See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/228763319

Smooth Spectra of Horizontal and Vertical Ground Motions

Article  in  Bulletin of the Seismological Society of America · May 2006

DOI: 10.1785/0120050062

CITATIONS READS
55 586

1 author:

Praveen K. Malhotra
StrongMotions Inc.
74 PUBLICATIONS   1,521 CITATIONS   

SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Performance-based seismic design of a liquid nitrogen tank in California. View project

Ground Motion Analysis and Interpretation View project

All content following this page was uploaded by Praveen K. Malhotra on 30 July 2015.

The user has requested enhancement of the downloaded file.

 Bulletin of the Seismological Society of America, Vol. 96, No. 2, pp. 506–518, April 2006, doi: 10.1785/0120050062

Smooth Spectra of Horizontal and Vertical Ground Motions

by Praveen K. Malhotra

Abstract An improved method of constructing a smooth-response spectrum from

peak values of ground acceleration, velocity, and displacement (PGA, PGV, and PGD)
is presented. Improved dynamic amplification factors are presented for applying
damping adjustments to the spectral accelerations or to backcalculate PGA, PGV, and
PGD from spectral accelerations. Horizontal-to-vertical spectral ratios are analyzed
for rock and soil sites to allow the construction of a vertical design spectrum from a
given horizontal design spectrum.

Introduction
Newmark and Hall (1969, 1982) proposed a smooth- ridge, California, earthquake. Located on a rock site, the
response spectrum, which had three regions: acceleration station is called “Castaic–Old Ridge Route” by the Califor-
(short period), velocity (medium period), and displacement nia Strong Motion Instrumentation Program (CSMIP, 1994).
(long period). These regions are constructed by applying dy- These histories were downloaded from the COSMOS (1999–
namic amplifications to design values of peak ground ac- 2005) web page. The peak values of ground acceleration,
celeration (PGA), peak ground velocity (PGV), and peak velocity, and displacement for the Castaic 90
motion are:
ground displacement (PGD). PGA ⳱ 0.568g, PGV ⳱ 51.5 cm/sec, and PGD ⳱ 8.47 cm.
Numerous researchers contributed to the development To test the mutual compatibility of processed accelera-
of the Newmark–Hall spectrum (e.g., Hall et al. 1975; tion, velocity, and displacement histories, we integrate and
Mohraz et al., 1972; Mohraz, 1976; Seed et al., 1976). The double-integrate the acceleration history (Fig. 1, top trace)
Newmark–Hall spectrum, with some modifications, has been to obtain the velocity and displacement histories shown in
adopted by building codes around the world. In the 1997 Figure 2. Significant differences between the processed and
Uniform Building Code (ICBO, 1998), the PGA (Ca) is used integrated displacement histories (Figs. 1 and 2, bottom
to construct the acceleration region and the spectral accel- traces) indicate that the response spectrum computed from
eration at 1 sec Cv is used to construct the velocity region. the acceleration history alone will not show correct asymp-
In the 2003 IBC (ICC, 2003) and ASCE 7 (2002), the spectral totic behavior at long periods.
acceleration at 0.2 sec SDS is used to construct the acceler- We use the method proposed by Malhotra (2001) to
ation region and the spectral acceleration at 1 sec SD1 is used compute the response spectrum. In this method, the short-
to construct the velocity region. Researchers continue to pro- period spectral values are derived from the processed accel-
pose site-specific response spectra that are similar in shape eration history and the medium- and long-period spectral
to the Newmark–Hall spectrum (e.g., Lam et al., 2000). The values are derived from the processed velocity and displace-
dynamic amplification factors proposed by Newmark and ment histories. The computed response spectrum shows cor-
Hall (1982) are the basis of damping adjustments applied to rect asymptotic behavior at both short and long periods, that
the 5% damping response spectra in most building codes. is, the spectral acceleration (SA) approaches PGA at short
These factors are also used to backcalculate PGA, PGV, and periods and the spectral deformation (SD) approached PGD
PGD from spectral accelerations for the purpose of selecting at long periods. Figure 3 shows the 5% damping response
spectrum-compatible ground-motion histories. spectrum of the Castaic 90
ground motion (Fig. 1) in a
The objectives of this article are to (1) re-examine the tripartite form. The spectral velocity (SV) is read along the
Newmark–Hall procedure and dynamic amplification factors vertical axis, the SA is read along the ⳮ45
axis, and the SD
using a new approach and a broader dataset and (2) analyze is read along the Ⳮ45
axis, with respect to the natural pe-
horizontal-to-vertical spectral ratios for constructing a vertical riod T along the horizontal axis. These quantities are related
design spectrum from a given horizontal design spectrum. to each other as follows:
2
Tripartite Response Spectrum SA • 冢2pT 冣 ⳱ SD ⳱ SV • 冢2pT 冣 . (1)
Figure 1 shows the processed acceleration, velocity, and
displacement histories of horizontal ground motion recorded The three lines corresponding to PGA, PGV, and PGD are
at a station in the 90
(east) direction, during the 1994 North- shown in Figure 3.

