Content uploaded by J. Y. Liu
Author content
All content in this area was uploaded by J. Y. Liu on Jan 22, 2014
Content may be subject to copyright.
545
TAO, Vol. 15, No. 3, 545-562, September 2004
Preliminary Results of the iSTEP Program on Integrated
Search for Taiwan Earthquake Precursors
Yi-Ben Tsai
1,
*, Jann-Yenq Liu
2,3
, Kuo-Fong Ma
1
, Horng-Yuan Yen
1
,
Kun-Shan Chen
3
, Yuh-Ing Chen
4
, and Chien-Ping Lee
1
(Manuscript received 12 May 2004, in final form 29 July 2004)
ABSTRACT
1
Institute of Geophysics, National Central University, Chung-Li, Taiwan, ROC
2
Institute of Space Science, National Central University, Chung-Li, Taiwan, ROC
3
Center for Space and Remote Sensing Research, National Central University, Chung-Li, Taiwan, ROC
4
Institute of Statistics, National Central University, Chung-Li, Taiwan, ROC
*
Corresponding author address:
Prof. Yi-Ben Tsai, Institute of Geophysics, Chung-Li, Taiwan, ROC;
E-mail: ybtsai@geps.gep.ncu.edu.tw
The predictability of earthquakes has been a hotly debated question in
earthquake science for some time. The answer to which begs another
question, “Are there credible earthquake precursors?” Intrigued by these
questions and encouraged by instrumental observations of conspicuous iono-
spheric and geomagnetic disturbances before the disastrous 1999 Chi-Chi
earthquake, we decided to undertake an integrated Search for Taiwan
Earthquake Precursors, called the iSTEP program. The multidisciplinary
program includes five major components aimed at identifying potential
seismological, geomagnetic, geodetic and ionospheric precursors, and sta-
tistical testing of any identified precursors. Since the program’s inception
in April 2002, some encouraging preliminary results have been obtained.
These includes precursory P wave travel-time changes six years before the
Chi-Chi earthquake, identifiable geomagnetic changes two years before
M
≥
6 earthquakes, consistent ionospheric anomalies three days before
M
≥
5 earthquakes. We have also developed high-resolution radar inter-
ferometric methods for monitoring crustal deformation. A method for fore-
casting aftershock distribution on the basis of stress transfer has also been
successfully tested on the Chi-Chi earthquake sequence. Aided by the newly
installed geomagnetic and ionospheric networks we are hopeful about iden-
tifying more earthquake precursors and developing relevant physical mecha-
nisms in the remaining two years of the iSTEP program.
(Key words: iSTEP, Precursor, Earthquake, Taiwan)
<Back to Index>
TAO, Vol. 15, No. 3, September 2004546
1. INTRODUCTION
In a region with high seismicity and dense population, Taiwan has suffered numerous
disastrous earthquakes in the past. The steady collision between the Philippine Sea and Eur-
asian plates means that Taiwan will inevitably face earthquake hazards in the future. The 1999
Chi-Chi earthquake, which took more than 2,500 lives and destroyed more than 100,000
dwellings, was only the latest case. For some natural hazards, such as typhoons, short-term
forecasting is an integral component of preparedness. In response, buildings can be secured,
equipment can be removed, emergency services can be put on alert, and populations can be
evacuated, if necessary. Unfortunately, earthquakes still cannot be forecast, much of this is
due to a lack of predictive capability.
The problem of earthquake prediction has been under intensive investigation for nearly
forty years in Japan, China, and the U.S. Yet relatively little progress has been made thus far.
Long frustration over the surprising difficult area of earthquake prediction has recently led to
serious debate within the seismological community regarding the predictability of earthquakes.
Some people argue that earthquakes are inherently unpredictable (e.g., Geller 1997). Yet oth-
ers believe that earthquake prediction is possible (e.g., Whiteside 1998). Our present inability
to predict earthquakes is partly due to inherent characteristics of earthquakes and seismic waves
and partly to incomplete understanding of the earthquake process (Silver 1998). At the heart of
the debate is a question of whether or not there are recognizable precursors of earthquake.
