Received 4 April 2006; received in revised form 13 September 2006; accepted 22 September 2006 Available online 9 November 2006

Transcription

1 Tectonophysics 429 (2007) Paleoseismological evidence for non-characteristic behavior of surface rupture associated with the 2004 Mid-Niigata Prefecture earthquake, central Japan Tadashi Maruyama a,, Katsutoshi Iemura b, Takashi Azuma a, Toshikazu Yoshioka a, Masaru Sato c, Riichiro Miyawaki b a Active Fault Research Center, National Institute of Advanced Industrial Science and Technology (AIST), Tsukuba Central 7, Higashi, Tsukuba, Ibaraki , Japan b Hanshin Consultants Co., Ltd., Satsumasendai, Kagoshima , Japan c Hanshin Consultants Co., Ltd., Toshima-ku, Tokyo , Japan Received 4 April 2006; received in revised form 13 September 2006; accepted 22 September 2006 Available online 9 November 2006 Abstract The 2004 Mid-Niigata Prefecture earthquake sequence (mainshock magnitude, M JMA 6.8), which occurred in an active fold-andthrust belt in northern central Japan, generated a small thrust surface rupture (b20 cm of vertical displacement) along a previously unmapped northern extension of the active Muikamachi Bonchi Seien fault zone, on the eastern margin of the epicentral region. To better understand past seismic behavior of the rupture, we conducted a paleoseismic trenching study across the 10-cm-high west-sideup surface rupture at the foot of a pre-existing 1.8-m-high east-facing scarp, which probably resulted from past earthquake(s). A welldefined west-dipping thrust fault zone accompanied by drag folding and displacing the upper Pliocene to lower Pleistocene strata and the unconformably overlying upper Pleistocene (?) to Holocene strata was exposed. The principal fault zone is connected directly to the 2004 surface rupture. From the deformational characteristics of the strata and radiocarbon dating, we inferred that two large paleoseismic events occurred during the past 9000 years prior to the 2004 event. These two pre-2004 events have a nearly identical fault slip (at minimum, 1.5 m), which is 15 times that of the 2004 event ( 10 cm). These paleoseismic data, coupled with the geological and geomorphological features, suggest that the 2004 event represented non-characteristic behavior of the fault, which can potentially generate a more destructive earthquake accompanied by meter-scale surface displacement. This study provides insight into the interpretation of past faulting events and increases our understanding of rupture behavior Elsevier B.V. All rights reserved. Keywords: Paleoseismology; Rupture behavior; Active fault; Surface rupture; Seismic hazard; The 2004 Mid-Niigata Prefecture earthquake 1. Introduction Corresponding author. Tel.: ; fax: addresses: tadashi-maruyama@aist.go.jp (T. Maruyama), iemura@hanshin-consul.co.jp (K. Iemura), t-azuma@aist.go.jp (T. Azuma), yoshioka-t@aist.go.jp (T. Yoshioka), m-sato@hanshin-consul.co.jp (M. Sato), miyawki@hanshin-consul.co.jp (R. Miyawaki). Characteristics of surface ruptures generated by shallow, moderate-to-large inland earthquakes, including their extent, style, geometry, and slip distribution, provide important clues for understanding rupture behavior and processes (e.g., Sieh, 1996; Yeats et al., 1997; Zhang et al., 1999). To assess seismic hazard and long /$ - see front matter 2006 Elsevier B.V. All rights reserved. doi: /j.tecto

