Principal Investigator(s) and
Institution(s)
Apostolos Papageorgiou, University at Buffalo
Objective
The specific objectives of this task are to provide a comprehensive review and study of
the factors that influence long period ground motions and their spatial variability that
are important for long span bridges, provide guidelines for incorporating these factors
explicitly in the estimation of seismic design ground motions, and to complement the
deficient data base of ground motion recordings close to large earthquakes with synthetic
time histories.
These objectives will be accomplished by an exhaustive survey of available pertinent
empirical data (i.e. strong motion recordings near the source of large earthquakes and
motions recorded on deep sedimentary deposits) and with the implementation of effective,
accurate and efficient simulation techniques.
Approach
It is widely recognized that there are two effects that have strong influence on long
period motions: (1) near-fault effects (forward rupture directivity), and (2) basin
response effects. Long period ground motions (T w 1 sec) and their spatial variability may
be modeled deterministically, unlike high frequency motions for which this kind of
modeling would be impractical. High frequency ground motions, and their spatial
variability, are stochastic in nature. In order to obtain reliable estimates of the
spatial incoherence of ground notion at a particular site, it may be necessary to use
accelerograms of a small earthquake or explosion recorded on a small array at the site.
For the purposes of this task, although the spatial variability of both long period and
high frequency ground motions will be reviewed, the focus in simulating such motions will
be long periods, in view of the fact that this range of periods is expected to affect the
response of long span bridges more significantly. Modeling/simulation of high frequency
motions, and their spatial variability, will be part of a future task.
Near-fault effects (forward rupture directivity) - The strong motion records that had by
far the most decisive effect on the analysis, modeling, and prediction of strong motion
near a rupturing fault are the Station No. 2 record obtained from the 1966 Parkfield
earthquake and the Pacoima Dam record obtained from the 1971 San Fernando earthquake. Both
records exhibit strong displacement velocity pulses of relatively long periods (~2 to 3
seconds). With their seminal study of damage of the Olive View Hospital, Bertero and
others were the first engineers to recognize and analyze the destructive potential of such
pulses on flexible structures. Such pulses may be detrimental for high rise buildings,
base isolated structures and long span bridges. The presence of relatively long period,
high amplitude pulses on the displacement and velocity records at near-source sites was
confirmed by the 1979 Imperial Valley earthquake. More recently, the occurrence of well
recorded major events such as the 1994 Northridge and the 1995 Hyogyo-Ken Nanbu (Kobe)
earthquake in the immediate vicinity of large metropolitan centers, highlighted the
tremendous destructive potential of near-source strong motions, and provided the necessary
data basis to study the effects of such motions on buildings.
Near-source ground motions are not adequately represented in modern codes, because these
codes historically were developed based on the experience of recorded motions not
sufficiently close to the causative fault. At present, there is little empirical guidance
to select the shape, amplitude, and duration of these "killer pulses." Attempts
to modify the amplitudes of empirical response spectra to account for near-source effects,
although pointing in the right direction, are not sufficient to describe the nature of
impulsive near-source ground motion, simply because synthetic motions generated by a
spectral matching technique cannot build a rupture directivity pulse where none is present
to begin with. Furthermore, the data base of recordings close to large earthquakes is
deficient and therefore any extrapolation of existing regression equations into such
regions of the Magnitude-Distance space is not controlled by data. Finally, ground
displacement is also a quantity poorly determined from the available processed
accelerometer data recorded with analog instruments. Thus, there is a clear need to
complement empirical data by theoretical models.
The source "fling" (or "killer pulse") that we mentioned above is the
result of forward rupture directivity. Directivity, in turn, is the inevitable result of
rupture propagation velocities that are close to the shear-wave velocity of the
propagation medium.
