Close
About
FAQ
Home
Collections
Login
USC Login
Register
0
Selected
Invert selection
Deselect all
Deselect all
Click here to refresh results
Click here to refresh results
USC
/
Digital Library
/
University of Southern California Dissertations and Theses
/
Observations and modeling of dynamically triggered high frequency burst events
(USC Thesis Other)
Observations and modeling of dynamically triggered high frequency burst events
PDF
Download
Share
Open document
Flip pages
Contact Us
Contact Us
Copy asset link
Request this asset
Transcript (if available)
Content
OBSERVATIONS AND MODELING OF DYNAMICALLY TRIGGERED HIGH
FREQUENCY BURST EVENTS
by
Adam David Fischer
A Dissertation Presented to the
FACULTY OF THE GRADUATE SCHOOL
UNIVERSTIY OF SOUTHERN CALIFORNIA
In Partial Fulfillment of the
Requirements for the Degree
DOCTOR OF PHILOSOPHY
(EARTH SCIENCES)
DECEMBER 2008
Copyright 2008 Adam David Fischer
ii
Acknowledgments
I would like to thank the members of my dissertation committee, Charlie
Sammis, Thorsten Becker, Yehuda Ben-Zion, Steve Nutt, and Leon Teng, for their
valuable suggestions and encouragement without which this dissertation would not be
possible.
Chapter 1 - I would like to thank the Taiwan Central Weather Bureau for
providing the strong-motion seismograms of the 1999 Chi-Chi, Taiwan, earthquake. I
would also like to thank Margaret Boettcher for providing data from the South Africa
mine, Joe Fletcher and Paul Spudich, and William Lee for helpful comments and
discussions. I thank Debi Kilb and one anonymous reviewer for helpful suggestions
and comments. This research was supported by the Southern California Earthquake
Center. SCEC is funded by NSF Cooperative Agreement EAR-0106924 and USGS
Cooperative Agreement 02HQAG0008. The SCEC contribution number for this paper
is 1129.
Chapter 2 - I thank Joe Fletcher, Paul Spudich, and Larry Baker for making the
waveforms of the UPSAR available. I thank Dominic Assimaki for pointing out the
work of Bonilla et al. [2005]. This work has benefitted from helpful reviews by John
Vidale, Paul Spudich, and Norm Sleep. This research was supported by the Southern
California Earthquake Center. SCEC is funded by NSF Cooperative Agreement EAR-
0106924 and USGS Cooperative Agreement 02HQAG0008. The SCEC contribution
number for this paper is 1201.
iii
Chapter 3 – We wish to thank Eric Dunham for making available the Spectral
Boundary Integral numerical code used in this study. We would also like to thank
Harsha Bhat for his patient advice. This research was supported by the Southern
California Earthquake Center. SCEC is funded by NSF Cooperative Agreement EAR-
0106924 and USGS Cooperative Agreement 02HQAG0008. The SCEC contribution
number for this paper is 1224.
iv
Table of contents
Acknowledgments
ii
List of figures
v
List of Tables
vii
Abstract
viii
Introduction
1
Chapter 1: Dynamic Triggering by Strong Motion P- and S-waves: Evidence
from the 1999 Chi-Chi, Taiwan Earthquake
3
Summary 3
1-1: Introduction 4
1-2: Analysis 10
1-3: Discusstion 34
1-4: Chapter 1 References
39
Chapter 2: Dynamic triggering of high-frequency bursts by strong motions
during the 2004 Parkfield earthquake sequence
42
Summary 42
2-1: Introduction 42
2-2: Analysis 46
2-3: Discussion 54
2-4: Supplementary Material 56
2-5: Chapter 2 References 62
Chapter 3: Body Wave Triggering of Small Shallow Events During Strong
Motion
65
Summary 65
3-1: Introduction 65
3-2: Analysis 70
3-3: Discussion 95
3-4: Chapter 3 References
98
Summary of Results
101
References 105
v
List of Figures
Figure 1-1: High frequency bursts in Taiwan 5
Figure 1-2: Filtered step function 13
Figure 1-3: Burst amplitude distribution 15
Figure 1-4: Colocated station bursts 17
Figure 1-5: Burst identification methods 19
Figure 1-6: Filtered noise 20
Figure 1-7: Burst amplitude vs. distance 22
Figure 1-8: Filtered mine earthquakes 24
Figure 1-9: Subarray bursts 26
Figure 1-10: Burst magnitudes 29
Figure 1-11: Triggering threshold 33
Figure 1-12: Triggering amplitude vs. period 35
Figure 2-1: Station map 45
Figure 2-2: UPSAR bursts 48
Figure 2-3: Station P11 bursts 50
Figure 2-4: Triggering threshold 53
Figure 2-5: Ringing correlation 58
Figure 2-6: Work Ranch bursts 61
Figure 3-1: Taiwan and Parkfield bursts 67
Figure 3-2: Maximum depth of tension 73
Figure 3-3: Strain geometry 76
vi
Figure 3-4: Station map 78
Figure 3-5: Three station stress 80
Figure 3-6: Normal, Shear, and Coulomb stress
82
Figure 3-7: Maximum Coulomb Stress with Burst Rate 84
Figure 3-7: Numerical model results 87
Figure 3-8: Event duration result
89
Figure 3-9: Burst magnitudes 93
Figure S-1: Triggering Thresholds 103
vii
List of Tables
Table 1: Triggering Stresses and Periods for Taiwan and Surface Waves 9
Table 2: Summary of Observations for Taiwan and Parkfield 104
viii
Abstract
A series of high-frequency (>20Hz) bursts of energy are observed on strong
motion records during the 1999 Chi-Chi, Taiwan Earthquake Mw7.6. We hypothesized
that these bursts originated near the individual stations as small, shallow events that
were dynamically triggered by the P- and S-waves generated by the Chi-Chi mainshock.
These bursts were originally interpreted as a mainshock source signal by Chen et al.,
[2006] but our observations of events on strong motion records recorded at stations up
to 170 km from the mainshock epicenter is consistent with the local triggering
hypothesis. If the bursts originated on the Chi-Chi fault plane, as hypothesized by Chen
et al. [2006] based on their analysis of recordings within 20Km from the Chelungpu
fault, then they should not be observable at this distance assuming any reasonable value
of crustal attenuation. The bursts on all strong motion stations in the Taiwan Central
Weather Bureau network (TWCB) were identified using a numerical algorithm
approach. This data set was analyzed in the context of local dynamic triggering which
resulted in a stress threshold for triggering in the range 0.03 to 0.05 MPa for S-wave
triggering and 0.0013 to 0.0033 MPa for P-wave triggering, consistent with prior
observations of surface wave triggering.
In an attempt to better characterize the nature of high frequency bursts, similar
analysis of strong motion records was performed on the records of the 2004 Parkfield,
CA earthquake (Mw6) at the USGS UPSAR array. The average array spacing was
relatively small compared to the instruments in Taiwan so that further constraint of the
location of bursts was possible. Bursts were found to be incoherent even for stations
ix
spaced 40m apart, suggesting that they occur in a region approximately 20m from the
stations. The triggering threshold was found to be ~0.02Mpa, consistent with the
observations from Taiwan.
To test the possibility of nucleating unstable slip events in the very shallow crust
we modeled what is now referred to as “dynamic driving” of high frequency burst
events through numerical simulation. Rate-and-state friction presented a paradox for
the nucleation of such shallow events but we have determined that tensile stresses due
to mode conversion at the free surface of the earth allow for nucleation of unstable slip
that can produce radiation recorded as high-frequency bursts.
1
Introduction
This thesis consists of observational analysis of high-frequency bursts recorded
during strong motion from the 1999 Chi-Chi, Taiwan (Mw7.6) and 2004 Parkfield, CA
(Mw6) earthquakes followed by quantitative analytical and numerical modeling which
further characterizes the nature of the bursts. Chen et al., [2006] originally proposed
that the source of the high-frequency bursts was due to slip on asperities located on the
Chelungpu fault, which broke during the Chi-Chi earthquake. This thesis proposes that
the source of the bursts is local and within 20m of the seismometers. Many of the
observational methods employed in this thesis are new or have been modified from
existing techniques and as such contribute to the field of observational seismology.
This thesis is divided into three chapters, each consisting of a separately
published paper. Chapter 1 details observations of high-frequency bursts from the 1999
Chi-Chi, Taiwan earthquake. Burst properties as a function of epicentral distance are
determined as well as their locations and frequency-magnitude characteristics. Some
care is taken to rule out the possibility that the bursts are instrumentally generated, and
are thus not a local seismic effect but are an artificial noise signal generated by some
imperfection of the recording device. A triggering threshold is determined and
compared with more commonly observed surface wave triggering. Surprisingly, the
triggering threshold for high-frequency bursts is found to occur at similar stresses
compared to surface wave triggering.
Chapter 2 expands the base of observations by analyzing seismic data from the
2004 Parkfield, CA earthquake. This work greatly improves and expands on the
2
analysis of Chapter 1. The principle improvement comes from the use of data from a
dense seismic array. Strong motion records from 12 USGS Parkfield Dense
Seismograph Array (UPSAR) with an average stations spacing of a few hundred meters
and a minimum station spacing of only 40 meters allows for a well defined constraint
on the size of the region in which high-frequency bursts are produced. Again, great
care is taken to rule out instrumental artifacts as the source of the bursts. The
magnitude of the largest burst is calculated. Finally, a triggering threshold is found to
be consistent with the observations from Taiwan.
Chapter 3 completes this work by developing a quantitative model of the burst
process. The possibility of nucleating unstable slip at very shallow depths is considered
in the framework of rate-and-state friction. Rate-and-state predicts that triggering by
period stresses should occur preferentially shallow but at the same time predicts that the
critical patch size for nucleation becomes unphysically large at depths shown to produce
bursts. This paradox is resolved by proposing and developing a model for nucleation by
simultaneous shear loading and normal stress unloading on pre-existing planes of
weakeness at depths around 10m. This is possible due to the mode conversion of SV to
P waves upon the free surface of the earth, which results in tensile stresses that can
occur at depths up to 70m for the strong motion amplitudes observed at Parkfield.
Finally we consider the energy conversion and effective attenuation produced by the
burst process.
3
Chapter 1: Dynamic Triggering by Strong Motion P- and S-waves:
Evidence from the 1999 Chi-Chi, Taiwan Earthquake
Summary
High-frequency band-pass filtering of acceleration records from the 1999 Chi-
Chi, Taiwan earthquake (Mw = 7.6) resolves the continuous signal into a series of
relatively short duration, discrete energy bursts. We hypothesize that these bursts
originate near the individual stations as small, shallow events that have been
dynamically triggered by the P- and S-waves generated by the Chi-Chi mainshock,
however, we cannot rule out the possibility that they originate from some non-seismic
source associated with the surface, such as buildings or trees. Bursts are observed only
during the seismic signal and not in the pre- or post-signal noise. The bursts are not
likely to be generated by the instruments since groupings of three co-located individual
instruments record identical bursts. Also, we show that the bursts are not due to the
band-pass filtering of instrumentally generated step functions. The hypothesis that they
are local events is supported by the observation of bursts in the 40 Hz frequency band at
distances up to 170 km from the epicenter. If the bursts originated on the Chi-Chi fault
plane, as hypothesized by Chen et al. [2006] based on their analysis of recordings
within 20Km from the Chelungpu fault, then they should not be observable at this
distance assuming any reasonable value of crustal attenuation. Assuming a local origin,
we estimate an average local event magnitude of Mw=0.2 and source-receiver distance
approximately 1km. We extended our analysis to lower stress levels by analyzing
records from a smaller (Mw=5.3) event that was recorded by many of the same
4
instruments used in the Chi-Chi analysis. For this event, bursts are observed only on
the accelerograms from stations relatively close to the mainshock hypocenter. Analysis
of the combined data set from both mainshocks suggests a stress threshold for triggering
in the range 0.03 to 0.05 MPa for S-wave triggering and 0.0013 to 0.0033 MPa for P-
wave triggering, consistent with prior observations of surface wave triggering.
1-1. Introduction
Aftershocks from large earthquakes have generally been assumed to be driven
by the increases in static stress associated with fault displacement (eg: King et al.,
1994). In this view, aftershocks are limited to areas of increased static stress while areas
of reduced static stress (stress shadows) show a reduction in seismicity (Harris and
Simpson, 1996). However, recent studies by Felzer and Brodsky (2005) and Ma et al.
(2005) have shown that static stress changes alone fail to account for the locations of all
aftershocks. In particular, aftershocks located in stress shadows appear to be
dynamically triggered by the passage of large amplitude seismic waves. There are also
many observations of earthquakes at distances greater than a few fault lengths from a
mainshock that appear to have been triggered by surface waves (See Freed, 2005 and
references therein). In this study, we investigate high-frequency bursts observed on
strong motion records from the 1999 Chi-Chi, Taiwan earthquake (Mw = 7.6) as
evidence for dynamic triggering by strong motion P- and S-waves.
5
−1000
−500
0
500
1000
Original
0 5 10 15 20 25 30 35 40
−10
−5
0
5
10
Band: 40-50Hz
Time (sec) after 1999/09/20,17:47:15.85
−10
−5
0
5
10
Band: 30-40 Hz
−10
−5
0
5
10
Band: 20-30 Hz
Acceleration (cm/s/s)
− 100
−50
0
50
100
Band: 10-20Hz
Figure 1-1: Example of bursts resolved by high frequency band pass filtering method. The top trace is
the original east-west component accelerogram of the Chi-Chi mainshock recorded at station TCU084.
The next four traces are the 10-20 Hz, 20-30 Hz, 30-40 Hz, and 40-50 Hz pass bands, respectively. The
continuous waveform is resolved into a series of discrete pulses, or bursts, of energy.
6
High-frequency band-pass filtering (10-20Hz, 20-30Hz, 30-40Hz, and 40-50Hz)
of acceleration records from the 1999 Chi-Chi, Taiwan earthquake (Mw = 7.6) resolves
the continuous signal into a series of relatively short duration discrete energy bursts as
illustrated in Fig. 1-1. Chen et al. (2006) assumed that the sources of these bursts were
small slip events on a strand of the Chelungpu fault and they were able to find locations
using a brute force trial-and-error algorithm. Their algorithm calculated the expected
arrival times at the 49 nearest stations (all within 20 km of the fault plane) from each 1
km square patch of the fault plane, where each patch generated an event every 30 μsec.
They located a total of 540 events, each of which satisfied the criteria that a burst
appears on at least 4 of the closest stations and within 0.2 sec of the predicted time. The
locations found in this way followed a sensible pattern in time and space on the fault
plane. The first bursts were located at shallow depths directly up-dip from the
hypocenter. Their origin times were consistent with their having been triggered by the P
wave. Later bursts spread out over the fault plane in a pattern consistent with a rupture
front traveling at about 2.3 km/sec, similar to the estimated rupture velocity of the Chi-
Chi earthquake (Ma et al., 2001). The relative magnitudes of the bursts were consistent
with a Gutenberg-Richter frequency magnitude distribution and the later bursts at each
distance were consistent with Omori’s law for aftershocks if the origin time at each
distance was taken as the arrival time of the rupture front.
In this work, we propose an alternative hypothesis that the high-frequency bursts
originate from very small slip events located very near the recording stations that are
dynamically triggered by the passage of large amplitude P and S waves generated by the
7
Chi-Chi mainshock. This new interpretation is motivated by observations of high-
frequency bursts at stations as far as 170 km from the epicenter, much too far for them
to have originated at small events on the Chelungpu fault plane. The proposed body
wave triggering yields a stress threshold similar to estimates from prior observations of
surface wave and tidal triggering.
Triggering of earthquakes by dynamic stress fluctuations during the
passage of surface waves is well documented [Hill et al. (1993), Brodsky et al. (2000),
Kilb et al. (2000), Gomberg et al. (2001), Gomberg et al. (2004), Pankow et al. (2004),
Prejean et al. (2004), Miyazawa & Mori (2006)]. The first such observation was by Hill
et al. (1993) who documented surface wave triggering as far as 1250 km from the
Landers epicenter, mainly in regions of high background seismicity and geothermal
activity. At such large distances, the static stress changes caused by slip on the Landers
fault were extremely small, falling below daily tidal stress fluctuations at distances
greater than about 250 km. Hill et al. (1993) used two methods to identify triggered
events: 1) a statistical increase of background seismicity and 2) high-frequency band
pass filtering that revealed high-frequency local events in lower frequency coda from
the distant mainshock. The dynamic stresses responsible for the local triggered events
were estimated from the mainshock waveform to be 0.6-0.82 MPa.
Numerous more recent observations suggest that dynamic triggering by surface
waves may be a general phenomenon associated with large earthquakes. Evidence for
triggering in both geothermal and non-geothermal areas was presented by Brodsky et al.
(2000) who analyzed surface waves from the 1999 Izmit, Turkey earthquake (Mw=7.4)
8
recorded on a seismic network in Greece. Band-pass filtered records of the surface
waves revealed local events that began immediately after the passage of the surface
waves. These observations were supported by significant increases in catalog seismicity
for a few hours following the surface waves.
