Imaging the So.bsurface Structure of the Northern Tip of the 1999 Chi-Chi Earthquake Fault in Central Taiwan Using the Electric Resistivity Method

In order to investigate the subsurface structure of the northern tip of the Chi-Chi earthquake fault, three electric resistivity image profilings were done in the northern Shihgang area where a large surface rupture was formed during the earthquake. The survey was conducted about three weeks after the Chi-Chi earthquake which occurred on 21September,1999 in central Taiwan. The pole-pole electrode configuration with electrode intervals of 6 meters was used for the profilings. Each profile consisted of 32 electrodes and 15 measured layers. The data were interpreted using the 2-D inversion method. The investigation depth was about 80-90 meters. The r�sults indicate that th� fault zone is clearly displayed in the pro­ files with a steep resistivity gradient zone. They also indicate that the rup­ ture is a reverse fault with a dip angle of about 60-80 degrees at the depth of 0-80 meters in the northern Shihgang area. The fault zone is about 30 meters wide on the ground surface and is about 10-15 meters wide at the depth of 30-80 meters. The rock sequences are similar on both sides of the fault. They are the Chinshui Shale overlain by layers of sand and gravel. It is inf erred that the fault in the northern Shih gang area is a new branch of the Chelungpu fault. A low resistivity zone (6-13 .Q-m) about 40-90 meters wide appeared adjacent to the fault zone on the footwall, and a high resis­ tjvity zone (36-100 Q-m) about 90 meters wide appeared adjacent to the fault zone on the hanging wall. Next to the high resistivity zone on the hang­ ing wall, a low resistivity zone and a high resistivity zone each about 50-100 meters wide appeared one after the other. This low and high resistivity zoning may be correlated to the strain brought on by the seismic stress released in the earthquake, and also implies that the formations were se­ verely and extensively disturbed on the hanging wall. 1 Institute of Geophysics, National Ci:mtral University, Chungli, Taiwan, ROC *Corresponding author addresS'. Prof. Ping-Hu Cheng, Institute of Geophysics, National Central University, Chung-Li, 320 Taiwan, ROC; E-mail: huh@geps.gep.ncu.edu.tw 721 722 TAO, Vol. 11, No. 3, September 2000 (


INTRODUCTION
The Chi-Chi earthquake occurred on September 21, 1999, with its epicenter near the town of Chi-Chi in central Taiwan. Extensive surface ruptures totaling about 100 km long were formed during the earthquake (Chang et al. 1999;Lee et al. 1999). The major ruptures ap peared along the Chelungpu fault running in a N-S direction between the city of Fongyuan in the north and the town of Chushan in the south, as well as in a belt about 1.5 km wide and about 20 km long trending in a NEE direction from Fongyuan to Cholan at the northern tip of the Chelungpu fault in central Taiwan (Fig. 1).
The Chelungpu fault had been known to be an active fault (Chang et al. 1998), and was regarded as the southern extension of the Sanyi fault (Lin 1957;Meng 1963). In the Chi-Chi earthquake, the major ruptures did not extend northwards from the Chelungpu fault to the Sanyi fault, but turned north-east-eastward to the town of Chalan at the northern tip where no faults had previously been found (Fig. 1). The fault in the NEE trending belt from Fongyuan to Chalan is regarded as a new branch of the Chelungpu fault formed during the 1999 Chi-Chi earthquake.
The Shihgang area is adjacent to the city of Fongyuan on the eastern side and is situated just at the turning part of the N-S trend segment to the NEE trend belt at the northern tip of the Chelungpu fault. Unlike the throw of 2-5 meters and the absence of any obvious strike-slip on the N -S trend segment of the Chelungpu fault, the ruptures in the Shihgang area show the largest vertical and horizontal displacements (both over 8 meters) from the Chi-Chi earth quake, although it is about 50 km away from the epicenter (Fig. 1).
In addition to the largest rupture displacement, a complex faulting and folding system was formed in the Shihgang area including two main faults with throws of over 4 meters (Chang et al. 1999). Of particular note, the fault has a throw of about 8 meters and a peculiar N-shaped turning between the Beefong Bridge and the Shihgang Dam in the northern Shihgang area (Fig. 1 ) . It destroyed the Bridge and the northern spillways and gates of the Shihgang Dam with a lift of about 8 meters on the southern block relative to the northern block.
The attitude of the fault between the Beefong Bridge and the Shihgang Dam was studied from the point of view of electric resistivity structures in this study.