The question that naturally arises is the following: are these near-source pulses a
universal phenomenon? The answer to this question is provided by the recording at station
Caleta de Campos obtained during the 1985 Michoacan, Mexico earthquake. This record
displays a simple ramp-function displacement which cannot be explained by dislocation
models but requires crack-like slip functions, in contrast to the Station No. 2 record
obtained from the 1966 Parkfield earthquake. Furthermore, the velocities recorded by the
Caleta de Campos station display a high frequency (~3 sec) component overriding on the
longer frequency (~10 sec) motion. The high frequency ripples were successfully
interpreted by the kinematics of the rupture front (i.e., by a series of changes of the
rupture velocity). Changes of the rupture velocity will result in emission of
high-frequency waves that have a very strong directivity.
Although "killer pulses" may be typical characteristics of near-source motions
of strike-slip faults, they are not a universal feature of near-source ground motions, as
has been demonstrated with the Caleta de Campos record of the 1985 Michioacan earthquake.
Near-source motions, being influenced primarily by source effects, may display a multitude
of characteristics reflecting the complexity of the source process.
Basin response effects - Sedimentary deposits in the form of basins, and in particular
deep basins, have a profound influence on long period ground motions. The strongest
effects occur when the earthquake occurs at the basin edge, outside the basin, because
entering body waves are trapped by the dipping edge of the basin. Such entrapment of
energy results in prolongation of the duration of long period ground motions and in
amplification of the amplitudes of velocities and displacements.
Basin effects at a given site strongly depend on the location of the earthquake source,
because such effects are controlled by the geometry of the basin between the site and the
earthquake source. Therefore simple modifications of existing attenuation relations may
not be possible and the only viable approach to the problem may be 2-D or 3-D simulations
of each specific case.
Based on the observations and facts described above, the following research will be
conducted under this task:
Subtask 1 - A comprehensive review and study of all the factors that influence long-period
ground motions and their spatial variability. This will result in a report which will
provide a comprehensive review of the state of knowledge (and state of practice) with
regard to the selection of long period ground motions for seismic design of long-span
bridges. One important aspect of such a report will be the thorough documentation of the
existing near-source recordings of moderate and large events. Although various collections
of such data have been published by various authors in the form of tables and lists of
recordings, no one has presented these data in a clear visual form, so that the position
of the recording station relative to the rupturing faults is clearly evident (such
information is currently scattered in the published literature).
Subtask 2 - Synthesis of the near-source wavefield for all those events for which the
slip-function has been determined by inversion of strong motion, teleseismic and/or
geodetic data (there are 18 such events for which the slip function has been inferred by
inversion, covering a very wide range of magnitudes). Stating this in layman's terms,
displacement and velocity time histories for a dense grid of points surrounding the
rupturing faults will be generated. Thus, the motion at sites where no recording
instrument existed during the event can be observed. These simulations will be performed
using the kinematic modeling approach of seismologists, which is based on the
elasto-dynamic representation theorem. Consistent with the inversion studies that provided
the slip-functions of the above events, the model of each earthquake event will consist of
a fault plane embedded in a layered half-space. The source wavefield will be expressed
using the Discrete Wavenumber Method because this is the most efficient method for ground
motion computations over an extended region.
A computer code that implements the above method has already been developed and
successfully used by the task Principle Investigator. The results of these simulations
will permit a more thorough investigation of the character of long-period near-source
ground motions. Also, with these simulations, it is anticipated that an improved
understanding of the spatial variability of long-period near-source ground motions
vis-à-vis the spatial distributions of slip on the fault plane, will be developed.
Hopefully, these results will form the basis for specific guidelines in selecting such
motions.
It should be mentioned at the outset that the motions synthesized by the simulations
described above will include frequencies up to 1Hz. Synthesis of broadband motions (i.e.,
motions containing both long-periods as well as high frequencies) will be part of a future
project task.
Specific simulations for basin models will not be conducted, even though codes for
simulating 2-D and 2.5-D basin response effects have been developed. Clearly, such effects
are very important for areas such as the Los Angeles basin region. However, a major effort
is under way by Southern California Earthquake Center (SCEC) researchers to study the site
amplification in the Los Angeles basin from 3-D modeling of ground motion, and thus
research on this could result in an unnecessary duplication of effort.
Anticipated Start Date and Duration
January 1, 1999 - 24 months
9/23/99 |