STRONG MOTION RECORD PROCESSING FOR THE PEER CENTER

BOB DARRAGH
WALT SILVA
NICK GREGOR
Pacific Engineering
311 Pomona Avenue
El Cerrito, California 94530
INTRODUCTION
The PEER Strong Motion Database provides access to important strong motion records that have
been reviewed and processed in a consistent manner. Acceleration, velocity and displacement time
histories, as well as, response spectra at multiple spectral damping levels are available from
recordings made during 175 earthquakes at over 3,500 sites, all in common format. An
accompanying catalog provides additional information, for example, earthquake magnitude, site
characterization, distance, filter parameters and peak values, among others.
Strong motion data processing has two major objectives to make the data useful for engineering
analysis: (1) correction for the response of the strong motion instrument itself, and (2) reduction of
random noise in the recorded signals. The PEER processing concentrates on extending both the
high- and low-frequency ranges of the useable signal in the records on an individual component
basis. More recent data, particularly from digital recorders, generally do not benefit from
additional processing and are entered into the database in a standard format after review.
The processing of the strong motion records in the PEER database (Figure 1) is in general different
than the processing done by the agency that collected the data. Although the processed records
may be different, the differences should be small within the frequency passband common to both
processing procedures. We present examples of PEER processing results at both high frequencies
and long periods.
PROCESSING EFFECTS AT HIGH FREQUENCIES AT HARD ROCK SITES
For hard-rock sites in tectonically active regions (e.g. western North America) and for rock sites in
stable areas (e.g. eastern North America), the generally higher frequency content present in the
motions places more stringent demands upon processing as well as instrumentation. In order to
illustrate the effects of low-pass antialias filters, sample rate, and instrumentation upon peak
acceleration, a close-in recording (Site #2), of the February 6, 1986 Painesville, Ohio aftershock
was examined in detail. More details are given in Silva [1].
This site was analyzed because the recordings of the horizontal components were made with a
GEOS digital event recorded with a sample rate of 400 samples/sec without antialias filters as
discussed in Borcherdt [2]. The sensors employed were 2-Hz geophones that produce voltage
proportional to ground velocity for frequencies exceeding approximately 2.5 Hz. The digital
velocity records were converted to acceleration in the frequency domain by removing the

displacement response of the geophone/amplifier system and doubly differentiating the resulting
displacement spectral density.
Table 1 shows the effects of decimation and filtering upon peak horizontal accelerations for eight
choices of these parameters. The effects of processing, that is, sampling rate and filtering show a
decrease in peak value with decreasing sample rate and lower filter corner frequency. The
processing parameters appropriate for an SMA-1 response at sample rates of 400 and 50
samples/sec are cases 7 and 8, respectively. Case 7, with a sample rate of 400 samples/sec,
approximates the peak accelerations (Volume I) recorded by an analog SMA-1 instrument and
shows that the instrument may reduce peak acceleration (Volume I) at very hard rock sites by
nearly a factor of 2. This is certainly an important consideration with significant implication for
instrument correction procedures. If the instrument parameters used in the correction procedure,
specifically the corner frequency and damping, are not very close to the actual values, the corrected
peak accelerations (Volume II) could be in error by a significant amount.
Table 1: Effects of Processing upon Peak Acceleration
Processing Parameter
Sample Rate
Butterworth Filter
Normalized arithmetic
Number
(Sec/Sample)
Corner Frequency
average peak
(Hz)
horizontal acceleration
1
0.0025
1.00
2
0.0050
0.94
3
0.0025
75.0
0.84
4
0.0050
75.0
0.81
5
0.0025
50.0
0.68
6
0.0050
50.0
0.65
7
0.0025
SMA-1*
0.56
8
0.0200
SMA-1*
0.18
* Convolved nominal SMA-1 response: corner frequency of 25 Hz, damping 0.6 critical
These results and others presented in Silva [1] indicate that for stations sited upon hard rock the
effects of processing, depending on the frequency range of interest and instrument type, could
result in an underestimate of response spectral ordinates at high frequencies (> 10 Hz), as well as in
peak acceleration. Since the proper analyses of response spectra and their shapes at hard rock sites
requires a bandwidth from several seconds to over 30 Hz (for M > 4.5) several strong motion data
sets have been reprocessed as described in Silva [1] and included in the PEER Strong Motion
Database. Figure 2 shows examples of the average horizontal 5% damped spectra computed from
the original data processing and PEER processing at four rock sites in western North America. In
general, the PEER processing has increased the response at these hard rock sites at high
frequencies.
BASELINE CORRECTION PROCEDURE FOR STATIC DISPLACEMENTS
The recent occurrence of several large magnitude earthquakes (i.e., magnitude > 7.0) has greatly
increased the number of near source (rupture distances less than 20 km) strong ground motion
recordings. Specifically, the 1999 Kocaeli Turkey (M 7.4), the 1999 Chi-Chi Taiwan (M 7.6), and
the 1999 Duzce Turkey (M 7.1) earthquakes have been well recorded on digital strong ground