506
Smooth Spectra of Horizontal and Vertical Ground Motions 507

Figure 1. Processed acceleration, velocity,

and displacement histories of ground motion
recorded during the Mw 6.7 1994 Northridge,
California, earthquake, at “Castaic–Old Ridge
Route” station in the 90
(east) direction. The
data were downloaded from the COSMOS web
page (http://db.cosmos-eq.org/).

Figure 2. Velocity and displacement his-

tories obtained from integration and double-
integration of the processed acceleration his-
tory of Castaic 90
motion (top trace in Fig. 1).

Correlation of SA with PGA, PGV, and PGD for periods up to 0.63 sec; after that the correlation reduces
significantly. This implies that the SAs for periods longer
PGA is determined by high frequencies in the ground
than 0.63 sec cannot be reliably predicted from PGA. Figure
motion, which also control the response of short-period sys-
6 shows a plot of correlation between SAs and PGA, PGV,
tems. Therefore, the short-period SAs should correlate well
and PGD. SAs for periods up to 0.62 sec correlate best with
with PGA. Figure 4a shows 5% damping SA(0.2 sec) plotted
PGA, those for periods between 0.62 sec and 2.6 sec correlate
versus PGA for 63 ground motions. (Thirty-three of these
ground motions were recorded on rock sites and the remain- best with PGV, and the rest correlate best with PGD. (For a
ing thirty were recorded on soil sites. The PGA for these different set of ground motions, the cutoff periods could be
ground motions ranged from 0.03g to 1.8g.) Figure 4b shows different.) Throughout the period range, the SAs are reason-
5% damping SA(1 sec) plotted versus PGA for the same set ably well correlated with PGA, PGV, or PGD. This implies
of ground motions. As expected, the correlation between that the entire response spectrum can be reliably constructed
SA(0.2 sec) and PGA is much better than the correlation be- from PGA, PGV, and PGD values.
tween SA(1 sec) and PGA. Hall et al. (1975) and Mohraz (1976) assumed that the
Figure 5 shows a plot of correlation between SA and SAs for periods 0.33 sec are sensitive to PGA, those for
PGA for 63 ground motions. The correlation is good (80%) periods between 0.33 sec and 3.33 sec are sensitive to PGV,
508 P. K. Malhotra

and the rest are sensitive to PGD. They used the ratios be- a priori assumption of the cutoff periods to construct a
tween spectral values and peak ground-motion parameters smooth-response spectrum. First, the spectra of various
(PGA, PGV, and PGD) to determine the average amplification ground motions are normalized such that their high- and
factors used in the construction of smooth-response spectra. low-frequency regions are aligned with each other. Next, the
Note that the cutoff periods used by Hall et al. (1975) and amplification factors are computed from statistical analysis
Mohraz (1976) are significantly different from those re- of the normalized spectra.
vealed by the correlation plots in Figure 6. Because the cut-
off periods can change from one set of ground motions to
Normalized Spectrum
another, the approach used in this study does not make
Figure 7a shows the normalized response spectrum of
the Castaic 90
ground motion, obtained as follows. First,
the “central” period of the ground motion is calculated:

冪
PGD
Tcg ⳱ 2p . (2)
PGA

Tcg marks the boundary between the high- and low-

frequency regions of the response spectrum (Malhotra,
2001). The period along the horizontal axis is normalized
with respect Tcg, causing a horizontal shift in the response
spectrum. The PGA changes to PGA ⳯ Tcg /2p; PGV remains
unchanged; and the PGD changes to PGD ⳯ 2p/Tcg . Next,
the SV and the PGV are normalized with respect to
冪PGA • PGD. This causes both PGA and PGD to become
unity and PGV and SV to take the following nondimensional
forms:

PGV
PGV ⳱ (3)
冪PGA • PGD

SV
Figure 3. Tripartite plot of 5% damping-response SV ⳱ . (4)
spectrum of Castaic 90
motion shown in Figure 1. 冪PGA • PGD

Figure 4. Spectral accelerations versus peak ground acceleration for 63 ground

motions. (a) SA(0.2 sec) versus PGA. (b) SA(1 sec) versus PGA.
Smooth Spectra of Horizontal and Vertical Ground Motions 509

Figure 5. Correlation of SA at various periods with PGA.