2 46 T. Maruyama et al. / Tectonophysics 429 (2007) term fault evolution, it is also important to evaluate whether or not the characteristics of surface ruptures associated with a certain earthquake event are representative of the activity of a given active fault ( characteristic slip ) through several earthquake cycles (e.g., Grant, 2002). On 23 October 2004, a shallow earthquake sequence with mainshock magnitude M JMA 6.8 (Mw 6.6) (called the 2004 Mid-Niigata Prefecture earthquake by the Japanese Meteorological Agency [JMA]) occurred in the Uonuma Hills, in one of the major active foldand-thrust belts of northern central Japan (e.g., Hikima and Koketsu, 2005; Hirata et al., 2005; Kato et al., 2005a) (Fig. 1). The earthquake sequence produced a small surface rupture extending for about 1 km along a previously unmapped northern extension of the Muikamachi Bonchi Seien fault zone (MBSFZ), the main west-dipping active reverse fault zone along the eastern margin of the Uonuma Hills (Maruyama et al., 2005, 2006) (Figs. 1 and 2). Several hypotheses concerning the origin of this small surface rupture and its relation to the deeper source fault have been proposed (e.g., Maruyama et al., 2005; Kato et al., 2005b; Maruyama et al., 2006; Watanabe et al., 2006; Ueta et al., 2006). The surface rupture, expressed morphologically by a west-side-up fault scarp and pressure ridge, is interpreted to reflect west-dipping reverse faulting due to E W and WNW ESE-trending compression, which is consistent with the west-dipping main source fault zone as constrained by focal mechanisms and aftershock distribution. Therefore, we previously inferred that the surface rupture was closely associated with the main source fault zone (Maruyama et al., 2005). Quantitatively, however, the surface slip of the 2004 event was very small (b20 cm of vertical displacement). In general, such a small rupture would be unlikely to be preserved on the landscape because of the rapid erosional depositional and anthropogenic processes observed in the study region. In fact, most of the surface rupture of the 2004 event was destroyed within several months after the earthquake by rainfall and cultivation. Furthermore, the surface rupture was limited spatially to the northern extension of the MBSFZ, which is shorter than expected considering the length of the source fault and the mapped length of the MBSFZ (Fig. 1). In contrast, geomorphic features suggestive of recurrent large earthquakes accompanied by significant surface ruptures, such as the distinct fault scarps that offset the late Quaternary fluvial terraces and alluvial fans, are developed along the entire fault zone (Kim, 2001; Watanabe et al., 2001). Historical earthquake data (e.g., Usami, 2003) include no evidence that the fault scarps along the fault zone were formed by the accumulation of small surface ruptures associated with repeated moderate earthquakes. Thus, the surface slip associated with the 2004 event seems scarcely to have contributed to the long-term progressive displacement recorded in the geomorphology and geology of the region. Such a discrepancy between long-term features and the 2004 event implies that a different type of earthquake with a large surface slip ( characteristic slip ) has occurred in the past, and suggests that the 2004 event was exceptionally small and non-characteristic (Maruyama et al., 2006). To verify this hypothesis, investigation of the past seismic behavior associated with the surface rupture was vital. Because no surface ruptures prior to the 2004 event have been documented historically, we conducted a paleoseismic trenching study across the 2004 surface rupture. 2. Tectonic setting Folds and thrusts that strike NNE SSW to NE SW are widely developed in the Neogene Quaternary sequences of the northern part of central Japan (Fig. 1). The Neogene sequences were deposited in rifts, which formed in the early Miocene period, concurrently with the opening of the Japan Sea, and they have subsequently been folded and faulted in a compressional stress field with an E W to WNW ESE orientation since Pliocene time (e.g., Okamura, 2003). This compression continued during the Quaternary and is expressed as deformed Quaternary strata, tectonic landforms, and contemporary earthquake activity, including the 2004 earthquake sequence. Studies of the late Quaternary geologic and geomorphic features have reported several active reverse faults, flexures, and folds that strike NNE SSW to NE SW in and around the epicentral region (e.g., Yanagisawa et al., 1986; Research Group for Active Faults of Japan, 1991; Watanabe et al., 2001; Nakata and Imaizumi, 2002) (Fig. 1). The paleoseismic activities associated with these active structures are, however, poorly known. Several historical moderate earthquakes have been recorded around the epicentral region (Usami, 2003) (Fig. 1), but the surface ruptures associated with these events have not been documented. The 2004 surface rupture, which strikes N S to NNW SSE, occurred along the northern extension of the MBSFZ, where no active faults had been mapped prior to the earthquake (Fig. 2). Geomorphologically,