Many precursors have been mentioned, but it is not clear which, if any, are reliable. Cer-
tainly any reliable scheme for practical prediction must be based on a combination of clues, so
that firm decisions are possible before warnings are issued. Several promising clues have been
mentioned in the literature, such as detection of strain in the Earth’s crust by geodetic surveys,
the identification of suspicious gaps in the regular occurrence of earthquakes in both time and
space, and the observation of foreshocks. In recent years, major earthquake prediction re-
search has emphasized more precise measurements of fluctuations in physical parameters of
crustal rocks in seismic areas. According to the dilatancy-diffusion model (Scholz et al. 1973),
the five parameters thought to be relevant are: seismic P velocities, the uplift and tilt of the
ground, the emission of radon gas from water wells, electrical resistivity in rocks, and the
number of earthquakes in the region.
An earthquake prediction research program in Taiwan was initiated in 1977, jointly by
Y. B. Tsai of the Academia Sinica and T. L. Teng of the University of Southern California, and
funded separately by the National Science Council and the U.S. Geological Survey (Tsai et al.
1983). Under this earlier program all the above five parameters were observed. Some of the
parameter observations, such as radon emission were discontinued due to funding and man-
power constraints. Others were upgraded to more advanced methods, such as replacement of
traditional trilateration measurements of crustal deformation with the new global positioning
system (GPS) methods. Still there are other areas that have continued using original instru-
ments until this day for observing changes in geomagnetic intensity and seismicity.
Recently, three new developments have prompted us to undertake a new program to con-
duct integrated search for Taiwan earthquake precursors, called iSTEP. Fortuitously, clear
changes in geomagnetic intensity were observed at one of the original stations at Liyutan at
Tsai et al. 547
least one month before and after the 1999 Chi-Chi earthquake (Yen et al. 2004). This was the
first major new development which encouraged us to undertake the present iSTEP program.
The second major new development prompting us to undertake the present program was the
observation of clear electromagnetic precursors in the ionosphere, both by ionosondes and
GPS data several days before large earthquakes (Liu et al. 2000, 2001).
The third major new development was the physical process proposed recently (Freund et
al. 1999; Freund 2000; Freund et al. 2004) to explain electromagnetic signals associated with
earthquakes observed both on the ground and in the ionosphere. According to this process,
igneous rocks contain positive hole pairs (PHP), i.e., dormant electronic charge carriers. The
PHP can be activated by microfracturing and/or dislocation movement. Upon activation, the
PHP release highly mobile charge carriers in the form of positive holes. Positive holes are
defect electrons in the oxygen sublattice of minerals that can conduct through rocks. Once
generated these charges spread out from the source volume. Further understanding of these
positive hole charge carriers and their manifestation will enable us to reevaluate electrical
phenomena associated with earthquake activity.
Considering the above background, the iSTEP program is set to achieve the following
three objectives:
1. To identify any possible, yet reliable precursory phenomena of earthquakes, including
seismological, geodetic, geomagnetic, and ionospheric variations.
2. To characterize feasible mechanisms behind the identified precursory phenomena in
order to give quantitative physical interpretations.
3. To correlate the identified precursory phenomena with existing or emerging earthquake
preparation models in order to lay a foundation for future earthquake prediction.
2. MAJOR COMPONENTS OF THE iSTEP PROGRAM
Given the above scientific objectives and taking into account the strength of our present
research team, we decided to include the following five major components in the iSTEP pro-
gram (Fig. 1).
A. Potential Earthquake Precursors Based on Seismological Variations
We use the excellent earthquake catalogs and high-quality digital seismic waveform data
obtained by the Central Weather Bureau (CWB) to study the following topics.
1. Precursory changes in P wave velocities and waveforms. If rock properties change
before an earthquake, then the P wave velocity and waveforms may also vary. We will as-
semble the P wave residuals and spectral contents at each of the Central Weather Bureau
Seismic Network (CWBSN) stations as function of time to see whether there are clear changes
(i.e., delays) preceding large earthquakes of M
≥
6.0 in the nearby areas.
2. Variations in seismicity rate: Significant changes in normal background seismicity were
observed for several large earthquakes. Another method is to measure the changes in b value.
We will perform a complete search of all M
≥
6.0 earthquakes to identify possible seismicity
rate changes.
TAO, Vol. 15, No. 3, September 2004548
3. Changes in shear wave splitting: Shear wave splitting can be used to monitor the effects
of stress build-up before earthquakes, and hopefully to forecast future large earthquakes
(Zatsepin and Crampin 1997; Crampin and Zatsepin 1997; Crampin 1998). The goal of this
study is to systematically examine possible anisotropy changes before earthquakes and look
for possible earthquake precursors. Especially, we explore possible correlation of these changes
with the electromagnetic disturbances in ionosphere.
