Impact of Basement High on the Structure and Kinematics of the Western Taiwan Thrust Wedge: Insights From Sandbox Models

Experimental modeling allows description of the development and ki­ nematics of structures in mountain belts formed during oblique convergence. In the collision geometry of the Taiwan mountain belt, the Chinese conti­ nental margin is oriented about N60°E, whereas the N16°E Philippine Sea plate backstop is moving toward the Eurasian plate in a N55°W direction. In addition to this oblique convergence mechanism, most of the foreland structures are strongly influenced by the shape of the backstop and struc­ tural highs. Sandbox experiments have been conducted to simulate the neotectonics of western Taiwan. The kinematics of deformation comprises a combination of compression and rotation, which results in a local parti­ tioning between thrusting and strike-slip movements. The results of spe­ cific analog models demonstrated that: (1) most of the tableland structures in the western Taiwan, such as the Tatu, Pakua, Chungchou and Chia-Yi tablelands can be interpreted as a hinge part of drag anticline formed by fault-propagating fold process; (2) most of the basin and plain structures in the western Taiwan, such as Taichung and Chianan basins, can be inter­ preted as a part of piggy back basins; (3) the frontal thrust may have the first appearance of rupture in front of and between the Peikang high and the Kuanyin high; (4) NW trending link faults may be developing within the transfer zones; and (5) an escape structure formed to the south of the Peikang high can be correlated with bathymetric map and models. (

The Taiwan mountain belt is an active curved thrust wedge (Figure 1) and represents an ideal tectonic setting to examine the process of active faulting in relation to oblique convergence, indentation and rotation tectonics because ( 1) the recent kinematics of plate movement around Taiwan is well known; (2) deformation is relatively simple as the Taiwan island is isolated from mainland China; and (3) the existing geodynamic understanding of the Taiwan thrust wedge is advanced.
This paper is an attempt to use 3-D sandbox modeling to better understand the kinematics of the western Taiwan thrust wedge and to interpret the structures produced by oblique con vergence, indentation and rotation tectonics.

PREVIOUS WORK
The Island of Taiwan is situated on the corner-shaped convergent boundary between the Eurasian plate and the Philippine Sea plate, resulted from the flipping of the subduction zone from the N-dipping Ryukyu Trench to the E-dipping Manila Trench.This results in the ob lique convergence and indentation tectonics setting.In the collision geometry of the Taiwan mountain belt, the Chinese continental margin is oriented about N60°E, whereas the Nl6°E Philippine Sea plate backstop is moving toward the Eurasian plate in a N55°W direction (Lu et al., 1997 and references there in).
Deformation and metamorphism of the Taiwan thrust wedge have been well studied, and different models are presently advocated (Suppe, 1980;Davis et al., 1983;Suppe, 1983;Barr and Dahlen, 1989;Dahlen and Barr, 1989).However all these models are constrained by frontal and normal convergence.Finite element methods have been conducted in order to reconstruct the paleostress fields (Huchon et al., 1986;Lee, 1986;Hu et al., 1996;Jeng et al., 1996;Hu et al., 1997) and to account for the observed transcurrent movement (Biq, 1972;Fitch, 1972;Bowin et al., 1978).However, these methods are limited to two dimensions and cannot easily take into account for the large displacements occurring on the long and narrow shear zones.  .PINGTUNG    For example: (1) the sigmoidal shape of the mountain front; (2) the Taiwan thrust wedge is much thinner to the south; and (3) the thrust spacing is smaller in the Western Foothills and relatively large in the Hsuehshan Range.These structural features were well illustrated by Big (1989) and are shown on the geological map of Taiwan.We consider the major discrepancies between the actual situation and this work to be the result of ignoring the existence of base ment structure to the west.
To better describe and understand the complex structures that may have resulted from the effect of basement structure, we used the same experimental modeling approach as Lu and Malavieille (1994) with emphasis on the important boundary conditions in the basement morphology.