Gomberg et al. (2001) re-evaluated the Landers and Hector mine events, taking
advantage of the relative similarity between the two events to demonstrate the role of
rupture directivity in remote triggering. They proposed two possible physical
mechanisms for failure by dynamic strains: weakening of a fault by changes in pore
pressure, or by cyclic fatigue. Both mechanisms predict a triggering threshold. More
recent studies have explored triggering by the 2002 Denali, Alaska earthquake
(Gomberg et al, 2004; Prejean et al., 2004; Pankow et al., 2004). Surface waves from
this event triggered seismicity across large areas of British Columbia and the western
United States, much of which occurred outside of regions with high background
seismicity or geothermal activity. Stress thresholds determined in the Denali studies are
consistent with those reported for Landers, Hector Mine, and Izmit earthquakes as
summarized in Table 1.
9
Study Trigger Mode
Trigger
Period (s)
Triggering
Stress (Mpa)
This Study P- and S- Wave 2 0.03 - 0.3
Brodsky et al. 2000 Surface Wave 20 0.18
Hill et al. 1993 Surface Wave 20 0.6
**
Gomberg et al. 2001 Surface Wave 50 0.05-0.36
Gomberg et al. 2004 Surface Wave 20 0.1-1.0*
Pankow et al. 2004 Surface Wave 15 0.12-0.35
Prejean et al. 2004 Surface Wave 30 0.01-0.09
Pankow et al. 2004 Surface Wave 15 0.12-0.35
Miyazawa & Mori 2006 (dilational) Surface Wave 25 0.0033-0.0045
Wilcock 2001 Tides 43,200 0.009-0.018*
Cochran et al. 2004 Tides 43,200 0.005-0.02
* - Estimated from figures
** - Reported in Brodsky et al. 2000
Table 1: – Observations of dynamic triggering. In cases where triggering stress was not specifically
given, an estimate was made using figures in the publication.
10
There is also evidence that tidal stress fluctuations can trigger local seismicity.
Wilcock (2001) describes results from an array of ocean-bottom-seismometers deployed
on the Endeavour segment of the Juan De Fuca Ridge over a period of 55 days. He
found a significant increase in seismicity rates during low tides in an ensemble of 1899
microearthquakes. Cochran et al. (2004) found similar results using a global set of
earthquakes from the Harvard Centroid Moment Tensor catalog. They found the most
significant increases in seismicity near boundaries between ocean basin and continental
crust where the ocean loading combines with solid earth tides to produce large stresses.
Even though each of these studies was in a different tectonic setting (ridge transforms
vs. convergent boundary), both found a similar triggering threshold (see Table 1).
These studies indicate that remote events can be triggered by dynamic stress
fluctuations over a wide range of frequencies from surface waves to 12-hour tides.
Surprisingly, the stress threshold for triggering does not appear to be very sensitive to
the frequency as predicted by rate-and-state models (Dieterich 1987 and 1994). Both
surface waves and tides appear to trigger seismicity down to stresses on the order of
0.005 MPa. In this paper we find that the body-wave strong motions from the Chi-Chi
earthquake also trigger small shallow events at a similar level.
1-2. Analysis
1-2.1 Data set
We analyzed 198 East-West component high quality records of the Chi-Chi
mainshock and 151 recordings of a small thrust event on July 7
th
, 1995 (Mw = 5.3) at a
11
depth of 13 Km. We limited the analysis to the E-W component because S wave
excitation was strongest on this component and also because the same high-frequency
bursts were observed on all three components of any given instrument. Data sets for
both earthquakes were recorded by the same strong motion stations, which have been
operated by the Taiwan Central Weather Bureau since 1994. The waveforms were
filtered in four pass bands: 10-20Hz, 20-30Hz, 30-40Hz, and 40-50Hz using a fourth-
order Butterworth filter. Bursts were identified using a criterion based on a minimum
amplitude and a minimum temporal separation between adjacent events. The minimum
amplitude was set as a multiple (5x, 10x, 20x) of the noise level before the P arrival in
the pass band of interest. The minimal separation between distinct events was set at 0.2
seconds. We also required that an event appear at the same time in at least 2 pass-bands
on the same record. The event search for all stations began with the P-wave arrival and
continued for 40 seconds thereafter.
1-2.2 Bursts as Instrumental or Processing Artifacts
Before proceeding with the data analysis, we first investigated the possibility
that the bursts were due to instrumental noise and also checked that the background
noise did not contain coherent bursts. If the bursts were caused by instrumentally
generated delta functions or step functions, then band-pass filtering should generate
energy bursts in all frequency bands as illustrated in Fig. 1-2. In contrast, the observed
bursts are generally correlated across a limited range of frequency bands. In many cases
the largest observed burst at a given time does not occur in the lowest frequency band,
12
as predicted for a step function in Fig. 1-2. A delta function would produce uniform
amplitude bursts in all pass-bands, also contrary to the observations.
13
−0.5
0
0.5
1
1.5
−0.5
0
0.5
Time (seconds)
−0.5
0
0.5
−0.5
0
0.5
−0.5
0
0.5
10-20 Hz
20-30 Hz
30-40 Hz
40-50 Hz
A m p l i t u d e
24 25 26
Figure 1-2: Identical band pass filtering procedure as in Figure 1-1 for a unit step function. The filtering
results in a series of time correlated bursts with decreasing amplitude for increasing frequency. Bursts
produced by filtering the Chi-Chi accelerograms do not always appear in all bands (10-20 Hz, 20-30 Hz,
30-40 Hz, and 40-50 Hz), and often the largest amplitude burst is observed in a high frequency band.
14
If the bursts are generated by some sort of mechanical event in the seismometer
such as a component striking the case, we might expect them to be correlated with the
amplitude of shaking. We tested for this by identifying 73 bursts on the three
components of a single station (TCU084), each burst having a S/N ratio greater than 50.
At the time of each burst, we determined the acceleration of the unfiltered signal. The
distribution of acceleration values that correspond to burst arrivals and the distribution
of randomly sampled acceleration values are almost identical (Fig. 1-3).
15
−400 −300 −200 −100 0 100 200 300
0
0.5
1
bursts
random sample
Amplitude (cm s )
-2
Number (normalized)
Figure 1-3: Normalized distributions of amplitudes associated with observed bursts. Bursts from station
TCU084 on all the components required signal-to-noise ratio of at least 50. The distribution of
amplitudes of the coseismic shaking on all three components that correspond to each individual burst
(dashed line) and a distribution of randomly chosen coseismic amplitudes (solid line) are shown for
comparison. The amplitudes associated with the arrival of bursts is nearly identical to the random sample
of amplitudes, indicating that there is no correlation between shaking of the instrument and occurrence of
bursts.
16
We conclude that the occurrence of bursts is uncorrelated with the acceleration in the
mainshock strong motion, so they are probably not generated by mechanical noise in the
seismometers.
At the time of the Chi-Chi earthquake, the network included 22 “collocated”
stations that contained both Teledyne Geotech model A-800 and A-900 model
instruments. While the records from different instruments at the same stations differed
in amplitude, frequency content, and baseline offset (Wang et al, 2003), we found the
same high-frequency bursts on collocated stations as illustrated in Fig. 1-4. Since each
instrument had its own digitizer, this observation rules out the possibility that the bursts
are an artifact from the digitization process, and makes it less likely that they are
generated mechanically within the individual instruments.
17
−4
−2
0
2
4
−4
−2
0
2
4
20 25 30 35 40 45 50
−4
−2
0
2
4
Time (s)
Amplitude (cm s )
-2
Station HWA
E-W Component
20 - 30 Hz
Station HWA2
E-W Component
20 - 30 Hz
Station H019
E-W Component
20 - 30 Hz
Figure 1-4: 20 – 30 Hz pass band filtered records at three colocated stations approximately 80 km from
the epicenter of the 1999 Chi-Chi, Taiwan earthquake. HWA is a Teledyne Geotech model A-800 digital
accelerograph while HWA2 and H019 are A-900A digital accelerographs. A few of the coherent bursts
are emphasized by dashed lines. While the amplitudes and shapes of the bursts are not identical due to
the differences in the instruments, the arrival times of all the bursts are the same and there is a strong
general coherency.
18
We eliminated the possibility that the bursts are artifacts of the filtering process
by showing that they can be observed in the unprocessed signals. Figure 1-5 shows a
narrow time window centered on a large burst identified in the filtered records. The raw
data contains a continuous, periodic, small amplitude signal superimposed on the large
amplitude, longer period signal of the mainshock. After detrending to remove the low
frequency components, the unprocessed burst is seen to be consistent in amplitude and
frequency with the output of the band-pass filter.
In Fig. 1-6 bursts that occurred during the signal are compared with bursts that
occurred during the pre-event noise and those that result from band-pass filtering a
random (Gaussian) noise time series. Bursts that occur during the signal are clearly
different from the other two cases in that they have a much larger amplitude and are
correlated across some or all frequency bands.
19
5 10 15 20 25 30 35 40
−1000
−500
0
500
1000
20 22 24 26 28
−200
−100
0
100
200
23 23. 5 24 24.5
−200
−100
0
100
200
23.45 23. 5 23.55 23.6
−200
−100
0
100
200
20 22 24 26 28
−1000
−500
0
500
1000
23 23.5 24 24.5
−1000
−500
0
500
1000
23.45 23.5 23.55 23.6
−400
−200
0
200
400
Time (seconds)
A m p l i t u d e ( c m / s / s )
Figure 1-5: Identification of triggered events does not always require the high frequency bandpass
filtering technique. The top trace is the East-West component accelerogram for station C028 from the
Chi-Chi mainshock. The left column shows the arrival time of a large burst in the 20-30 Hz band in 3
consecutively shorter time windows. The right column shows the unprocessed accelerogram at the same
time scales. A few cycles of high frequency energy can be seen superposed on the mainshock waveform
in the unfiltered record on the right. The dot on the plots indicates where the burst is on the record.
20
−5
0
5
−1
0
1
−1
0
1
−2
0
2
−1
0
1
7.5
8
8.5
−0.1
0
0. 1
−0.1
0
0.1
−0.1
0
0.1
−0.1
0
0.1
0 1 2 3 4 5 6 7 8 9 10 0 1 2 3 4 5 6 7 8 9 10
Time (seconds)
A m p l i t u d e
(A) Preseismic Noise (B) Random Noise
Figure 1-6: Comparison of the pre-mainshock ground noise and computer generated white noise. A) Ten
seconds of ground noise from station TCU084 (E-W component). B) Ten seconds of white noise. Note
that in both cases the bursts are uncorrelated across the frequency bands. The seismic noise differs from
white noise only in that it has relatively less energy in the high frequency bands
21
1-2.3 Decay of Burst Amplitudes with Epicentral Distance
The strongest evidence supporting our hypothesis that the bursts do not originate
on the Chelungpu fault is the observation of bursts at stations up to 170 km from the
fault epicenter, with similar observations for the smaller (Mw 5.3) event in 1995. In
Fig. 1-7 the maximum burst amplitude observed at these distant stations is plotted as a
function of epicentral distance, and compared with the amplitude decay predicted by
geometric spreading and anelastic attenuation. While the epicentral distance gives only
an approximate measure of the actual distance to small events distributed on the large
Chelungpu fault plane, the errors associated with the path calculation do not weaken our
argument because the high-frequency bursts are expected to attenuate over distances
that are short in comparison to the epicentral distance. Assuming Q = 100 at a frequency
of 25 Hz and v
S
= 3.3 km/s, the amplitude of a 25 Hz burst should decay by a factor of
about 1000 over a distance of only 25 km. Bursts originating on the Chelungpu fault
plane should not be observable at distances beyond about 170km except for unphysical
large Q values approaching 2000. Since an average Q as low as 43 has been reported for
the top 200 km in Taiwan (Liu et al. 2005), it is very unlikely that the burst observed at
distant stations originated on the Chelungpu fault.
22
0 50 100 150 200
10
0
10
5
0 50 100 150 200
10
0
10
5
0 50 100 150 200
10
0
10
5
0 50 100 150 200
10
0
10
5
10−20Hz 20−30Hz
Q = 100
Q = 500
Q = 2000
30−40Hz 40−50Hz
Q = 100
Q = 500
Q = 2000
Q = 100
Q = 500
Q = 2000
Epicentral Distance (Km)
A m p l i t u d e ( c m / s / s )
Q=100
Q=500
Q=2000
Figure 1-7: Observations of maximum recorded burst amplitude as a function of epicentral distance for
Chi-Chi mainshock data set (open circles). Lines indicate the decrease in amplitude with distance R
predicted by geometrical spreading and inelastic attenuation calculated using (1/R)exp[-πfr/Qβ] for the
indicated values of the quality factor Q. In this expression f is the average frequency of the pass band, R
is the epicentral distance, and βis the shear wave velocity. If the source of the largest observed burst is on
the Chelungpu fault, observation at distant stations would effectively require Q >= 2000, a result
inconsistent with current knowledge of shallow crustal attenuation.
23
1-2.4 Burst Locations and Magnitudes
If the bursts do not originate as small events on the Chelungpu fault plane, what
are their sources? Our hypothesis is that they are very small events that occur very close
to the stations. We could not use the P- and S-wave travel time difference to estimate
distances because we were not able to confidently resolve P- and S-wave arrivals within
the 50 Hz upper frequency limit of the data. The identification of P- and S-wave onsets
was also complicated by the arrivals from a large number of triggered events in a short
time interval.
The difficulty in identifying P and S arrivals is illustrated in Figs. 1-8A and 1-
8B which show a magnitude 0 earthquake recorded at hypocenter distances of 1.45 km
and 1.38 km by two stations in a 2 km deep South African gold mine at a sampling rate
of 3000 samples/sec (Richardson and Jordan 2002). Note that frequency of the P and S
waves are close to 100 Hz. For comparison with the Taiwan bursts, we band-pass
filtered the South African data at 10-20 Hz, 20-30 Hz, 30-40 Hz, and 40-50 Hz. In
these pass bands, the durations of the bursts from this event are almost identical to those
we observe in Taiwan. At one of the stations, the earthquake appears as a sequence of
two distinct bursts associated with the P and S arrivals (Fig. 1-8A), while in on the other
station the P wave arrival is very weak (Fig. 1-8B) and cannot be identified. We
conclude that the dynamically triggered events in Taiwan may be similar in size and
hypocentral distance to the mine earthquakes, but P and S are poorly resolved due to the
50 Hz aliasing limit of the Taiwan strong motion data.
24
−5
0
5
x 10
−5
− 5
0
5
x 10
−8
− 1
0
1
x 10
−7
− 2
0
2
x 10
−7
− 1 − 0. 5 0 0. 5 1 1. 5 2
− 5
0
5
x 10
−7
−1
0
1
x 10
−5
− 2
0
2
x 10
−8
− 5
0
5
x 10
−8
− 1
0
1
x 10
−7
− 0. 2 0 0. 2 0. 4 0. 6 0. 8 1 1. 2 1. 4 1.6 1.8
− 2
0
2
x 10
−7
Time (seconds)
A m p l i t u d e ( m / s / s )
(A)
(B)
Raw
10-20Hz
20-30Hz
30-40Hz
40-50Hz
Raw
10-20Hz
20-30Hz
30-40Hz
40-50Hz
Figure 1-8: Examples of a small event recorded in a South Africa mine. This Mw = 0 event was
recorded by 7 seismometers located within the mine. A) The top trace is the raw waveform as recorded
by a station 1450 meters from the event hypocenter. The 4 traces below are 10-20Hz, 20-30Hz, 30-40Hz,
and 40-50Hz band pass records, respectively. This recording shows a clear separation of P- and S-wave
arrivals in the high frequency pass as a series of two distinct bursts.. B) A different station recording the
same event as A. The recording shows that in some cases the P-wave arrival is completely lost in the
filtering process.
25
Our attempts to locate the sources by identifying arrivals at several nearby
stations also failed. The problem is illustrated in Figure 1-9 which compares
seismograms at seven of the most tightly clustered instruments in the array. It is not
possible to identify the arrival time of any single high-frequency burst on all stations,
even though on average they are only about two kilometers apart. Several of the clear
bursts (stations B, C, D at ~ 42s) in this figure appear to contradict this statement, but
corresponding burst arrivals are missing on records from the other four stations in the
cluster. Another apparently coherent signal can be seen in the high frequency bands
from stations C and G at ~53 seconds. This must be two events with coincidental
arrival times because it does not appear on a station located directly between the two.
This result is not surprising since small-aperture arrays of surface seismometers show a
loss of coherency at a distance of a few hundred meters for frequencies above 10 Hz
(Vernon et al. 1991, Vernon et al 1998, Ebel 1989, Wilson and Pavlis 2000).
26
Figure 1-9: We isolate a subset of seven stations with a relatively high spatial density (map and map
inset). The differences in high frequency energy amongst the stations is illustrated by showing
waveforms (E-W component) from all seven stations. An identification Letter is used to mark stations on
the map correspond with letters marking the waveforms. The lowpass filtered waveforms, filtered at
1Hz, are extremely similar. The 10-20 Hz pass bands are markedly different from the lowpass
waveforms as well as from each other. Bursts in the higher frequency bands are not coherent at more than
a single station.
