**3. Paleoseismologic analysis**

Although the Taiwan seismic catalogs provide a historical record of strong earthquakes over the past three centuries, there is no record of any large earthquake associated with the Shihkang-Chelungpu fault (Chen *et al*., 2004). Paleoseismic study is one of the best methods for characterizing the earthquake behavior along the Shihkang-Chelungpu fault. Over the past ten years, we have excavated numerous sites and continuously cored borings on the Chelungpu fault, where the surface rupture shows surface deformation within the Holocene deposits; our results have provided quantitative data for late Holocene slip rates (Fig. 1; Chen *et al.*, 2001a, b, 2004, 2007a, b; Ota *et al.*, 2001, 2005; Streig *et al.*, 2007). Bedrock exposed along the Shihkang fault has not been characterized in previous paleoseismic studies. Here, we integrate paleoseismologic data derived from four excavations of the Shijia, Siangongtemple, Pineapple-field, and Chushan sites to understand the characteristics of coseismic deformation of the Chelungpu fault.

#### **3.1 Shijia site**

The Shijia site is located along the frontal foothills; the surface features of the site before the earthquake show a gently alluvial fan slope of approximately 3°, which was an undeveloped terrace scarp. The 1999 fault runs along the fan slope with a hangingwall uplift of gently dipping coseismic fold scarps that are about 1 to 2 m higher than those of the footwall. We excavated a 7-m-deep, 27-m-long trench across the coseismic fold scarp formed during the Chi-Chi earthquake. The excavation showed clear exposures of well-sorted silty sand interbedded with mud and humic soil; these exposures represent overbank deposits. The exposures show three depositional units of silty sand layers (cw1, cw2, and cw3) which are defined by the onlap of H1 and H2 humic soils (Fig. 2). The upper sequence includes two wedge-shaped deposits of cw1 and cw2 units associated with deposition across the scarp; these deposits are defined by the onlap of H1 and H2 humic soils across the forelimb. The onlapped relation indicates the occurrence of two folding events after the deposition of the H1 and H2 humic soils (Fig. 2). Detrital charcoal that was collected from the alluvial

Fig. 2. The contact of alluvial deposits and humic soil (H1 and H2) occurs a distinct onlapped structure interpreted to represent a paleoearthquake event. Radiocarbon dates are used to constrain the timing of S1 and S2 paleoseismic events. Fault dip of fault depends on correlation between the borehole S1 and S2 of 49° dips. The Pliocene formation shows inclined bedding plane and is weakly sheared. The fluvial cobble deposits between boreholes S2 and S3 offset about 6-m high in both sides of the fold scarp.

deposits yields eleven radiocarbon ages; the upper part of cw3 unit: 800–660 yr BP and 940– 700 yr BP, thus indicating the depositional age of 940–660 yr BP; cw2 unit: 790–640 yr BP, 740–550 yr BP, and 710–620 yr BP, indicating the depositional age of 790–550 yr BP; cw1 unit: 280–0 yr BP (Chen *et al*., 2007b).

Borehole S1 and S2, which were drilled on the hangingwall, showed the location of the fault zone at a depth of 20.7 m and 30 m in borehole S1 and 13 m in borehole S2, where Pliocene shale is displaced over gravel deposits (Fig. 2). Based on the depth of the shear zones in both the boreholes and location of fold scarp in the excavated exposures, the fault must dip about 49º along the shallowest portion of the thrust ramp (Fig. 2). The S2 and S3 borehole logging of the top of gravel bed on both sides of the main thrust reveal a total vertical offset of 6 m, indicating that repeated large earthquakes have occurred. However, based on onlapping feature of H1 and H2 humic soils and wedge-shaped deposits of cw1 and cw2 units, we can identify two paleoearthquake events which occurred at >280 yr BP (S1 event) and 790–680 yr BP (S2 event).