TECTONIC SETTING: THE TAIWAN MOUNTAIN BELT
The Taiwan Mountain-Belt (Figure lB) is an active curved collision belt and thrust wedge (Chai, 1972;Suppe, 1981 and1987;Barrier, 1985;Angelier, 1986;Ho, 1986 and1988;Lu and Hsu, 1992), which developed as a result of the late Cenozoic oblique convergence be tween the Philippine Sea plate and the Eurasian plate (Figure lA).A characteristic feature of the Taiwan Mountain Belt is the S shape virgations of the general structural trends (Lu, 1994): (1) Northern Taiwan (i.e., to the north of 24.5°N, see Figure lB).The difference in strikes between these segments of the curved belt ranges from 40° in the outer zones (Western Foot hills, unit The stratigraphy of the Coastal Plain and the Western Foothills is known as a passive margin shallow marine elastic sequence of the Oligocene-Miocene-Quaternary age.The strata are fossilferrous and little metamorphosed, and they have been deformed by folding and fault ing.However, their stratal continuity has not been severely disrupted.The stratigraphy is di vided naturally into two sequences by a regional lower Oligocene unconformity (Huang, 1982).The units near the unconformity function widely as the basal decollement in the Western Foothills fold-and-thrust belt (Suppe, 1987).They are succeeded by a Pliocene and Pleis tocene a four-kilometer thick molassic sequence of foreland basin sediments, derived from the eastern mountain chains.No elastic sediment was fo und encroaching from the east before 3 Ma (NN15) (Huang, 1976).The Backbone Range Slate Belt (unit III) is interpreted as a part of the Miocene accretionary wedge.The boundary between the Backbone Range and the Hsuehshan Range (unit II) is considered as the boundary between the Philippine Sea plate and the Eurasian plate (Lu and Hsu, 1992).According to these interpretations, the western bound ary of the Paleozoic/ Mesozoic basement is regarded as the crystalline leading edge of the overriding plate and the Backbone Range Slate Belt is inferred to represent the remnants of a Miocene accretionary wedge.The contact between the Paleozoic/Mesozoic basement and the Backbone Range is marked by a mylonitic fault zone associated with a series of thrust and ductile shear structures in the Backbone Range Slate Belt.The boundary between the Hsuehshan Range and the Backbone Range on the Central Cross Island Highway is presently mapped as a major west-dipping back thrust (the Lishan fault), based on up dip stretching lineations and reverse movement sense as indicated by asymmetric pressure shadows in the slate units (Clark et al., 1993).Actually this fault separates the active part of the Taiwan Mountain Belt from the older metamorphic domain structured earlier than the current collision.Our study is mainly concerned with the post-collision deformations to the west of the Lishan fault, in the foreland thrust belt.Therefore the Mesozoic/ Paleozoic metamorphic tectonic units situated east of this boundary fault may be regarded as a relatively rigid backstop during the active foreland wedge building.

EXPERIMENTAL SET-UP
The principle of our apparatus is similar to that used by Davis et al. (1983), except that our device is large enough to simulate oblique convergence boundary conditions.Sand constitutes a good analog to the brittle Mohr-Coulomb behavior during the superficial deformation of sedimentary materials (Hubbert, 1951;Horsfield, 1977;Byerlee, 1978;Krantz, 1991).In our study we used dry cohesionless quartz sand.It exhibits a Navier-Coulomb rheology and has a friction angle of about 30, similar to many sedimentary rocks.This sand has a density of 1.6 gt cm3 and a fairly homogeneous grain size, mostly ranging from 150 µm to 300 µm based on sieve analysis.Liquid dye was used to mark the sand layers without altering its mechanical properties.
In addition to those used in the to Lu and Malavieille (1994) experiments, three additional kinds of boundary conditions (Figure 2 A) were built by laying horizontal layers of colored sand onto a plastic plate (Figure 2 B-D).A rigid mobile backstop pushes the sand over the basement plate and generates a Coulomb thrust wedge.The amount of shortening is large enough to reach the critical taper of the wedge, to which a shortening of 20 cm is required.A sufficient width, greater than 30 cm, minimizes the influence of boundary effect on the varia tion oflateral deformation to be investigated.In experiments C and D, the 3D basement shapes were built mainly on the basis of aeromagnetic profiles (Bosum et al., 1970) and other pub lished profiles (Tang, 1964;Chiu, 1973;Chou and Yang, 1986).A grid of colored lines was lain on the models allowing observation of the kinematics of surface deformation.The experi ments were video recorded and photographed at small shortening intervals.After deformation, the models were impregnated with water and cut along directions perpendicular, parallel and oblique to the backstop in order to reconstruct the 3-D geometry of structures.

