The Detection of Active Faults on Taiwan Using Shallow Reflection

The purpose of this study is to detect some active faults in western Taiwan using the newly developed shallow seismic reflection method. Most of these faults are in urban areas or close to the populous regions, hence, it is important with respect to earthquake hazard mitigation. The position, fracture attitude, faulting mechanism and even activity of the fault are under investigation. The revealed fault images from some s.eismic ·sections are so sharp that many details can be measured. It is found that every fault has its own charac­ teristics. They all behave differently. The following conclusions are obtained: (1) Hsinchuang fault (Taipei basin) is obviously a normal fault with a verti­ cal displacement of 50 m within the 150 m distance range. The fault may have slid down the basement rock in a sequence of stages. The seismic section obtained here reveals only a part of it. Many small back�thrust faults are found on the hanging wall, which has cut through the young sediment. This fault could be potentially active. · (2) Hsintien fault (Taipei basin) is an old thrust fault with a large offset on.Tertiary rocks. Obvious reflections are obtained on the hanging wall side, but not on the other. This is due to lithological variations across the fault. The apparent difference in reflection signals can be easily used to delineate the fault. · (3) Shihtan fault (1935 earthquake) is an interesting fault. The breakage on the surface is attributed to the effect of squeezing on the vertically dip­ ping unconsolidated sandstone layer (Shangfuchi Sandstone) by the com­ pressional pressure which induced the earthquake. This coarse-grained sandstone 'flowed' out, protruded the overlaying alluvial deposit, and ·caused the surface rupture. The fault resulted in a reverse type. (4) Tunzuchiao fault (1935 earthquake) shows a complicated sense of move­ ment along its fault during the earthquake. A twisted near-surface curved layer caused by the earthquake is used to explain this fault. The · rupture happened at a place where the layer had large curvature. The position and the direction of rupture are random within a narrow zone surrounding the fault depending on where the layer is distorted most. 1 Institute of Geophysics, National Central University, Chung-Li, Taiwan, R.O.C.

(1) Hsinchuang fault (Taipei basin) is obviously a normal fault with a verti cal displacement of 50 m within the 150 m distance range.The fault may have slid down the basement rock in a sequence of stages.The seismic section obtained here reveals only a part of it.Many small back�thrust faults are found on the hanging wall, which has cut through the young sediment.This fault could be potentially active. -• (2) Hsintien fault (Taipei basin) is an old thrust fault with a large offset on.Tertiary rocks.Obvious reflections are obtained on the hanging wall side, but not on the other.This is due to lithological variations across the fault.The apparent difference in reflection signals can be easily used to delineate the fault.
• (3) Shihtan fault (1935 earthquake) is an interesting fault.The breakage on the surface is attributed to the effect of squeezing on the vertically dip ping unconsolidated sandstone layer (Shangfuchi Sandstone) by the com pressional pressure which induced the earthquake.This coarse-grained sandstone 'flowed' out, protruded the overlaying alluvial deposit, and • caused the surface rupture.The fault resulted in a reverse type.
(4) Tunzuchiao fault (1935 earthquake) shows a complicated sense of move ment along its fault during the earthquake.A twisted near-surface curved layer caused by the earthquake is used to explain this fault.The • rupture happened at a place where the layer had large curvature.The position and the direction of rupture are random within a narrow zone surrounding the fault depending on where the layer is distorted most.