27
Without clear P- and S-wave arrivals or coherent arrivals at multiple stations we
cannot locate the source of the bursts. However, the lack of correlated arrivals on
nearby stations makes it possible to constrain the maximum magnitude and source-
receiver distance. We first compared the acceleration spectra of the observed bursts
with that expected from an idealized Brune (1970) source model, including attenuation,
in order to constrain the ratio r/Q in the exponential term of the attenuation function
A=A
0
exp(-πfr/Qβ) (1)
In this equation, A
0
is the source amplitude, f the average frequency, r the hypocentral
distance, Q the quality factor, and β the S-wave velocity. The theoretical source
displacement spectrum is flat below a corner frequency, f
c
, above which it decays as
1/ω
2
. We assume the data from 10-50 Hz is below the corner frequency based on
observed scaling laws for microearthquakes that estimate f
c
~ 100 Hz for magnitude 0
events (Abercrombie 1995 and Hiramatsu et al. 2002).
Six events were visually identified on seismograms from stations in the cluster
shown in Fig. 1-9. The acceleration in each 10 Hz pass-bands between 10 to 50 Hz was
converted to a displacement and plotted in Fig. 1-10a. These displacements were
corrected for attenuation using a range of r/Q from 0.01 to 50 as illustrated for one
event in Fig 1-10b. For each event, we found the value of r/Q that gave the flattest
spectrum by minimizing the standard deviation between the corrected spectrum and a
horizontal line as shown in Fig. 1-10c. The amplitude of the flattest spectrum was used
28
to calculate the moment magnitude M
o
using the relation (Richardson and Jordan,
2002):
M
0
=4πρβ
3
Ω
0
Ψ
−1
(2)
where ρ is crustal density, Ω
ο
is the long period spectral amplitude, and Ψ a correction
for radiation pattern (approximately 2 /5 for S-waves). Assuming ρ = 3000 kg/m
3
and
β=0.5 km/s, Fig. 1-10d shows the resulting moment magnitude values for the 6 events.
Note that M
o
ranges from –0.5 to 0.7 and increase linearly with the ratio r/Q. There
are two possibilities. Either Q is constant and the larger events are systematically more
distant from the station. Intuitively we expect larger events to be seen from larger
distances, and the lack of large events closer to the stations may be a result of the small
sample size. The other possibility is that these events have a corner frequency in the
range 10 to 50 Hz where the acceleration and that we are systematically over correcting
the larger events making them appear to be further away.
29
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
−7
−6
−5
−4
1.1 1.2 1.3 1.4 1.5 1.6 1.7 1.8
− 10
− 5
0
5
10
15
20
25
30
35
0 50 100 150 200 250 300 350
0
1
2
0 50 100 150 200 250
− 1
−0.5
0
0.5
1
1.5
2
2.5
Log (Frequency)
Log (Frequency)
r/Q
r/Q
Log (Amplitude) [meters]
Log (Amplitude) [meters]
Standard Deviation
Moment Magnitude
0.5 km/s
3 km/s
o - 0.5 km/s
x - 3 km/s
(A)
(A)
(A)
(A)
Figure 1-10: (A) Uncorrected displacement spectra for all six events identified in figure 1-9. (B) A
displacement spectra for one of the events corrected for a range of r/Q values to find the correction that
gives the flattest spectrum. (C) The standard deviation of the displacement spectra of all nine events from
a horizontal flat line measured as a function of r/Q. (D) Magnitude variation with r/Q.
30
Having constrained the ratio r/Q, we can now use the fact that none of these 6
events is seen on more than one station in the cluster to further constrain the event-
station distance. To quantify this constraint, consider station H013, which is roughly in
the center of the cluster, and designate it station 1. Let A
1
be the amplitude of the largest
burst recorded by station 1. Let station 2 be any one of the surrounding stations in the
cluster a distance D from station 1, and let N
2
be the amplitude of the noise at station 2.
The amplitude at the source A
s
can be found from the observed amplitude A
1
using
A
s
=A
1
exp(πfr/Qβ) and the ratio r/Q = 11 found by flattening the amplitude spectrum of
the burst. Assume the burst is located between the two stations. Because it is not
recorded by station 2, we can determine the maximum source-receiver distances for
which the burst amplitude is attenuated below an observation threshold (which we
define as 5 times the noise level at station 2). Note that as the event distance r from
station 1 increases, Q must also increase (since r/Q is fixed) and the event becomes
more likely to be observed at station 2. The amplitude at station 2 also increases
because the distance D-r between the event and station 2 is decreasing with increasing
r.
We found that the maximum distance the burst can be from station 1 such that it
was not observed at any of the surrounding stations ranges from 745m to 2900m with
corresponding maximum Q between 67 and 263. It is possible that the event was closer
to station 1, in which case Q would be even smaller. The assumption that the event is
between the two stations may not be true for any given observing station, but it is most
31
likely roughly true for one of the surrounding stations, for which the analysis is then
valid.
As a check, we estimated the moment magnitude of one of the six events
previously identified using spectral analysis. The burst on station D (Figure 1-9) at
approximately 42 seconds was detrended and windowed from the unfiltered record. We
then scaled the acceleration spectrum to produce a displacement spectrum and
calculated the moment from the long period spectral amplitude. Assuming
ρ = 3000 kg/m
3
, the event had a moment magnitude -0.28 (when β = 0.5 km/s). This
magnitude estimate is not corrected for attenuation. However, since the magnitude
determined from spectral analysis of a detrended burst is comparable to those
determined using the attenuation model, the bursts are probably located very close to
the stations.
1-2.5 Triggering Threshold
We assume that each burst was triggered by the maximum stress, independent of
frequency, in the strong motion recorded at the station prior to its arrival. Stress was
calculated from the velocity seismogram of the P and S waves according to
σ=ν
s
ρβ or σ=ν
p
ρα (3)
where ν
p
and ν
s
are the amplitude of the P- and S-wave particle motions, respectively, α
the P-wave velocity, β is the S-wave velocity. This is a slightly modified version taken
from Brodsky et al. (2000) where we have replaced the moduli using the relation
between seismic velocity, density, and modulus. Velocity seismograms were generated
by integrating the original broadband acceleration records. If no bursts were identified
32
on a record, we noted the maximum stress for that record. In general, each individual
station recorded multiple unique events, and we computed the triggering threshold for
each.
We observed a wide range of stresses where some stations showed triggering
and some stations did not, suggesting a strong dependence on site conditions. We
defined the threshold as the stress level above which 90% of the triggering occurred.
The choice of 90% eliminates the low-level outliers and reflects the steep increase in
triggering in Fig. 1-11. We found a threshold triggering stress between about 0.03 and
0.05 MPa for S-waves (β=0.5km/s, ρ=3000kg/m
3
) and between 0.0013 and 0.0033 MPa
for P-waves (α=0.9km/s, ρ=3000kg/m
3
) although the P-wave triggering is less
prevalent and only occurs at stations a few kilometers from the fault.
33
− 3 − 2 − 1 0 1
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Cumulative percent of triggered events
s/n 5
s/n 10
s/n 20
(A)
0
0.02
0.04
0.06
0.08
0. 1
0.12
s/n 5
s/n 10
s/n 20
− 3 − 2 − 1 0 1
Stress [log(MPa)]
(B)
Stress [log(MPa)]
Figure 1-11: Analysis of the stress threshold including the 1999 Chi-Chi mainshock data set as well as
the Mw 5.3 event from 1995. If a station records at least one burst, the peak dynamic stress prior to the
arrival time is assumed to be the triggering stress. A) For bursts identified during the S-wave window of
the strong shaking. B) For bursts identified during the P-wave window.
34
1-3. Discussion
Triggering by surface waves is well documented [Hill et al. (1993), Brodsky et
al. (2000), Kilb et al. (2000), Gomberg et al. (2001), Gomberg et al. (2004), Pankow et
al. (2004), Prejean et al. (2004), Miyazawa & Mori (2006)] and evidence for statistical
increases in seismicity due to changes in tidal stress is growing [Wilcock (2001) and
Cochran (2004) et al.]. However, as far as we know, dynamic triggering by strong
motion P- and S-waves has not been documented prior to the conclusion of this study.
We present evidence that high-frequency bursts revealed by the band-pass filtering of
strong motion from the Chi-Chi earthquake (Mw 7.6), as well as a relatively smaller
(Mw 5.3, 1995) earthquake in the same region, are the result of triggered events very
near the accelerometers in the Taiwan strong motion array. The minimum triggering
stress we calculate is comparable to the triggering levels found in studies of surface
waves and tides (Figure 1-12). These three modes of dynamic triggering represent a
huge range in loading period from about 1 sec for body waves to 12 hours for tidal
loading. However the minimum value of stress required for triggering only varies by
about 1 order of magnitude.
35
0.001
0.01
0.1
1
10
0.1 1 10 100 1000 10
4
10
5
Taiwan (B = 3.3 km/s)
Taiwan (B = 0.5 km/s)
Brodsky et al. (2000)
Gomberg et al. (2001)
Gomberg et al. (2004)
Hill et al. (1993)
Kilb et al., (2000)
Miyazawa & Mori (2006)
Prejean et al. (2004)
Pankow et al., (2004)
Cochran et al. (2004)
Wilcock (2001)
Stress (MPa)
Period (s)
Figure 1-12: Observed earthquake triggering stresses from this study, surface wave studies, and tidal
stress studies as indicated. Surface wave and tides appear to trigger at levels approximately one order of
magnitude lower than strong motion. Static stress generally associated with large earthquakes and
aftershocks is indicated by a solid line around 0.1 MPa (Freed 2005).
36
While the hypothesis proposed by Chen et al (2006) is consistent with their
observations at stations within 20 km of the fault plane, it is refuted by our observations
at more distant stations. It is interesting that their observations are also consistent with
our hypothesis of local triggering. For example, they observed that the earliest bursts
were triggered at shallow depths, directly up slope from the hypocenter by the arrival of
the mainshock P-wave. This is consistent with our hypothesis because dynamic stresses
generated by the P-wave are above the triggering threshold for these stations very near
to the Chelungpu fault. The origin times of events following the early P-wave
triggering were interpreted by Chen et al. (2006) to be triggered by passage of a rupture
front traveling at approximately 2.3 Km/s along the fault plane. Again this is also
consistent with our hypothesis that bursts are locally triggered by strong shaking, and
therefore arrive after the direct S. If the sources are constrained to lie on the fault plane,
they would appear to originate from a disturbance traveling slightly slower than the S
waves, which Chen et al. (2006) interpreted as the rupture front. Their observation of
Gutenberg-Richter and Omori law distributions at the nearest stations are also consisted
the triggering of a large number of events and possible aftershocks near each station.
We have assumed reasonable parameter values to calculate burst magnitude and
dynamic triggering stress. The magnitudes we calculate by two different methods range
from about Mw -0.5 to Mw 0.7. The burst magnitude calculation depends strongly on
the cube of the shear wave velocity and the choice of this parameter makes the largest
contribution to the uncertainty in the reported values. Throughout our analysis we use a
shear wave velocity of 500 m/s, representative of the highly weathered layer that makes
37
up the top few hundred meters of crust and is characterized by low shear wave
velocities and very high attenuations (e.g., Hanks 1982, Aster and Shearer 1991,
Fletcher et al., 1992). Likewise, the triggering stress calculation depends on this choice
of shear wave velocity. We find a minimum stress threshold for triggering ranging
from 0.03 to 0.05 MPa for S-wave triggering and 0.0013 to 0.0033 MPa for P-wave
triggering. These results are comparable with observations of triggering by surface
waves and daily earth tides.
We have ruled out the possibility that the origin of the bursts is somewhere on
the Chelungpu fault due to their appearance on records at large distances. Bursts are
further confirmed to be local by their absence on multiple stations which we also use to
constrain the horizontal distance from source to receiver. We do not constrain the depth
extent of the bursts, but it is reasonable to assume they occur in the top few hundred
meters of highly weathered and highly heterogeneous crust where many failure surfaces
exist. This is also an indication that our results are not affected by background
seismicity which exists almost entirely at depths below 1km. We cannot rule out the
possibility that the bursts are instrumental artifacts or a local, non-earthquake source.
Many of the strong motion stations in the Taiwan CWB network are located in urban
areas. It is conceivable, and has been suggested to us, that the bursts are falling roof
tiles or building damage of some sort resulting from the mainshock strong motions.
Occurrences like this very near the stations might produce the discrete bursts of energy
we attribute to locally triggered earthquakes. However, the instruments included in the
data set used in this study are located over a wide region and supplied by multiple
38
geotechnical manufacturers; consequentially we do not expect to see such ubiquitous
instrumental artifacts (Willie Lee, private communication). The possibility remains that
the source of the bursts is local, but originates from some poorly known process. Rate-
and-state friction has been shown to be a poor candidate for earthquake nucleation and
triggering by periodic loading (Dieterich 1987) at depths below a few hundred meters,
but may be a viable mechanism for triggering of the very shallow events found in this
study. High frequency seismology is plagued by the extremely heterogeneous top 100m
or so of crust on which almost all seismometers sit and the nature of this environment
may produce a variety of signals due to scattering, reflection, trapped waves, etc. and
very dense seismic arrays should provide greater insight into the nature of dynamically
triggered high-frequency bursts.
39
Chapter 1 References
Abercrombie, R.E. (1995). Earthquake source scaling relationships from -1 to 5 M
L
using seismograms recorded at 2.5 km depth, J. Geophys. Res., 100, 24015-24036
Aster, R. C., and P. M. Shearer (1991b). High-frequency borehole seismograms
recorded in the San Jacinto fault zone, southern California. Part 2. Attenuation and site
effects, Bull. Seism. Soc. Am., 81, 1081-1100
Brodsky, E. E., V. Karakostas, and H. Kanamori (2000). A new observation of
dynamically triggered regional seismicity: Earthquakes in Greece following the August,
1999 Izmit, Turkey earthquake, Geophys. Res. Lett., 27, 2741–2744
Brodsky, E. E., and S. G. Prejean (2005). New constraints on mechanisms of remotely
triggered seismicity at Long Valley Caldera, J. Geophys. Res., 110, B04302,
doi:10.1029/2004JB003211.
Brune, J. (1970). Tectonic stress and the spectra of seismic shear waves from
earthquakes, J. Geophys. Res., 75, 4997-5009.
Chen, Y., C. G. Sammis, and T.-L. Teng (2006). A High-Frequency View of the 1999
Chi-Chi, Taiwan, Source Rupture and Fault Mechanics, Bull. Seism. Soc. Am., 96, 807-
820.
Cochran, E. S., J. E. Vidale, and S. Tanaka (2004). Earth tides can trigger shallow
thrust fault earthquakes, Science 306, no. 5699 (20041112): 1164-1166.
Dieterich, J. H. (1987). Nucleation and triggering of earthquake slip: effect of periodic
stresses, Tectonophysics, 144, 127-139.
Dieterich, J. H., (1994). A constitutive law for rate of earthquake production and its
application to earthquake clustering, J. Geophys. Res., 99, 2601-2618.
Ebel, J. (1989). The effect of crustal scattering on observed high-frequency earthquake
seismograms, Geophys. J. R. Astr. Soc. 98, 329-341.
Felzer, K. R., R. E. Abercrombie, and E. E. Brodsky (2005). Testing the stress shadow
hypothesis, J. Geophys. Res. 110, doi:10.1029/2004JB003277.
Freed, A. (2005), Earthquake triggering by static, dynamic, and postseismic stress
transfer, Annu. Rev. Earth Planet. Sci., 33, 335–368.
40
Gomberg, J. S., P. A. Reasenberg, P. Bodin, R. A. Harris (2001). Earthquake triggering
by seismic waves following the Landers and Hector Mine earthquakes, Nature 411, no.
6836 (20010524): 462-466
Gomberg, J., P. Bodin, K. Larson, and H. Dragert (2004). Earthquake nucleation by
transient deformations caused by the M = 7.9 Denali, Alaska, earthquake, Nature, 427,
621–624
Hill, D., et al. (1993). Seismicity remotely triggered by the magnitude 7.3 Landers,
California, earthquake, Science, 260, 1617–1623.
Kilb, D. L., J. Gomberg, and P. Bodin (2000). Triggering of earthquake aftershocks by
dynamic stresses, Nature, 408, 570–574.
King G. C., R. S. Stein, and J. Lin (1994). Static stress changes and the triggering of
earthquakes, Bull. Seism. Soc. Am., 84, 935-953.
Harris, R. A. and R. W. Simpson (1996). In the shadow of 1857 – the effect of the great
Ft. Tejon earthquake on subsequent earthquakes in southern California, Geophys. Res.
Lett., 23, 229-232.
Hiramatsu, Y., H. Yamanaka, K. Tadokoro, K. Nishigami, and S. Ohmi (2002). Scaling
law between corner frequency and seismic moment of microearthquakes: Is the
breakdown of the cube law a nature of earthquakes?, Geophys. Res. Lett., 29
(8),10.1029/2001GL013849.
Liu Y., Teng T-L. Teng, and Y. Ben-Zion (2005). Near-surface seismic anisotropy,
attenuation and dispersion in the aftershock region of the 1999 Chi-Chi earthquake,
Geophysical Journal International, 160 (2), 695–706. doi:10.1111/j.1365-
246X.2005.02512.
Ma, K.-F., J. Mori, S.-J. Lee, and S. B. Yu (2001). Spatial and temporal distribution of
slip for the 1999 Chi-Chi, Taiwan, earthquake, Bull. Seism. Soc. Am., 91, 1069 –1087.
Ma, K.F., C. H. Chan, and R. S. Stein (2005). Response of seismicity to Coulomb
stress triggers and shadows of the 1999 Mw = 7.6 Chi-Chi, Taiwan, earthquake, J.