4. Stress transfer among large earthquakes and the physics on earthquake dynamic trig-
gering will be addressed. The correlation among these observations will be examined. Seis-
mological observations will further link with other non-seismological observations from other
iSTEP program components to find possible earthquake precursors and to look for feasible
physical mechanisms of these observations.
B. Potential Earthquake Precursors Based on Variations of Geomagnetic and Gravity
Fields
After the Chi-Chi earthquake, we examined data from a geomagnetic network of eight
stations operational since early 1980’s to provide continuous observations of total intensity
Fig. 1. The five major components of the iSTEP program.
Tsai et al. 549
over Taiwan. It was found that variations in the eight stations generally yield a similar tendency.
At the Lunping station, which served as the base-station of the geomagnetic network, no dis-
turbance was recorded before and after the 1999 Chi-Chi earthquake. However, during the two
months prior to the earthquake, data recorded at Liyutan geomagnetic station (24.2°N, 120.4°E),
which is only about six kilometers north of the northern end of the Chelungpu fault, fluctuated
significantly. It was surprising to find that these fluctuations disappeared on 22 October 1999,
when a strong earthquake M = 6.2 occurred near the southern end of this fault. These anoma-
lous variations, more than 150 gammas in amplitude, appeared to be associated with the earth-
quake (Yen et al. 2004). As far as we know, this type of geomagnetic change has never before
been reported in scientific literature. Accordingly, we include geomagnetic variations as part
of the iSTEP program.
We will first analyze the continuous geomagnetic data observed over the last 20 years at
the eight permanent stations. We will also upgrade the existing geomagnetic network (Fig. 2).
In addition to geomagnetic methods, repeated microgravity surveys along profiles across known
active faults will also be made to detect crustal deformation, subsurface material movement
and subsequent mass redistribution.
C. Potential Earthquake Precursors Based on Crustal Deformation Detected by Radar
Interferometry
The use of interferometric synthetic aperture radar (InSAR) for monitoring crustal defor-
mation processes has received considerable attention lately. Several studies suggested great
potential for the InSAR technique for mapping coseismic movement associated with strong
earthquakes (Massonnet et al. 1993; Zebker et al. 1994). Microwave frequency allows for
surface changes to be detected on a scale of centimeters. A number of studies have demon-
strated the potential and capability of satellite radar interferometry for mapping ground sur-
face deformation. For small-scale surface height movements, differential radar interferometry
is usually adopted.
The objectives of this iSTEP program component can be summed up as follows:
1. To integrate atmospheric and ionospheric observations from very high frequency (VHF)
and GPS into radar interferometry for path length calculations, and thus achieving higher
estimation accuracy of terrain height and height changes.
2. To detect surface movements and pre-, co-, and post-seismic crustal deformations.
3. To analyze and investigate the correlation of radar interferograms with the temporal
variations and spatial distributions of seismic electromagnetic phenomena in the ionosphere
and atmosphere.
4. To process and analyze the thermal infra-red (IR) satellite imagery as a precursor be-
fore an earthquake.
D. Potential Earthquake Precursors Based on Ionospheric Variations
Although seismic waves are the most obvious manifestation, earthquakes may be accom-
panied or preceded by signals of a different nature – electric, electromagnetic, or luminous –
TAO, Vol. 15, No. 3, September 2004550
that may help forecasting impending seismic activity (Uyeda 2004; Freund 2000). After the
disastrous 1995 Kobe earthquake in Japan, the Science and Technology Agency of Japan
called for the “Earthquake Integrated Frontier Research” project. It started in 1996 to study
extensively the whole spectrum of seismo-phenomena taking place in the atmosphere and
ionosphere/magnetosphere and then to investigate energy transfer from the lithosphere to the
atmosphere and ionosphere/magnetosphere (Uyeda 2004).
Fig. 2. Locations of new magnetometer and ionosonde stations deployed under
the iSTEP program.
Tsai et al. 551
We analyzed ionospheric foF2 and found three clear precursors appearing about 1, 3, and
4 days prior to the Chi-Chi earthquake. In this iSTEP program component, we will systemati-
cally search through ionospheric variations of all large earthquakes in Taiwan to see whether
there are persistent precursors. Both ionosonde and GPS data will be used (Fig. 2).