INTRODUCTION
Active fault detection is among one of the important problems for preparation of the earthquake risk analysis of an area (Aki, 1988).It is often beneficial to know the exact location of near-siirface faults to guide more detailed studies.According to the elastic rebound theory, eljrthquakes tend to occur repeatedly at places where faults have been displaced before.After accumulating enough distortion stresses due to tectonic plate movement, fault breakages occur easily at these old 'wounds'.Taiwan is located along the Neotectonic belt surrounding the Pacific Ocean, and high earthquake activities have marked this island .Hundreds of faults are distributed over the mountains as well as in the plain areas (Shih, 1986).Among them, some have been identified as active faults (Bollina, 1977;Hsu, 1989).A detailed study of these active faults will provide a basis which may greatly reduce the damage when the unavoidable large earthquakes come.
The shallow reflection method has been used in fi nding previously unknown faults, and to provide additional details in mapping known faults (Steeples and Miller, 1990).Many succes�ful examples have been obtained in recent years (Hunter et al., 1984;Treadway et al., 1988;Shedlock, et al., 1990;Stephenson et al., 1993;etc.).By using the specific resolution power of reflection signals, this method can provide much information about the fault that other methods may not provide (Miller et aL, 1990).It can map faulted bedrock under alluvium at a 200 m depth (Miller et al., 1992) as well as at unconsolidated sedimentary contacts for the upper 50 m (Treadway et al., 1988) or even a very shallow 5 m depth (Goforth and Hayward, 1992).The pictures of underground structures that seismic signals provide are so clear that the fault can be identified without much ambiguity.This 'direct mapping' property of the reflection method has caused it to be widely applied in the exploration for oil, and may be equally practical for shallow structure study.
In the past few years, the shallow seismic reflection method has attracted more and more attention in this nation (Wang et al.,l991a).Many large construction projects such as highway systems, the high-speed railroad, nuclear power plants and dams are scheduled in the nation's developm�nt program for the coming decade.Intensive investigation of basement bedrock related to local geology is urgently required for such large scale constructions.In combination with the electric method and the refraction method, reflection seismics has proven its special ability to map underground structures, particularly around a fault.This paper will discuss five cases of active fault detection that we have conducted in the past years.Besides providing seismic images of these important faults, the cases also describe in some detail this new fault detecting technique.Some experiences are cited, that can be used as references for applying this method in similar eases.The vertical resolution power of the seismic sections in this paper is estimated to be about 2 meters by using 150 Hz dominant frequency and 1500 m/sec velocity according to one-fourth wavelegnth rule of Widess (1973).

DATA ACQUISITION
Three basic requirements must be fulfi lled to ensure the success of the shallow seismic reflection method.The first of these is a seismograph with high dynamic range and analog filter capability.These are needed in order to suppress the-low frequency, high amplitude groundroll noises by means of a low-cut filtering before AID digitizing, and to preserve tll.e weak reflection signals by storing the data in the high dynamic range memory (long bits).The second requirement is a high frequency receiver (50 Hz or 100 Hz) which is used to suppress unwanted low frequency noises and to record the desired signals with better resolution.Due to the small electric current output from the high frequency geophone, two or three geophones grouped together within 1 meter are recommended.It may not be very efficient for the deletion of groundroll, but is helpful in enhancing the signals.
The third requirement is an eligible source.This is perhaps the most difficult part.A capable source for shallow seismics should have properties such as sufficient power, high frequency, repeatability, portability, safety, low price, and ease to operate (Steeples and Miller, 1990).Gun instruments including .30-caliber or .50-caliberrifles and buffalo guns are typical sources in some developed countries, but are not popular in others due to security restrictions.A continuous search for better sources seems to be a fundamental problem.
In this study, a standard CDP (Common Depth Point) technique for data acquisition and processing, which has been widely used in oil exploration was employed.The CDP method has been proven efficient for improving signal-to-noise ratio under the noisy conditions which are usually met with shallow seismics (Miller and Steeples, 1990).It may take much more labor and expense by CDP survey which is however worthwhile considering so weak a signal in such a noisy background.The instruments implemented in this study were Geometrics, ES-2401x, 48 channel seismograph, 100 Hz OYO geophone, and a firecracker source developed by ourselves (Tsai et al., 1990).The fi eld work used a 1 m geophone interval (2 geophone a group), a near-trace offset of 40-60 m, a far-trace offset of 90-110 m, pre-AID low cut filter 100 Hz, and keeping at least 12 folds.This survey geometry may not pay sufficient attention to very shallow part (above 10 m), however, overall structure above 200 m can be mapped.The produced seismic profile is suitable for a near-surface fault study with a wide enough span and a depth extension.Knapp and Steeples (1986) pointed out a main factor which affects the depth preparation of shallow seismic profiling, i.e., higher frequencies for shallower depths.Thus if the depth of interest is 5-60 m, the field parameters could be taken at a geophone interval of 0.5 m, a geophone spread between 5 m and 60 m, and a low cut of 200 Hz or 300 Hz.Different acquisition parameters need to be used for detecting different depths.It should also be noted that it had better keep one survey geometry for one seismic line to avoid inconsistent signal distortion during data processing.This is especially important for signals at shallow depths.
After experimenting with many cases using a firecracker source (Wang et al., 1994 ), we actually became more flexible about the selection of sources.Ground surface conditions definitely determine the quality of the data, and the source should be used in accordance with it.It is certainly of help to bury the source or even the receivers (at least 50 cm) to improve coupling or source energy release.However, if surface conditions make it too hard to dig a hole, eg. on a paved road or gravel area, impact sources like Dynasource or sledge hamm ers may not be a bad alternative.The hammer can strike directly on bare ground, not necessarily on a plate, if the ground is stiff enough.Furthermore, dynamite is always a good option if security problems can be overcome.'Not big but deep' is probably a key factor for using dynamite.Thirty gram plastic explosives bwied in a hole• 1 m deep.has been demonstrated to be efficient for shallow seismic purposes.Nevertheless.experience still needs to be accwnulated by using appropriate sources to obtain suitable data.
The data acquired in the field need processing before a readable final section could be prepared.We used an HP-735 workstation with 4 GB hard disk and the seismic data processing software SSS developed by ourselves to treat the data.The processing procedures are essentially similar to those used in oil exploration.These procedures include: 1) pre-stack: geometry build-up.amplitude balance, trace edit, air-wave removing, frequency filtering, automatic gain control (AGC), deconvolution (if necessary), 20 filtering, 2) dwing-stack: velocity analysis, residual static correction, NMO stretch mute by 70%, NMO sta�k.3) post stack: frequency filtering, AGC.Due to great signal variation over a record, it is not suitable for using deconvolution.The migration after stack was not included either, owing to shallow depth and small seismic section available.
In the following sections, we will describe the seismic investigation of active faults in three different regions of western Taiwan (Figure 1).Five faults with a total of nine seismic sectiOns are included.These examples not only provide clear insight into the faults under study but may also help in other investigation in similar geologic areas.Further studies on other active faults, especially those on the plain, the populous region in Taiwan, should be.carried out as soon as possible.