Geophys. Res., 110, doi:10.1029/2004JB003389.
Miyazawa, M., and J. Mori (2006). Evidence suggesting fluid flow beneath Japan due
to periodic seismic triggering from the 2004 Sumatra-Andaman earthquake, Geophys.
Res. Lett., 33, L05303, doi:10.1029/2005GL025087.
41
Pankow, K. L., W. J. Arabasz, J. C. Pechmann, and S. J. Nava (2004), Triggered
seismicity in Utah from the November 3, 2002, Denali fault earthquake, Bull. Seism.
Soc. Am., 94, S332–S347.
Prejean, S. G., D. P. Hill, E. E. Brodsky, S. E. Hough, M. J. S. Johnston, S. D. Malone,
D. H. Oppenheimer, A. M. Pitt, and K. B. Richards-Dinger (2004). Remotely triggered
seismicity on the United States west coast following the M (sub w) 7.9 Denali Fault
earthquake The 2002 Denali Fault earthquake sequence, Bull. Seism. Soc. Am., 94, no.
6, 348-359.
Richardson, E. and T.H. Jordan (2002). Seismicity in deep gold mines of South Africa:
Implications for tectonic earthquakes, Bull. Seism. Soc. Am., 92, 1766-1782.
Vernon F. L., J. Fletcher, L. Carroll, A. Chave, and E. Sembera (1991). Coherence of
seismic body waves from local events as measured by a small-aperature array, J.
Geophys. Res. 96, 11981-11996.
Vernon, F. L., G. L. Pavlis, T. J. Owens, D. E. McNamara, and P. N. Anderson (1998).
Near-surface scattering effects observed with a high-frequency phased array at Pinyon
Flats, California, Bull. Seism. Soc. Am. 88, 1548 -1560.
Wang, G.-Q., D. M. Boore, H. Igel, and X.Y. Zhou (2003). Some Observations on
Colocated and Closely Spaced Strong Ground Motion Records of the 1999 Chi-Chi,
Taiwan, Earthquake, Bull. Seism. Soc. Am. 93, 674-693.
Wilock, W.S.D. (2001). Tidal triggering of microearthquakes on the Juan de Fuca
Ridge: Geophys. Res. Lett., 28, 3999-4002.
Wilson, D. C., and G. L. Pavlis (2000). Near-surface site effects in crystalline bedrock:
a comprehensive analysis of spectral amplitudes determined from a dense, three-
component seismic array, Earth Interactions 4.
42
Chapter 2: Dynamic triggering of high-frequency bursts by strong
motions during the 2004 Parkfield earthquake sequence
Summary
High-pass filtering (>20Hz) of acceleration records from the USGS Parkfield
Dense Seismograph Array (UPSAR) reveals a series of bursts that occur only during
strong shaking from the 2004 Mw6 Parkfield, California, earthquake and its immediate
aftershocks. Because there is no correlation between these high frequency bursts
observed at closely spaced stations, we hypothesize that they are associated with
dynamically triggered events occurring within 20 meters of the stations in the highly
fractured shallow crust. The triggering threshold was found to be ~0.02 MPa, consistent
with a previous estimate based on a similar analysis of high-frequency bursts observed
in strong motion data from the 1999 Chi-Chi earthquake in Taiwan [Fischer et al.,
2008]. The consistent observation of high-frequency bursts at both Parkfield and
Taiwan suggest that they may be a common phenomenon associated with strong motion
in the very shallow crust.
2-1. Introduction
The high-frequency content of seismic signals recorded by surface seismometers
can be attributed to a number of processes. These include high-frequency radiation
from the source region [e.g., Peng et al., 2006], scattering [e.g., Ebel, 1989; Wilson and
Pavlis, 2000], reverberation in near-surface layers [Blakeslee and Malin, 1991], and
dynamic triggering of local earthquakes and non-volcanic tremor during the passage of
43
regional and teleseismic surface waves [e.g., Hill et al., 1993; Prejean et al., 2004; Hill
and Prejean, 2007; Gomberg et al. 2008; Miyazawa and Brodsky, 2008].
Many studies have shown that the top few hundred meters of the crust can produce
and attenuate high-frequency waveforms [Hanks, 1982; Shearer and Orcutt, 1987;
Cranswick, 1988; Ebel, 1989; Blakeslee and Malin, 1991; Vernon et al. 1991, 1998;
Wilson and Pavlis, 2000]. This layer is extremely fractured and weathered, and is
characterized by low shear wave velocity (~200-400 m/s) and a very high attenuation
(Q ~1-10) [e.g., Aster and Shearer, 1991]. Other recent studies have found high-
frequency spikes (or cusped waveforms) in the latter portions of the strong-motion
seismograms from stations on soft soils for several large earthquakes [e.g., Holzer et al.,
1989; Frankel et al., 2002; Bonilla et al., 2005]. Bonilla et al. [2005] suggested that the
high frequency spikes in these acceleration records could be related to an increase of
effective stress and strain hardening due to an increase of pore pressure in the granular
material under large strains.
Recently, band-pass filtered acceleration records of strong motion P- and S-waves
generated by the 1999 Mw7.6 Chi-Chi, Taiwan, earthquake were observed to contain
high-frequency (>20 Hz) bursts at distances up to 170 km from the epicenter [Fischer et
al., 2008]. These bursts differ from anything previously observed in that they are small,
discrete higher-frequency packets superimposed on the much larger lower-frequency
strong motion, and they are only apparent through either extreme magnification and
detrending or high-pass filtering [Chen et al.. 2006, Fischer et al., 2008]. Fischer et al.
[2008] suggested that they are local triggered events located in the highly damaged and
44
weathered top few hundred meters of the crust, a hypothesis supported by their
observation at large distances from the fault plane and a lack of correlation at
neighboring stations. That work challenged the original interpretation by Chen et al.
[2006], who argued that the sources of the bursts were located on the Chelungpu fault
plane that ruptured during the Chi-Chi earthquake. By comparing the seismic signals at
stations that recorded bursts with those that did not, Fischer et al. [2008] found a
threshold stress for dynamic triggering by S-waves to range from 0.03 to 0.05 MPa,
depending on the signal-to-noise ratio (SNR) used to define the bursts.
45
-120˚36' -120˚24' -120˚12'
35˚48'
36˚00'
36˚12'
0 5 10
km
UPSAR
Mw6.0
Creeping
SAF
Locked
SAFOD
Pacific
Ocean
California
(a)
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
S-N Distance (km)
-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3
W-E Distance (km)
P01
P02
P03
P04
P05 P06
P07
P08
P09
P10
P11
P12
P13
P14
(b)
Figure 2-1: (a) Map view of San Andreas Fault near Parkfield, CA. The 2004 M6 Parkfield earthquake
and large early aftershocks are indicated by stars and other aftershocks are shown as dots. The thick gray
and light lines denote surface traces of SAF and nearby roads, respectively. The background is shaded
topography with white being low and dark being high. The locations of the USGS Parkfield Dense
Seismograph Array (UPSAR), and the SAFOD are denoted by the inverted triangle and square,
respectively. The inset shows the study area on a map of California. (b) Small scale view of UPSAR
stations. The distances are relative to station P06.
46
In this study we document similar high-frequency bursts observed by the USGS
Parkfield Dense Seismograph Array (UPSAR) during the strong shaking from the 2004
Mw6 Parkfield earthquake (Figure 2-1) and its largest immediate aftershocks. Our
primary goals are to document high-frequency bursts in a region other than Taiwan and
to better constrain their locations and sizes using the closer station spacing at Parkfield.
2-2. Analysis
2-2.1 Observation of Bursts on the UPSAR Records During the 2004 Parkfield
Earthquake
The Parkfield mainshock occurred at 17:15:14 UTC on 28 September 2004 with an
epicenter 11 km southeast of Parkfield, California [Langbein et al., 2005], and was
followed by numerous aftershocks. The UPSAR array is located about 10 km SE of the
Parkfield segment of the San Andreas Fault (SAF), and consists of 14 stations within an
area of about 1 km
2
(Fig. 2-1). Each station has a three-component weak-motion
velocity transducer and a strong-motion accelerometer, which are recorded at 200
samples per second [Fletcher et al., 1992]. These recordings are ideal for searching for
dynamically triggered high-frequency bursts during the strong shaking associated with
the Parkfield earthquake sequence.
We searched for bursts by applying a 4
th
order Butterworth high-pass filter with a
cutoff frequency of 20 Hz to all the strong motion recordings from the 12 active stations
during the first 900 seconds of the Parkfield earthquake sequence (Fig. 2-2). We chose
20 Hz as a lower cutoff because seismic energy above this frequency attenuates very
quickly and is less likely to be contaminated by radiation from the mainshock source
47
region, and because this is the frequency range in which high-frequency bursts were
observed in Taiwan [Chen et al., 2006; Fischer et al., 2008]. We observed high-
frequency bursts on the filtered records from all three components of all 12 stations
during the strong shaking from the Parkfield mainshock (Figs. 2-2,2-3). The amplitudes
and durations of these bursts are similar to those observed in Taiwan [Chen et al., 2006;
Fischer et al., 2008]. Bursts were also observed during strong shaking associated with
three largest immediate aftershocks (Mw ≥ 4) at Parkfield, but not during the relatively
quiet periods between these events.
48
1 2 3 4 5 6 7 8 9 10
0
5
10
15
20
25
30
35
40
P01
P02
P03
P05
P06
P07
P08
P09
P10
P11
P12
P13
Amplitude (m/s )
-2
360
(a)
P-wave
S-wave
1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
5
6
P-wave S-wave
360
(b)
Time (seconds after 9/28/2004 17:15:27.4) Time (seconds after 9/28/2004 17:15:27.4)
Figure 2-2: Raw (a) and 20-Hz high-pass-filtered (b) north-component acceleration seismograms
recorded by the UPSAR during the first 10 s of the 2004 Parkfield earthquake. The station name is
marked on the left. The high-frequency bursts picked by our automated algorithm with a signal-to-noise
ratio of 50 are indicated in black in (b). Records are lined up with the origin time of the Parkfield
mainshock as indicated. P- and S-wave arrivals are indicted on the figure for reference.
49
Although Peng et al. [2006] used a similar high-pass filtering technique to identify
many early aftershocks in the source region that were not in the standard earthquake
catalogs, it is evident from Fig. 2-2 that the bursts observed here are different in that
they are not clearly correlated at different stations in the array. Even the largest
amplitude burst observed during the Parkfield mainshock cannot be clearly identified at
any nearby station, although there are many small amplitude events at about the same
time that might be correlated (see, for example, the large burst at station P10 in Fig. 2-2
at ~8 s). This first-order observation suggests that the high-frequency bursts are most
likely due to a phenomenon located very near each station.
50
0
5
10
15
20
25
1 2 3 4 5 6 7 8 9 10
0
1
2
3
4
Amplitude (m/s )
-2
UP
360
90
UP
360
90
P-wave
S-wave
Time (seconds after 9/28/2004 17:15:27.4)
P11
Figure 2-3: Raw (top) and 20-Hz high-pass-filtered (bottom) three-component seismograms recorded by
station P11 during the first 10 s of the Parkfield mainshock. The high-frequency bursts picked by our
automated algorithm with a signal-to-noise ratio of 50 are indicated in black.
51
2-2.2 Burst Identification Algorithm
Our automated algorithm for identifying bursts generally follows that in Fischer et
al. [2008] and is briefly described here. The accelerograms were first high-pass filtered
at 20 Hz using a 4
th
order Butterworth filter. Bursts were identified using an algorithm
that imposed an amplitude threshold based on the SNR. Local minima and maxima
were found by differentiation and local extrema that exceeded the amplitude threshold
were saved. If any two saved picks occurred within 0.2 seconds of each other, they
were considered as a single event. We also required that a burst appear on all three
components.
This algorithm picked bursts that were obvious by visual inspection, but it also
picked events that were not obvious or appeared as a relatively continuous “tremor-like”
signal [e.g., Obara, 2002]. Most of these questionable picks were eliminated by
increasing the SNR threshold and paying the penalty of reducing the total number of
identified bursts. We found that a SNR of 50 gave the best agreement with visual
inspection.
We used this algorithm to search for bursts on the records from all 12 operating
UPSAR stations during a 900 second interval that include the Parkfield mainshock and
its early aftershocks. We found 648 high-frequency bursts using the SNR = 10, 136
bursts using SNR = 50, and 44 bursts using SNR = 100.
2-2.3 Triggering Threshold
If the high-frequency bursts originate in the shallow crust very near the stations and
are dynamically triggered by strong ground motions [Fischer et al., 2008], then the
52
triggering threshold can be estimated by calculating the maximum dynamic stress in the
seismic waves just prior to the arrival of each burst. For plane waves, the peak dynamic
stress
d
σ is equal to β / u G & [Hill and Prejean, 2007], where G is the shear modulus, u & is
the peak particle velocity, and β is the shear-wave velocity. Values of u & were obtained
by integrating the strong motion acceleration records. The triggering stress of a burst
was defined as the maximum value of
d
σ , on all three components, within a 1 second
window prior to its arrival. Similar instantaneous triggering was observed during the
passage of surface waves generated by the 2002 Denali earthquake [Prejean et al.,
2004; Gomberg et al. 2008], and 2004 Sumatra earthquake [e.g., Miyazawa and
Brodsky, 2008]. Repeating the calculations with longer time windows did not produce a
significant change in the threshold.
Figure 2-4 shows the distributions of triggering stresses for different SNR. Taking
G = 0.3 GPa and β = 326 m/s [Fletcher et al., 2006] for the shallow crust, gives a stress
amplitude threshold of 0.02 MPa for SNR = 50. We define the threshold as the stress
level above which 95% of bursts are triggered. These stresses are consistent with the
range of 0.03 to 0.05 MPa obtained from the high-frequency triggering analysis of
bursts observed during strong motion from the Chi-Chi earthquake [Fischer et al.,
2008].
53
10
−3
10
−2
10
−1
10
0
SNR = 10 (648 bursts)
SNR = 20 (389 bursts)
SNR = 30 (250 bursts)
SNR = 40 (192 bursts)
SNR = 50 (136 bursts)
SNR = 100 (44 bursts)
Triggering Stress (MPa)
Percent with lower Threshold
100
50
0
5%
Figure 2-4: Triggering threshold for a range of signal-to-noise ratios (SNR). The horizontal line
indicates the 95% Threshold. The intersection of this line with the curves indicates the triggering stress.
For our preferred SNR = 50, the threshold is 0.02 MPa. For SNR > 30, triggering thresholds cluster
about our preferred value of 0.02 MPa at SNR = 50.
54
2-2.4 Location and Size of the Bursts
In the Taiwan study [Fischer et al., 2008], observed a lack of correlation across a
sub-array with average station spacing of 3 km. The failure to identify a burst on more
than one station implied they were within 750 meters of the station and that the
magnitude had to be below Mw = 0. The nearest neighbors in the UPSAR array are 40
meters apart. Lack of correlated bursts at UPSAR implies that they are within 20
meters of a station and the magnitude of the largest amplitude burst has Mw = −1 and
corresponding seismic potency Po = 1.2x10
-5
km
2
cm, where this magnitude was
determined from the spectral amplitude in the low frequency limit. Because the
spectrum is contaminated by energy from the Parkfield mainshock at low frequency, we
assumed that the spectral amplitude at 10 Hz gives an adequate approximation to the
magnitude. It is of interest to note that for vertically propagating SH-waves, peak
dynamic stresses are found at the ¼ wavelength depth [Sleep and Ma, 2008]. The
dominant frequency from the average spectrum of all the records is about 2 Hz.. The
corresponding wavelength with the near-surface S-wave velocity β = 326 m/s is about
140 m, and the ¼ wavelength is about 35 m. This value is close to the distance where
we think the bursts are produced.
2-3. Discussion
In this study we observed many high-frequency bursts during strong motion from
the 2004 Parkfield earthquake and its immediate largest aftershocks recorded by the
UPSAR array. After investigating and ruling out several possible artifacts associated
55
with instruments (See Appendix A in the Supplementary Material), we hypothesize that
these signals are generated by sources within 20 m of the receivers that were
dynamically triggered by the strong ground motion. This hypothesis is supported by
three lines of evidence. First, the high-frequency bursts were only observed during the
strong shakings (Fig. 2-4). Second, although an individual burst was recorded on all
three components of a given station, it could not be identified at other nearby stations
(Figs. 2-2, 2-3). This lack of correlation is most likely due to the relatively small
amplitudes and high-frequencies of the bursts, and to the extremely high attenuation in
the highly fractured and weathered uppermost crust. Third, the threshold stress required
for triggering found here (0.02 MPa) is consistent with a previous estimate for bursts
triggered by 1999 Chi-Chi earthquake in Taiwan [Fischer et al., 2008].
Recent studies using repeating earthquakes have found reduced seismic velocities in
the top few hundred meters of the crust associated with strong motions from nearby
large earthquakes [e.g., Rubinstein and Beroza, 2005; Peng and Ben-Zion, 2006]. The
reductions are most likely caused by the opening of pre-existing cracks during dynamic
shaking, and are typically followed by logarithmic recovery. We speculate that this
cracking process is accompanied by frictional loss from discrete events that reradiate
energy to higher frequencies, some of which are recorded as high-frequency bursts by
nearby stations. Large, but very rare bursts have been identified in the Parkfield region
[Sleep and Ma, 2008] and have implications for theoretical limits of ground motion. In
contrast, the numerous small high-frequency events observed in our study may be
56
ubiquitous and play an important role in the attenuation of strong motion in the shallow
crust.