E. Statistical Studies of Potential Earthquake Precursors
Although extensive and reliable data on possible precursors of major earthquakes are
expensive to collect, and may take a long time to accumulate (Vere-Jones 1995), we have
observed some surprising anomalous fluctuations of the maximum plasma frequency and total
electron content (TEC) of the ionosphere which may possibly be related to strong earthquakes
in Taiwan area (Liu et al. 2000, 2001, 2004a). These anomalies reveal not only the temporal,
but also the spatial information about the earthquakes to follow. Therefore, according to the
empirical data we have obtained so far, we will develop appropriate statistical process control
(SPC) techniques (Stoumbos et al. 2000) to identify the variations of the relevant measure-
ments in electromagnetic fields.
In order to validate the reliability of precursory parameters, the Monte Carlo simulations
(Musson 1997; Stark 1997) of some reasonable prediction processes based on either the occur-
rence of precursors or previous earthquake catalogs will be compared with the earthquakes of
interest. The generalized linear model (McCullogh 2000) is further introduced to describe the
relationship between the occurrence or strength of precursors and the associated earthquake
characteristics, for instance, magnitude, depth and the distance of propagation from the epi-
center to the ionosonde station.
3. PRELIMINARY RESULTS OF iSTEP PROGRAM
The iSTEP program started officially in April 2002 for four years. Since then an upgraded
geomagnetometer network has been installed. Several ionosonde stations have also been added.
Figure 2 shows the locations of these instruments. Since data obtained by the new observa-
tional facilities are still in the early stage, we concentrated our research efforts in the first two
years on analyzing the existing data for past earthquakes. Detailed results are presented in
several papers both in this special issue and elsewhere (Lee and Tsai 2004; Chan and Ma 2004;
Chen et al. 2004; Chang et al. 2004; Liu et al. 2004b; Chen et al. 2004). We will only give a
summary below to highlight the results obtained so far.
A. Seismological Results
1. P wave velocity varitations before and after the Chi-Chi earthquake
P wave velocity variations may provide precursory information before large earthquakes.
We present results on variations of P wave travel-time residuals before and after the 1999 Chi-
Chi earthquake (Lee and Tsai 2004). The short-period seismic network deployed by the CWB
(CWBSN) is routinely used for earthquake location in Taiwan. Since the CWB used a 1D
TAO, Vol. 15, No. 3, September 2004552
velocity model, the travel-time residuals would show lateral and vertical inhomogeneities of
the velocity structures. Also, the travel-time residuals may change with time if the velocity
structure changes temporally. We used the CWBSN data from 1991 to 2002 to obtain the P
wave travel-time residuals. Our purpose is to find the mean value of the P wave travel-time
residuals at each station and its temporal changes at stations near the Chelungpu fault.
The results show that the mean P wave travel-time residuals at stations 40 km east of the
Chelungpu fault changed very little both before and after the Chi-Chi earthquake. But the
mean P wave travel-time residuals at the stations immediately west of the Chelungpu fault
changed significantly before and after the Chi-Chi earthquake. The anomalous zone is bounded
by stations in the black frame, as shown in Fig. 3. These temporal changes of P wave travel-
time residuals can be attributed to changes in the velocity structure, which in turn might be
caused by crustal deformation east of the Chelungpu fault beginning about six years before the
Chi-Chi earthquake. In other words, there was a long-term six-year precursor of P wave veloc-
ity decrease before the Chi-Chi earthquake which may be associated with dilatancy due to
development of cracks.
Tsai et al. 553
2. Forecasting aftershock distribution from stress changes following large earthquakes
Non-uniform spatial slip dislocation models from seismic waveform inversions of several
past earthquakes in Taiwan are used to calculate possible stress transfer associated with the
aftershock distribution. Our results show that the positive stress changes calculated for opti-
mal orientation of fault planes after the mainshock are correlated with the aftershock distribution.
In order to assess the possibility for forecasting aftershock distribution from stress changes of
the mainshock, we used a homogeneous fault model on the basis of the earthquake scaling law
to make rapid stress change calculations.
Comparison of the stress changes from both the homogeneous and heterogeneous fault
models show similar patterns with good correlation with aftershock distribution, including the
complex fault rupture of the Chi-Chi earthquake (Fig. 4). Our results, thus, indicate good
possibility for forecasting aftershock distribution from stress changes of the mainshock. Once
the location, magnitude and focal mechanism of a strong earthquake become available, stress
change calculations can be carried out to forecast aftershock distribution for timely aftershock
hazard mitigation.