### **3.2 Siangong-temple site**

128 Earthquake Research and Analysis – Seismology, Seismotectonic and Earthquake Geology

defined by the onlap of H1 and H2 humic soils (Fig. 2). The upper sequence includes two wedge-shaped deposits of cw1 and cw2 units associated with deposition across the scarp; these deposits are defined by the onlap of H1 and H2 humic soils across the forelimb. The onlapped relation indicates the occurrence of two folding events after the deposition of the H1 and H2 humic soils (Fig. 2). Detrital charcoal that was collected from the alluvial

Fig. 2. The contact of alluvial deposits and humic soil (H1 and H2) occurs a distinct onlapped structure interpreted to represent a paleoearthquake event. Radiocarbon dates are used to constrain the timing of S1 and S2 paleoseismic events. Fault dip of fault depends on correlation between the borehole S1 and S2 of 49° dips. The Pliocene formation shows inclined bedding plane and is weakly sheared. The fluvial cobble deposits between boreholes S2 and S3 offset about 6-m high in both sides of the fold

scarp.

The earthquake ruptures cut through the foot of an alluvial fan and produced a gentle 1-mhigh monoclinal scarp of approximately 11°. The pre-existing ground surface exhibits a gentle westward-dipping slope of approximately 2°. We excavated a 4- to 6-m-deep, 38-mlong trench across the earthquake scarp; the excavation exposed six wedge-shaped alluvial units: aw1, aw2, aw3, aw4, aw5, and aw6 (Fig. 3). A distinctive humic soil overprints the top of each colluvial unit, which consists of a well-sorted sand layer with a channelized gravel bed. The wedge-shaped colluvial deposits were well-defined by Os1, Os2, Os3, Os4, and Os5 humic soils, which were unconformably overlain by colluvial deposits after a large earthquake. Unit aw2, for example, pinched out over the Os3 soil of the forelimb or was eroded on the uplifted side of the fold crest. However, based on onlapping relations, it was deduced that the colluvium was deposited across the forelimb of a coseismic fold scarp and that at times, the colluvium onlapped and overlapped against the forelimb. The dips of soil in the forelimb progressively decrease toward the ground surface from 38° (Os5) to 11° for the Chi-Chi earthquake-induced slope (Fig. 3). Upward change in dips, hindward thinning, and angular unconformites between soils and overlying colluvium are interpreted as indicating the occurrence of repeated large earthquakes on the Chelungpu fault. The radiocarbon ages obtained from detrital charcoal that was collected from units aw1, upper aw4, lower aw4, and aw5 are <300 yr BP, 1960–1810 yr BP, 3000–2840 yr BP, and 3100–2920 yr BP, respectively.

Boreholes T1 and T2 were drilled to a depth of 40 m on the hangingwall and to a depth of 50 m on the footwall. Borehole T1 constrains the location of the fault zone at a depth of 22.8 m between the Pliocene shale and Holocene colluvial deposits. We link the fault zones in borehole T1 and the synclinal fold axis in the excavated exposures; the results suggest a 30° east-dipping blind fault (Fig. 3). A corresponding gravel bed between boreholes T1 and T2 is offset to a height of 40 m on both sides of the fold scarp, indicating that several large earthquakes have occurred. Based on the syntectonic sedimentary structure, we infer that five paleoearthquake events including the Chi-Chi earthquake occurred near 1960–1810 yr BP (G3 event) and 3160–2840 yr BP (G4 event); however, we have not been able to obtain age control for the G1 and G2 events (Chen *et al*., 2007b).

Fig. 3. Sketch of trench-wall emphasizing the top of beds of soil (Os) and gravel depositional wedges (aw). Alluvial gravels onlap onto each soil bed forming a wedge-shaped depositional unit. Borehole T1 reveals a shear zone at the bottom of Pliocene formation which shows inclined bed and slightly shearing at the excavated site. Correlation with a shear zone at depth in the borehole T1 and the synclinal fold axis in the excavated exposures identifies a fault-plane dipping of 30°.