3 T. Maruyama et al. / Tectonophysics 429 (2007) Fig. 1. Map of the epicentral region of the 2004 Mid-Niigata Prefecture earthquake, showing major active faults and folds and earthquake epicenters (M 5.5) since AD 1738 (simplified from Kato and Yamazaki, 1979; Research Group for Active Faults of Japan, 1991; Watanabe et al., 2001; Nakata and Imaizumi, 2002). Epicenters of the earthquakes from 23 to 27 October 2004 are also shown. A surface rupture about 1 km long occurred on the northern extension of the Muikamachi Bonchi Seien fault zone, which marks the topographic boundary between the Uonuma Hills on the west and the Muikamachi Basin on the east. The black rectangle shows location of Fig. 2b. the surface rupture lies at the topographic boundary between the hilly land to the west and lowlands to the east (Fig. 2b). Near the trench site, an 15-cm-high west-side-up surface rupture occurred at the base of a 1.8- to 2-m-high pre-existing east-facing scarp on a fluvial terrace (terrace L4, described below) (Figs. 2a and 3a). The surface rupture is also located at the geological boundary between the upper Pliocene to lower Pleistocene Uonuma Formation on the west and late Pleistocene to Holocene fluvial terrace and landslide deposits on the

4 48 T. Maruyama et al. / Tectonophysics 429 (2007) 45 60

5 T. Maruyama et al. / Tectonophysics 429 (2007) Fig. 3. a) View looking southwest of the surface rupture at the trench site. The rupture (red arrows), which strikes N S and has a west-side-up vertical component, occurred at the base of a pre-existing east-facing scarp. The trench was excavated across both the surface rupture and the pre-existing east-facing scarp. b) Close-up view, looking west, of the rupture at the trench site. Tracks of a tractor (blue arrows) perpendicular to the surface rupture (red arrows) display no lateral offset, indicating nearly E W horizontal compression. Both photos were taken in November Fig. 2. a) Geomorphologic map of the region around the surface rupture showing the location of the trench and of the east-facing warping and fault scarps on the fluvial terraces. The base topographic contour map was redrawn from maps published by the former Hirokami Village (present Uonuma City), Niigata Prefecture. Contour interval is 2 m. Topographic profiles across the tectonic scarp are also shown. Measurement was carried out with a total station. Black dots indicate measurement points. Note that the higher terraces have larger scarp heights. Terrace L1 is developed on the east of terrace L2. b) Map showing the relationship between the 2004 surficial deformation and previously mapped active tectonic features. Note that the surficial deformation is distributed within a narrow zone that trends nearly N S, where a topographic contrast between hilly land to the west and fluvial lowland to the east is developed (compiled from Yanagisawa et al., 1986; Watanabe et al., 2001; Maruyama et al., 2005). The base map is the 1/ 25,000 topographic map Obiro of the Geographical Survey Institute of Japan.

6 50 T. Maruyama et al. / Tectonophysics 429 (2007) 45 60

7 T. Maruyama et al. / Tectonophysics 429 (2007) Fig. 5. Close-up photographs of the a) north and b) south walls, showing the principal slip surfaces associated with event 1 (red arrows) and event 2 (green arrows). Bold red numerals denote stratigraphic units. Note that the base of unit 1 slipped 10 cm during event 1. east (e.g., Yanagisawa et al., 1986). Such west-side-up features along the surface rupture are concordant with those of the MBSFZ to the south (Fig. 1). We therefore inferred that the topographic and geologic contrasts along the surface rupture might have resulted from recurrent west-side-up faulting during the Quaternary and that the surface rupture occurred along the northernmost part of the MBSFZ (Maruyama et al., 2006). These interpretations were confirmed by shallow seismic profiling conducted in the transition zone between the Fig. 4. Trench logs of the a) north and b) south walls across the surface rupture and pre-existing east-facing scarp, showing fault traces, stratigraphic units, and radiocarbon dates. The principal fault zone is shown by heavy red lines. Bold red numerals denote stratigraphic units. Calibrated calendar age ranges (within 2σ, except sample Nb, which is within 1σ conventional age) based on 14 Cdatingarealsoshown(seeTable 1). Grid interval is 1 m. Location of the trench site is shown in Fig. 2a. The topographic profile surveyed immediately after the earthquake is also projected (solid blue line), showing that the principal fault zone connects directly to the 2004 surface rupture. The log of the south wall is flipped horizontally to facilitate comparison with the north wall. The trench walls slope at ; we projected the stratigraphic features onto a vertical plane during logging. The walls were logged at a scale of 1:20. Black arrows in Fig. 4a indicate the location of the event horizons (i.e., the stratigraphic unit that was the ground surface at the time of the earthquake). c)lower hemisphere stereo projection of the fault surfaces of the principal fault zone, striations, and bedding surfaces of unit 12. Striations were measured at the place indicated by a white star in Fig. 4a. Note that the trench orientation is nearly parallel to the slip direction, which enabled us to conduct a retrodeformation study of the walls, as shown in Fig. 6.