2-4. Supplementary Material
It is possible that the occurrence of high-frequency bursts is merely an artifact from
instrument noises. We have investigated and ruled out several possible instrumental
artifacts.
First, the UPSAR accelerometers recorded the Parkfield mainshock and aftershocks
on scale, so the bursts were not generated by band-pass-filtering of clipped waveforms.
Secondly, it has been pointed out that a possible source of the bursts identified on
acceleration records by the UPSAR is from the coupling of velocity transducer
instruments with the accelerometers housed on the same foundation (J. Fletcher, per.
comm., 2007). If the velocity instrument is sufficiently accelerated in one direction then
it may be possible for physical contact between moving and stationary parts of the
instrument to take place. This is different from signal clipping that is present on all the
velocity records at UPSAR during the mainshock. Signal clipping happens when the
ground motion exceeds the dynamic range of the instrument. Physical contact between
the moving part of the transducer and its stopping mechanism would register as a large
acceleration but zero velocity. Hence, this type of contact could occur at zero crossing
of the velocity records.
If contact is made, and the instrument continues to be accelerated in the same
direction following the first contact, it will continue to make contact much like a ball
57
bouncing under the influence of gravity. This will show up as a short series of zero
crossings on the velocity records (Fig. 2-5).
58
−500
500
−2
2
−500
500
−2
2
−500
500
5 6 7 8 9 10
−2
2
Time (seconds)
- - Velocity instrument
*
Ringing
Station P11
Burst
__
Accelerometer
Amplitude (m/s )
-2
UP
UP
360
360
90
90
Figure 2-5: Correlation between ringing events (red triangle) where velocity instrument makes physical
contact with its stopping mechanism and the arrival of bursts (asteroid) on the accelerometers. The raw
waveforms are plotted on the top panel for all three components (gray) and the 20 Hz high-pass-filtered
waveforms (black) are shown directly below along with the velocity transducer recording that is off scale.
We find no significant correlation between burst occurrence time and ringing occurrence time.
59
We have determined several instances where physical contact could be made by
visually inspecting 15 seconds of velocity data during the Parkfield mainshock. We
chose all instances where a series (at least 2) of repeating zero crossings or “ringing”
events within 0.5 s occurred and compared their onset times with the arrival times of
manually identified characteristic bursts (Fig. 2-5). The manually identified bursts are a
subset of the automatically picked bursts and correspond to large amplitude, short
duration bursts well separated from any other high frequency energies. The results
show that there is no clear correlation between the ringing events and high-frequency
bursts, suggesting that the observed bursts are unlikely to be caused by the physical
contact within the velocity sensor.
Furthermore, the GEOS station Work Ranch (WFU) about 500 m from UPSAR has
an accelerometer decoupled from the velocity transducer instrument (P. Spudich, per.
comm., 2007) and shows at least one burst (Fig. 2-6). Unfortunately, the record from
this instrument is only 9 seconds long, preventing us from identifying more bursts at
latter period.
Finally, it is possible that the bursts may reflect signals on and above the surface, such
as falling objects [e.g., Frankel et al., 2002], cracking in buildings, or other cultural
noise during strong shaking. Since the UPSAR array is deployed on top of a local
ridge, and is relatively far from commercial or residential buildings, the bursts are
unlikely to be associated with building cracking or cultural noise. Although we cannot
totally rule out falling objects or other movements on the surface as the cause, the fact
60
that high-frequency bursts are observed in many different environments in California
and in Taiwan [Chen et al., 2006; Fischer et al., 2008] suggests they are most likely
generated in the shallow crust.
61
0 1 2 3 4 5 6 7 8 9
0
5
10
15
20
0 1 2 3 4 5 6 7 8 9
0
0.5
1
1.5
2
Time (seconds)
Amplitude (m/s )
-2
UP
360
90
Work Ranch (WFU)
UP
360
90
Figure 2-6: Raw and high-pass-filtered (20 Hz) waveforms for the GEOS station at Work Ranch. Three
components are as indicated for the raw waveforms (top 3 traces) and high-pass-filtered waveforms
(bottom 3 traces).
62
Chapter 2 References
Aster R. C. and P. M. Shearer (1991), High-frequency borehole seismograms recorded
in the San Jacinto fault zone, southern California. Part 2. Attenuation and site effects,
Bull. Seismol. Soc. Am., 81, 1081–1100.
Blakeslee S. and P. Malin (1991), High-frequency site effects at two Parkfield
downhole and surface stations, Bull. Seismol. Soc. Am., 81, 332–345.
Bonilla, L. F., R. J. Archeleta, and D. Lavellee (2005), Hysteretic and dilatant behavior
of cohesionless soils and their effects on nonlinear site response: field data observations
and modeling, Bull. Seism. Soc. Am. 95, 2373 –2395.
Cranswick, E. (1988), The information content of high-frequency seismograms and the
near-surface geologic structure of "hard rock" recording sites, Pure Appl. Geophys.,
128, 333–363.
Chen, Y., C. G. Sammis, and T.-L. Teng (2006), A high frequency view of the 1999
Chi-Chi, Taiwan, Source Rupture and Fault Mechanics, Bull. Seismol. Soc. Am., 96,
807–820.
Ebel, J. E. (1989), The effect of crustal scattering on observed high-frequency
earthquake seismograms, Geophy. J. Int., 98, 329–341.
Fischer A. D., C. G. Sammis, Y. Chen, and T. L. Teng (2008), Dynamic Triggering by
Strong Motion P- and S-waves: Evidence from the 1999 Chi-Chi, Taiwan Earthquake,
Bull. Seism. Soc. Am., 98, 580–592.
Fletcher, J. P, P. Spudich, P. Goldstein, J. Sims, and M. Hellweg (1992), The USGS
Parkfield, California dense seismograph array – UPSAR, Bull. Seismol. Soc. Am., 82,
1041–1070.
Fletcher, J. P, P. Spudich, and L. M. Baker (2006), Rupture Propagation of the 2004
Parkfield, California, Earthquake from Observations at the UPSAR, Bull. Seismol. Soc.
Am., 96, S129–S142; doi: 10.1785/01200050812.
Frankel, A. D., D. L. Carver, and R. A. Williams (2002). Nonlinear and linear site
response and basin effects in Seattle for the M6.8 Nisqually, Washington, Earthquake,
Bull. Seismol. Soc. Am. 92, 2090–2109.
Gomberg, J., J. L. Rubensteing, Z. Peng, K. C. Creager, J. E. Vidale, and P. Bodin
(2008), Widespread Triggering of Nonvolcanic Tremor in California, Science, DOI:
10.1126/science.1149164.
63
Hanks T. C. (1982). f
max
, Bull. Seismol. Soc. Am., 72, 1867–1880.
Hill, D., et al. (1993), Seismicity remotely triggered by the magnitude 7.3 Landers,
California, earthquake, Science, 260, 1617–1623.
Hill, D.P. and S.G. Prejean (2007), Dynamic Triggering, in Treatise on Geophysics (ed.
G. Schubert), v. 4, Earthquake Seismology (ed. H. Kanamori), 257–292, Elsevier,
Amsterdam.
Holzer, T. L., T. L. Youd, and T. C. Hanks (1989). Dynamics of liquefaction during
1987 Superstition Hills, California, earthquake, Science 244, 56 –59.
Langbein, J., R. Borcherdt, D. Dreger, J. Fletcher, J. L. Hardebeck, M. Hellweg, C. Ji,
M. Johnston, J. R. Murray, R. Nadeau, M. J. Rymer, and J. A. Treiman (2005),
Preliminary report on the 28 September 2004, M6.0 Parkfield, California earthquake,
Seismol. Res. Lett., 76, 10–26.
Miyazawa, M., and E. E. Brodsky (2008), Deep low-frequency tremor that correlates
with passing surface waves, J. Geophys. Res., 113, B01307,
doi:10.1029/2006JB004890.
Obara, K. (2002), Nonvolcanic deep tremor associated with subduction in southwest
Japan, Science, 296, 1699–1681.
Peng, Z., and Y. Ben-Zion (2006), Temporal changes of shallow seismic velocity
around the Karadere-Duzce branch of the north Anatolian fault and strong ground
motion, Pure Appl. Geophys., 163, 567–599.
Peng, Z., J. E. Vidale, and H. Houston (2006), Anomalous early aftershock decay rates
of the 2004 M6 Parkfield earthquake, Geophys. Res. Lett., 33, L17307,
doi:10.1029/2006GL026744.
Prejean, S. G., D. P. Hill, E. E. Brodsky, S. E. Hough, M. J. S. Johnston, S. D. Malone,
D. H. Oppenheimer, A. M. Pitt, K. B. Richards-Dinger (2004), Remotely triggered
seismicity on the United States west coast following the Mw 7.9 Denali Fault
earthquake The 2002 Denali Fault earthquake sequence, Bull. Seismol. Soc. Am., 94,
348–359.
Rubinstein, J. L. and G. C. Beroza (2005), Depth constraints on nonlinear strong
ground motion from the 2004 Parkfield earthquake, Geophys. Res. Lett., 32, L14313,
doi: 10.1029/2005GL023189.
64
Shearer, P. M., and J. A. Orcutt (1987), Surface and near-surface effects on seismic
waves–theory and borehole seismometer results, Bull. Seismol. Soc. Am., 77, 1168–
1196.
Sleep, N. H. and S. Ma (2008), Production of brief extreme ground acceleration pulses
by nonlinear mechanisms in the shallow subsurface, Geochem. Geophys. Geosyst., 9,
Q03008, doi:10.1029/2007GC001863.
Vernon, F. L., J. Fletcher, L. Carroll, A. Chave, and E. Sembrera (1981), Coherence of
Seismic Body Waves from Local Events as Measured by a Small-Aperture Array, J.
Geophys. Res. 96, 11981–11996.
Vernon, F. L., G. L. Pavlis, T. J. Owens, D. E. McNamara, and P. N. Anderson (1998),
Near-surface scattering effects observed with a high-frequency phased array at Pinyon
Flats, California, Bull. Seismol. Soc. Am. 88, 1548–1560.
Wilson, D. C. and G. L. Pavlis, (2000), Near-surface site effects in crystalline bedrock:
a comprehensive analysis of spectral amplitudes determined from a dense, three-
component seismic array, Earth Interactions, 4, 1–31.
65
Chapter 3: Body Wave Triggering of Small Shallow Events During
Strong Motion
Summary
High-pass filtering (>20Hz) of acceleration records from the 1999 Chi-Chi Taiwan and
2004 Parkfield, California earthquakes reveal a series of bursts that occur only during
strong shaking. Initially interpreted as originating from asperity failure on the
Chelungpu fault, bursts observed during the Chi-Chi earthquake were subsequently
determined to be a local effect within about 1 km of the seismic stations. Similar bursts
were observed at the UPSAR array during the Parkfield earthquake and were
constrained to originate less than 20m from the instruments. Such small shallow events
can not result from the triggered release of stored elastic energy because rate-and-state
friction rules out stick-slip instability on such small, shallow patches. Our hypothesis is
that the bursts are not triggered, but are driven by simultaneous shear and tensile
stresses near the surface during the strong motion. At 2 Hz, SV to P wave mode
conversion at the free surface produces tensile stresses to depths of 70m. Where
standard triggering releases stored elastic energy and adds to the incident wavefield, this
new driving mechanism takes energy out of the 2Hz strong motion and reradiates it at
high frequencies. It is thus an attenuation mechanism which we estimate can contribute
3% to the net attenuation in the very shallow crust.
3-1. Introduction
The high-frequency content of seismic signals recorded by surface seismometers
can be attributed to a number of processes. These include high-frequency radiation
66
from the source region [e.g., Peng et al., 2006], scattering [e.g., Ebel, 1989; Wilson and
Pavlis, 2000], reverberation in near-surface layers [Blakeslee and Malin, 1991], and
dynamic triggering of local earthquakes and non-volcanic tremor during the passage of
regional and teleseismic surface waves [e.g., Hill et al., 1993; Prejean et al., 2004; Hill
and Prejean, 2007; Gomberg et al. 2008; Miyazawa and Brodsky, 2008]. Recently,
high-frequency (>20 Hz), short duration (0.2s) bursts of energy were observed on strong
motion instruments during the strong motion shaking from the 1999 Chi-Chi, Taiwan
earthquake [Chen et. al., 2006; Fischer et. al., 2008a] and the 2004 Parkfield, California
earthquake [Fischer et. al., 2008b] (see Figure 3-1 for examples of bursts).
67
0
500
0 2 4 6 8 10 12 14 16 18 20
−50
0
50
0
0 5 10 15 20 25 30 35 40
−20
0
20
Acceleration (gals) Acceleration (gals)
1000
-1000
Chi-Chi Earthquake, Taiwan - Station TCU084 EW component
Time (seconds after 1999/09/20, 17:47:15.85)
Time (seconds after 2004/09/28, 17:15:27.4)
Parkfield Earthquake, California - Station P06 EW component
Figure 3-1: Example of bursts observed in Taiwan (top panel) and Parkfield (bottom panel). The top
record in each case is the raw strong motion while the lower record is the result of high-pass filtering at
>20Hz. High-frequency bursts are clearly visible on the filtered records appearing as spikes on these
timescales.
68
Band-pass filtered acceleration records of strong motion P- and S-waves
generated by the 1999 Mw7.6 Chi-Chi, Taiwan, earthquake were observed to contain
high-frequency bursts at distances up to 170 km from the epicenter [Fischer et al.,
2008a]. These bursts differed from anything previously observed in that they were
small, discrete higher-frequency packets superimposed on the much larger lower-
frequency strong motion, and they were only apparent after extreme magnification and
detrending or high-pass filtering [Chen et al.. 2006, Fischer et al., 2008a]. Fischer et al.
[2008a] suggested that they were local triggered events located in the highly damaged
and weathered top few hundred meters of the crust, a hypothesis supported by the
observation of bursts at large distances from the fault plane and a lack of correlation at
neighboring stations. That work challenged the original interpretation by Chen et al.
[2006], who argued that the bursts originated on the Chelungpu fault plane that ruptured
during the Chi-Chi earthquake. By comparing the seismic signals at stations that
recorded bursts with those that did not, Fischer et al. [2008a] found a threshold stress
for dynamic triggering by S-waves to range from 0.03 to 0.05 MPa, depending on the
signal-to-noise ratio used to define the bursts.
High-frequency bursts were also observed by Fischer et. al. [2008b] on strong
motion seismograms from the 2004 Parkfield, CA earthquake (Mw6) recorded by the
USGS UPSAR array situated roughly 10 km from the surface trace of the San Andreas
fault near Parkfield [Fletcher et. al., 1992]. This is a dense array with some stations
approximately 40m apart. As in the Chi-Chi case, neighboring instruments recorded
uncorrelated bursts above 20 Hz, which constrained both the size of the bursts and their
69
distance from the stations. In the case of Parkfield, the largest amplitude burst measured
Mw=-1.0 and had to originate less than 20m from the individual stations that recorded
them.
These observations raise interesting questions about the type of mechanism that
can generate a high-frequency burst of seismic energy in the top 20 meters of the crust.
While it is well established that the very shallow crust is highly fractured and that
motion on these fractures results in low shear wave velocities (200-400 m/s) and high
attenuation (Q~1-10) [e.g., Aster and Shearer, 1991], it is not obvious that slip on these
shallow fractures can be sufficiently unstable to radiate high-frequency seismic energy.
In fact, we show in the next section that it is impossible to nucleate unstable slip on
small fractures within the top 20 meters if that slip is controlled by rate- and state-
dependent friction.
However, we also show that mode conversion of P to SV and SV to P at the free
surface produces a combination of shear and tensile loading on shallow cracks that can
result in unstable slip, regardless of the frictional rheology. In fact, when the strong
motion produces a simultaneous increase in shear stress and decrease in normal stress, it
can produce a shear instability in the absence of any tectonic pre-stress. We term this
process “dynamic driving” to distinguish it from the more common notion of “dynamic
triggering” in which the seismic waves trigger the release of stored tectonic stress.
While dynamic triggering adds energy to the total wavefield from the tectonic release,
dynamic driving removes energy from the strong motion by reradiating it at higher
frequencies. Dynamic driving can thus be viewed as an attenuation mechanism.
70
The primary goal of this paper is to develop a model for the dynamic driving of
small events at shallow depths and to assess the model by measuring the dynamic stress
field during the 2004 Parkfield strong motion at the UPSAR array using a multiple
station technique. We also assess the significance of the dynamic driving process as an
effective attenuation mechanism for strong motion.
3-2. Analysis
3-2.1 Constraints on shallow nucleation imposed by rate- and state-dependent
friction
Dieterich [1987] used a rate- and state-dependent friction law to explore the
conditions under which seismicity can be triggered by diurnal tidal stress fluctuations.
He found that the fractional rate of increased seismicity, R, is in phase with the periodic
loading and scales with the ratio of shear stress perturbation τ to normal stress σ acting
on a distribution of faults as
1
2
A
R
σ
τ
=
(1)
where A
1
is a constitutive parameter in the rate- and state-dependent friction law. If the
threshold for detecting a rate change in a catalog is R>10%, and assuming a typical
value of A
1
~ 0.003, then for a shear stress fluctuation τ= 0.002 MPa (typical for
diunrnal tides) equation (1) gives a maximum normal stress of 8 MPa corresponding to
a maximum depth of about 800m.
