NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

Module 5
LOCAL SITE EFFECTS AND DESIGN GROUND
MOTIONS
(Lectures 23 to 26)

Lecture 26

Topics
5.5 DEVELOPMENT OF GROUND MOTIONS TIME
HISTORIES
5.5.1 Modification of actual ground motions records
5.5.2 Time-domain generation
5.5.3 Frequency-domain generation
5.5.4 Green’s function Techniques
5.5.5 Limitations of artificial ground motions

5.5 DEVELOPMENT OF GROUND MOTIONS TIME

HISTORIES
On many occasion, ground motion parameters along do not adequately describe the
effects of ground shaking. For analysis of nonlinear problems such as the response of
inelastic structures or the permanent deformation of an unstable slope, time histories
of motion are required. Time histories can also be required in the development of
site-specific design ground motions. In these cases, time histories that match target
ground motion parameters such as peak acceleration, velocities, or spectral ordinates
are required.

In some cases, the local and regional geologic and tectonic conditions of the site of
interest may be so similar to those of sites were actual strong motions have
previously been measured that those strong motion records can be sued directly.
Usually, this is not the case, and artificial ground motions must be developed.
Artificial ground motions can be developed in a number of different ways. The main
challenges in their development are to ensure that they are consistent with the target
parameters and that they are realistic (i.e., that their characteristics are consistent
with those of actual earthquakes). This is not as easy as it might appear; many
motions that appear reasonable in the time domain may not when examined in the
frequency domain, and vice versa. Many reasonable-looking time histories of
acceleration produce, after integration, unreasonable time histories of velocity and/or
displacement. The quality of an artificial ground motion is very difficult to discern

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 1

NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

by eye.

The most commonly used methods for generation of artificial ground motions falls
into four main categories: (a) modification of actual ground motion records, (b)
generation of artificial motion in the time domain, (c) generation of artificial motions
in the frequency domain, and (d) generation of artificial motions using Green’s
function techniques.

5.5.1 Modification of actual ground motions records

Perhaps the simplest approach to the generation of artificial ground motions is the
modification of actual recorded ground motions. Maximum motion levels, such as
peak acceleration and peak velocity, have been used to rescale actual strong motion
records to higher or lower levels of shaking (figure 20). Krinitszky and Chang
(1979) recommended that the scaling factor (the ratio of the target amplitude to the
amplitude of the record being scaled) should be kept as close to 1 as possible and
always a between 0.25 and 4.0, and that analyses be conducted with several scaled
records. Vanmarcke (1979), noting that simple amplitude scaling fails to account for
differences in important characteristics such as frequency content and duration,
suggested that limits on the scaling factor should be related to the type of problem to
which the resulting motion is to be applied. For analysis of linear elastic structures,
the limits of Krinitszky and Chang (1979) were considered suitable, but for
liquefaction a scaling factor range of 0.5 to 2.0 was recommended.

This type of rescaling procedure requires careful selection of the actual motion that
is to be used. A desirable ground motion record will not only have a peak
acceleration or velocity close to the target value, but will have magnitude distance,
and local site characteristics that are similar to those of the target motion. Such a
record is most likely to have a similar frequency content and duration to the target
motion. Computer programs (e.g., Dussome et al., 1991; Ferritto, 1992) that contain,
or at least interact with strong motion data-bases are available to aid in the selection
of actual ground motions for rescaling.

Rescaling of the time scale has been used to modify the frequency content of an
actual ground motion record. This is usually accomplished by multiplying the time
step of a digitized actual record by the ratio of the predominant period of the target
motion to the predominant period of the actual motion (figure 5.23). Since this
approach changes the frequency content over the entire spectrum as well as the
duration of the resealed record, it should be used carefully to avoid unintended
consequences.

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 2

NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

Figure 5.23 (a) Original acceleragroam from actual earthquake; (b) rescaled
version of original acceleragraom in which time scale was scaled upward by a
factor of 1.3 to match target predominant period. Note that the duration has also
been increased by a factor of 1.3

To generate artificial ground motions of long duration without significantly

changing the frequency content, some (e.g., Seed and Idriss, 1969) have spliced
parts of actual ground motion records together. Procedures of this type must also be
used with caution. Careful examination of the reasonableness of spliced motions in
both the time and frequency domains is advised.

5.5.2 Time-domain generation

The resemblance of ground motion time histories to transient stochastic processes
was noted years ago (Housner, 1947). Since then, a number of procedures that treat
ground motions as stochastic processes have been developed. Many of these operate
entirely in the time domain.

A stationary stochastic process is one whose statistics remain constant with time. a
stationary accelerogram, for example, would have a constant mean acceleration,
constant standard deviation of acceleration, and a constant frequency content-the
accelerations would continue indefinitely. The fact that the acceleration amplitude of
actual ground motion varies with time (ground motions have a beginning and an end,
after all) renders their amplitudes nonstationary. Studies have also shown that the
frequency content of a typical ground motion is also nonstationary-it changes over
the duration of shaking.

Generation of an artificial ground motion time history in the time domain typically
involves multiplying a stationary,, filtered white noise (or filtered Poisson process)
signal by an envelope function that describes the buildup and subsequent decay
(nonstationarity) of ground motion amplitude (Shinozuka and Deodatis, 1988), as
illustrated in (figure 5.24). More recently models that consider the nonstationary of
both amplitude and frequency content (e.g., Sharma and Shah, 1986; Shinozuka and
Deodatis, 1988) have been developed. The use of autoregressive moving average
(ARMA) models (e.g., Chang et al., 1982; Kozin, 1988) has also increased in recent
years.