2.METHOD
The technique of electric resistivity image profiling (RIP) with the pole-pole eletctrode configuration was used in this study because it has a high data density for high resolution interpretation. Basically, it is a four-electrode configuration with one current electrode and one potential electrode, called the measuring electrodes, being set on the surface of the profile to be investigated, and the other two electrodes including one current electrode and one poten tial electrode, called the remote electrodes, being fixed at distant places. The remote elec trodes are far from the profile and are far from each other with distances greater than ten times  the largest measuring electrode-spacing (Fig. 2a).
In the field practice, many electrodes were arranged on the surface of the profile at equal intervals to enable the automatic changing of the measuring electrodes (Fig. 2b). To start, the first electrode was used as the measuring current electrode, and the second electrode, the third electrode, ... , and eventually the (N+l)th electrode were used in tum as the measuring potential electrodes. Then the first sequence of N data was obtained, where N was the number of mea sured layers. The largest measuring electrode-spacing was NZ for an electrode interval l. After wards, the second electrode was used as the measuring current electrode, and the third elec trode; the fC?uth electrode� ... , and eventually the (N+2)th electrode were used in turn as the measuring potential electrodes for N measured layers. In this way, the second sequence of N data was'obtairied. Similarly, the third electrode, the fourth electrode, ... , and the (M-l)th elec trode were used irt turri as0the measuring current electrodes, and their successive N electrodes (if adequate) were used in tum as the measuring potential electrodes. Accordingly, a set of RIP data with N measured layers and M measuring electrodes was obtained.
A set of RIP data is usually displayed in the form of apparent resistivity pseudosection, by which each apparent resistivity is plotted at the mid-point of the measuring electrodes, which serve as the abscissa, and the measuring electrode-spacing, which serve as the ordinate (pseudo depth). Theoretically, the depth of the investigation is proportional to electrode-spacing. Hence, the apparent resistivities for the shorter electrode-spacings are the responses of shallow strata, those of the larger electrode-spacings are the responses of deeper Strata. An apparent resistiv ity pseudosection looks like an image of the total resistivity distribution of the formations. The apparent resistivity is not the true resistivity at that place, but rather an equivalent resistivity of the formations in that electrode geometry. The true resistivities of the strata can be obtained with the proper interpretation.
The RIP data were interpreted following the 2-D inversion method because a fault can be regarded as a 2-D structure. The forward part of the 2-D inversion program used in this study is based on the: finite element method and the inverse part is based on the least-squares optimi zation tecb)�iqtie (deGroot-Hedlin and Constable 1990; Lake and Barker 1996; Tong and Yang 1990).

RESULTS··· ,·.
Three RIP data setS were obtpined in the northern Shihgang area about three weeks after the 1999 Chi�Chi earthquake. The profiles were located between the Shihgang Dam in the east and the Beefong Bridge in the west on the northern side of the Tachiahsi stream. They are labeled A-A', B-B', and C-C' in Fig. 1.
The measured apparent resistivities of the RIP are shown in Fig. 3. They were obtained using 32 �1ectrod�s.at electrode intervals of 6 meters in the pole-pole array.
The interp'reteq results of the RIP are shown in Figs. 4-6. There are three sections in each figure. The top is the measured apparent resistivity pseudosection, the middle is the calculated apparent resistivity .ps�udcisectiori, while the bottom is the model resistivity section as inter preted from the measured RIP data. The number of iterations and the root mean square errors are also shown in each figure. (a) The mea suring electrodes C1 and P 1 were arranged on the surface o� the profile to be investigated. They have a electrode-spacing varying from l to Nl. The remote electrodes C2 and P 2 are fixed at distant places greater than ten times of the largest measuring electrode-spacing. (b) A number of elec trodes were arranged on the surlace of the profile to enable the automatic changing of the measuring electrodes. ..c: ;:::l 11) r:ri  Calcul.aled Apparan� Resistivity Ps.eudnsicticm    5.2 r::: ::=: :: :: :: :: :: :: :: :: :: :: :: :: :: :: :: :: :: :'.: :: :: :: :: :: :: :: :: :: :: :: :: ::

Profile A-A'
Profile A-A' is located on the eastern side of the Beefong Bridge. It is 186 meters long on the ground surface and is spread in the Nl40° E direction. The profile crosses an irregular sloping fault zone transversely at 48-78 meters from the northwestern end of the profile (point A, Fig. 1). The sloping fault zone was formed during the 1999 Chi-Chi earthquake with a trend in the N50° E direction. The southeastern block was lifted about 5 meters relative to the north western block. A large rupture dipping steeply southeastwardly appears at 60 meters from point The consistency between the locations of the resistivity-disturbed zone and the sloping fault zone on the ground surface implies that the SRGZ-1 is the fault zone formed during the Chi-Chi earthquake. The southeast dipping of the SRGZ-1 is consistent with the outcrop which appeared in the fault zone. This implies that the lifted block is the hanging wall and the rupture is a reverse fault, and that the masses in the forepart of the sloping zone (54-60 meters from point A) are the materials which collapsed from the advanced hanging wall.