Figure 6. Correlation of SA at various periods with PGA, PGV, and PGD.

The normalized peak ground velocity PGV is shown by by a single frequency of period Tcg. Because earthquake
a horizontal line in Figure 7a. The higher the PGV, shorter ground motions, in general, are composed of many frequen-
the width of the horizontal line. In the limiting case, when cies, PGV is usually less than 1. The smaller the PGV, the
PGV ⳱ 1, the PGA and the PGD lines meet and the hori- wider the bandwidth of frequencies present in the ground
zontal line disappears. The ground motion is then dominated motion. Mohraz et al. (1972) (and Mohraz 1976) used a
510 P. K. Malhotra

Figure 7. Normalized 5% damping response spectra. (a) Castaic 90
ground motion.
(b) Sixty-three horizontal ground motions.

similar parameter PGA • PGD/PGV2 to characterize ground and displacement amplification factors are ␣A ⳱ 2.14,
motions on rock and soil sites. They showed that ground ␣V ⳱ 1.63, and ␣D ⳱ 1.84. The control periods (where the
motions on soil sites have lower PGA • PGD/PGV2 (therefore straight-line segments meet) are T1 ⳱ 0.035Tcg, T2 ⳱
higher PGV). This implies that soil ground motions have 0.2Tcg, T3 ⳱ 0.5Tcg, T4 ⳱ 1.7Tcg, T5 ⳱ 4.6Tcg, and T6 ⳱
narrower bandwidths of frequencies. The high- and low- 17Tcg. The smooth spectra were also computed for several
frequency regions of the normalized spectrum are to the left other damping ratios. Table 1 lists the amplification factors
and right of T/Tcg ⳱ 1 (Fig. 7a). for several damping ratios.
Next, the rock and soil records were separated and the
amplification factors computed separately for these two
Smooth Spectrum of Horizontal Motion groups. The differences between the rock and soil amplifi-
Figure 7b shows the normalized 5% damping response cation factors were statistically insignificant according to the
spectra of 63 horizontal ground motions. The high- and low- analysis of variance by MATLAB routine ANOVA1
frequency regions of various spectra are aligned in this plot. (MathWorks, 2004). Therefore, only one set of horizontal
The central period of these ground motions Tcg ranged from amplification factors is proposed in this article.
0.39 sec to 2.3 sec and the PGV ranged from 0.35 to 1.
Figure 8a shows the median normalized spectrum ob- Smooth Spectrum of Vertical Motion
tained by averaging log SV at each normalized period:
Figure 9a shows the median 5% damping normalized
N spectra of 17 rock Ⳮ 13 soil motions in the vertical direc-
llog SV ⳱
1
N
• 兺 log SVi , (5) tion. The shaded area corresponds to Ⳳ1 standard deviation
i⳱1 about the median. The central period Tcg of vertical motions
ranged from 0.28 sec to 1.9 sec and the PGV ranged from
where N ⳱ 63 is the number of records. The median spec- 0.27 to 0.77, with a median value of 0.47. In general, the
trum is a plot of exp(llog SV) versus the normalized period. central period of vertical motions is shorter than that for the
The shaded area corresponds to Ⳳ1 standard deviation about horizontal motions, implying a bias toward high frequencies
the median. The PGV ⳱ 0.66 shown in Figure 8a is the in the vertical motions. A similar observation has been made
median value for the 63 ground motions. in the past (e.g., Campbell and Bozorgnia, 2003). The PGV
Figure 8b shows a smooth-response spectrum obtained for the vertical motions is much smaller than that for the
by least-squares fitting of straight-line segments through the horizontal motions, implying that the vertical motions are
median curve shown in Figure 8a. The acceleration, velocity, more wide-banded than the horizontal motions.
Smooth Spectra of Horizontal and Vertical Ground Motions 511

Figure 8. Normalized 5% damping median spectrum of 63 horizontal ground mo-

tions. (a) Actual median spectrum (shaded area corresponds to Ⳳ1 standard deviation
about the median). (b) Smooth median spectrum.