HSINCHUANG FAULT AND HSINTIEN FAULT IN THE TAIPEI BASIN
The Hsinchuang fault is closely related to the origin of the Taipei basin.Geologists have considered the Taipei basin as a tectonic graben which is formed by the normal faulting of a pre-existent reverse fault along its western border.Figure 1, adapted from Wang Lee et al. (1978), illustrates the triangular shape of the Taipei basin and the faults around it.
Among them, the Chinshan fault which is a large offset reverse fault to the north of Taipei, changes to a normal fault.renamed as the Hsinchuang fault.which cuts through the western boundary of the basin.This normal faulting happened during the expanding period of upper Pliocene and Pleistocene and is supposed to be the main mechanism for the sinking of the Taipei basin.The study of the Hsinchuang fault is of great significance for determining the geologic evolution of the Taipei basin as well as aCtivity on this fault which is located in the highly populated Taipei neighborhood.the main river flowing across the basin.Four apparent faults: the Chinshan fault (also the Hsinchuang fault), the Kanchiao fault, the Taipei fault, and the Hsintien fault.all trend in the northeastern direction.These faults belong to a series of southeast-dipping imbricated thrust faults in Taiwan's western foothills probably formed during the Pliocene or Pleistocene.
The small figure inserted in Figure 1 points out all known active faults on Taiwan (adapted from Hsu, 1987).on which also denotes the fi ve seismic lines to be dealt with in this paper.
Except for those taken as geologic boundaries, many faults are of purely earthquake originate.Among them.faults in the plain area especially need attention.Figure 2 shows the results of seismic surveys on the Hsinchuang fault.The line HCl obtained using a Dynasource impact source is located 4 km north of line HC2 which was obtained using the firecracker source.Except for the higher frequency content and more detail at HC2, these two seismic sections are essentially similar.The main fault appears at the location 100 m dipping toward the east, and its western side (foot wall) has apparently bent layers which could be drug by sliding down along the normal fault.The reflection layers on the eastern side (hanging wall) have been warped and dislocated by some back-thrust faults.The seismic velocities on the eastern side indicate that they come from unconsolidated layers (smaller than 1800 m/sec) which must correspond to some Quaternary formations.The back thrust faults just cut through these young sediments.In this respect, it may imply that the Hsinchuang fault may still have action during the deposition of recent sediment.It can be ranked as a potentially active fault.The disturbed zone spans approximately 200 meters.
These back-thrust faults can also be explained as antithetics accompanied with the normal faulting.Due to the reverse type of antithetics, the corresponding main normal fault could concave upward in the deeper place (Wu, 1965), which may push the mass above it to rotate and cause a series of breakages on the hanging wall block.The disturbed zone spans about 150 meters wide.The vertical displacement of the main fault found in Figure 2 is not more than 50 meters.It is known that the Tertiary basement of the Taipei basin is at a depth more than 400 meters (Wang et al., 1994).Thus, the Hsinchuang fault may have slid down the basement layers in a sequence of stages.The seismic sections obtained here may just represent a part of it.This fault could be much more complicated than originally thought.
As to illustrate the geologically defined old fault, we also take a seismic section HT along the Hsintien fault (Figure 1). Figure 3   formation has good layering dipping to the south and the Taliao fonnation, probably due to the interruption of tuff, lack of refl ection events.The near surface deposits at the top of these two units are also different.This may be a result of the erosion on different lithology.Due to structural distinction, it may not be too difficult to delineate this old fault.This obviously differs from other active fault detection.Many 'young' active faults generally do not have enough lithologic variation to distinguish them.This point is worth consideration when studying these two different kinds of faults.