Since R is inversely proportional to the normal stress in equation (1), dynamic
triggering by a periodic source like tides or seismic waves becomes increasingly likely
71
as one approaches the surface. However, the nucleation of unstable slip at shallow
depths is severely limited by a minimum critical patch size inherent in rate- and state-
dependent friction. Dieterich [1986] estimated the critical radius r
c
of this minimum
patch to be
) ( 24
7
A B
GD
r
c
c
−
=
σ
π
(2)
where G is the shear modulus, D
c
is the characteristic slip distance, and A and B are
constitutive parameters in the friction law. Note that r
c
grows as 1/σ and is very large
near the surface. At a depth of 20m, the critical patch radius is 15m for the smallest
value of D
c
measured in the laboratory, and becomes larger than the depth closer to the
surface. The events observed in Taiwan and Parkfield [Fischer et al. 2008a and 2008b]
are too shallow to be stick-slip events on fractures controlled by rate- and state-
dependent friction.
We show in the next section that fluctuations in normal stress associated with
mode conversion between P and SV waves at the free surface can produce tension on
very shallow fractures, thus obviating the restrictions on nucleation usually imposed by
rate- and state-dependent friction.
3-2.2 Shallow stress fields generated by SV to P and P to SV mode conversion at
the free surface
In general, a P or an SV wave incident on a free surface of an elastic half-space
will be reflected as both P- and SV-waves. Aki and Richards [2002] give analytic
expressions for the horizontal and vertical displacements, u
i
and u
j
, for an incident P and
72
an incident SV wave at arbitrary incident angle. The spatial derivatives ∂u
i
/∂x
j
of this
displacement field can be used to find the 2D strain tensor
+ =
i
j
j
i
ij
x
u
x
u
∂
∂
∂
∂
ε
2
1
(3)
which, in turn yields the corresponding stress tensor
σ
ij
= λΔδ
ij
+ 2με
ij
(4)
The stress tensor can be used to find the normal and shear stress on any arbitrarily
oriented surface in the medium according to σ'= TσT
−1
where T is the 2D rotation
tensor
T =
sinθ cosθ
−cosθ sinθ
(5)
We varied the angle θ in 1 degree increments from 0 to 90 degrees, representing all
possible fault dip angles from vertical to horizontal, and calculated the maximum depth
of tension assuming a lithostatic stress that increases linearly with depth. As a result of
mode conversion, the reflected P wave may produce dilatational stresses sufficiently
large to overcome any lithostatic stresses and produce tension along the pre-existing
fault planes. For the strong motion amplitudes observed at Parkfield, tension is possible
to depths of approximately 70m for most fault orientations (Figure 3-2).
73
0 10 20 30 40 50 60 70 80 90
0
10
20
30
40
50
60
70
80
8
11
14
17
20
Fault dip angle (degrees from vertical)
Maximum depth of Tension (m)
Incident Angle
Figure 3-2: Maximum depth of tension as a function of fault dip angles for a range of SV incident angles.
The maximum depth of tension does not depend strongly on the incident angle and is between 60-70m for
most fault dips.
74
3-2.3 Measuring the dynamic stress in seismic waves
In order to better understand the slip instability that produces an observed high-
frequency burst, we require both the shear and normal components of the transient
strong-motion stress on an arbitrary slip plane near the instrument. Using a method
described by Spudich et al. (1995), the horizontal components of strain can be
determined using a minimum of 3 instruments if the station spacing is sufficiently small
compared to the wavelength of the transient deformation. The 2D strain tensor is
determined by solving the following set of 6 linear equations in the 4 unknown
displacement derivatives
∂u
i
∂x
j
where i and j now are two orthogonal horizontal
directions taken as i = West = x and j = North = y corresponding to the orientation of
the horizontal seismographs in the subarray in Figure 3-3.
Δu
y
12
= u
y
2
− u
y
1
=
∂u
y
∂y
Δy
12
+
∂u
y
∂x
Δx
12
(a)
Δu
x
12
= u
x
2
− u
x
1
=
∂u
x
∂x
Δx
12
+
∂u
x
∂y
Δy
12
(b)
Δu
y
13
= u
y
3
− u
y
1
=
∂u
y
∂y
Δy
13
+
∂u
y
∂x
Δx
13
(c)
Δu
x
13
= u
x
3
− u
x
1
=
∂u
x
∂x
Δx
13
+
∂u
x
∂y
Δy
13
(d)
Δu
y
23
= u
y
3
− u
y
2
=
∂u
y
∂y
Δy
23
+
∂u
y
∂x
Δx
23
(e)
Δu
x
23
= u
x
3
− u
x
2
=
∂u
x
∂x
Δx
23
+
∂u
x
∂y
Δy
23
( f )
(6)
75
In these equations u
x
s
is the East component of displacement at station s, u
y
s
is the N
component of displacement at station s. The E-W and N-S distances between two
stations r and s are Δx
rs
= x
s
− x
r
and Δy
rs
= y
s
− y
r
as illustrated in Figure 3-3.
76
y
x
U
y1
U
x1
U
y2
U
x2
U
y3
U
x3
Δx
12
Δy
12
Δy
13
Δx
13
Δy
23
Δx
23
Figure 3-3: Geometry of strain measurement method using displacements measured at three stations
represented as black dots
77
In order for this method to work, the spacing between the stations must be much
less than the wavelength of the seismic waves. For the strong motion at Parkfield, the
wavelength of the 2.5 Hz S-waves with a velocity of 326 m/s is about 130m. Of the 12
seismic stations in the UPSAR array used by Fischer et al. (2008), only the 3 station
grouping P05-P06-P07 with an average spacing of about 40m meets this requirement
(Figure 3-4).
78
-120˚36' -120˚24' -120˚12'
35˚48'
36˚00'
36˚12'
0 5 10
km
UPSAR
Mw6.0
Creeping
SAF
Locked
SAFOD
Pacific
Ocean
California
(a)
-0.3
-0.2
-0.1
0.0
0.1
0.2
0.3
0.4
S-N Distance (km)
-0.4 -0.3 -0.2 -0.1 0.0 0.1 0.2 0.3
W-E Distance (km)
P01
P02
P03
P04
P05 P06
P07
P08
P09
P10
P11
P12
P13
P14
(b)
Figure 3-4: (a) Map view of San Andreas Fault near Parkfield, CA. The 2004 M6 Parkfield earthquake
and large early aftershocks are indicated by stars. Other aftershocks are shown as dots. The thick gray
and light lines denote surface traces of SAF and nearby roads, respectively. The background is shaded
topography with white being low and dark being high. The locations of the USGS Parkfield Dense
Seismograph Array (UPSAR), and the SAFOD are denoted by the inverted triangle and square,
respectively. The inset shows the study area on a map of California. (b) Small scale view of UPSAR
stations. The distances are relative to station P06.
79
The 2D strain tensor (eqn. 3) was calculated as a function of time from the
spatial derivatives determined from equations (6). The corresponding stress tensor was
calculated according to eqn. (4), and the three independent components are plotted as a
function of time in Figure 3-5. The elastic constants λ = 6.7 GPa and μ = 0.24 GPa
were calculated from V
p
= 537 m/s V
s
= 326 m/s and ρ = 2300 g/cm3 appropriate to the
shallow crust (Fletcher et al., 2006). Note that the normal stress σ
xx
on vertical planes
with normal vectors pointing East and the normal stress σ
yy
on vertical planes with
normal vectors pointing North often have negative (tensile) values. We have shown in
the previous section that such tension during the S wave train is a shallow phenomenon
associated with mode conversion upon reflection at the surface. At times when the
normal stress is tensile, the constraints imposed by rate- and state-dependent friction do
not apply and a shear instability of any size is possible.
80
0 2 4 6 8 10 12
−1
−0.5
0
0.5
1
1.5
2
2.5
Time (seconds after 9/28/2004 17:15:27.4)
Stress (MPa)
σ
xx
σ
yy
σ
xy
Figure 3-5: Stress tensor calculated from displacements recorded at stations P05, P06, and P07.
81
Figure 3-6 shows the normal, shear and coulomb stresses on a fault oriented in
the EW direction. In these figures we have assumed a lithostatic stress on the fault
corresponding to a depth of 10m. The Coulomb stress peaks around 8 seconds and is
caused by a combination of large shear stress and tension.
82
−0.5
0
0.5
0
0.2
0.4
0 2 4 6 8 10 12
−0.5
0
0.5
Time (seconds after 9/28/2004 17:15:27.4)
Stress (MPa)
Normal Stress
Shear Stress
Coulomb Stress
Figure 3-6: Normal, Shear, and Coulomb stresses on a plane oriented in the EW direction as determined
by the displacement field measured by stations P05, P06, and P07 during the 2004 Parkfield Earthquake.
Times of tensile stress are indicated in grey on the shear stress and coulomb stress plots.
83
The stress field σ’ (equation 5) was calculated for fault orientation ranging from
N to S in 5
o
increments. The maximum coulomb stress on any of these planes is plotted
as a function of time in Figure 3-7 along with the temporal distribution of burst
production. The peak in coulomb stress corresponds to the peak in burst production.
84
0 2 4 6 8 10 12
0
0.5
0 2 4 6 8 10 12
0
10
0 2 4 6 8 10 12
0
2
4
6
Time (seconds)
Coulomb Stress (MPa)
BPS (#bursts/sec)
BPS
coulomb
stress
Figure 3-7: Maximum Coulomb stress on any plane. Grey shows times when the plane of maximum
Coulomb stress was under tension. Plotted over the coulomb stress are the number of bursts observed
across the array in 1 second time bins. The peak of the burst production rate corresponds well to the peak
in coulomb stress.
85
3-2.4 Spectral Boundary Integral Model of Driven Cracks
We use a 2D boundary integral crack model to test the hypothesis that the
reduction in normal stress in the presence of shear stress generated by the strong motion
can drive a dynamic rupture capable of radiating seismic energy. The Boundary Integral
Method has been used widely to study the spontaneous propagation of cracks in elastic
materials [Das and Aki, 1977; Andrews 1985; Geubelle and Rice, 1995]. The dynamic
rupture calculations presented here are performed using MDSBI, the spectral boundary
integral code developed by Dunham [2005] and Noda et al. [2008]. MDSBI
accommodates a number of fault plane rheologies including velocity weakening friction
and rate-and-state friction. For our simulation we chose the following rate-and-state
friction parameters: A=0.001, B=0.005, Dc =5mm. Since B>A, friction is velocity
weakening. We control the crack length by setting the reference coefficient of friction to
an unrealistically high value everywhere in the model except for the crack which has a
value of 0.6. The length of a slip patch that we can model is limited by the value of the
critical displacement D
c,
which places an upper limit on the grid spacing Δx. For the
calculation to produce physical results an additional condition must be met which links
the time step Δt to the grid size Δx
V
s
Δt
Δx
<1 (7)
where V
s
is the shear wave velocity. Therefore, very small grid sizes require very small
time steps. For D
c
=5mm, we chose a grid spacing of 1mm, which for V
s
=326 m/s
requires approximately 100,000 timesteps to model 1 second of real time. This problem
is similar to that encountered when modeling an earthquake cycle where long periods of
86
stable sliding are computed with large time steps which are then refined when dynamic
rupture starts [Lapusta et. al, 2000]. Similarly, we run fully dynamic simulations only
during those parts of the strong motion cycle when a slip instability begins.
The entire model domain is first loaded with 1x10
4
Pa shear stress and 0.27x10
6
Pa normal stress, corresponding roughly to the minimum dynamic shear stress found to
trigger a burst event at Parkfield and the lithostatic stress at 10m depth, respectively.
The normal stress in the entire model domain is then reduced following a 1Hz sinusoid
with an amplitude just below the lithostatic stress so that at maximum dynamic tension
the fault approaches 0 normal stress. Figure 3-8 shows slip velocities and accelerations
for a range of crack lengths. The results for both peak velocity and peak acceleration
show a strong dependency on crack length. While we have not modeled the larger
meter scale crack sizes using rate-and-state friction due to the limitations described
above we have included some simulations using a slip weakening friction law. Slip
weakening friction does not require extremely fine resolution grids so that larger scale
problems can be modeled very quickly.
87
10
−3
10
−2
10
−1
10
0
10
1
10
2
10
−6
10
−5
10
−4
10
−3
10
−2
10
−8
10
−7
10
−6
10
−5
10
−4
Fault Length (m)
Peak Acceleration (m/s
-2
)
0 1 2 3
−1
0
1
2
3
4
5
6
x 10
−7
Time (milliseconds)
Acceleration (m/s
2
)
0.005m
0.01m
0.1m
0.5m
(c)
0 1 2 3
0
1
2
3
4
5
6
x 10
−4
Time (milliseconds)
Slip Velocity (m/s)
0.005m
0.01m
0.1m
0.5m
Fault Length (m)
Peak Velocity (m/s)
(a) (b)
Rate and State
Slip Weakening
Rate and State
Slip Weakening
10
−3
10
−2
10
−1
10
0
10
1
10
2
Figure 3-8: Velocities and accelerations produced by the crack model. Panel (a) shows slip velocity as a
function of time for range of crack lengths. Panel (b) shows the scaling of peak velocity with fault length.
Panel (c) shows slip acceleration as a function of time for a range of crack lengths. Panel (d) shows the
scaling of peak acceleration with fault length. Parameter values: A=0.001, B=0.005, D
c
=5mm; grid
spacing, Δx=1mm, μ=0.3GPa, V
s
=326m/s, V
p
/V
s
= 3 ; initial shear stress τ
0
=0.01MPa; initial normal
stress, σ
0
=-0.27MPa.
88
The results generally show increasing maximum slip velocities and maximum
accelerations with increasing fault length. The peak velocities found by rate and state
versus slip weakening agree for the 0.2m fault. It is interesting to note that while the
maximum accelerations for rate-and-state faults increases with fault length, the slip
weakening cases show constant acceleration. With no rate controls on the fault the only
limit on acceleration are inertial effects.
This model represents a study in the plausibility of unstable slip occurring on a
fault in the very shallow subsurface, which we have shown may occur for faults of very
small length. The slip found here is unstable and accelerates on a time scale much
smaller than the loading (Figure 3-9). Since we have not calculated the entire slip
distribution due to intense computational demand, we assume that the slip distribution is
roughly triangular. Slip reaches maximum velocity on the order of milliseconds and the
entire event lasts on the order of tenths of a second. It is clear that the modeling
produces unstable slip events.
89
−0.3
−0.2
−0.1
0
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45 0.5
0
200
400
600
Time (s)
Normal Stress (MPa) Slip Velocity (microns/s)
4
1
2
3
4 Δt 0.03s, 2a = 0.005m
1 Δt 0.05s, 2a = 0.5m
2 Δt 0.03s, 2a = 0.1m
3 Δt 0.03s, 2a = 0.01m
Figure 3-9: Comparison between the 1 Hz strong motion reduction in normal stress above and the slip
instabilities below. The model results for slip velocity (bottom panel) give large accelerations over the
time scale of a few milliseconds. We assume triangular slip distributions based on the peak velocity from
the model which is reached in milliseconds. The area under each triangle is determined by the total slip
given by the analytical solution of a stress free crack in an elastic solid. When compared with the time
scale of loading, the crack accelerates instantaneously so our assumption that shear stress is constant over
the time scale of crack acceleration is valid.
90
3-2.5 Attenuation of strong motion by bursts
We have shown that 1 Hz strong motion in the upper 70 meters of the crust can
drive sliding instabilities on fractures and faults that radiate bursts of energy into higher
frequency bands. This reradiation of strong motion energy can be viewed as an
attenuation mechanism and parameterized by a quality factor Q defined as
s
b
E
E
Q
π 2
Δ
=
(8)
where ΔE
b
is energy density per cycle of the seismic wave converted by the bursts and
Es is the peak energy density per cycle in the seismic wave. To calculate Q, we first
estimate ΔE
b
and Es.
3-2.6 Energy of bursts
The total elastic energy released by a burst ΔE
b
is partitioned into seismic
radiation E
r
, surface energy E
s
, and frictional dissipation E
f
as
ΔE
B
= E
r
+ E
s
+ E
f
(9)
While E
r
can be measured seismologically, the other two energies are more difficult to
determine. Their contribution is included by multiplying the seismic energy by an
efficiency factor η
E
r
= ηE
b
= ησ
M
o
μ
(10)
91
Where ησ is the apparent stress [Wyss and Brune, 1968], M
o
the seismic moment, and μ
the shear modulus. The apparent stress can be shown to equal ½ the stress drop
Δσ during the earthquake
σ ησ Δ =
2
1
(11)
Therefore, if we can estimate the seismic moment and the stress drop, the total energy
released by a burst can be calculated.
For a discrete ensemble of bursts in a volume V
b
, the energy loss in time interval
τ is found by summing the energy E
i
of the N(τ) individual bursts
E
b
= E
i
i=1
N(τ )
∑
(12)
The energy E
i
of an individual burst can be found by combining equations (10) and (11)
μ
σ
2
0i
i
M
E
Δ
=
(13)
For comparison with seismic energy associated with the strong motion, we define the
burst energy density,
E
b
*
, per cycle
E
b
*
=
E
b
V
b
(τf )
(14)
Where V
b
is the volume in which the observed bursts originate, τ = 12s is the time
interval in which bursts occur, and f=2Hz is the predominant frequency of the seismic
waves at Parkfield.