Table 1 ␣(y) ⳱ a Ⳮ b • ln y Ⳮ c • (ln y)2, (6)

Acceleration-, Velocity-, and Displacement-Amplification Factors
for Median Horizontal Spectrum of Different Damping Ratios where ␣ ⳱ ␣A, ␣V, or ␣D and y ⳱ percentage of critical
(% of Critical)
damping. The coefficients a, b, and c in equation (6) are
Damping Ratio determined through least-squares fitting of the data points
(%) ␣A ␣V ␣D
(circles, stars, and triangles in Fig. 10). The functional forms
0.5 3.56 2.46 2.47 of the horizontal amplification factors are presented in the
1 3.17 2.27 2.33 second column of Table 3. Similarly, the functional forms
2 2.75 2.02 2.15
5 2.14 1.63 1.84
of vertical amplification factors were also determined. These
10 1.70 1.30 1.56 are presented in the third column of Table 3.
20 1.31 0.97 1.29
30 1.12 0.78 1.14
Comparison with Previously Published Values
Figure 11 compares the horizontal amplification factors
The smooth median spectrum of the vertical motions is obtained in this study with those reported by Mohraz (1976)
shown in Figure 9b. Table 2 lists the amplification factors and Newmark and Hall (1982). The Mohraz values are the
for various damping ratios. average of those for various soil conditions. The ␣A values
The rock and soil records in the vertical direction were obtained in this study are up to about 10% higher than
separated and the amplification factors computed separately Newmark–Hall and about 10% lower than Mohraz. These
for these two groups. The differences between the rock and differences are considered to be negligible. The ␣V values
soil amplification factors were statistically insignificant obtained in this study are up to about 10% lower than
according to the analysis of variance by MATLAB routine Newmark–Hall and about 20% higher than Mohraz. The dif-
ANOVA1 (MathWorks, 2004). Only one set of vertical am- ferences from the Mohraz values are somewhat significant.
plification factors, therefore, is proposed in this article. The ␣D values obtained in this study are up to about 30%
higher than Newmark–Hall and about 10% lower than
Functional Form of Amplification Factors Mohraz. The differences from the Newmark–Hall values are
significant. The possible causes of differences are (1) a
Figure 10 displays the variation of horizontal amplifi- priori assumption of cutoff periods in previous studies and
cation factors with damping (from Table 1). The following (2) asymptotically incorrect behavior of response spectra
function is fitted through the data points: used in those studies.
512 P. K. Malhotra

Figure 9. Normalized 5% damping median spectrum of 30 vertical ground motions.

(a) Actual median spectrum (shaded area corresponds to Ⳳ1 standard deviation about
the median). (b) Smooth median spectrum.

Table 2 1. Obtain amplification factors ␣A, ␣V, and ␣D from Table

Acceleration-, Velocity-, and Displacement-Amplification Factors 3 for appropriate damping.
for Median Vertical Spectrum of Different Damping Ratios 2. Compute the control periods T3 and T4 as follows:
(% of Critical)
Damping Ratio PGV • ␣V
(%) ␣A ␣V ␣D T3 ⳱ 2p • (7)
PGA • ␣A
0.5 3.73 2.16 2.87
1 3.18 1.95 2.78
2 2.59 1.70 2.56 PGD • ␣D
5 1.96 1.35 2.17 T4 ⳱ 2p • (8)
10 1.55 1.06 1.77
PGV • ␣V
20 1.22 0.801 1.39
30 1.07 0.664 1.18 3. Compute control periods T1 and T2 as follows:

T1 ⳱ 冦TT /10
3
3
/14 for horizontal
for vertical
(9)
Procedure for Constructing a Smooth Spectrum
An improved procedure is presented for constructing a T2 ⳱ 冦TT /2.5
3
3 /2.7
for horizontal
for vertical
(10)
smooth-response spectrum from the PGA, PGV, and PGD val-
ues (Fig. 12). This procedure is similar to the Newmark–
Hall procedure, but there are some important differences. 4. Compute control periods T5 and T6 as follows:
First, the amplification factors are different. Second, the con-
trol periods T1, T2, T5, and T6 are not absolute—they are
relative to the “central” period of the ground motion, rec-
T5 ⳱ 冦1.5T
2.7T 4
4
for horizontal
for vertical
(11)

ognizing that the frequency-content can vary significantly

from one ground motion to another. In the next section, the
accuracy of this procedure is tested for various ground mo-
T6 ⳱ 冦10T
5T
4
4
for horizontal
for vertical
(12)

tions.
The following steps should be used to construct a 5. Draw the smooth spectrum (Fig. 12) on a tripartite paper
smooth spectrum from PGA, PGV, and PGD, for desired as follows: SA ⳱ PGA for T  T1. SA ⳱ ␣A • PGA for
damping ratio: T2  T  T3. Join SA(T1) ⳱ PGA and SA(T2) ⳱ ␣A PGA
Smooth Spectra of Horizontal and Vertical Ground Motions 513