SHIHTAN FA ULT OF 1935 EA RTHQUAKE
The 1935 Hsinchu-Taichung Earthquake (M=7.1) was the most damaging earthquake (3279 causalities) in Taiwan's history.The earthquake was shallow (5-10 km) with its epi center located near to the present Liyutan darn site, Maoli.Due to the intensive investigation after the earthquake by the Japanese who occupied Taiwan at that time, many earthquake fault details were reported (Otuka, 1936).Figure 4 describes the location of the earthquake, the surface rupture, and land slides.This earthquake induced two faults: (1) Shihtan fault at dipping vertically with a maximwn horizontal displacement of more than 1 m and vertical displacement of less than 60 cm.There is a gap between these two faults, where only randomly distributed land slides were reported.The epicenter was located in this gap, which has a focal mechanism (dip=83, slip=190, strike=225 ), i.e., an almost purely right-lateral strike-slip fault oriented in the same direction as the Tunzuchiao fault (Figure 4 ).The Shihtan and Tunzuchiao faults both have a length of 20 km, but behave quite differently.At Shihtan fault, there were groups of parallel and steep ground fractures accompanying the main fault.On the other hand, Tunzuchiao fault had several segments disappear along its length, and which was the uplift side of the fault is uncertain.The Shihtan fault is of the compression type, and the Tunzuchiao fault is of the shear type.Much larger damage was found around the Tunzuchiao fault as a consequence of shear type shaking.It would be interesting to compare these two faults by their underground breakage features.
Figure 5 shows the final sections of seismic lines, STl and ST2, for the Shihtan fault.These two sections have been described in Wang       Under enormous earthquake compressional pressure, the soft Shangfuchi sandstone was squeezed and 'flowed' out to protrude the overlying alluvial fill, and split the ground surface.The main distortion zone is about 50 m wide, surrounded by many small fractures.The Shihtan fault can be thought of as a result of breakdown of a soft structural layer under the great compressional pressures which induced the earthquake.
At the same locations of STl and ST2 we also conducted a seismic refraction survey.The results are shown in Figure 6.The refraction method is supposed to be sensitive to the investigation of fracture zones either by travel time curve anomalies or refractor velocity variations.This method, however, is only effective in• detecting the bedrock surface.Figure 6 indicates exactly the same fault positions as the reflection method does in Figure 5.A shape of bulge is also visible.The lateral velocity changes in the bedrock imply that the material next to the fault is greatly disturbed, but outside the bulge fault zone it may return to normal.The results of both methods are in good match.The refraction method is much easier to carry out than the refl ection method and can be used as a supplement for fault detection as this example shows.