8 52 T. Maruyama et al. / Tectonophysics 429 (2007) 45 60

9 T. Maruyama et al. / Tectonophysics 429 (2007) surface rupture and the northernmost part of the previously mapped MBSFZ, where an emergent thrust dipping gently to the west was imaged (Kato et al., 2005b). 3. Paleoseismological trenching 3.1. Geomorphic and geologic features in and around the trench site Along the surface rupture, the geomorphic expression of active faulting is subtle, but sporadically developed N S-striking east-facing warping and fault scarps on the several steps of the late Quaternary fluvial terraces may be evidence of recurrent faulting; the higher terraces have larger scarp heights: 4 7 m on terrace L2 and 2 mon terrace L4 (Fig. 2a). A distinct surface rupture and surface deformation of man-made structures associated with the 2004 event were observed along or close to these tectonic geomorphic features (Fig. 2a). To better understand past seismological behavior of the surface rupture, we excavated a 16-m-long and 4-mdeep trench across the ~10 cm-high west-side-up rupture at the base of a 1.8-m-high east-facing scarp, which had probably resulted from one or more past earthquakes, on fluvial terrace L4 (Figs. 2 4). Because the trenching site was located where (i) two sub-parallel scarps, present both to the north and the south of the trench site, become a single fault scarp, (ii) the 2004 surface rupture was superimposed on the scarp, and (iii) terraces younger than terrace L4 do not occur in the fault zone (Fig. 2a), we inferred that there was little possibility of missing recent events since the formation of terrace L4 on other fault strands beyond the trench site. The trenching study was conducted during late October to early December 2005, 12 months after the 2004 event. A distinct principal fault zone striking N S and dipping gently (20 30 ) to the west was exposed on both the north and south walls of the trench (Figs. 4 6) and was connected directly to the 2004 surface rupture (Figs. 4 and 5). Drag folding associated with the thrust faulting, comprising a hanging wall anticline and a footwall syncline, was also observed (Fig. 4), and accommodates a large portion of the net displacement, as described below. The trench was oriented E W, almost parallel to the fault slip direction, which was inferred from (i) reconstruction of piercing points displaced by the surface rupture (Fig. 3) and (ii) striations that developed in the principal fault zone (Fig. 4c). Therefore, we were able to conduct a retrodeformation study of the trench walls (e.g., McCalpin, 1996) (Figs. 4 and 6) Stratigraphy The strata exposed on the trench walls comprise the weakly consolidated upper Pliocene to lower Pleistocene Uonuma Formation, unconsolidated upper Pleistocene (?) to Holocene fluvial terrace deposits, overbank deposits, paleosols, and cultivated soils. Based on their sedimentological characteristics, radiocarbon ages (Table 1), and relationship with the faults, these strata were divided into 12 units (Fig. 4). The oldest exposed stratigraphic unit, which was limited to the hanging wall of the principal fault zone, is composed chiefly of weakly consolidated dark gray silt, sand, and granules with abundant wood fragments (unit 12). On the basis of its sedimentary characteristics and nearby distribution pattern, this unit is interpreted as the Uonuma Formation. A number of fractures and small faults are found in this unit, and there is a general tendency for the number of fractures to increase toward the principal fault zone. Along the principal fault zone, a b5-cm-thick fault breccia and gouge zone is developed. These stratigraphic and structural features are cut unconformably by the overlying unit 11. Unit 11 is a clast-supported gravel layer of terrace L4. The gravel layer is composed of weakly stratified subrounded to sub-angular clasts with a maximum diameter of 30 cm in a matrix composed of granules and coarse sand. A 14 C age of 38,110±360 y BP obtained from a small lens of organic sediment in the basal part of this unit suggests that this unit is late Pleistocene in age. However, we question the reliability of this age because the gravel has not been subjected to the considerable oxidation and weathering that is observed in gravels deposited at ka in central Japan, and the sedimentary characteristics suggest that it is unlikely Fig. 6. Retrodeformation of the north wall log. A similar slip history can also be reconstructed for the south wall. a) Restoration of the 2004 slip (0.1 m of dip slip) along the principal fault zone shows residual slips in units 3 12, which are attributed to earlier faulting event(s). The difference in slip between units 5 3 and units 7 and 6 across the principal fault zone is probably due to mainly (i) partitioning of slip along the principal fault zone into series of spray faults, (ii) thickening of the strata along the principal fault zone, and (iii) removal of the equivalent hanging wall units across the fault that slipped during the 2004 event by erosion and cultivation after event 2. b) Restoration of the slips of the 2004 event (0.1 m) and event 2 (1.5 m) along the principal fault zone reveals the approximate original geometry of the colluvial wedge deposit (unit 7) that lies directly above the fault zone, but still leaves a residual offset of about 1.5 m of units 10 and 9 and the base of the trench wall, which represents the slip of event 3. Note that the restoration does not account for any displacement due to drag folding and subsidiary faulting. For details, see text.