Figure 10. Median amplification factors versus damping ratio for horizontal
ground-motion spectra.

Table 3
Functional Forms of Amplification Factors for Horizontal and Vertical Spectra
(f ⳱ Percent of Critical Damping)
Horizontal Vertical

␣A 3.148 ⳮ 0.628 ln f Ⳮ 0.007 (ln f)2

3.14 ⳮ 0.832 ln f Ⳮ 0.065 (ln f)2
␣V 2.254 ⳮ 0.337 ln f ⳮ 0.029 (ln f)2 1.935 ⳮ 0.353 ln f ⳮ 0.008 (ln f)2
␣D 2.321 ⳮ 0.253 ln f ⳮ 0.029 (ln f)2 2.747 ⳮ 0.26 ln f ⳮ 0.062 (ln f)2

by a straight line. SV ⳱ ␣V • PGV for T3  T  T4. SD Damping Adjustment to the Design-Response

⳱ PGD for T  T6. SD ⳱ ␣D PGD for T4  T  T5. Join Spectrum
SD(T6) ⳱ PGD and SD(T5) ⳱ ␣D • PGD by a straight line.
The design-response spectrum of horizontal ground mo-
tion in the 2003 IBC (ICC, 2003) and ASCE 7 (2002) is con-
structed from 5% damping SAs at 0.2 sec SDS, and at 1 sec
Validation of Proposed Procedure SD1, which are based on probabilistic seismic hazard analysis
(Cornell, 1968; Reiter, 1991; Frankel et al., 1996; Petersen
The smooth spectra of numerous horizontal ground mo- et al., 1996; Leyendecker et al., 2000; McGuire, 2004). First,
tions were computed from their PGA, PGV, and PGD values, the control period defining the transition between the accel-
by the method described in the previous section. The actual eration- and velocity-sensitive regions is computed:
spectra of these ground motions were also calculated by the
method described by Malhotra (2001). Figure 13a shows the SD1
ratios between the 5% damping actual and smooth spectra TS ⳱ • 1 sec. (13)
SDS
at various natural periods. Figure 13b shows the median
spectral ratio; the shaded area corresponds to Ⳳ1 standard The control period marking the beginning of the “flat”
deviation about the median. The fact that the median ratio acceleration-sensitive region is assumed to be:
between the actual and smooth SVs is close to unity for all
natural periods confirms the accuracy of the procedure. T0 ⳱ 0.2TS . (14)
514 P. K. Malhotra

Figure 11. Comparison of horizontal dynamic amplification factors obtained in this

study with those obtained in previous studies. The values obtained in this study are
shown as the percent difference from the previous studies.

and T3 in the smooth spectrum of a ground motion (Fig. 12).

The ratio T0 /TS ⳱ 0.2 is smaller than the average T2/T3 ⳱
1/2.5 ⳱ 0.4 (equation 10). This is because the code spectrum
is conservative and it is a composite of several seismic
events—it does not correspond to one ground-motion his-
tory. Note that the code spectrum assumes PGA ⳱ 0.4SDS,
which is lower than the average ratio of 0.47 (1/2.12) be-
tween the maximum SA and PGA. This is slightly unconser-
vative, but it is not likely to affect the design of most struc-
tures.
To obtain a spectrum for a damping value other than
5%, we first need to obtain SDS and SD1 for that damping
ratio,

SDS(f) ⳱ SDS /bS (16)

SD1(f) ⳱ SD1 /b1 , (17)

where bS is the ratio between ␣A at 5% damping and that at

the desired damping, and b1 is the ratio between ␣V at 5%
damping and that at desired damping. bS and b1 for various
Figure 12. Construction of smooth-response damping ratios are listed in Table 4. These values are up to
spectrum from PGA, PGV, and PGD. 22% different from those in the FEMA 356 (2000), which