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 3

NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

Figure 5.24 Example of time-domain generation of synthetic time history; (a)

time history of white noise is filtered in the time domain to produce (b) time
history of filtered white noise. Filtered white noise is multiplied by envelope
function in (c) to produce the artificial ground motion
5.5.3 Frequency-domain generation
Ground motions can be generated quite conveniently in the frequency domain by
combining a Fourier amplitude spectrum with a Fourier phase spectrum. The
amplitude spectrum may be computed from an actual ground motion spectrum or
may be represented by some theoretical means, such as a Brune spectrum or a power
spectral density function. The phase spectrum may be obtained from an actual
ground motion or may be computed from a time history given by the product of
white noise and an envelope function (figure 5.25). Some investigators (e.g.,
Ohsaki, 1979, Abrahamson and Singh, 1986) have used phase difference
distributions as an indicator of phase structure to develop nonrandom, artificial phase
spectra.

Figure 5.25 Example of frequency-domain generation of synthetic time history; (a)

time history of white noise is shaped by envelope function to produce (b) time
history of envelope white noise. Fourier transform of enveloped white noise is
performed to obtain (c) phase spectrum is combined with (d) amplitude spectrum to
produce (e) synthetic time history

Frequency-domain methods are particularly useful for generating motions that are
consistent with target response spectra. Computer programs such as EQGEN (Chang
et al., 1986b) and RASCAL (Silva, 1987) assume initial Fourier amplitude and phase

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 4

NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

spectra, and then iteratively adjust the ordinates of the Fourier amplitude spectrum
until a motion consistent with the target response spectrum is produced. The origin
of the target response spectrum must be kept in mind when generating spectrum-
compatible motions. Constant risk spectra, for example, represent the aggregate
effect of potential earthquakes of many different magnitudes occurring at many
different distances. Because a constant risk spectrum does not correspond to any
particular seismic event, a motion generated from a constant risk target spectrum
should not be expected to correspond to a particular seismic event (Naiem and Lew,
1955).

5.5.4 Green’s function Techniques

The Green’s function approach to ground motion modeling is based on the idea that
the total motion at a particular site is equal to the sum of the motions produced by a
series of individual ruptures of many small patches on the causative fault. Obtaining
the site motion requires defining the geometry of the earthquake source, dividing the
source into a finite number of patches, defining the sequence in which the patches
rupture, defining the slip functions (functions describing the variation of slip
displacement with time for each patch) across the source, and defining Green’s
functions (functions that describe the motion at the site due to an instantaneous unit
slip at the source; see (figure 5.26) across the source. Combining the Green’s
function with the slip function gives the motion at the site due to slip of each
individual patch. Summing the effects of the slips of each patch while accounting for
the order in which they rupture produces the overall ground motion t the site.
Obviously, the summation procedure assumes that all materials remain linear.

Figure 5.26 Schematic of Green’s functions for a fault divided into N patches.
Differences in the Green’s functions for the different patches are due to differences
in focal depth, epicentral distance, and geologic, structure along the source-the-site
path.

Calculation of Green’s functions requires knowledge of the velocity structure of the

crustal material between the source and site. However, estimation of the velocity

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 5

NPTEL- GEOTECHNICAL EARTHQUAKE ENGINEERING

structure, particularly with respect to heterogeneities responsible for the scattering

that produces latearriving coda waves, is a very difficult problem. Considerable
computational effort is also required to calculate Green’s function finite-element,
finite-difference, ad ray theory techniques are usually used for this purpose. Hatzell
(1978) bypassed these computations by using the weak motions of small earthquakes
as empirical Green’s functions to simulate the strong motion of large earthquakes.
Empirical Green’s functions have the benefit of automatically retaining the effects of
the crustal velocity structure.

The Green’s function approach is particularly useful for generating near-field

motions that is, motion at sites close enough to the fault that the fault dimensions
become significant (for far-field sites, the fault can be treated as a point source
without undue loss of accuracy). The nature of the rupture pattern including the
general direction in which ruptures progresses, and the site azimuth (relative to the
fault) can strongly influence ground motion in the near field. The Green’s function
approach allows phenomena such as directivity and fling to be reflected in artificial
motions.

5.5.5 Limitations of artificial ground motions

As discussed in Module 3, actual ground motions are complicated-they are
influenced by, and consequently reflect, characteristics of the seismic source, the
rupture process, the source-site travel path, and local site conditions. Although it is
convenient to characterize them with a small number of parameters, it is important to
remember that such characterizations can never by complete.

Artificial motions that match a small number of target parameters are not unique;
many different motions can produce the same target parameters. If such a set of
motions are used to analyze problems for which damage correlates will to the target
parameters, the predicted damage is likely to be consistent. For example, a set of
different motions with the same peak acceleration will produce similar base shears in
a stiff, linear elastic structure founded on rock. The same set of motions, however,
might produce a broad range of base shears in a flexible and/or inelastic structure or
in a structure founded on soft soil. They could also produce significantly different
estimates of permanent slope movement or liquefaction potential. When using
artificial motions, the eventual use(s) of the motions must always be reconciled with
the criteria from which they were developed.

Dept. of Civil Engg. Indian Institute of Technology, Kanpur 6