EXPERIMENT RESULTS
A detailed description of the experiment for the genesis and the evolution of the thrust wedge has been given in Lu and Malavieille (1994).Thi s paper focused on the impact of the basement highs on structures observed in thrust wedges and then compared with the structures described in Taiwan.Western Taiwan is essentially studied and discussed, since the previous model (Figure 2A) did not involve these under-lain basement highs.
The experimental results are described as follows:

Experiment B
We began to modify the previous model (Figure 2A) by using a dome-shaped glass to represent the Peikang High (Figure 2B).A thrust wedge built up at the beginning of the colli sion.At 15 cm of shortening, the wedge was constructed by -2 cm-spaced imbricate thrusts.As the thrust front approached the basement high, the propagation of the thrust front was retarded in front of the basement high (see Lallemand et al., 1994).Nevertheless, the thrust fronts advanced more to the south of the basement high upon subsequent shortening.As a result, a curvature of the thrust front in association with an escape structure in the southwest ern part of model was yielded (Figure 3A).

Experiment C
This experiment series (Figure 2C) comprises triplicate experiments with the same testing conditions, to determine the reproductivity of the experiment.The results of each experiment are consistent with one another.A curved thrust wedge built up at the beginning of the collision.At 15 cm of shortening, the wedge was constructed by -1 cm-spaced imbricate thrusts (Figure 4A).While approaching the basement high underneath the sand layer, the thrust front ad-  1996).
vanced more in the region around the Peikang high and the Kuanyin high.The sudden appear ance of a thrust segment occurs in this area during the subsequent compression (marked as F. A.T.S., first appearance of thrust segment, in Figure 4A).Subsequently this segment was fur ther elongated along the strike, while the segment, which developed last, was elongated in the south.A northwest oriented link fault was then (after 25 cm of shortening) developed between these two thrusts (Figure 5A).

Experiment D
This experiment series (Figure 20) comprises triplicate experiments with the same test ing conditions.Because experiment C was not large enough to accommodate the whole moun tain range in western Taiwan, we then continued with a larger sandbox, 1.8m X 2m in size, to verify whether the neotectonics of the northern and the southern part of the western Taiwan deformation belt can be simulated.
Figure 7 shows the evolution stages of the thrust wedge.The results indicate that the beginning stages were similar to those in experiment C, i.e. curved thrust wedge, with the first appearance of a thrust segment in front of and between the two basement highs and link faults.In contrast to the previous experiment, the final stages formed an asymmetrical mountain belt with a narrower southern end and a sharp elevation change.(Figure 7) 6. APPLICATION TO REGIONAL GEOLOGY OF WESTERN TAIWAN

Escape Tectonics in the Southwestern Region Offshore From Taiwan
The Peikang high is beneath the middle part of western Taiwan and the adjacent Taiwan Strait The drilling data and the leading edge of the foothills thrust front sketch out the semi circular shape of this structural high (Figure 1).The earthquakes with epicenters in the inner part of this thrust-rimmed semicircle are apparently fewer than those with epicenters both in the peripheral zone and the areas immediately on the other side of the boundary thrust.This significant map-pattern shows clearly that the pre-Tertiary Peikang high underlies the coastal plain and is distinguished by its steadfastness in a mobile belt.Such a crustal mass of great rigidity is very likely to play an important part in determining the superficial strain pattern high above.The Chinese continental margin is oriented in a N55°E direction in this sector, and thus a large transfer zone is associated with the increasing sediment thickness to the south, during the plate convergence (Figure 3B).The thrusts and folds along the deformation front change their trend continuously into a curve form convex to the WSW (Figure 3B).This dem onstrates a large dextral transfer zone around the shelf-slope break of the SE China continental margin, the structure to the south being dragged to the SW.This indicates that the thrust to the SW migrates faster than those around the Peikang basement high.As a result, the sediments in southwest Taiwan are now escaping to the SW.Along the shelf-slope break three sets of systematic fractures are evident: two of them are conjugated fractures, the third one is parallel to the convergent direction and bisects the direction of the previous fracture sets.Sandbox models have been performed to demonstrate the kinematics of this escaping tectonics in this paper.The southern boundary of the Peikang high is underneath the shelf-slope transition of the Chinese continental margin.To the south of the Peikang high, there is a large transtension zone (Yang et al., 1991� Biq, 1992Yang et al., 1994).Earthquake focal mechanisms indicate some extension events and geomagnetic data show a diachronic clockwise rotation in this area (Yeh, 1991;Horng, 1991 ).This suggests that a dextral shear zone bound the southern edge of the Peikang high.As the conjugate part, the boundary between the Western Foothills and the eastern backstop (Backbone Range) is a sinistral shear zone (Figure 3B; Biq, 1989).The shift in velocity of GPS stations (Yu et al., 1997) of the southwestern Backbone Range to the E-W direction instead of the general NW convergent direction also suggests that there might be left lateral movement along this boundary.In the Kaohsiung-Pingtung coastal area, the station velocities are even directed toward the southwest.These arguments suggest that the area be tween the two shear zones mentioned above is escaping toward the SW (Lu, 1994).