Fig. 3. A contour map of mean P
wave travel-time residuals
in Taiwan for three differ-
ent time periods: (a) Mean
P wave travel-time residu-
als from 1991 to 1993. (b)
Mean P wave residuals
from 1994 to Chi-Chi
earthquake. (c) Mean P
wave residuals from Chi-
Chi earthquake to 2002.
(After Lee and Tsai 2004).
TAO, Vol. 15, No. 3, September 2004554
Fig. 4. The changes of Coulomb failure stress based on (a) the heterogeneous
slip model by Ji et al. (2003), (b) three-segments model, and (c) one-
segment model, respectively. The three-month aftershocks (green circles)
are considered. The amount of Coulomb stress change is shown by the
colored bar. (After Chan and Ma 2004).
Tsai et al. 555
B. Geomagnetic Total Field Variations Associated with Earthquakes in Taiwan
Electromagnetic phenomena associated with seismic activity have been extensively dis-
cussed (Pulinets 1998; Molchanov and Hayakawa 1998; Hayakawa 1999; Freund 2000). Zeng
et al. (2001) studied a large numbers of cases and found about 80% of M
≥
6.0 earthquakes
occurred within nine months to 2.5 years after appearances of the zero isoporic zone (ZIZ),
which is defined as the annual change rate of geomagnetic parameters to be less than or equal
to
± 5 nT yr
-1
.
We have examined variations in the geomagnetic total field recorded by the eight magne-
tometer stations which might be related to occurrence of M
≥
6.0 earthquakes in Taiwan from
1989 to 2001. It is found that the annual change rates of the field strength reduced down to a
zero isoporic value of
± 5 nT yr
-1
after 1997 and the following M
≥
6.0 earthquakes often
occur within the ZIZ. Our results, as shown in Fig. 5, suggest that the spatial and temporal ZI
signatures tend to lead the occurrence of M
≥
6.0 earthquakes by about two years.
C. Application of Space-borne Radar Interferometry for Monitoring Crustal
Deformation in Taiwan
The orogen of Taiwan is young and active as revealed by frequent earthquakes. Some
strong earthquakes, with the accompanying crustal deformation often caused severe damages.
We have reviewed recent results from application of the radar interferometry technique to
monitor crustal deformation in Taiwan. Results from five deformation events have been
obtained. They are the coseismic deformation of the Chi-Chi earthquake, uplift of the Tainan
area, active deformation of the Hukuo area, rapid land subsidence of the Chungli area, and
seasonal land subsidence in the Pingtung plain’s area. As shown in Fig. 6, for the uplift in the
Tainan area, the results show that the radar interferometry is a useful high-resolution tool for
monitoring crustal deformation of different characteristics.
D. Ionospheric foF2 and TEC Anomalies During M
≥
5.0 Earthquakes in Taiwan
We have examined variations in the critical frequency foF2 recorded by an ionosonde,
and the TEC derived from a network of 5 ground-based receivers of the GPS, and correlated
them with the occurrences of 144 M
≥
5.0 earthquakes in Taiwan during 1997 - 1999. Results
in Fig. 7 show that the foF2 and TEC yield similar patterns, and often concurrently register
pronounced depression anomalies four days before the earthquakes. An unbiased investiga-
tion of all anomalies before and after the earthquakes positively confirms that the anomalous
depression in the foF2 and TEC are mostly premonitory anomalies.
E. Statistical Tests on Pre-earthquake Ionospheric Anomalies
Anomalous depression of the maximum plasma frequency in the ionosphere (foF2) ap-
pears consistently within 1 - 5 days before many M
≥
5.0 earthquakes in the Taiwan area
during 1994 - 1999 (Liu et al. 2003), as shown in Fig. 8. We have made two statistical tests
TAO, Vol. 15, No. 3, September 2004556
Fig. 5. Temporal variations of the geomagnetic total field during 1988-2001.
The solid and dashed lines denote the yearly value of the observed and
IGRF field, respectively. The error bars are the standard deviation of the
yearly values. (After Chen et al. 2004).