92
We estimate
E
b
*
from the observed distribution of burst magnitudes observed at
all 12 stations at UPSAR (Figure 3-10). The bursts were identified using an algorithm
developed by Fischer et. al. [2008b]. For an individual burst, all three components of
the raw seismogram on which the burst was identified were windowed for 0.2s centered
on the picked burst time. The windowed seismogram was integrated and the resulting
velocity record was then highpass filtered at 10Hz. The spectra were computed and the
mean amplitude at 10Hz for all three components was used to calculate the scalar
moment. Energy for all 136 events was calculated using equation (12) and (13). The
total energy of the bursts (equation 14) in a volume V
b
=4.2x10
5
m
3
(12 hemispheres of
radius 20m) ranges from 0.005 to 0.9 J/m
3
depending on the stress drop which ranged
from Δσ = 0.01 to 2 MPa. These assumed stress drop values were chosen because they
are the minimum and maximum values of stress measured which were determined to
trigger the burst events from Fischer et al., 2008b. All other bursts were triggered at
stresses somewhere between these two extremes, so we chose the maximum and
minimum values to represent limits on the energy released by the bursts while knowing
that the real value is somewhere inbetween.
93
−2.2 −2 −1.8 −1.6 −1.4 −1.2 −1 −0.8 −0.6 −0.4
10
0
10
1
10
2
Magnitude
Cumulative Number
b = 1
b-value -0.95 +/- 0.21 (MC LS)
b-value -2.38 +/- 0.19 (MC MLE)
b-value -1.80 +/-1.5 (TGR MLE)
Figure 3-10: Cumulative distribution of burst magnitudes (136 unique events). The distribution cannot be
fit with standard Gutenberg-Richter models. We attempted to fit the curve with least-squares (MC LS
[Zmap]), Maximum Likelihood (MC MLE [Zmap]) and a Tapered Gutenberg-Richter model (TGR MLE
[Kagan 2000]). All three fits are very poor. For comparison with an expected Gutenberg-Richter
distribution we have drawn a line of b=-1. Perhaps this unusual distribution of event sizes is further
evidence of the non-tectonic nature of the “driven” bursts.
94
3-2.7 Energy of seismic waves
An estimate of Q using equation 8 requires that the energy converted by the
burst process be compared with the seismic energy through the same volume. For a
harmonic elastic wave in a medium with density ρ, the peak strain energy density per
cycle, E
s
, depends on the square of the maximum particle velocity ν as:
2
ρν =
s
E
(18)
Since the dominant frequency in the strong motion is about 1-2 Hz (Fig. 3-1), we made
the seismograms more harmonic by narrow-band filtering the velocity records between
1-2 Hz. The result is a set of quasi-harmonic seismograms (3 components for each of
the 12 stations). The vector sum of the three components is formed at each station and
the peak velocities were picked during the time interval t=12 s. The peak velocities at
the 12 stations were averaged to find the average peak velocity ν = 0.16 m/s +/- 0.04
m/s which we used in eqn, (18) (with ρ = 2300 g/cm
3
) to calculate Es = 55.6 J/m
3
+/-
28.6 J/m
3
.
3-2.8 Effective attenuation
Our calculations show that Q may be as large as 7.5x10
4
+/- 3.9x10
4
and as low
as 394 +/- 202.4 (note that Q as stated here is the conventional Q
-1
). As mentioned
earlier in this work, Q in the very shallow crust may be as low as 1-10, so the burst
process represents at most approximately 3% of the total attenuation.
95
3-3. Discussion
The assumption that the high-frequency bursts observed in Taiwan and
California are examples of dynamic triggering by strong motion analogous to larger
events that are triggered by surface waves leads to a paradox. The problem is that
bursts appear to originate in the very shallow crust where rate- and state-dependent
friction predicts an unphysical large minimum radius for unstable slip. We propose that
this paradox may be resolved if the instability that produces the bursts involves a
simultaneous increase in shear stress and decrease in normal stress associated with the
strong motion. Since the shear energy comes from the strong motion, we use the term
dynamic driving as opposed to dynamic triggering which usually refers to the release of
stored tectonic stress. The requisite reduction in normal stress on a shallow fracture at a
time of maximum shear stress can be shown to occur at very shallow depths during the
mode conversion of the strong motion SV waves to P waves upon reflection at the
Earth’s surface. The strong motion observed during the 2004 Parkfield earthquake
produced tension across pre-existing fractures as deep as 70m. Numerical modeling of
a fracture governed by rate- and state-dependent friction subjected to a transient shear
load followed by a rapid unloading of the normal stress produces an unstable slip event
with seismic radiation that converts the relatively long period strong motion energy to
much higher frequency bursts. Since the normal stress becomes tensile it is reasonable
to assume that unstable slip will occur regardless of the frictional rheology. It is
interesting to note that our results show that failure occurs well before tension is
reached. Failure does occur at a level consistent with simple static/dynamic frictional
96
considerations, suggesting that the rate and state effects are not important under these
conditions.
We have considered the energy conversion process under some reasonable
assumptions to show the dynamic driving of high frequency bursts produce an effective
attenuation of relatively low frequency strong motion waves. The total energy must be
conserved but the conversion to higher frequency energy can be important for seismic
hazard for two reasons: 1) High frequency energy attenuates faster as it propagates
through the earth and 2) Structural resonance is less likely to occur for waves with
frequencies above 20 Hz compared to waves that oscillate at 1-2 Hz. It has long been
hypothesized that attenuation through anelastic effects, sometimes referred to as internal
friction, is due to microscopic mechanisms such as stress-induced migration of defects
in minerals, vibration of dislocations, and frictional sliding on crystal grain boundaries
[Stein and Wysession, 2003]. The effective attenuation due to the high-frequency burst
mechanism may play a similar but macroscopic role in reducing the amplitude of strong
motion. Our calculations show a small, but observable attenuation due to dynamically
driven burst events in the very near the surface.
The energy taken out of the strong motion by dynamic driving depends on a
number of parameters which we have constrained through observation including but not
limited to the assumed stress drop and the frequency-magnitude distribution of events.
It is very likely that some larger events were missed and not included due to the
extremely limited spatial resolution of the UPSAR array which only captures strong
motion over an area roughly 1 Km
2
. It is possible to get a smaller value Q than
97
determined here if there are a few larger events which contain much more energy that
may not have been observed in the UPSAR region. For example, Ma and Sleep [2008]
hypothesize that a large burst-like event occurred in the fault zone during the 2004
Parkfield earthquake. The magnitude of this event is roughly 0.8, which is an entire
magnitude larger than the events we see at UPSAR.
98
Chapter 3 References
Aki K. Richards P. G. (1980). Quantitative Seismology, Volume I, W. H. Freeman, San
Francisco
Andrews D. J. (1985). Dynamic plane-strain shear rupture with a slip-weakening
friction law calculated by a boundary integral method, Bull. Seism. Soc. Am. 75, 1-21.
Aster R. C. and P. M. Shearer (1991), High-frequency borehole seismograms recorded
in the San Jacinto fault zone, southern California. Part 2. Attenuation and site effects,
Bull. Seismol. Soc. Am., 81, 1081–1100.
Blakeslee S. and P. Malin (1991), High-frequency site effects at two Parkfield
downhole and surface stations, Bull. Seismol. Soc. Am., 81, 332–345.
Chen, Y., C. G. Sammis, and T.-L. Teng (2006), A high frequency view of the 1999
Chi-Chi, Taiwan, Source Rupture and Fault Mechanics, Bull. Seismol. Soc. Am., 96,
807–820.
Das S. Aki K. (1977a). A numerical study of two-dimensional spontaneous rupture
propagation, Geophys. J. 50, 643-668.
Dieterich J. H. (1986) A Model for the nucleation of earthquake slip. Geophsyical
Monograph 37 (AGU)
Dieterich J. H. (1987) Nucleation and triggering of earthquake slip: effect of periodic
stresses. Tectonophysics, 144, 127-139
Dunham, E. M. (2005), Dissipative interface waves and the transient response of a three
dimensional sliding interface with Coulomb friction, Journal of Mechanics and Physics
of Solids, 53(2), 327-357, doi:10.1016/j.jmps.2004.07.003.
Ebel, J. E. (1989), The effect of crustal scattering on observed high-frequency
earthquake seismograms, Geophy. J. Int., 98, 329–341.
Fischer A. D., C. G. Sammis, Y. Chen, and T. L. Teng (2008a), Dynamic Triggering by
Strong Motion P- and S-waves: Evidence from the 1999 Chi-Chi, Taiwan Earthquake,
Bull. Seism. Soc. Am., 98, 580–592.
Fischer A. D., Z. Peng, C. G. Sammis (2008b), Dynamic triggering of high-frequency
bursts by strong motions during the 2004 Parkfield earthquake sequence, Geophys. Res.
Lett., vol. 35, L12305, doi:10.1029/2008GL033905
99
Fletcher, J. P, P. Spudich, P. Goldstein, J. Sims, and M. Hellweg (1992), The USGS
Parkfield, California dense seismograph array – UPSAR, Bull. Seismol. Soc. Am., 82,
1041–1070.
Fletcher, J. P, P. Spudich, and L. M. Baker (2006), Rupture Propagation of the 2004
Parkfield, California, Earthquake from Observations at the UPSAR, Bull. Seismol. Soc.
Am., 96, S129–S142; doi: 10.1785/01200050812.
Gomberg, J., J. L. Rubensteing, Z. Peng, K. C. Creager, J. E. Vidale, and P. Bodin
(2008), Widespread Triggering of Nonvolcanic Tremor in California, Science, DOI:
10.1126/science.1149164.
Geubelle P. H. Rice J. R. (1995). A spectral method for three-dimensional
elastodynamic fracture problems, J. Mech. Phys. Solids 43, 1791-1824.
Hill D. P. Reasenberg P. A. Michael A. Arabasz W. J. Beroza G. Brune J. N.
Brumbaugh D. Castro R. Davis S. dePolo D. Ellsworth W. L. Gomberg J. Harmsen S.
House L. Jackson S. M. Johnston M. Jones L. Keller R. Malone S. Munguia L. Nava S.
Pechmann J. C. Sanford A. Simpson R. W. Smith R. S. Stark M. Stickney M. Vidal A.
Walter S. Wong V. Zollweg J. (1993). Seismicity in the western United States remotely
triggered by the M 7.4 Landers, California, earthquake of June 28, 1992, Science 260,
1617-1623.Hill, D.P. and S.G. Prejean (2007), Dynamic Triggering, in Treatise on
Geophysics (ed. G. Schubert), v. 4, Earthquake Seismology (ed. H. Kanamori), 257–
292, Elsevier, Amsterdam.
Lapusta, N., J. R. Rice, Y. Ben-Zion, and G. Zheng (2000). Elastodynamic analysis for
slow tectonic loading with spontaneous rupture episodes on faults with rate-and state-
dependent friction, J. Geophys. Res. 105, 23,765-23,789.
Miyazawa, M., and E. E. Brodsky (2008), Deep low-frequency tremor that correlates
with passing surface waves, J. Geophys. Res., 113, B01307,
doi:10.1029/2006JB004890.
Noda, H., E. M. Dunham, and J. R. Rice (2008), Earthquake ruptures with thermal
weakening and the operation of faults at low overall stress levels, in preparation.
Peng, Z., and Y. Ben-Zion (2006), Temporal changes of shallow seismic velocity
around the Karadere-Duzce branch of the north Anatolian fault and strong ground
motion, Pure Appl. Geophys., 163, 567–599.
100
Prejean, S. G., D. P. Hill, E. E. Brodsky, S. E. Hough, M. J. S. Johnston, S. D. Malone,
D. H. Oppenheimer, A. M. Pitt, K. B. Richards-Dinger (2004), Remotely triggered
seismicity on the United States west coast following the Mw 7.9 Denali Fault
earthquake The 2002 Denali Fault earthquake sequence, Bull. Seismol. Soc. Am., 94,
348–359.
Sleep, N. H. and S. Ma (2008), Production of brief extreme ground acceleration pulses
by nonlinear mechanisms in the shallow subsurface, Geochem. Geophys. Geosyst., 9,
Q03008, doi:10.1029/2007GC001863.
Spudich P. Steck L. K. Hellweg M. Fletcher J. B. Baker L. M. (1995). Transient stresses
at Parkfield, California, produced by the M 7.4 Landers earthquake of June 28, 1992:
observations from the UPSAR dense seismographic array, J. Geophys. Res. 100, 675-
690.
Stein S. and M. Wysession (2003), An Introduction to Seismology, Earthquakes, and
Earth Structure, Blackwell, Malden, MA.
Wilson, D. C. and G. L. Pavlis, (2000), Near-surface site effects in crystalline bedrock:
a comprehensive analysis of spectral amplitudes determined from a dense, three-
component seismic array, Earth Interactions, 4, 1–31.
Wyss M. Brune J. N. (1968). Seismic moment, stress and source dimensions for
earthquakes in the California-Nevada region, J. Geophys. Res. 73, 4681-4694.
101
Summary of Results
• High-frequency bursts triggering by strong motion are a general phenomenon.
Observations of bursts have been made in two different regions, Taiwan and
Parkfield, CA.
• High-frequency bursts are a very shallow crustal phenomenon. Observations
from Taiwan and Parkfield show the ubiquitous nature of burst triggering in this
highly weathered upper layer of crust and constrain their sources to be within
20m of the recording seismometers.
• The triggering threshold for production of bursts is approximately 0.02 MPa.
This value is consistent at both Taiwan and Parkfield and is within the range of
triggering thresholds observed for dynamic triggering by surface waves (Figure
S-1).
• Bursts are not generated by instrumental noise or signal processing artifacts.
• Simultaneous shear loading and normal stress reduction can produce instabilities
in the very shallow crust where bursts are produced. The combination of shear
and normal stress is a direct result of mode conversion of P and SV waves at the
free surface of the earth.
• The burst process can be viewed as an attenuation mechanism and may be
responsible for up to 3% of the total wave attenuation in the very shallow crust.
• Bursts are numerous and their magnitudes is very small. The magnitudes
measured in Taiwan vary from Mw-0.28 to Mw0.7 and Mw-1.8 to Mw-0.8 in
102
Parkfield. This is consistent with the observation that bursts are uncorrelated on
multiple stations.
• Bursts can be observed at relatively large distances from the source of strong
motion waves that trigger burst production. Bursts were observed 10km from
the fault trace in Parkfield and up to 170km from the epicenter in Taiwan.
103
0.001
0.01
0.1
1
0.1 1 10 100 1000 10
4
10
5
Taiwan (P-wave)
Taiwan (S-wave)
Brodsky et al. (2000)
Gomberg et al. (2004)
Gomberg et al. (2001)
Hill et al. (1993)
Kilb et al., (2000)
Miyazawa & Mori (2006) (dilational strain)
Prejean et al. (2004)
Pankow et al., (2004)
Cochran et al. (2004)
Wilcock (2001)
Period (s)
Triggering Stress (MPa)
Static Triggering
Freed (2005)
Parkfield
Figure S-1: Same as figure 1-12 but includes results from UPSAR recordings of 2004 Parkfield, CA
earthquake.
104
Dominant
Frequency
Seismic
waves (Hz)
Triggering
Stress
Threshold
(MPa)
Maximum
distance from
source to
receiver (m)
Burst
Magnitudes
(Mw)
Taiwan Strong
Motion
1-2 0.03 to 0.05 745 to 2900 -0.28 to 0.7
Parkfield Strong
Motion
1-2 0.02 20 -1.8 to -0.8
Surface Waves 0.02 to 0.06 0.0033 to 0.6 N/A <6
Tides 2.3x10
-5
0.005 to 0.02 N/A <5
Table 2: – Summary of some important results from observation of high-frequency bursts at Taiwan and
Parkfield. Also given are results from literature for surface wave triggering.
105
References
Abercrombie, R.E. (1995). Earthquake source scaling relationships from
-1 to 5 ML using seismograms recorded at 2.5 km depth, J. Geophys. Res.,
100,24015-24036
Aki K. Richards P. G. (1980). Quantitative Seismology, Volume I, W. H.
Freeman, San Francisco
Andrews D. J. (1985). Dynamic plane-strain shear rupture with a
slip-weakening friction law calculated by a boundary integral method,
Bull. Seism. Soc. Am. 75, 1-21.
Aster, R. C., and P. M. Shearer (1991b). High-frequency borehole
seismograms recorded in the San Jacinto fault zone, southern California.
Part 2. Attenuation and site effects, Bull. Seism. Soc. Am., 81, 1081-1100
Blakeslee S. and P. Malin (1991), High-frequency site effects at two
Parkfield downhole and surface stations, Bull. Seismol. Soc. Am., 81,
332–345.
Bonilla, L. F., R. J. Archeleta, and D. Lavellee (2005), Hysteretic and
dilatant behavior of cohesionless soils and their effects on nonlinear
site response: field data observations and modeling, Bull. Seism. Soc.
Am. 95, 2373 –2395.
Brodsky, E. E., and S. G. Prejean (2005). New constraints on mechanisms
of remotely triggered seismicity at Long Valley Caldera, J. Geophys.
