low values of H/2, at which the topographic step is small compared to the wavelength, the slope has little effect on the response. Effect of S V (In-Plane) Waves on a Vertical Slope For the analysis of the response to S V waves, both a horizontal and vertical component need to be considered. Since the input motion only consists of horizontal motion, the transfer functions for the vertical response are given relative to the horizontal input motion, and the vertical amplification is relative to the free-field horizontal response. The results of the horizontal response due to an incident S V wave are shown in Figure 4 for frequencies ranging from 0.1 to 10 Hz. The results are very similar to those obtained for the S H waves. The first peak amplification occurs at HI2 = 0.2, and the second peak occurs at H/2 = 1.0. The magnitudes of both amplification peaks are on the order of 50%, which are higher than those observed for S H waves, with the second peak significantly so. The pattern of attenuation and amplification with increasing distance away from the slope is also similar to the SH-wave case, though the magnitudes are greater for the S V case. The results showing the vertical response are presented in Figure 5. The vertical response is most pronounced at the crest of the slope, and at H/2 > 0.2, it is greater than the free-field horizontal response. Though not presented herein, the amplitude at the crest does not seem to be affected by damping (Ashford and Sitar, 1994). The amplification of the vertical response away from the crest does not exceed about

Analysis of a Stepped L a y e r over a Half-space The parametric study of the stepped half-space provides a basic understanding of the influence of topography on site response. The next step is to evaluate the relationship be-

2
z
m

O
< 1.5 (..)
LL ,--.I 13-

....~.

rl

~.'~./
/
I

<
.,-I

1~

< I.z O N 0.5


E:

'1I I I I i I I

0.01

0.03 0.1 0.3 SLOPE H E I G H T / W A V E L E N G T H

Figure 4. Horizontal amplifications for vertically incident SV wave on a stepped half-space for various distances behind crest, fl = 1%.

706
2

S.A. Ashford, N. Sitar, J. Lysmer, and N. Deng

< L)
..1 Q.

1.5

/
0.03 O. 0.3 SLOPE HEIGHT 'WAVELENGTH

<
_J

(J
n~ 0 . 5
LU

<

>

O.01

.Figure 5. Verticalamplifications for vertically incident SV wave on a stepped halfspace for various distances behind crest, fl = 1%.

tween the natural frequency of the site and topographic amplification. A vertically stepped layer over a half-space was used in the analyses, as shown in Figure 1. The layer has the same material properties as used in the half-space study discussed previously, while the underlying half-space has properties such that the resulting specific impedance is three times that of the layer. The natural frequency of the model behind the crest of the step is varied by changing the thickness of the layer, Z, from H to 5H. The thickness Z is treated as an independent variable because the topographic effect is normalized as a function of HI2, which is dependent on Vs. Changing Vs of the layer, therefore, would not have allowed for the separation of the topographic effect and resonance effect at the natural frequency. The results are presented in Table 1 in the form of the horizontal transfer functions for the free field behind the crest and at the crest of the slope. The transfer functions shown are those at the natural frequency of the free field behind the crest, fn, defined as
Vs L = --

4Z

(3)

and at the topographic frequency, ft, defined as f, Vs

5H

(4)

since the peak effect of topography occurs at about H/2 = 0.2. For the transfer functions discussed below, the subscript "n" refers to the transfer function occurring at the natural frequency, "t" refers to that at the topographic frequency,