The Potential of Earthquake in Western Taiwan
Friction of faults is often unstable, and slip occurs rapidly as a rupture dynamically propa gates over the fault surface.These sudden motions generate seismic waves and this is the mechanism of the most common and important type of earthquake (Scholz, 1990).The geolo gist G. K. Gilbert (1884) first clearly made connection between earthquakes and dynamic faulting and its relationship to tectonic processes.He had seen the immediate aftereffects of earthquakes and had remarked on the fresh-appearing scarps that so often front the mountain ranges.He concluded that repeated ruptures along these faults produced the elevation of the mountain.
The energy released from the suddenly occurring thrust movement and the link fault de formation can be reasonably assume to provides the source of the major local earthquake in this area, namely, the first appearance of thrust segment in front of and between the Peikang high and the Kuanyin high.Coincidentally, a major earthquake occurred in 1935 located in this specific area.Figure 4B shows the magnitude contour of the 1935 earthquake around Taiwan (Hsu, 1978), in which the largest magnitude contour (black area in Figure 4B) coin cides with the first appearance of the thrust segment.The development of the northwest-di rected link faults can be correlated with the transform fault or transfer fault described in other papers (B iq, 1976;Biq, 1992;Deffontaines et al., 1997).In this paper we emphasize the sig nificance of the forming mechanism of these transfer faults.Because these faults remain ac tive during the thrust front migration, they might cumulate the strain energy, discontinuously release the energy through local earthquakes and consequently give rise to earthquake concen tration zones.Figure 5B shows the distribution of earthquakes in Taiwan based on the events recorded in the year 1992 (Shin and Chang, 1992), in which the NW-directed earthquake concentration zones are well-illustrated (arrows).

From Analog Models
Interpretation of natural tectonic situations with the help of model experiments must be done with caution.Sand box experiments focus mainly on brittle to brittle-ductile structures in the thrust wedge, to the exclusion of equally important ductile deformation.The influence of anisotropic layering on the deformation of sediments is also not accounted for.However, if one merely looks at geometric and kinematic aspects of the coulomb wedges, analogies be tween figure lB and figure 7E warrants attention.
In kinematics, basin and tableland topography in association with the 3-D morphologic map of western Taiwan can be interpreted as piggy back basins and anticlinal terraces.For instance, Figure 6A shows the morphology to be similar and can be interpreted as the Taichung Basin, the Tatu Terrace and the Pakua Terrace (Figure 6B). Figure 6C shows the vertical section in association with interpretations (Figure 6B) within the model.

CONCLUSIONS
The results of these analog models suggest that: ( 1) the geometrical similarity of the struc ture of the model and the regional structures suggests that the same mechanism can be apply to the actual tectonic setting as we performed in the analogic models; (2) the structural units of the foreland thrust wedge and the tectonic style are strongly influenced by the shape of the backstop and the geometry of basement highs; (3) the kinematics of deformation comprises a combination of compression and rotation, which together result in a local partitioning between thrusting, transcurrent and strike-slip faulting; (4) the contraction against structural highs is the main cause for the curvature of the thrust front and escape tectonics of western Taiwan; (5) in fault kinematics, the first appearance of thrust segments, and the NW link transcurrent faults between these segments, might be correlated with the earthquake distribution in western   the National Science Council (grant NSC 87-2116-M002-003).