motion instruments in the near source region of the fault rupture. For these recent earthquakes, the
near source recordings (i.e., distances less than 20 km) have been reprocessed using a procedure
developed to preserve static (permanent or tectonic) displacements.
The standard procedure used by PEER to process strong ground motion time histories includes
bandpass filtering of the recorded motions based on the frequency range with substantial signal to
noise ratios (Figure 1). The time histories in the PEER Strong Motion Database have been
analyzed and discussed in Abrahamson [3].
Due to recent advances in instrumentation (digital recording) the ability exists to recover
permanent or static displacements (i.e., tectonic displacements) recorded on strong ground motion
instruments located close to the rupturing fault. The standard data processing procedure applied to
the PEER Strong Motion Database cannot preserve static displacements due to the application of
the high pass filters. However, caution must be used when analyzing displacement time histories
with apparent static displacements since they may be caused by non-seismic or non-tectonic effects
(Boore [4, 5]) such as instrument malfunction as well as local tilts due to soil deformation.
In a recent study [6], to correct for instrument malfunctions that distort tectonic displacements, we
have used the methodology proposed by Grazier [7] to perform the baseline correction of the
recorded motions. For this procedure, a least-squares fit is performed on the original integrated
velocity time history using three different functional forms. The first functional form is a simple
linear trend in velocity. The second function is a bilinear function which is piecewise continuous
and the last function is a simple quadratic in velocity. The best fitting velocity function is then
differentiated to produce an acceleration trace, which is then removed from the recorded
acceleration time history.
The fitting process has been automated such that after the staring time point for the velocity fit has
been selected, all linear, bilinear, and quadratic segments are fitted and their standard errors ranked.
The best fit is then viewed for reasonableness. A suite of starting times is marched through
resulting in a fairly rapid and exhaustive evaluation of the most appropriate correction function for
each component. The corrected time histories are then low-pass filtered to remove potential high
frequency noise. The low-pass filters are causal Butterworth, using the same corner frequencies as
in the PEER processing (typically near 50 Hz).
An illustration of the static baseline correction procedure is given in Figure 3. The recorded
velocity time history is TCU068 site (north-south component) from the 1999 ChiChi earthquake
and is plotted in Figure 1a along with the preferred linear least squares fit to this time history
(dashed line). This station was located at a rupture distance of 1.1 km from the fault. The static
baseline corrected acceleration, velocity, and displacement time histories are shown in Figure 4.
For comparison, the acceleration, velocity, and displacement time histories for the north-south
component of motion for station TCU068 using the standard PEER database processing are shown
in Figure 5a. The north-south component was bandpass filtered between 0.02 seconds and 50.0 Hz
using both causal (Figure 5a, standard PEER database processing) and acausal (Figure 5b)
Butterworth filters for comparison. This bandpass filtering of the time histories does not allow for
the displacements to have a static offset (i.e., frequency=0 Hz) and the peak ground displacement
values are lower than for the static baseline corrected cases. Interestingly, a comparison of the