"c" refers to the crest location, and ,)o, refers to the free field behind the crest location. A review of the transfer functions for the response at the crest shows that T~c (i.e., the transfer function at the topographic frequency at the crest location) is never greater than Tnc (i.e., the transfer function at the natural frequency at the crest location). The results also show that Tnf (the transfer function atfn in the free field) remains relatively constant. However, at the crest, Tnc increases asfn approaches ft (i.e., as Z/H approaches 1.00). This trend is clearly shown in Figure 9, in which the ratio of the transfer functions, T,JTnf, is plotted versus the ratio of the frequencies, fn/fp At low values o f f , / f , where slope height is small compared to the wavelength at the natural frequency, the transfer function at the crest, T,~, is approximately equal to the free-field transfer function, Tnf. However, when the natural frequency of the site occurs near the topographic frequency, the free-field motion is amplified by over 50%. This amount of amplification is similar to the amount observed at the topography frequency of the stepped half-space. It would seem, therefore, that the effects of the natural frequency and those of topography may work independently. Also presented in Figure 9 are results from Sitar et al. (1980) and Bachus et al. (1981) for a finite-element analysis of a homogeneous isotropic stepped layer supported by a rigid base excited by a transient motion. These results, for an impedance ratio between the base and layer of infinity, confirm the trend of the current study for higher levels of impedance ratio. It seems reasonable that the effect of the natural frequency would be less pronounced for much lower values of impedance ratio, with the limiting case being simply the stepped half-space shown previously. In general, the results of the frequency-domain analysis of a stepped layer over a half-space indicate two important

Topographic Effects on the Seismic Response of Steep Slopes

707

1.5

O i F< L)
LL ,.d ft.

90 75 60 45

degree slope v deg~e slope degr~.e slope d e r ~ slope


L~..t~..~t" s~,=q-~"~-.~ "~

<

0.5

0.01

0.03

0.1

0.3

SLOPE HEIGHT/WAVELENGTH Figure 6. Horizontal amplification at the crest for a vertically incident SH wave on an inclined slope, fl = 1%. 2

1.5

O F<
LL

90 75 60 45 30

I degr~ee slope v decl~.e slope degr~..e,slope degrge slope I deg.r~_eslope ~

i
"Y'%V...r~i

Q.

< 0.5

0.01

0.03 0.1 0.3 SLOPE HEIGHT/WAVELENGTH

Figure 7. Horizontal amplification at the crest for a vertically incident SV wave on an inclined slope, fl = 1%.

points. First, the natural frequency of the site has a greater effect on surface amplification than does the topography for high levels of impedance ratio. Second, it appears that the topographic amplification can be added onto the amplification caused by the natural frequency. Conclusions Based on the parametric study of the seismic response of a stepped half-space and a stepped layer over a half-space, several conclusions can be made. The topographic effect of

a steep slope on the seismic response of that slope can be normalized as a function of the ratio of the slope height (H) and the wavelength of the motion (2). The relationship between slope height and wavelength was also noted by May (1980) for horizontally propagating SH waves incident on a vertical scarp, and similar relationships were observed between structure dimension and wavelength by others (e.g., Boore, 1972; Geli et aL, 1988; Dakoulas, 1993). For both SH and SV waves, the magnitude of the response at the crest of the slope is significantly reduced by increased damping, particularly at higher frequencies. How-

708

S.A. Ashford, N. Sitar, J. Lysmer, and N. Deng

O < (D
ii / n

1.5

I 90 degree slope 75 deg~e slope 60 degr~.e slope 45 degree slope 30 decj..re.e,slope

< 0.5
s~r~

,~
.~.~,. /lt,/~"

~.,,~"

z~"

..,e ''e''

0.01

0--

0.03 O. 1 0.3 SLOPE HEIGHT/WAVELENGTH

Figure 8. Vertical amplification at the crest for a vertically incident SV wave on an inclined slope, fl = 1%.