Lu
and Malavieille (1994)  used 3-D sandbox modeling experiments (Figure 2A) to illus trate the kinematic processes of the Taiwan thrust wedge development during oblique indenta tion.The major result obtained from this model is the development of an asymmetrical thrust wedge with different tectonic domains.The kinematics of deformation comprises a combina tion of compression, rotation and extension, which results in a local partitioning between thrust ing and wrenching.This experimental model shows that the faults or shear zones went through a rotation centered at the indentation point of the backstop and induced a series of transcurrent and bookshelf faulting.However, this simple preliminary work cannot account several features of western Taiwan.

Fig. 2 .
Fig. 2. Experimental boundary conditions of models: A. Oblique convergence and collision model (Lu and Malavieille, 1994), backstop dip =30, low basal friction and free borders.Thickness of sand cake= 1.5 cm.B. Di ameter of basement high = 12 cm, low basal friction, thickness of sand cake= 1.5 cm to right end and decrease to 0 at left end.C. Backstop dip =15, low basal friction thickness of sand cake= 1.1 cm.D. backstop dip =15, low basal friction and free borders, thickness of sand cake= 1.5 cm.
I in Figure lB) to about 90° in the inner ones (the eastern boundary of the Backbone Range -unit III), as Figure 1B shows.On average, the mountain ranges of Taiwan trend NNE SSW (azimuth 020) south of 24.5°E, whereas they trend ENE-WSW (azimuth 070) in north eastern Taiwan, near 25°N -12 1.5°E, which represents an average difference of 50 in strike (Figure lB).(2) Central Taiwan (between 23.7°N and 24.5°N).The mountain ranges of the Taiwan foreland trend NNE-SSW (azimuth 020) south of 24.5°E, whereas they trend NW-SE (azimuth 330) to the south of Taichung.Azimuth 350 , on average, along the contact of Hsuehshan Range (unit II) and Western Foothills, and nearly N-S along the contact of the Backbone Range and the Hsuehshan Range).(3) In southern Taiwan (to the south of23.7°N) a complicated transfer zone occurs (to be discussed later).(a) Along the contact of the Wes tern Foothills and the Coastal Plain, the mountain ranges trend NE-SW north of 23.5°N, azimuth 030 on average, whereas they trend N-S south of 23.5°N.(b) Along the contact of the Hsuehshan Range and the Western Foothills the mountain ranges trend NNW-SSE north of 23.5°N, azi muth 345 on average, whereas they trend NE-SW, azimuth 030 on average, south of23.5°N.( c) Along the contact of Backbone Range and Hsuehshan Range the mountain ranges trend NNE-SSW ncnth of 23.5°N, azimuth 345 on average, whereas they trend N-S, azimuth 005 on average, south of 23°N.

Fig. 3 .
Fig. 3. A. Experimental result of experiment B (Figure 2 B) shows tectonic es cape to south of basement high.B. Structural sketch map of southwest ern Taiwan based on bathymetric map of ACT cruise (Lallemand et al., 1996).

A
Fig. 4. A. Experimental result of experiment C (Figure 2 C) shows first appear ance of thrust segment (F.A.T.S.) in front of and between the two base ment highs.B. Magnitude contour of 1935 earthquake around Taiwan(after, Hsu, 1978).
The 3-D topography of the deformed model was replicated by pouring silicon over the model to form a mold, casting with plaster, and finally digitizing and processing using a com puter (Figure SA), The digitized data were than used to generate 30 oblique views with differ ent shading which help with structural interpretation.For example, in Fig SA the light source is from the N30°E direction and the large NS trending structure units are well defined.As shown in Figure SB, this topography exhibits the following features: (1) the major structure units of part of the Hsuehshan Range and the W estem Foothills are consistent with the actual geological units; (2) most of the basins in western Taiwan can be interpreted as piggy back basins formed during fault-propagating folding; (3) the thrust and backthrust have intensively developed in the eastern part; (4) wide-spaced thrusts and strike-slip faults have developed in the middle part; and (5) narrower spaced thrusts have developed in the Western Foothills.

Fig. 6 .
Fig. 6. A. Experimental result of experiment C (Figure 2 C) demonstrates the similarity of morphology in areas indicated by arrows with inset of B. B 3D morphology of Taichung Basin, Tatu Terrace and the Pakua Terrace from DTM (digital terrain model) data.C. Schematic sketch of the ver tical sections

Fig. 7 .
Fig. 7. Evolution stages of experimental result of experiment D number 2 (Fig ure 2 D)