Table 4
Next, the SAs in various period ranges are computed from: Damping Adjustment Factors for Use in Equations (16) and (17)
Damping Ratio
SA(T) (15) (%) bS b1

冦
0.4SDS Ⳮ 3SDS • T/TS for T  T0
0.5 0.602 0.662
⳱ SDS for T0  T  TS .
1 0.676 0.719
SD1 • 1 sec/T for T  TS 2 0.781 0.806
5 1 1
The 5% damping-response spectrum obtained by the 10 1.26 1.25
20 1.63 1.67
preceding procedure is shown in Figure 14a. Note that con-
30 1.91 2.08
trol periods T0 and TS are analogous to control periods T2
Smooth Spectra of Horizontal and Vertical Ground Motions 515

Figure 13. Ratios between 5% damping smooth spectrum (from this study) and
actual spectrum of 63 horizontal ground motions. (a) Actual ratios. (b) Median ratios
(shaded area corresponds to Ⳳ1 standard deviation about the median).

Figure 14. Design-response spectrum for 5% damping (a) and different damping
ratios (b).

Design Spectrum for Vertical Motion

were adopted from Naeim and Kircher (2001). The response
spectrum for any damping can be constructed from SDS and The design-response spectrum for vertical motion is
SD1 for that damping using equations (13) to (15), except usually not obtained from a probabilistic seismic hazard
that the spectral acceleration at 0 sec (PGA) is 40% of SDS analysis. It is usually assumed that the vertical spectrum is
for 5% damping. Figure 14b shows the design spectra for two-thirds of the horizontal spectrum (e.g., FEMA-356,
three values of damping. 2000). This implies that the frequency distribution in the
516 P. K. Malhotra

vertical direction is similar to that in the horizontal direction. is higher than that for the rock motions. The scatter can
However, the analyses in this and previous studies have probably be reduced by using a broader dataset and dividing
shown that the frequency distribution in the vertical direction records into different soil categories. The median spectral
is quite different from that in the horizontal direction. ratio at 0.2 sec is 0.73 (0.75) and that at 1 sec is 0.28
Vertical-to-horizontal spectral ratios were computed for (0.3). The period dependence of vertical-to-horizontal
33 rock motions Ⳮ 30 soil motions. Figure 15a shows the spectral accelerations is more pronounced for soil motions
5% damping spectral ratios for rock motions. The median than for rock motions. It appears that the soil amplification
spectral ratio is shown by a solid line in Figure 15b; the of short-period (high-frequency) SAs is greater in the vertical
shaded area represents Ⳳ1 standard deviation about the me- direction than in the horizontal direction, therefore, higher
dian. The vertical-to-horizontal spectral ratio is period de- vertical-to-horizontal spectral ratio at short periods on soil
pendent. This has been shown previously (e.g., Campbell sites. The soil amplification of 1-sec period SA is higher in
and Bozorgnia, 2003). The logarithmic standard deviation the horizontal direction than in the vertical direction, there-
does not change significantly with the natural period. The fore, lower vertical-to-horizontal spectral ratio at 1-sec pe-
median spectral ratio at 0.2 sec is 0.53 (0.55) and that at riod on soil sites.
1 sec is 0.38 (0.4). By using the vertical-to-horizontal spectral ratios, we
Figure 16a shows the vertical-to-horizontal spectral ra- can construct a vertical ground-motion spectrum from a
tios for soil motions. The median and Ⳳ1 standard deviation given horizontal ground-motion spectrum, by first comput-
values of spectral ratio are shown in Figure 16b. The scatter ing the 0.2-sec and 1-sec SAs of vertical motion, as follows:

Figure 15. Vertical-to-horizontal spectral ratios of rock motions. (a) 5% damping

spectral ratios for 33 pairs of rock motions. (b) Median spectral ratio (shaded area
corresponds to Ⳳ1 standard deviation about the median).