Profile B-B'
Profile B-B' is located at about 400 meters west of the northern end of the Shihgang Dam.
It is 186 meters long on the ground surface and is spread in the N135° E direction. The profile crosses an irregular sloping fault zone transversely at 30-60 meters from the northwestern end of the profile (point B, Fig. 1). The sloping fault zone was formed during the Chi-Chi earth quake with a trend in the N45° E direction. The southeastern block was lifted about 8 meters relative to the northwestern block. The southeastern block is covered with gravel and sand, whereas the northwestern block is covered with sand. There are several ruptures in the sloping zone and on the northwestern margin of the southeastern block. On the outcrops, a large rup-ture dips southeastwardly suggesting that the fault is a reverse fault. The measured data and the interpreted results are shown in Figs. 3b and 5. The interpreta tive results indicate that the formations are divided into three parts by two steep resistivity gradient zones. One of the steep resistivity gradient zones (SRGZ-1, Fig. 5) exists beneath 42-60 meters from point B. It dips southeastwardly at an angle of about 80 degrees at the depth of 20-80 meters. The resistivity is 6-10 !1-m for the formations at the 20-to 80-meter depth on the northwestern side of the SRGZ-1 but is higher than 36 0-m for the formations in a zone of about 70-100 meters wide on the southeastern side of the SRGZ-1. The other steep resistivity gradient zone exists beneath about 160-172 meters from point B (SRGZ-2, Fig. 5). The SRGZ-2 separates a high resistivity (greater than 36 0-m) block on the northwestern side and a low resistivity (less than 10 .Q-in) block on the southeastern side. It dips southeastwardly at an angle of about 45-80 degrees at the 5-to 20-meter depth, and dips northwestwardly at an angle of about 45-80 degrees at the depth of 30-80 meters.
In agreement with the findings for Profile A-A', the consistency between the locations of the SRGZ-1 and the sloping fault zone for Profile B-B' indicates that the SRGZ-1 is the. fault zone which was formed in the Chi-Chi earthquake. It is a reverse fault with a 10-to 30-meter wide zone located at 42-60 meters from point B (Fig. 5). The fault dips southeastwardly at an angle of about 60 degrees at the depth of 0-25 meters and at an angle of about 80 degrees at the depth of 30-80 meters. The masses in the forepart of the sloping zone (30-42 meters from point B) are the materials which collapsed from the advanced hanging wall.

Profile C-C'
Profile C-C' is located at about 200 meters west of the northern end of the Shihgang Dam. It is 186 meters long on the ground surface and spreads in the Nl35° E direction (Fig. 1). There is no rupture on the surface, but a peculiar N-shaped turning of the fault zone is evident around the profile. The northwestern end of the profile (point C, Fig. 1) is about 84 meters east of the fault zone.
The measured data and the interpreted results are shown in Figures 3c and 6. The interpre tative results indicate that there are two steep resistivity gradient zones, SRGZ-3 and SRGZ-2', beneath the profile (Fig. 6). The SRGZ-3 is located at 66-78 meters from point C. The resistiv ity is 6-20 !1-m for the strata below the 30-meter depth on the northwestern block and is 28-56 !1-m for that on the southeastern block. The SRGZ-3 is about 10 meters wide and dips southeastwardly at an angle of about 80 degrees at the depth of 30-80 meters. The other steep resistivity gradient zone, SRGZ-2', is located at 16-24 meters from point C (Fig. 6) and is about 100-110 meters from the fault zone. It separates a high resistivity block on the north western side and a low resistivity block on the southeastern side. The SRGZ-2' dips southeast wardly at an angle of about 30-60 degrees at the depth of 5-20 meters and dips northwest wardly at an angle of about 30-60 degrees at the depth of 25-40 meters.