TUNZUCHIAO FAULT OF 1935 EARTHQUAKE
Besides the Shihtan fault, the 1935 earthquake also created another east-west trending, right-lateral strike-slip fault, i.e., Tunzuchiao fault.The Tunzuchiao fault is thought to be the major fault associated with this earthquake.According to Neotectonics of Taiwan (Seno, 1977;Biq, 1981), the Philippine island arc collides with the Eurasian continental margin in a northwest direction at a speed of 7 cm/year.Represented by the eastern Coastal Mountain Range which is thought of as a part of the Philippine-sea plate, the arc-continental collision started about 6 million years ago (Teng, 1990) and still persists now.Plate collision in such a direction could induce the left-lateral strike-slip fault trending in north-south such as the Coastal Mountain Range fault in Longitudinal Valley, and the east-west right-lateral strike slip fault such as the Meishan fault in the Chianan plain.The Tunzuchiao fault studied here is a typical example belonging to the latter.
The Tunzuchiao fault, as opposed to the Shihtan fault, is quite tender in its seismic appearance (Figure 7). Figure 4 denotes the location of seismic lines 1Zl and TZ2.The fault is situated on the Houli terrace which is composed of Pleistocene lateritic deposits with flat wiconsolidated gravel beds intercalated by sandy or silty lenses.The seismic sections in Figure 7 reflect much of the character of such a gravel layer.The-seismic signals are difficult Tunzuchiao Fault Fig. 7.The seismic section of the Tunzuchiao fault.The Tunzuchiao fault has little vertical displacement (80 cm), there appears on the sections a small breakage offset together with a long wavelength swing on the upper gravel layer surface.The deposit sequences on the two sides of fault seem different, which may form a weak zone where the earthquake rupture occurred.
to propagate through gravels (Hunter et al., 1984).In order to detect this tiny vertical offset (60 cm) fault under such unfavorable conditions, we chose a dynamite source (40 gm plastic explosive), 1 m interval, 24 folds, and 200 Hz low-cut.As presented in Figure 7, TZI and TZ2 have similar data quality.The seismic signals may not penetrate deeply but the layers near the surface are still clear enough to describe the associated fault.
Since the Tunzuchiao fault is mainly a strike-slip fault with a very small vertical offset, the fracture may not be easy to see in a seismic section.However, Figure 7(a) still gives an apparent rµpture at the middle, and more interestingly, the gravel layer (first reflection layer) shows a slightly up-and-down wavy shape with a wavelength of about 70 m.One explanation for this observation is that the shear force of earthquake may have twisted this area to make its 'skin layer' (gravel layer) 'crinkle'.The ruptures may have happened at the places of largest curvature with a certain degree of randomness.Figure 7(b), seismic line TZ2, shows two ruptures with the upthrusted side to either the south or the north.This makes the surface ruptures look like being composed of several segments distributed along a zone centering on the main fault line.Which side uplifted depends on where the rupture sits on the up-or down-wave side of wavy bedrock surface.The Tunzuchiao fault provides an example of shear distortion due to an earthquake.The ruptures discovered by seismic images may not be large, however, the twist effect of shear faulting should be taken into account to recognize their seismic meaning.

HSINHUA FAULT OF 1946 EARHTQUAKE
The Hsinhua fault was caused by the 1946 Tainan earthquake (M=6.3).Its epicenter was near the town of Hsinhua, Ta inan county, with a rupture length of 6 km trending in the direction of N80E.The inset figure of Figure 8 sketches the surface disturbances after the earthquake (Hsu, 1985).The ground breakage appears segment by segment, but can be lined up.Many places where muddy water was ejected were found distributed within 500 m around the fault.The fault has a maximum right-lateral slippage of 2 m and a vertical displacement of 76 cm with the northern side upthrown.Many people have explained east west trending earthquake faults in western Ta iwan as oblique faults.Big (1991) proposed a different mechanism of 'stop-and-tum' when rupture is approaching the edge of a fault.
Several fault branches, either old or new, can be induced and slide to balance the dislocated mass.The Hsinhua fault is probably.of this kind.�Figure 8 also depicts the locations of two seismic lines, HHI and HH2, conducted for this study.There is some earthquake damage still visible today.The bending of the irrigation waterway and the Taisugar railroad (indicated by small arrows on Figure 8) are still apparent.
These old earthquake damage relics may provide us with some check points for the seismic results.The breakage positions as indicated by the small arrows in Figure 8 are not parallel to each other.The fracture zone seems to expand toward the east.A charity house in this direction had been reported being destroyed almost completely during this earthquake.
Due to unfavorable surface conditions (dry and coarse-grained), the signals in the seis mic section do not penetrate too deep (Figure 9).Near-surface top reflection layers, however, still s.how obvious earthquake traces.HHl (Figure 9(a)), the eastern line, has the northern side thrown up within two disturbance zones, denoted as zones Land R .Zone L corresponds to the main fault and zone Ris a branch extending toward the east.The layers in each zone are twisted along two fractures both obviously with the north side up.The reflection lay ers are generally continuous, which are distorted but do not slip away.The regions outside the it may not be too unreasonable to assume that zones L and R come close together to make this special 'lost-signal' zone at seismic line HH2.The Hsinhua fault could have several branches.This is probably due to the mechanism of 'stop-and-tum' as it comes close to the Chukou fault, one of the big boundary faults in the western foothills of Taiwan (Figure 1).