Figure 16. Vertical-to-horizontal spectral ratios of soil motions. (a) 5% damping

spectral ratios for 30 pairs of soil motions. (b) Median spectral ratio (shaded area
corresponds to Ⳳ1 standard deviation about the median).
Smooth Spectra of Horizontal and Vertical Ground Motions 517

V
SDS ⳱ 冦0.75S
0.55S DS
DS
for rock
for soil
(18)
4. The procedure proposed in this article improves the ac-
curacy of the smooth Newmark–Hall spectrum con-
structed from the PGA, PGV, and PGD values because
V
SD1 ⳱ 冦0.4S
0.3S
D1
D1
for rock
for soil .
(19)
the control periods are not assumed to be fixed, they are
allowed to change from one ground motion to another.
5. The damping adjustment factors proposed in this study
The design spectrum for vertical motion is obtained by sub- are up to 23% different from those adopted by FEMA
V V
stituting SDS and SD1 with SDS and SD1 , respectively, in equa- 356.
tions (13) to (15) Figure 17a and 17b show the 5% damping 6. The central period Tcg and normalized velocity PGV are
spectra of vertical motion for rock and soil sites, respec- useful frequency parameters of ground motion. The cen-
tively. tral period of vertical motions is lower than that of hor-
izontal motions, implying a bias toward high frequencies
Conclusions in the vertical motions. The normalized velocity of ver-
tical motions is lower, implying wider frequency bands
1. The acceleration- and velocity-amplification factors ob- for vertical motions.
tained in this study are close to the widely used values 7. The vertical spectrum does not have the same shape as
of these factors. The displacement-amplification factors the horizontal spectrum.
are significantly different. These differences are possibly 8. The period dependence of vertical-to-horizontal spectral
due to a priori assumption of cutoff periods in previous ratios is more pronounced for soil sites than for rock
studies, or asymptotically incorrect behavior of response sites.
spectra used in previous studies. 9. The soil amplification of short-period SAs is higher in
2. The acceleration-amplification factors of vertical motion the vertical direction than in the horizontal direction.
are nearly same as those for horizontal motion. The The soil amplification of 1-sec period SAs is higher in
velocity-amplification factors of vertical motion are the horizontal direction than in the vertical direction.
roughly 15% lower than those for horizontal motion and The short-period SAs in the vertical direction are
the displacement-amplification factors of vertical mo- roughly 0.55 and 0.75 times those in the horizontal di-
tion are up to 20% higher than those for horizontal rection for rock and soil sites, respectively. The 1-sec
motion. period spectral accelerations in vertical direction are
3. The effect of soil on amplification factors ␣A, ␣V, and roughly 0.4 and 0.3 times those in the horizontal direc-
␣D is found to be statistically insignificant for both hor- tion for rock and soil sites, respectively.
izontal and vertical motions. 10. The correlation of SAs with PGA drops sharply after

Figure 17. Design-response spectra of vertical ground motion derived from 0.2-sec
and 1-sec period SAs of horizontal ground motion. (a) Rock sites. (b) Soil sites.
 518 P. K. Malhotra