10 54 T. Maruyama et al. / Tectonophysics 429 (2007) Table 1 Results of radiocarbon analyses Lab. no. Sample no. Unit Material Method Measured age (y BP, ±1σ) δ 13 C ( ) Conventional age (y BP, ±1σ) Calendric age range (AD/BC, ±2σ) Beta S1 1 Organic sediment AMS 790± ±40 AD 1170 AD 1280 Beta S2 2 Organic sediment AMS 3700± ± BC 1990 BC Beta S7 2 Organic sediment AMS 3950± ± BC 2350 BC Beta S4 3 Organic sediment AMS 3220± ± BC 1440 BC Beta N1 3 Organic sediment Radiometric 3790± ± BC 2040 BC Beta S5 3 Organic sediment AMS 3980± ± BC 2460 BC IAAA C15 3 Organic sediment AMS 4100± ± BC 2590 BC IAAA C14 3 Organic sediment AMS 4120± ± BC 2630 BC Beta S3 5 Organic sediment AMS 5090± ± BC 3800 BC Beta S6 5 Organic sediment AMS 5280± ± BC 4000 BC IAAA Na 5 Organic sediment AMS 5360± ± BC 4050 BC Beta S11 6 Charcoal AMS 5710± ± BC 4380 BC Beta S12 7 Organic sediment AMS 5440± ± BC 4180 BC Beta N11 8 Charcoal AMS 7220± ± BC 5930 BC Beta S13 8 Organic sediment AMS 7000± ± BC 5740 BC Beta S8 10 Wood AMS 8080± ± BC 6830 BC Beta S9 10 Organic sediment AMS 8570± ± BC 7530 BC IAAA Nb 11 Organic sediment AMS 38,260± ,110±360 N.A. Lab. No.: Beta, Beta Analytic Inc., Miami, Florida, USA; IAAA, Institute of Acceleratory Analysis Ltd., Shirakawa, Fukushima, Japan. Samples are listed in descending stratigraphic order. See Fig. 4 for sampling locations. Method: AMS, accelerator mass spectrometry; radiometric, conventional β-counting. Conventional 14 C ages were corrected by δ 13 C and calculated using the Libby half-life of 5568 years. Calendar years were determined from dendrochronologically calibrated probable age ranges with confidence limits of 2σ and rounded to nearest decade. Calibration was carried out using Calib Rev Program ( with calibration data set IntCal04 (Reimer et al., 2004). that a significant depositional hiatus exists between unit 11 and the overlying Holocene units 10 to 8. We thus infer that the dated sediment is derived from older strata exposed uphill, north of the trench site, and that the age of unit 11 is considerably younger than this 14 C age. Units10to8areoverbanksiltandsanddeposits that conformably overlie the terrace gravels (unit 11). Unit 10 is composed of white to pale gray silt with a lensofpeatandwoodfragmentsdatedincalendar years at and BC, respectively. Unit 9 consists of yellow to ivory well-laminated medium to coarse sands containing a continuous, white silty sand layer (shown as purple in Figs. 4 and 6 for emphasis). Unit 8 is composed of white silt with lenses of medium to coarse sand. Similar to the white silty sand layer in unit 9, a thin, continuous pink silt layer within this unit provides a reliable marker for tracking the deformation. Organic sediments within the pink silt layer and a sand lens of this unit were dated at BC and BC, respectively (Fig. 4). Inversion of these two ages with respect to the sampling positions is a possible result of inclusion of reworked material. Unit 7 is a wedge-shaped deposit that uncomformably overlies units 9 and 8 at the base of the fault scarp (Figs. 4 and 6). This unit is composed of non-fabric, gray silt and sand containing blocks of white silt and sand derived from units 9 and 8. On the footwall side of the north wall, an injected vein of sand from unit 8 was observed. Based on its depositional location and morphological and sedimentary characteristics, we interpret this unit as a colluvial wedge deposit, formed immediately after a surface-rupturing earthquake as a result of the collapse and degradation of a steep fault scarp (e.g., Carver and McCalpin, 1996). As shown in Fig. 4, the degree of stratal deformation along the principal fault zone is remarkably different between the units above and below this colluvial wedge deposit. Weakly humic sediment from this unit was dated at BC (south wall). Unit 6, which is also composed of overbank sand, consists of massive orange medium to coarse sand with occasional sub-angular to angular granules. A fragment of charcoal retrieved from this unit was dated at BC. The lowermost part of this unit is composed of massive orange silt. Unit 6 is overlain by a light brown paleosol (unit 5) dated from to BC. The sedimentary sequence of units 4 and 3 is similar to that of units 6 and 5. Organic sediments from unit 3 were dated from to BC. Unit 2, a ditch-shaped unit of black humic soil containing yellow silt, is cut into unit 3 (Figs. 4b and 5b) and is seen only on the footwall side of the south wall. Organic sediments from unit 2 were dated at and