DISCUSSION AND CONCLUSIONS
The shallow seismic reflection method can be useful for detecting near-surface active faults as discussed in this paper.The revealed fault pictures in some cases are so sharp that many details can be measured.The signal-detectable depth is between 10 and 200 meters, however, an earthquake fault might span a much larger distance than this.Nevertheless, faults near the surface mostly agree with the focal mechanism of the earthquake.A detail study using shallow .re. flection seismics would be able to discover the main characteristics of the earthquake and its rupture.
The refraction method is traditionally used in fracture zone investigation for engineer ing purposes.In the Shihtan case we have compared different contribution of reflection and refraction methods.By using the first arrival time, we can map very shallow refractors (as bedrock) in the refraction approach.Due to refracted ray geometry, the method is more accu rate for velocity measurement.Although the refraction method worked well for engineering purpose, the bedrock surface more or less represented a barrier which hinders the discovery of deeper refraction layers.The reflection method, the other approach,.collects reflected signals directly from the inner structure.It seems to take a photograph of underground.The rays penetrate into the bedrock and bring back plentiful structural information, though with weak signals.The enhancement of such weak signals has made the method sophisticated and expensive to use.Considering this, we feel that both the reflection and the refraction methods have their own advantages and disadvantages.A combined use of both methods could be more fruitful, especially in fault detection.
Although the sites studied in this paper are confined to places with evident fault either from earthquake history or.geologic certification, seismic reflections still reveal many details about the faulting.The h'orizontal length of fault affected area in most cases is less than 100 meters in which the most seriously distorted zone is about 50 meters.This size may depend on local geology, although in western Taiwan it must actually be around this.The seismic images indicate not only the breakage of layers but variations of reflection 'strength, character, and disturbance pattern.All these can be used to settle-down the fault and its possible derivations.Two close parallel seismic lines are designated in this study.This seems necessary as to provide related data to confirm the fault images as well as to measure the extent of the rupture.In poor data areas or for important projects, parallel line surveys might be of some help.They could be used to establish a basis for the drawing of reliable conclusions after the examination and comparison of both data sets.
The seismic sections of the Tu nzuchiao fault and the Hsinl!ua fault in this study come out quite clean.Massive disturbances do not appear to suggest the possibility of repeated earthquakes.It is probably diffic:: ult to conclude that these young earthquake faults have occurred once and for all, however, considering their geographic locations we cannot exclude this implication.The earthquake faults on the plain area of Taiwan have their own complicated 'simplicity'.This topic is interesting to study further.
In this study, we find that every fault has its own characteristics.They all behave differently.The properties of theses faults can be summ arized as following: (1) Hsinchuang fault (Taipei basin) is obviously a normal fault with a vertfoal displacement of at least 50 m.Many small back-thrust faults (or antithetics) are found on-the hanging wall , which cut through the young sediment.This fault could be potentially active.
(2) Hsintien fault (Taipei basin) is an old thrust type with a large offset.The seismic signals on both sides of the fault come out quite different.Obvious reflections are obtained on the hanging wall side, but not on the other.This is due to lithological changes across the fault.The difference in the signal's appearance can be easily used to define the position of the fault.
(3) Shihtan fault (1935 earthquake) is an interesting fault.The breakage on the surface is attributed to the squeezing effect on the vertically dipping unconsolidated sandstone layer (Shangfuchi Sandstone) by the earthquake compression force.The coarse-grained sandstone 'flowed' out and pushed the overlaying alluvial deposit causing the surface rupture.
(4) Tunzuchiao fault (1935 earthquake) showed a complicated movement along the fault during the earthquake.A near-surface twisted layer is used to explain this fault.The rupture happens at the place where the layer was mostly distorted.The break position and the direction of movement are random in a narrow zone surrounding the fault.
(5) Hsinhua fault (1946 earthquake) has the developing type of fractures.Two fracture zones were found, which are separated more toward the east.Large disturbances oc curred within a 60 m range where these two fracture zones came together in the west.
This 60 m zone has reflection signals that are totally lost due to elimination of the layering by mass mixtures on the fault.But outside the fault zone, the layer is quite flat and free of disturbance.