about 0.6 sec natural period. Therefore, reliable esti- Mohraz, B. (1976). A study of earthquake response spectra for different
mates of SAs cannot be obtained from PGA alone for geological conditions, Bull. Seism. Soc. Am. 66, no. 3, 915–935.
Mohraz, B., W. J. Hall, and N. M. Newmark (1972). A study of vertical
periods longer than 0.6 sec The addition of PGV extends and horizontal earthquake spectra, AEC Report WASH-1255, Nathan
the range of reliable estimates to about 2.6 sec natural M. Newmark Consulting Engineering Services, Urbana, Illinois.
period. Beyond 2.6-sec period, the PGD is needed to Naeim, F., and C. A. Kircher (2001). On the damping adjustment factors
reliably estimate the spectral accelerations. for earthquake response spectra, Struct. Des. Tall Build. 10, 361–369.
Newmark, N. M., and W. J. Hall (1969). Seismic design criteria for nuclear
reactor facilities, in Proc. World Conf. on Earthquake Eng. 4th San-
tiago, Chile, B-4, 37–50.
Newmark, N. M., and W. J. Hall (1982). Earthquake Spectra and Design,
Acknowledgments Earthquake Engineering Research Institute, Oakland, California.
Petersen, M., W. Bryant, C. Cramer, T. Cao, M. Reichle, A. Frankel, J.
Two anonymous reviewers provided several helpful suggestions. Lienkaemper, P. McCrory, and D. Schwartz (1996) Probabilistic seis-
Louis Gritzo and Franco Tamanini provided the encouragement. mic hazard assessment for the state of California, Calif. Div. Mines
Geol. Open-File Rept. 96-08 and U.S. Geol. Surv. Open-File Rept.
96-706.
References Reiter, L. (1991). Earthquake Hazard Analysis—Issues and Insights, Co-
lumbia University Press, New York, 254 pp.
American Society of Civil Engineers (ASCE) (2002). Minimum design Seed, H. B., C. Ugas, and J. Lysmer (1976). Site-dependent spectra for
loads for buildings and other structures, in SEI/ASCE 7-02, Reston, earthquake-resistant design, Bull. Seism. Soc. Am. 66, no. 1, 221–243.
Virginia.
Campbell, K. W., and Y. Bozorgnia (2003). Updated near-source ground
motion (attenuation) relations for the horizontal and vertical compo-
nents of peak ground acceleration and acceleration response spectra, Appendix
Bull. Seism. Soc. Am. 93, 314–331.
Consortium of Organizations for Strong-Motion Observation Systems ␣A Acceleration amplification factors (Fig. 12)
(COSMOS) (1999–2005). COSMOS virtual data center, http://db. ␣D Displacement amplification factors (Fig. 12)
cosmos-eq.org (last accessed 6 January 2004). ␣V Velocity amplification factors (Fig. 12)
Cornell, A. C. (1968). Engineering seismic risk analysis, Bull. Seism. Soc. f Percent of critical damping
Am. 58, no. 5, 1583–1606.
PGA Peak ground acceleration
California Strong Motion Instrumentation Program (CSMIP) (1994).
CSMIP strong-motion records from the Northridge, California earth- PGD Peak ground displacement
quake of January 17, 1994, California Geological Survey, Sacra- PGV Peak ground velocity
mento, California. PGV Normalized peak ground velocity (equation 3)
Federal Emergency Management Agency (FEMA) (2000). Prestandard and SA Spectral acceleration
commentary for the seismic rehabilitation of buildings, FEMA-356,
SAV Spectral acceleration of vertical motion
Washington, D.C.
Frankel, A., C. Mueller, B. Theodore, D. Perkins, E. V. Leyendecker, N. SD Spectra deformation
Dickman, S. Hanson, and M. Hopper (1996). National seismic hazard SDS Short (0.2 sec) period spectral acceleration of hor-
maps, June 1996 documentation, U.S. Geological Survey, Denver, izontal design ground motion
Colorado. V
SDS Short-period spectral acceleration of vertical de-
Hall, W. J., B. Mohraz, and N. M. Newmark (1975). Statistical studies of
sign ground motion (equation 18)
vertical and horizontal earthquake spectra, Nathan M. Newmark Con- V
sulting Engineering Services, Urbana, Illinois. SDS 1-sec period spectral acceleration of vertical de-
International Conference of Building Officials (ICBO) (1998). 1997 Uni- sign ground motion (equation 19)
form building code, International Conference of Building Officials, SD1 1-sec period spectral acceleration of horizontal
Whittier, California, International Code Council (ICC), 2003, Inter- design ground motion
national Building Code (IBC) 2003, Building Officials and Code Ad-
SV Spectral velocity
ministrators International, Inc., Country Club Hills, Illinois; Interna-
tional Conference of Building Officials, Whittier, California; and SV Normalized spectral velocity (equation 4)
Southern Building Code Congress International, Inc., Birmingham, T Natural period of structure
Alabama. Tcg Central period (equation 2)
Lam, N., J. Wilson, A. Chandler, and G. Hutchinson (2000). Response TS Control period in design spectrum (equation 13)
spectrum modelling for rock sites in low and moderate seismicity
T0 Control period in design spectrum (equation 14)
regions combining velocity, displacement and acceleration predic-
tions, Earthquake Eng. Struct. Dyn. 29, no. 10, 1491–1525. T1, . . . , T6 Control periods in smooth response spectrum
Leyendecker, E. V., R. J. Hunt, A. D. Frankel, and K. S. Rukstales (2000). (Fig. 12)
Development of maximum considered earthquake ground motion
maps, Earthquake Spectra 16, no. 1, 557–578.
Malhotra, P. K. (2001). Response spectrum of incompatible acceleration, FM Global
velocity and displacement histories, Earthquake Eng. Struct. Dyn. 30, 1151 Boston-Providence Turnpike
Norwood, Massachusetts 02062-9102
no. 2, 279–286.
[email protected]
MathWorks (2004). One-way analysis of variance (ANOVA1), Statistics
Toolbox, MATLAB, Natick, Massachusetts.
McGuire, R. K. (2004). Seismic Hazard and Risk Analysis, Earthquake
Engineering Research Institute, Oakland, California. Manuscript received 31 March 2005.

View publication stats