Tsai et al. 557
Fig. 6. Interferograms of the Tainan area. Top: (a) Pair-1: 1997/2/20~1998/2/5
(year/month/day); (b) Pair-2: 1998/1/1 ~ 1998/11/12; (c) Pair-3: 1998/1/
1 ~ 1999/1/21; (d) Pair-4: 1998/11/12 ~ 2000/1/6; (e) Pair-5: 1999/1/21
~ 2000/1/6; (f) Pair-6: 2000/1/6 ~ 2000/11/16. Bottom: (a) Slant range
displacement along the profile AA’ in the top panels. OP: observing point;
RP: reference point. (b) Estimated cumulative displacement between OP
and RP in (a). Error bar is estimated from the ratio of the topographic
altitude (ht) to the altitude of ambiguity (ha
~ 9416/vertical baseline offset)
(error = wave length * ht / ha). (c) Displacement rate between the OP
and RP of each image pair in (a). (After Chang et al. 2004).
TAO, Vol. 15, No. 3, September 2004558
against the foF2 anomaly as a candidate precursor of earthquakes based on several objective
criteria, including the successful rate, alarm rate, probability gain, and R score. One statistical
test is designed to compare the foF2 anomaly-based method with competitive alternatives for
predicting the earthquakes under study. The concerned alternatives are a naive prediction based on
coin-tossing experiments and a simple prediction method constructed from the current M
≥
5.0
earthquake catalog during 1994 - 1999. The other statistical test is to investigate the significance
of the identified foF2 anomalies, among all possible foF2 anomalies, related to the recorded
M
≥
5.0 earthquakes in the Taiwan area during 1994 - 1999. The results demonstrate that the
foF2 anomaly is superior to the alternatives for temporal prediction of the M
≥
5.0 earth-
quakes in the Taiwan area during 1994 - 1999.
Fig. 7. The daily changes of foF2 and
TEC observed during Septem-
ber 1999. (a) foF2 (dark curves)
and TEC (gray curves). (b) foF2
and the associated upper and
low bounds (thin curves). (c)
TEC and the associated upper
and low bounds (thin curves).
(After Liu et al. 2004).
(a)
(b)
(c)
Tsai et al. 559
4. CONCLUSIONS
The iSTEP program is aimed at searching for earthquake precursors using multidisciplinary
approaches. Since its inception in April 2002, some preliminary results have been obtained
and summarized below.
From seismological studies, we obtained two main results. First, we examined the P wave
velocity variations before and after the Chi-Chi earthquake by analyzing the P wave travel-
time residuals. The result shows P wave velocity changes in the footwall of the Chelungpu
fault beginning about six years before the Chi-Chi earthquake. This appears to signal precur-
sory phenomena due to development of cracks and resultant dilatancy of rocks. Second, a
model for forecasting aftershock distribution from stress transfer following large earthquakes
was tested by the Chi-Chi earthquake. The results indicate good possibility for forecasting
aftershock distribution from stress changes due to the mainshock. If we have sufficient timely
information of a large earthquake, we can forecast its aftershock distribution for effective
aftershock hazard mitigation.
In geomagnetic studies, we examined variations in the geomagnetic total field recorded
by eight magnetometer stations which may be related to occurrence of M
≥
6.0 earthquakes in
Taiwan. The results show that the annual change rates of the field strength reduced down to a
Fig. 8. Cross-correlation coefficient of the foF2 anomalies as a function of time
relative to the origin time of earthquakes. (After Liu et al. 2003).
TAO, Vol. 15, No. 3, September 2004560
zero isoporic value of
± 5 nT yr
-1
after 1997. The following M
≥
6.0 earthquakes often occur
within the ZIZ. Moreover, the spatial and temporal ZI signatures tend to lead the occurrence of
M
≥
6.0 earthquakes by about two years.
By processing the radar interferometry images, we observed several types of ground
deformation, i.e., uplift of the Tainan area, active deformation in the Hukuo area, rapid land
subsidence of the Chungli area, and seasonal land subsidence of the Pingtung plain’s area.
Thus radar interferometry is demonstrated to be a useful high-resolution tool for monitoring
crustal deformation of different characteristics, including that of tectonic origin.
From ionospheric observations, we examined variations in terms of foF2 and TEC and
correlated these variations with the occurrence of earthquakes with M
≥
5.0. The variations of
both foF2 and TEC show similar patterns, and often concurrently register pronounced depres-
sion anomalies four days before large earthquakes. Unbiased statistical tests of all anomalies
before and after the earthquakes positively confirm that the anomalous depression in the foF2
and TEC are mostly premonitory anomalies.