Figure 1
Figure 1 also shows the location of seismic lines HCl and HC2 across the Hsinchuang fault, and line HT across the Hsintien fault.The geologic units on the north and the southeast border of the Taipei basin consist of Tertiary sedimentary rocks, and on the west deposits of Quaternary Linkou gravel which was supposedly uplifted after Taipei basin subsidence along the Hsinchuang fault.Two volcanos extruded in the north on both sides of the Tamshui river.

Fig. 2 .
Fig. 2. Reflection seismic sections of the Hsinchuang fault.Two sections (a) HCl and (b) HC2 are separated by 4 km, but their faulting types are quite similar.HCl was obtained using the Dynasource impact source, and HC2 used the fi recracker source.The main fault at the 100 m location appears as a normal fault with a slippage of 50 meters.It is interesting to find some back-thrusts (antithetics?)which cut through the unconsolidated sediment on hanging wall side.This fault can be explained to form by consecutive sliding in successive steps and is still active even very recent.
displays such a profile, in which 3(b) is an interpretation of 3(a).The Hsintien fault separates the Ta liao formation to the south from the younger Nanchuang formation to the north.Both formations are of Tertiary age.It is ap parent that the dip angles of layers on the two sides of the fault are different.The Nanchuangs

Fig . . 3 .
Fig . .3. A seismic section across the Hsintien fault.This fault is obviously a boundary between the Taliao fonnation (in the south, left-hand side) and the younger Nanchuang fonnation (in the north, right-hand side).The subsurface structures are totally different on either side of the fault.

Fig. 4 .
Fig. 4. The map shows ground breakage of the 1935 Hsinchu-Taichung ear-th quake (M=7.1 ) (modified from Miyamura, 1985).This earthquake killed more than 3000 people and is the most hazardous earthquake in .Taiwan's history.Two surface ruptures: Shihtan fault in the north trending in a N30E direction and the Tunzuchiao fault in the southwest trending in S30W direction.The survey Jines STl, ST2, 121.and 122 are also indicated.
et al. (1 991b) and are reproduced here for comparison with the Tunzuchiao fault.The geologic units in the Shihtan area are composed of later Miocene rocks, including shelf facies deposited Nanchuang formation and shallow marine deposited Kueichulin formation.The structural layers have high dip angles and there is a Shihtan syncline to the west and a Pakuali anticline to the east.The Shihtan fault has a close relationship with the white sandstone member, called Shangfuchi sandstone, which belongs to the upper Nanchuang formation.In most places, the fault follows this sandstone layer.The Shangfuchi sandstone layer is composed of coarse-grained white sandstone containing rounded quartz grains loosely cemented by clay and sand.It can be easily scratched off by the fingers and is used as an indexed layer for the Nanchuang formation in this area.From the viewpoint of rock, Sungfuchi sandstone is thought to be very 'soft' and easy to reshape under pressure.� fl l'l ;j <l � � !:: ig ;:I � i;i l!i

Fig. 5 .
Fig. 5. Two seismic sections (a) STI and (b) ST2 across the Shihtan fault.(c) gives an explanation of the constitution of this fault.The earthquake compressional force has squeezed the soft Shangfuchi sandstone, causing it to flow out and distort the overlying alluvial layers.A rupture zone of about 60 meters is visible.

Fig. 6 .
Fig. 6.Seismic refraction sections on the lines STl taJ and ST2 (b).These sections express obvious distortions of the bedrock surface.A squeeze type of faulting is visible on (a) with disturbance zone spreading over a range of 50 meters.The reflection (Figure 5) and the refraction (this Figure) results are consistent.

Fig. 8 .
Fig. 8.The map shows the Hsinhua fault and two seismic survey lines, HHl and HH2.The arrows indicate the old earthquake caused damages still visible today.

Fig. 9 .
Fig. 9. Seismic sections HHl(a) and HH2(b) across the Hsinhua fault.Two rup ture zones, denoted as L and R, stand out.Along HHl, there is an apparent twisted uplift of north side, each zone with two fractures.The top layer was distorted without much breakage.On HH2, L and R zones come closer and form a region, denoted as C zone, on which there is no seismic signal received.The structural distortions are more serious on HH2.