DISCUSSION
The root mean square errors are 30.2%, 33% and 6.6% for the interpretative results of Profiles A-A', B-B' and C-C', respectively. The root mean square error is low for the interpre tative results of Profile C-C', implying that the fitting is good and the results are acceptable. The root mean square errors are high for the interpretative results of Profiles A-A' and B-B', largely due to noise in the data caused by the irregular topography of the sloping fault zone and perhaps partly by the geologic heterogeneity disturbed by the earthquake. Although the inter pretative results are not fitted well with the measured data, they are acceptable because the characteristic patterns which appeared on the measured apparent resistivity pseudosections do not deviate from the calculated pseudosections derived from the interpretative results.
Some outcrops on the hanging wall indicate that the rock sequences are shales overlain by thin layers of gravel and sand. On the basis of the textures of the rocks on the hanging wall, the shales can be recognized to be Miocene Chinshui Shale and the gravel and sand layers to be Holocene deposits. Normally, undisturbed and water-saturated Chinshui Shale has a resistiv ity of 9-15 .Q-m (You et al. 1999) and is the formation oflowest resistivity beyond the coastal area. The roclcs at the depth of 5-80 meters of the footwaJl are interpreted as being Chinshui Shale, for they have a resistivity of 6-13 0-m. It is inferred that the fault was newly formed since the rock sequences are similar and the overlying Holocene deposits have equivalent ranges of thickness and resistivity on both sides of the fault. The dip angle is about 60-80 degrees for the new branch, which is greater than the previously-investigated value of 25-60 degrees on the N-S segment of the Chelungpu fault (Lin 1957;Meng 1963;Pan 1967;Chang 197 1;Chang et al. 1998;You et al. 1999).
The SRGZ-2 on Profile B-B' is located about 100-110 meters southeast of the fault zone. It dips southeastwardly at an angle of about 45-80 degrees at the depth of 5-20 meters and dips northwestwardly at an angle of about 45-80 degrees at the depth of 30-80 meters. The SRGZ-2' on Profile C-C' is located about 100-110 meters southeast of the fault zone and dips south eastwardly at an angle of about 30-60 degrees at the depth of 5-20 meters and dips northwest wardly at an angle of about 30-60 degrees at the depth of 25-40 meters. The consistency of the locations and the similarity in the dipping trends between the SRGZ-2 and SRGZ-2' imply that the SRGZ-2 and the SRGZ-2' ar� correlated to the same structure. Combining Profile C-C ' with Profiles A-A' and B-B', three vertical resistivity zones are found on the hanging wall ( Fig. 7). They appear in a wave�like form of high and low in alternation and each has a width of about 50-100 meters. The high resistivity is about three times or more greater than that of the undisturbed state. Archie's law, written as p = a</Jm P w for water-saturated stratum, indi cates that the bulk resistivity p increases (or decreases) with decreasing (or increasing) poros ity ¢,if the parameters a and m (being positive) and the pore water resisti vity P w are un changed. From the evidence, it is reasoned that the high and low resistivity zoning which appears around the fault zone is a result of an increase and decrease in porosity caused by seismic stress. The displacement of the blocks which occurred during the earthquake observed by the GPS method on the ground surface indicates that the direction of the displacement is to the northwest (Lee et al. 1999;Chang et al. 1999), which is consistent with the direction of the zoning sequence in this area. This supposts the inferrence that the resistivity zoning was caused by the seismic stress released with the Chi-Chi earthquake. Cross-section of the structure across the major fault zone in the northern Shihgang area. The steep resistivity gradient zone-I (SRGZ-1) is the major fault zone formed during the Chi-Chi earthquake. The wave-like resis tivity zoning which appears on the hanging wall demonstrates that the rocks were severely disturbed.

SE
The electric resistivity structures indicate that the major fault between the Shihgang Dam and the Beefong Bridge in the northern Shihgang area is a reverse fault. It runs in a northeast direction and dips southeastwardly at an angle of about 60-80 degrees at the depth of 0-80 meters_ The rock sequences are similar on both of the footwall and the hanging wall. They are Chinshui Shale overlain unconformably with gravel and sand layers. The gravel and sand layers are Holocene deposits and are about 5 meters thick on both sides of the fault zone. It is inferred thatthe fault was newly formed at the time of the 1999 Chi-Chi earthquake.
The fault zone is about 10-30 meters wide and has a steep gradient in resistivity from 13 to 23 Q-m or mo re. The resistivity of the strata at the depth of 20-80 meters on the footwall is 6-13 Q-m and is slightly lower than that in the undisturbed state indicative of a slight increase in porosity and a slight dilation of the rocks. On the other hand, the resistivity of the strata at the depth of 20-80 me ters in a zone of about 70-100 meters wide adjacent to the fault zone on the hanging wall is mainly in a range of 36-80 Q-m. It is three times or more greater than that in the undisturbed state implying that the porosity was reduced and the rocks were compressed. Next to the high resistivity zone, a low resistivity zone and another high resistivity zone each about 50-70 meters wide appear successively on the hanging wall. This high and low resistiv ity wave-like zoning in alternation indicates that the rocks on the hanging wall were com pressed and dilated in alternation. The formations on the hanging wall were severely and extensively disturbed by the seismic stress released during the earthquake.