Two statistical tests were made against the foF2 anomaly as a candidate precursor of
earthquakes. One is designed to compare the foF2 anomaly-based method with competitive
alternatives for predicting the earthquakes under study. The other is to investigate the signifi-
cance of the identified foF2 anomalies, among all possible foF2 anomalies, related to the
recorded earthquakes with M
≥
5.0. The results show that the foF2 anomaly is superior to the
alternatives for temporal prediction of the M
≥
5.0 earthquakes in Taiwan area.
The results summarized above, even though still preliminary, appear to have identified
several promising precursory changes, ranging from a few years to a few days prior to large
earthquakes. In the case of forecasting aftershock distribution, we in fact treat a mainshock as
a certain precursor. The iSTEP program will continue for at least two more years. We hope to
identify more earthquake precursors from observations of our new instrumental networks in
the near future. In the meantime, we will pursue relevant physical mechanisms. Similar pro-
grams are in progress in other countries, such as Japan, China, and the U.S. International
cooperation with these and other countries are important to us. Through integrated search for
earthquake precursors and positive establishment of related physical mechanisms, we hope to
lay the foundation for eventual earthquake prediction.
Acknowledgments We thank all the research staff and assistants of the iSTEP Program Team
for their dedicated work. We also thank Y. H. Yeh and an anonymous reviewer for their com-
ments and suggestions. The program is supported under the Program for Promoting University
Academic Excellence No. A-91-N-FA07-7-4 by the Ministry of Education, Taiwan, ROC.
REFERENCES
Chan, C. H., and K. F. Ma, 2004: Possibility on forecasting aftershock distributions from
stress change: a case study of inland Taiwan. TAO, 15, 503-521.
Chang, C. P., K. S. Chen, C. T. Wang, J. Y. Yen, T. Y. Chang, and C. W. Lin, 2004: Applica-
tion of space-borne radar interferometry on crustal deformations in Taiwan: a perspec-
Tsai et al. 561
tive from the nature of events. TAO, 15, 523-543.
Chen, C. H., J. Y. Liu, H. Y. Yen, X. Zeng, and Y. H. Yeh, 2004: Changes of geomagnetic
total field and occurrences of earthquakes in Taiwan. TAO, 15, 361-370.
Chen, I. Y., J. Y. Liu, Y. B. Tsai, and C. S. Chen, 2004: Statistical tests for pre-earthquake
ionospheric anomaly. TAO, 15, 385-396.
Crampin, S., 1998: Shear-wave splitting in a critical crust: the next sep. Rev. Inst. Fr. Petrol.,
53, 749-763.
Crampin, S., and S. V. Zatsepin, 1997: Modelling the compliance of crustal rock. II. Re-
sponse to temporal changes before earthquakes. Geophys. J. Int., 129, 495-506.
Freund, F., 2000: Time-resolved study of charge generation and propagation in igneous rocks.
J. Geophys. Res., 105, 11001-11019.
Freund, F., 2004: Stress-induced changes in the electrical conductivity of igneous rocks and
the generation of ground currents. TAO, 15, 437-467.
Freund, F., A. Gupta, S. J. Butow, and S. Tenn, 1999: Molecular hydrogen and dormant
charge carriers in minerals and rocks. In: M. Hayakawa (Ed.), Atmospheric and iono-
spheric electromagnetic phenomena associated with earthquake. TERRAPUB, Tokyo,
859-871.
Geller, R. J., 1997: Predictable probability. Seism. Res. Lett., 68, 477-480.
Hayakawa, M., 1999: Atmospheric and ionospheric electromagnetic phenomena associated
with earthquakes. TERRAPUB, Tokyo, 996 p.
Ji, C., D. V. Helmerger, D. J. Wald, and K. F. Ma, 2003: Slip history and dynamic implica-
tions of the 1999 Chi-Chi, Taiwan, earthquake. J. Geophys. Res., 108, B9, 2412, doi:
10. 1029/2002JB001764.
Lee, C. P., and Y. B. Tsai, 2004: Variations of P wave travel-time residuals before and after
the 1999 Chi-Chi, Taiwan, earthquake. Bull. Seis. Soc. Am. (in press).
Liu, J. Y., Y. I. Chen, and Y. J. Chuo, 2003: Seismo-ionospheric precursors. Abstract of
XXIII Gen. Assem. of the Inter. Union of Geodesy and Geophys., A 189.