11 T. Maruyama et al. / Tectonophysics 429 (2007) BC, dates close to those obtained from the underlying unit 3. The morphology of this unit and its radiocarbon age indicate that it probably consists of fill, composed of unit 3 materials, in an artificial ditch. Thus, the true age of unit 2 formation is probably significantly younger than the 14 C ages obtained. Unit 1 is modern cultivated soil. Organic sediment in the lowermost part of this unit was dated at AD Paleoearthquakes All stratigraphic units exposed are displaced by the principal fault zone (Fig. 4). Based on several lines of evidence for paleoearthquakes, such as multiple fault terminations at a single stratigraphic horizon, tilted or folded strata overlain by less deformed strata, and the colluvial wedge deposit (e.g., McCalpin, 1996; Yeats and Prentice, 1996), we were able to identify multiple faulting events, including three well-defined events since deposition of the fluvial terrace gravel of unit 11 (events 1 to 3, where 1 is the 2004 event). Event 1 (the 2004 event) is demonstrated on both the north and south walls by an 10 cm slip of the base of unit 1 (Figs. 4a, b, and 5). No offsets are observed on the ground surface (top of unit 1) because of agricultural modification after the earthquake (Fig. 5). Projection of the topographic profile (Profile 2 in Fig. 2a) surveyed immediately after the 2004 event onto the north wall shows that the vertical displacement at the ground surface roughly corresponded to that of the base of unit 1(Figs. 4a and 5a). Event 2 is defined by (i) truncation of one of the principal faults by unit 2, on the south wall (Figs. 4b and 5b), and (ii) presence of residual slip in units 3 12 after restoration of the 2004 slip (10 cm of slip) along the principal fault zone (Figs. 4 and 6a). On both walls, the principal faults flatten upward and displace the lower part of unit 3 (Figs. 4 and 5). Thus, the relationship between the upper part of unit 3 and the faults responsible for this event is unclear. These findings indicate that event 2 postdates at least the deposition of the lower part of unit 3 and predates the development of unit 2. Because of both the (i) poor age control of unit 2, as described above, and (ii) limited age information from the displaced part of unit 3, the timing of this event is loosely constrained to sometime after 2890 BC (the 14 C age from the clearly displaced part of unit 3; sample No. C14) (Fig. 4; Table 1). The magnitude of slip associated with event 2 could be roughly measured by matching the displaced correlative units that straddle the principal fault zone on the restored log (Fig. 6). This procedure shows the minimum slip, however, because the restoration does not take into account the effects of the drag folding and subsidiary faulting (discussed later). Matching of the bases of units 7 and 6 across the principal fault zone suggests that the slip was about 1.5 m (as described below, unit 7 is a scarp-derived colluvial wedge deposit associated with the earlier event 3, which was subsequently displaced by events 2 and 1) (Fig. 6a). On the other hand, on the north wall, the displacement of units 5, 4, and 3 (lower part) is smaller than that of units 7 and 6, which suggests that an additional slip event might have occurred between units 6 and 5. However, the following observations suggest that this difference in the amount of displacement is only apparent: (i) units 6 to 3 (lower part) have nearly identical slips