peak-to-peak value for the standard PEER processed time histories gives approximately the same
zero to peak displacement values as the static baseline corrected time histories. This suggests that
the standard processing, which does not preserve static fields, may result in similar dynamic loads
to structures.
As noted by Boore [4, 5] the difference in the acceleration response spectra between time histories
which have been processed using a standard approach compared to those using a static baseline
correction approach are relative small, for periods less than about 20 seconds. For comparison, the
acceleration response spectra (5% spectral damping) are shown for standard PEER processing,
acausal filtering and the static baseline correction for the north-south component of motion in
Figure 6. The acceleration response spectra are very similar for the three different methods of data
processing. This is not completely unexpected because of the large frequency range used in the
standard processing (i.e., 0.02 seconds to 50 Hz). Additional examples are given in Gregor [6].
PEAK GROUND MOTIONS FROM RECORDS WITH STATIC DISPLACEMENTS
The static baseline correction procedure was applied to all strong ground motion recordings with
rupture distances of less than 20 km from the 1999 Kocaeli Turkey, 1999 Chi-Chi Taiwan, and the
1999 Duzce Turkey earthquakes.
A comparison between the peak values for the dynamic and static datasets is shown in Figures 7 to
8. This comparison reflects those strong ground motion stations that were processed using both the
PEER standard and the static baseline correction procedures. As one would expect, the PGA
(figure 7) values are very similar between the two processing procedures (i.e. the data points
generally fall along the 1:1 line). The PGV (Figure 8) and especially the PGD [6] values show a
larger scatter with the static values being greater than the corresponding dynamic values, in
general.
REFERENCES
1. Silva WJ, Darragh RB. "Engineering characterization of earthquake strong ground motion
recorded at rock sites." Electric Power Research Institute, Palo Alto, California 1995; TR-102261.
2. Borcherdt RD. Preliminary report on aftershock sequence for earthquakes of January 31, 1986
near Painesville, Ohio. U.S. Geological Survey Open File Report 86-181.
3. Abrahamson NA, Silva WJ. Empirical Response Spectral Attenuation Relations for Shallow
Crustal Earthquakes. Seismological Research Letters 1997; 68(1): 94-127.
4. Boore DM. Effect of Baseline Corrections on Displacements Response Spectra for Several
Recordings of the 1999 Chi-Chi, Taiwan, Earthquake. Bulletin of the Seismological Society of
America 2001; 91(5): 1199-1211.
5. Boore DM. Effect of Baseline Corrections on Response Spectra for Two Recordings of the
1999 Chi-Chi, Taiwan, Earthquake. U.S.Geological Survey Open File Report 99-945.

6. Gregor N, Silva W, Darragh R. Development of attenuation relation for peak particle velocity
and displacement. A PEARL report to PG&E/CEC/Caltrans 2002.
7. Grazier VM. Determination of the true ground displacement by using strong motion records.
Izvestiya Academy of Sciences, USSR, Physics of the Solid Earth 1979; 15(12): 875-885.

Figure 1. PEER processing flowchart

Figure 2. Plots of average 5%-damped spectral shapes computed from the original (dotted
lines) and PEER (solid lines) Volume II processed horizontal acceleration time histories for
selected data recorded in WNA tectonic environments. The PEER reprocessing was performed
on the original Volume I data.

Figure 3. Time history plot showing the observed velocity record (north-south component) for the
TCU068 station from the 1999 Chi-Chi Earthquake and the least square fit using a linear function
(dashed line).

Figure 4. Baseline corrected acceleration, velocity, and displacement time histories for the northsouth component for the TCU068 station.

Figure 5a. Acceleration, velocity, and displacement time histories for the TCU068 station (north-south
component) based on the standard PEER strong ground motion database processing procedure using causal
Butterworth filters.

Figure 5b. Acceleration, velocity, and displacement time histories for the TCU068 station (north-south
component) using acausal Butterworth filters in the processing.

Figure 6. Comparison between the acceleration response spectra (5% spectral damping) for the
static baseline corrected time history (solid line) and the PEER database time history for the northsouth component at the TCU068 station using causal (dotted line) and acausal (dashed line)
Butterworth filters. Note that there is little difference between the static baseline corrected (solid
line) and the acausal spectra (dashed line).

Figure 7. Comparison between the PGA values for the static baseline corrected time histories and the
PEER database time histories.

Figure 8. Comparison between the PGV values for the static baseline corrected time histories
and the PEER database time histories.