Liu, J. Y., Y. I. Chen, Y. J. Chuo, and H. F. Tsai, 2001: Variations of ionospheric total elec-
tron content during the Chi-Chi earthquake. Geophy. Res. Lett., 28, 1383-1386.
Liu, J. Y., Y. I. Chen, S. A. Pulinets, Y. B. Tsai, and Y. J. Chuo, 2000: Seismo-ionospheric
signatures prior to M
≥
6.0 Taiwan earthquakes. Geophys. Res. Lett., 27, 3113-3116.
Liu, J, Y., Y. J. Chuo, S. J. Shan, Y. B. Tsai, Y. I. Chen, S. A. Pulinets, and S. B. Yu, 2004a:
Pre-earthquake ionospheric anomalies registered by continuous GPS TEC measurements,
Ann. Geophys., 22, 1585-1593.
Liu, J. Y., Y. I. Chen, H. K. Jhuang, and Y. H. Lin, 2004b: Ionospheric foF2 and TEC anoma-
lous days associated with M5.0 earthquakes in Taiwan during 1997-1999. TAO, 15,
371-383
Massonnet, D., M. Rossi, C. Carmona, F. Adragna, G. Peltzer, K. Feigl, and T. Rabaute,
1993: The displacement field of the Landers earthquake mapped by radar interferometry.
Nature, 364, 138-142.
McCullogh, C. E., 2000: Generalized linear models. J. Am. Stat. Ass., 95, 975-978.
TAO, Vol. 15, No. 3, September 2004562
Molchanov, O. V., and M. Hayakawa, 1998: Subionospheric VLF signal perturbations possi-
bly related to the earthquake. J. Geophys. Res., 103, 17489-17504.
Musson, R. M. W., 1997: Testing earthquake prediction results statistically. Seism. Res. Lett.,
68, 944-946.
Pulinets, S. A., 1998: Seismic activity as a source of ionospheric variability. Adv. Space Res.,
22, 903-906.
Scholz, C. H., L. R. Skyes, and Y. P. Aggarwal, 1973: Earthquake prediction: a physical
basis. Science, 181, 803-810.
Silver, P., 1998: Why is earthquake prediction so difficult? Seism. Res. Lett., 69, 111-113.
Stark, P. B., 1997: Earthquake predictions: the null hypothesis. Geophys. J. Inter., 131, 495-
499.
Stoumbos, Z. G., M. R. Reynolds, Jr. T. P. Ruan, and W. H. Woodall, 2000: The state of
statistical process control as we proceed into the 21st century. J. Am. Stat. Ass., 95,
992-998.
Tsai, Y. B., T. L. Teng, Y. H. Yeh, S. B. Yu, K. K. Liu, and J. H. Wang, 1983: Status of
earthquake prediction research in Taiwan, ROC. Bull. Inst. Earth Sci., Acad. Sinica, 3,
1-25.
Uyeda, S., 2004: A report from the RIKEN international frontier research project on earth-
quakes (IFREQ). TAO, 15, 269-310.
Vere-Jones, D., 1995: Forecasting earthquakes and earthquake risk. Inter. J. Forecasting, 11,
503-538.
Whiteside, L. S., 1998: Earthquake prediction is possible. Seism. Res. Lett., 69, 287-288.
Yen, H. Y., C. H. Chen, Y. H. Yeh, J. Y. Liu, C. R. Lin, and Y. B. Tsai, 2004: Geomagnetic
fluetuations during the 1999 Chi-Chi earthquake in Taiwan. Earth Planets Space, 56,
39-45.
Zatsepin, S. V., and S. Crampin, 1997: Modelling the compliance of crustal rock. I. Response
of shear-wave splitting to differential stress. Geophys. J. Inter., 129, 477-494.
Zebker, H. A., P. A. Rosen, R. M. Goldstein, A. Gabriel, and C. L. Werner, 1994: On the
derivation of coseismic displacement fields using differential radar interferometry: The
Landers earthquake. J. Geophys. Res., 99, 19,617-19,634.
Zeng, X., Y. Lin, C. Xu, and S. O. Yang, 2001: Turning changes in evolution of geomagnetic
field and infrastructural analysis of earthquake prediction. Kybernetes- Inter. J. Sys. &
Cybernetics, 30, 365-377.