Table 1 Transfer Funcdonsfor Stepped Layer over Half-space fo


(Hz) 2.0 2.0 2.0 2.0 2.0 2.0 2.0

ZTH 1.~ 1.30 1.60 2.00 3.~ 4.~ 5.00

f,
(nz) 2.5 1.92 1.56 1.25 0.83 0.62 0.5 T.~ 5.2 5.0 4.9 4.3 3.3 3.0 3.0

Crest Tt~ 4.3 4.8 3.0 2.6 1.6 2.1 2.3

Free Field T.s 2.9 2.9 3.0 3.0 2.9 2.8 2.8 T 2.2 2.81.9 1.2 1.2 2.1 1.0

1.75

This_~udy Sitar et { , (1980) Bachus etAal. (1981)

1.5

1,25

10

0.25

0,5

0.75

1.25

1.5

fn/ft
Figure 9. Comparison of transfer function ratio, TnlT4, as a function of frequency ratio, fn/f,

ever, the amplification of the motion at the crest over that in the free field behind the crest is relatively unaffected by damping. The fact that amplification is relatively unaffected by damping in a homogeneous system was also observed by Boore (1972). In addition, for both SH and SV waves, the peak topographic effect occurs at H/2 = 0.2. This amplification is on the order of 25% for SH waves and 50% for SV waves. This first peak, at H/2 = 0.2, approximately corresponds to the first mode of vibration of a soil column of thickness H (H/2 = 0.25), which is the frequency at which Boore (1972) and Geli et aI. (1988) observed the peak response in their studies of ridges. Secondary peaks occur near H/2 = 0.7 for SH waves and H/2 = 1.0 for SVwaves. The vertical component of the topographic effect occurs independently of the natural frequency of the site. Considering the effect of slope angle on response, the topographic effect is most apparent for slopes steeper than 60 and tends to decrease with slope angle. The initial peak horizontal response occurs at the topographic frequency and tends to decrease with increasing slope angle. The vertical response increases with normalized frequency, but this increase is independent of the topographic frequency. For a stepped layer over a half-space, the natural frequency of the site behind the crest dominates the response, which agrees with observations by Sitar and Clough (1983). If the natural frequency of the site is approximately equal to the topographic frequency, i.e.,fn = ft, then that response is amplified. In no case is the topographic effect greater than the response at the natural frequency. Based on these results, it appears that the effect of topography can be handled sep-

Topographic Effects on the Seismic Response of Steep Slopes

709
Idriss, I. M. and H. B. Seed (1967). Response of earthbanks during earthquakes, J. Soil Mech. Found. Div. ASCE, 93 (SM3), 61-82. Idriss, I. M. (1968). Finite element analysis for the seismic response of earth banks, J. Soil Mech. Found. Div. ASCE, 94 (SM3), 617-636. Kansel, E. and J. L. Tassoulas (1981). Transmitting boundaries: A closedform comparison, Bull. Seism. Soc. Am. 71, 143-159. Kovacs, W. D., H. B. Seed, and I. M. Idriss (1971). Studies of seismic response of clay banks, J. Soil Mech. Found. Div. ASCE, 97 (SM2), 441-455. Lysmer, J. and R. L. Kuhlemeyer (1969). Finite dynamic model for infinite media, J. Eng. Mech. Div. ASCE, 95 (EM4), 859-877. Lysmer, J. and G. Waas (1972). Shear waves in plane infinite structures, J. Eng. Mech. Div. ASCE 98, 85-105. Lysmer, J., M. Tabatabaie, F. Tajirian, S. Vahdani, and F. Ostadan (1981). SASSI, a System for Analysis of Soil-Structure Interaction, UCB/GT/ 81-02, University of California, Berkeley. May, T. W. (1980). The effectiveness of trenches and scarps in reducing seismic energy, Ph.D. Thesis, University of California at Berkeley, Berkeley, California. Mooney, H. M. and R. A. Bolt (1966). Dispersive characteristics of the first three Rayleigh modes for a single surface layer, Bull. Seism. Soc. Am. 56, 43-67. Ostadan, F. (1983). Dynamic analysis of soil-pile-structure systems, Ph.D. Thesis, University of California at Berkeley, Berkeley, California. S~inchez-Sesma, F., I. Herrera, and J. Aviles (1982). A boundary method for elastic wave diffraction: application to scattering SH waves by surface irregularities, Bull. Seism. Soc. Am. 72, 473-490. Sitar, N., G. W. Clough, and R. Bachus (1980). Behavior of weakly cemented soil slopes under static and seismic loading, Report No. 44, The John A. Blume Earthquake Engineering Center, Stanford University. Sitar, N. and G. W. Clough (1983). Seismic response of steep slopes in cemented soils, J. Geotech. Eng. ASCE, 109, 210-227. Smith, W. D. (t975). The application of finite element analysis to elastic body wave propagation problems, Geophys. Z R. Astr. Soc. 42, 747768. Stoneley, R. (1957). The attenuation of Rayleigh waves with depth in a medium with two surface layers, Roy. Astr. Soc., Geoph. Suppl. 7, no. 5. Department of Applied Mechanics and Engineering Science University of California, San Diego La Jolla, California 92093-0085 (S .A.A.) Department of Civil Engineering University of California at Berkeley Berkeley, California 94720 (N.S., J.L.) Bechtel Corporation P.O. Box 193965 San Francisco, California 94119-3965 (N.D.) Manuscript received 19 October 1995.