Res., 110, B04302, doi:10.1029/2004JB003211.
Brodsky, E. E., V. Karakostas, and H. Kanamori (2000). A new observation
of dynamically triggered regional seismicity: Earthquakes in Greece
following the August, 1999 Izmit, Turkey earthquake, Geophys. Res.
Lett., 27, 2741–2744
Brune, J. (1970). Tectonic stress and the spectra of seismic shear waves
from earthquakes, J. Geophys. Res., 75, 4997-5009.
Chen, Y., C. G. Sammis, and T.-L. Teng (2006). A High-Frequency View of
the 1999 Chi-Chi, Taiwan, Source Rupture and Fault Mechanics, Bull.
Seism. Soc. Am., 96, 807-820.
106
Cochran, E. S., J. E. Vidale, and S. Tanaka (2004). Earth tides can
trigger shallow thrust fault earthquakes, Science 306, no. 5699
(20041112): 1164-1166.
Cranswick, E. (1988), The information content of high-frequency
seismograms and the near-surface geologic structure of "hard rock"
recording sites, Pure Appl. Geophys., 128, 333–363.
Das S. Aki K. (1977a). A numerical study of two-dimensional spontaneous
rupture propagation, Geophys. J. 50, 643-668.
Dieterich J. H. (1986) A Model for the nucleation of earthquake slip.
Geophsyical Monograph 37 (AGU)
Dieterich, J. H. (1987). Nucleation and triggering of earthquake slip:
effect of periodic stresses, Tectonophysics, 144, 127-139.
Dieterich, J. H., (1994). A constitutive law for rate of earthquake
production and its application to earthquake clustering, J. Geophys.
Res., 99, 2601-2618.
Dunham, E. M. (2005), Dissipative interface waves and the transient
response of a three dimensional sliding interface with Coulomb friction,
Journal of Mechanics and Physics of Solids, 53(2), 327-357,
doi:10.1016/j.jmps.2004.07.003.
Ebel, J. E. (1989), The effect of crustal scattering on observed
high-frequency earthquake seismograms, Geophy. J. Int., 98, 329–341.
Felzer, K. R., R. E. Abercrombie, and E. E. Brodsky (2005). Testing the
stress shadow hypothesis, J. Geophys. Res. 110, doi:10.1029/2004JB003277.
Fischer A. D., C. G. Sammis, Y. Chen, and T. L. Teng (2008a), Dynamic
Triggering by Strong Motion P- and S-waves: Evidence from the 1999
Chi-Chi, Taiwan Earthquake, Bull. Seism. Soc. Am., 98, 580–592.
Fischer A. D., Z. Peng, C. G. Sammis (2008b), Dynamic triggering of
high-frequency bursts by strong motions during the 2004 Parkfield
earthquake sequence, Geophys. Res. Lett., vol. 35, L12305,
doi:10.1029/2008GL033905
107
Fletcher, J. P, P. Spudich, and L. M. Baker (2006), Rupture Propagation
of the 2004 Parkfield, California, Earthquake from Observations at the
UPSAR, Bull. Seismol. Soc. Am., 96, S129–S142; doi: 10.1785/01200050812.
Fletcher, J. P, P. Spudich, P. Goldstein, J. Sims, and M. Hellweg
(1992), The USGS Parkfield, California dense seismograph array – UPSAR,
Bull. Seismol. Soc. Am., 82, 1041–1070.
Frankel, A. D., D. L. Carver, and R. A. Williams (2002). Nonlinear and
linear site response and basin effects in Seattle for the M6.8
Nisqually, Washington, Earthquake, Bull. Seismol. Soc. Am. 92, 2090–2109.
Freed, A. (2005), Earthquake triggering by static, dynamic, and
postseismic stress transfer, Annu. Rev. Earth Planet. Sci., 33, 335–368.
Geubelle P. H. Rice J. R. (1995). A spectral method for
three-dimensional elastodynamic fracture problems, J. Mech. Phys. Solids
43, 1791-1824.
Gomberg, J. S., P. A. Reasenberg, P. Bodin, R. A. Harris (2001).
Earthquake triggering by seismic waves following the Landers and Hector
Mine earthquakes, Nature 411, no. 6836 (20010524): 462-466
Gomberg, J., J. L. Rubensteing, Z. Peng, K. C. Creager, J. E. Vidale,
and P. Bodin (2008), Widespread Triggering of Nonvolcanic Tremor in
California, Science, DOI: 10.1126/science.1149164.
Gomberg, J., P. Bodin, K. Larson, and H. Dragert (2004). Earthquake
nucleation by transient deformations caused by the M = 7.9 Denali,
Alaska, earthquake, Nature, 427, 621–624
Hanks T. C. (1982). fmax, Bull. Seismol. Soc. Am., 72, 1867–1880.
Harris, R. A. and R. W. Simpson (1996). In the shadow of 1857 – the
effect of the great Ft. Tejon earthquake on subsequent earthquakes in
southern California, Geophys. Res. Lett., 23, 229-232.
108
Hill D. P. Reasenberg P. A. Michael A. Arabasz W. J. Beroza G. Brune J.
N. Brumbaugh D. Castro R. Davis S. dePolo D. Ellsworth W. L. Gomberg J.
Harmsen S. House L. Jackson S. M. Johnston M. Jones L. Keller R. Malone
S. Munguia L. Nava S. Pechmann J. C. Sanford A. Simpson R. W. Smith R.
S. Stark M. Stickney M. Vidal A. Walter S. Wong V. Zollweg J. (1993).
Seismicity in the western United States remotely triggered by the M 7.4
Landers, California, earthquake of June 28, 1992, Science 260,
1617-1623.Hill, D.P. and S.G. Prejean (2007), Dynamic Triggering, in
Treatise on Geophysics (ed. G. Schubert), v. 4, Earthquake Seismology
(ed. H. Kanamori), 257–292, Elsevier, Amsterdam.
Hill, D.P. and S.G. Prejean (2007), Dynamic Triggering, in Treatise on
Geophysics (ed. G. Schubert), v. 4, Earthquake Seismology (ed. H.
Kanamori), 257–292, Elsevier, Amsterdam.
Hiramatsu, Y., H. Yamanaka, K. Tadokoro, K. Nishigami, and S. Ohmi
(2002). Scaling law between corner frequency and seismic moment of
microearthquakes: Is the breakdown of the cube law a nature of
earthquakes?, Geophys. Res. Lett., 29 (8),10.1029/2001GL013849.
Holzer, T. L., T. L. Youd, and T. C. Hanks (1989). Dynamics of
liquefaction during 1987 Superstition Hills, California, earthquake,
Science 244, 56 –59.
Kilb, D. L., J. Gomberg, and P. Bodin (2000). Triggering of earthquake
aftershocks by dynamic stresses, Nature, 408, 570–574.
King G. C., R. S. Stein, and J. Lin (1994). Static stress changes and
the triggering of earthquakes, Bull. Seism. Soc. Am., 84, 935-953.
Langbein, J., R. Borcherdt, D. Dreger, J. Fletcher, J. L. Hardebeck, M.
Hellweg, C. Ji, M. Johnston, J. R. Murray, R. Nadeau, M. J. Rymer, and
J. A. Treiman (2005), Preliminary report on the 28 September 2004, M6.0
Parkfield, California earthquake, Seismol. Res. Lett., 76, 10–26.
Lapusta, N., J. R. Rice, Y. Ben-Zion, and G. Zheng (2000). Elastodynamic
analysis for slow tectonic loading with spontaneous rupture episodes on
faults with rate-and state-dependent friction, J. Geophys. Res. 105,
23,765-23,789.
Liu Y., Teng T-L. Teng, and Y. Ben-Zion (2005). Near-surface seismic
anisotropy, attenuation and dispersion in the aftershock region of the
1999 Chi-Chi earthquake, Geophysical Journal International, 160 (2),
695–706. doi:10.1111/j.1365-246X.2005.02512.
109
Ma, K.-F., J. Mori, S.-J. Lee, and S. B. Yu (2001). Spatial and temporal
distribution of slip for the 1999 Chi-Chi, Taiwan, earthquake, Bull.
Seism. Soc. Am., 91, 1069 –1087.
Ma, K.F., C. H. Chan, and R. S. Stein (2005). Response of seismicity to
Coulomb stress triggers and shadows of the 1999 Mw = 7.6 Chi-Chi,
Taiwan, earthquake, J. Geophys. Res., 110, doi:10.1029/2004JB003389.
Miyazawa, M., and E. E. Brodsky (2008), Deep low-frequency tremor that
correlates with passing surface waves, J. Geophys. Res., 113, B01307,
doi:10.1029/2006JB004890.
Miyazawa, M., and J. Mori (2006). Evidence suggesting fluid flow beneath
Japan due to periodic seismic triggering from the 2004 Sumatra-Andaman
earthquake, Geophys. Res. Lett., 33, L05303, doi:10.1029/2005GL025087.
Noda, H., E. M. Dunham, and J. R. Rice (2008), Earthquake ruptures with
thermal weakening and the operation of faults at low overall stress
levels, Submitted to J. Geophys Res.
Obara, K. (2002), Nonvolcanic deep tremor associated with subduction in
southwest Japan, Science, 296, 1699–1681.
Pankow, K. L., W. J. Arabasz, J. C. Pechmann, and S. J. Nava (2004),
Triggered seismicity in Utah from the November 3, 2002, Denali fault
earthquake, Bull. Seism. Soc. Am., 94, S332–S347.
Peng, Z., and Y. Ben-Zion (2006), Temporal changes of shallow seismic
velocity around the Karadere-Duzce branch of the north Anatolian fault
–599.−and strong ground motion, Pure Appl. Geophys., 163, 567
Peng, Z., J. E. Vidale, and H. Houston (2006), Anomalous early
aftershock decay rates of the 2004 M6 Parkfield earthquake, Geophys.
Res. Lett., 33, L17307, doi:10.1029/2006GL026744.
Prejean, S. G., D. P. Hill, E. E. Brodsky, S. E. Hough, M. J. S.
Johnston, S. D. Malone, D. H. Oppenheimer, A. M. Pitt, K. B.
Richards-Dinger (2004), Remotely triggered seismicity on the United
States west coast following the Mw 7.9 Denali Fault earthquake The 2002
Denali Fault earthquake sequence, Bull. Seismol. Soc. Am., 94, 348–359.
Richardson, E. and T.H. Jordan (2002). Seismicity in deep gold mines of
South Africa: Implications for tectonic earthquakes, Bull. Seism. Soc.
Am., 92, 1766-1782.
110
Rubinstein, J. L. and G. C. Beroza (2005), Depth constraints on
nonlinear strong ground motion from the 2004 Parkfield earthquake,
Geophys. Res. Lett., 32, L14313, doi: 10.1029/2005GL023189.
Shearer, P. M., and J. A. Orcutt (1987), Surface and near-surface
effects on seismic waves–theory and borehole seismometer results, Bull.
Seismol. Soc. Am., 77, 1168–1196.
Sleep, N. H. and S. Ma (2008), Production of brief extreme ground
acceleration pulses by nonlinear mechanisms in the shallow subsurface,
Geochem. Geophys. Geosyst., 9, Q03008, doi:10.1029/2007GC001863.
Spudich P. Steck L. K. Hellweg M. Fletcher J. B. Baker L. M. (1995).
Transient stresses at Parkfield, California, produced by the M 7.4
Landers earthquake of June 28, 1992: observations from the UPSAR dense
seismographic array, J. Geophys. Res. 100, 675-690.
Stein S. and M. Wysession (2003), An Introduction to Seismology,
Earthquakes, and Earth Structure, Blackwell, Malden, MA.
Vernon F. L., J. Fletcher, L. Carroll, A. Chave, and E. Sembera (1991).
Coherence of seismic body waves from local events as measured by a
small-aperature array, J. Geophys. Res. 96, 11981-11996.
Vernon, F. L., G. L. Pavlis, T. J. Owens, D. E. McNamara, and P. N.
Anderson (1998). Near-surface scattering effects observed with a
high-frequency phased array at Pinyon Flats, California, Bull. Seism.
Soc. Am. 88, 1548 -1560.
Vernon, F. L., J. Fletcher, L. Carroll, A. Chave, and E. Sembrera
(1981), Coherence of Seismic Body Waves from Local Events as Measured by
a Small-Aperture Array, J. Geophys. Res. 96, 11981–11996.
Wang, G.-Q., D. M. Boore, H. Igel, and X.Y. Zhou (2003). Some
Observations on Colocated and Closely Spaced Strong Ground Motion
Records of the 1999 Chi-Chi, Taiwan, Earthquake, Bull. Seism. Soc. Am.
93, 674-693.
Wilock, W.S.D. (2001). Tidal triggering of microearthquakes on the Juan
de Fuca Ridge, Geophys. Res. Lett., 28, 3999-4002.
111
Wilson, D. C. and G. L. Pavlis, (2000), Near-surface site effects in
crystalline bedrock: a comprehensive analysis of spectral amplitudes
determined from a dense, three-component seismic array, Earth
Interactions, 4, 1–31.
Wyss M. Brune J. N. (1968). Seismic moment, stress and source dimensions
for earthquakes in the California-Nevada region, J. Geophys. Res. 73,
4681-4694.
Abstract (if available)
Abstract
A series of high-frequency (>20Hz) bursts of energy are observed on strong motion records during the 1999 Chi-Chi, Taiwan Earthquake Mw7.6. We hypothesized that these bursts originated near the individual stations as small, shallow events that were dynamically triggered by the P- and S-waves generated by the Chi-Chi mainshock. These bursts were originally interpreted as a mainshock source signal by Chen et al., [2006] but our observations of events on strong motion records recorded at stations up to 170 km from the mainshock epicenter is consistent with the local triggering hypothesis. If the bursts originated on the Chi-Chi fault plane, as hypothesized by Chen et al. [2006] based on their analysis of recordings within 20Km from the Chelungpu fault, then they should not be observable at this distance assuming any reasonable value of crustal attenuation. The bursts on all strong motion stations in the Taiwan Central Weather Bureau network (TWCB) were identified using a numerical algorithm approach. This data set was analyzed in the context of local dynamic triggering which resulted in a stress threshold for triggering in the range 0.03 to 0.05 MPa for S-wave triggering and 0.0013 to 0.0033 MPa for P-wave triggering, consistent with prior observations of surface wave triggering.
Linked assets
University of Southern California Dissertations and Theses
Conceptually similar
PDF
Dynamic rupture processes and seismic radiation in models of earthquake faults separating similar and dissimilar solids
PDF
Analysis of waveform and catalog data of aftershocks for properties of earthquakes and faults
PDF
Heterogeneity of earthquake stress drops, focal mechanisms and active fault zones
PDF
Statistical analyses of ambient seismic noise spectra with applications to detecting imperfectly diffuse wave field and deriving attenuation and phase velocity information
PDF
Multi-scale imaging and monitoring of crustal and fault zone structures in southern California
PDF
Applying automated techniques to large seismic datasets for systematic analyses of phases, source, and structure
PDF
High-resolution imaging and monitoring of fault zones and shallow structures: case studies in southern California and on Mars
PDF
Multi-scale imaging of the fault zone velocity structure: double-difference tomography, inversion of fault zone headwaves, and fault zone sensitivity kernels
PDF
Elements of seismic structures near major faults from the surface to the Moho
PDF
Stress-strain characterization of complex seismic sources
Asset Metadata
Creator
Fischer, Adam David
(author)
Core Title
Observations and modeling of dynamically triggered high frequency burst events
School
College of Letters, Arts and Sciences
Degree
Doctor of Philosophy
Degree Program
Geological Sciences
Publication Date
10/31/2008
Defense Date
10/17/2008
Publisher
University of Southern California
(original),
University of Southern California. Libraries
(digital)
Tag
attenuation,bursts,Chi-Chi,dynamic triggering,earthquake interaction,Earthquakes,high frequency,Modeling,OAI-PMH Harvest,observation,Parkfield,secondary radiation,seismic waves,seismology,Taiwan
Place Name
California
(states),
Chi-Chi
(city or populated place),
faults: Chelungpu fault
(geographic subject),
Parkfield
(city or populated place),
Taiwan
(countries)
Language
English
Contributor
Electronically uploaded by the author
(provenance)
Advisor
Sammis, Charles G. (
committee chair
), Becker, Thorsten W. (
committee member
), Ben-Zion, Yehuda (
committee member
), Nutt, Steven R. (
committee member
), Teng, Ta-Liang (
committee member
)
Creator Email
adfische@usc.edu,geophysicsman@gmail.com
Permanent Link (DOI)
https://doi.org/10.25549/usctheses-m1738
Unique identifier
UC1420094
Identifier
etd-Fischer-2506 (filename),usctheses-m40 (legacy collection record id),usctheses-c127-116807 (legacy record id),usctheses-m1738 (legacy record id)
Legacy Identifier
etd-Fischer-2506.pdf
Dmrecord
116807
Document Type
Dissertation
Rights
Fischer, Adam David
Type
texts
Source
University of Southern California
(contributing entity),
University of Southern California Dissertations and Theses
(collection)
Repository Name
Libraries, University of Southern California
Repository Location
Los Angeles, California
Repository Email
cisadmin@lib.usc.edu
Tags
attenuation
bursts
dynamic triggering
earthquake interaction
high frequency
observation
secondary radiation
seismic waves
seismology