arately from amplification due to the natural frequency of the layer behind the crest of the slope. This concept of separating the amplification caused by topography from that caused by the natural frequency is advantageous to the development of a simplified method to estimate topographic effects. Acknowledgments
This research was supported by the U.S. Geological Survey under USGS Award Number 14-08-0001-G2127. The views and conclusions contained in this article are those of the authors and should not be interpreted as necessarily representing the official policies, either expressed or implied, of the U.S. government. Partial support was also provided by the California Department of Transportation. The financial support of the donors is greatly appreciated by the authors.

References
Ashford, S. A. and N. Sitar (1994). Seismic Response of Steep Natural Slopes, Report No. UCB/EERC 94-05, Earthquake Engineering Research Center, College of Engineering, University of California at Berkeley. Bachus, R. C., G. W. Clough, N. Sitar, N. Shafii-Rad, J. Crosby, and P. Kaboli (1981). Behavior of weakly cemented soil slopes under static and seismic loading. Vol. II, Report No. 52, The John A. Blume Earthquake Engineering Center, Stanford University. Bard, P.-Y. (1982). Diffracted waves and displacement field over two-dimensional elevated topographies, Geophys. J. R. Astr. Soc. 71, 731760. Barlow, N. (1933). Charles Darwin's Diary of the Voyage of H.M.S. Beagle, Nora Barlow (Editor), Cambridge U Press, New York. Boore, D. M. (1972). A note on the effect of simple topography on seismic SH waves, Bull. Seism. Soc. Am. 62, 275-284. Celebi, M. (1987). Topographic and geological amplification determined from strong-motion and aftershock records of the 3 March 1985 Chile earthquake, Bull. Seism. Soc. Am. 77, 1147-1167. Celebi, M. (1991). Topographic and geological amplification: case studies and engineering implications, Struct. Safety 10, 199-217. Chen, J.-C., J. Lysmer, and H. B. Seed (1981). Analysis of Local Variations in Free Field Seismic Ground Motion, Report No. UCB/EERC-81/03, Earthquake Engineering Research Center, College of Engineering, University of California at Berkeley. Dakoulas, P. (1993). Earth dam-canyon interaction effects for obliquely incident SH waves, J. Geotech. Eng. ASCE, 119, 1696-1716. Deng, N. (1991). Two-dimensional site response analyses, Ph.D. Thesis, University of California at Berkeley, Berkeley, California. Gazetas, G. and P. Dakoulas (1992). Seismic analysis and design of rockfill dams: state-of-the-art, Soil Dyn. Earthquake Eng. 11, 27-61. Geli, L., P.-Y. Bard, and B. Jullien (1988). The effect of topography on earthquake ground motion: a review and new results, Bull. Seism. Soc. Am. 78, 42-63. Hudson, M., I. M. Idriss, and M. Beikae (1994). User's Manual for QUAD4M, Center for Geotechnical Modeling, Department of Civil and Environmental Engineering, University of California, Davis.