on the south wall (1.3 m, 1.2 m, m, and 1.0 m for the bases of units 6, 5, 4, and 3, respectively) (Fig. 4b); (ii) the principal fault zone diverges and becomes several faults within units 6 3, which resulted in the partitioning of the slip from the principal fault zone; (iii) the thicknesses of units 6 and 5 (and 4?) increase toward the principal fault zone on the footwall side, which resulted from either stacking of the strata by reverse faulting along the subsidiary faults or ductile deformation, or both; (iv) no scarp-derived colluvial wedge deposits or angular unconformity is developed between units 6 and 5; and (v) the hanging wall equivalents of units 5 3 along the fault that slipped during the 2004 event were probably stripped by erosion and cultivation after event 2. In light of these observations, we interpret the approximately 1.5-m slip of the bases of units 7 and 6 to be representative of the slip during event 2. It is unclear whether a colluvial wedge, such as formed during event 3 (described below), was formed during event 2, because the upper parts of units 3 and 2 around the fault zone had been removed for agricultural purposes (Figs. 4 and 5). Event 3 is defined by (i) reconstruction of the approximate original geometry of the scarp-derived colluvial wedge deposit (unit 7) that lies directly and unconformably above both the principal fault zone and the steeply inclined units 9 and 8, and (ii) the presence of a residual offset of about 1.5 m of the lower units 10 and 9 and the base of the trench wall after restoration of the slips associated with the two recent events (0.1 m+1.5 m) (Fig. 6b). The absence (north wall) and thinning (south wall) of unit 8 in the hanging wall are probably due to erosion and the formation of unit 7 subsequent to development of the fault scarp associated with this event (Fig. 4). Both (i) the truncation of multiple fault strands (including one of the principal faults) by unit 7 on the south wall and (ii) the remarkable difference in the degree of stratal deformation below and above unit 7 are also evidence for this event (Fig. 4). These observations

12 56 T. Maruyama et al. / Tectonophysics 429 (2007) Fig. 7. Timing of recent faulting events as inferred from the trench exposure. Bold red numerals denote stratigraphic units. indicate that event 3 occurred after the deposition of unit 8 and before the development of the colluvial wedge (unit 7). The 14 C age from unit 7 is younger than that of the overlying strata (unit 6), probably owing to either (i) incorporation of young carbon into unit 7 or (ii) of older carbon into unit 6. We suppose that the former is plausible because of the following stratigraphic features. First, as described above, unit 7 is interpreted as a colluvial wedge deposit derived from underlying strata, so the 14 Cage from unit 7 is expected to be close to that of underlying units. However, the 14 C age from unit 7 is about years younger than the ages from unit 8. Second, it is unlikely that the 14 C age of unit 7 overlaps with the age of the overlying unit 5 (Fig. 7), taking into account that unit 6, a thick deposit consisting of sand and silt, is between units 7 and 5. We, thus, inferred that this unexpectedly young age of unit 7 is due to contamination by young materials such as intruded roots, and we excluded the 14 C age from unit 7 from our interpretation of the timing of event C ages from units 8 and 6 indicate that event 3 occurred between 5970 and 4380 BC (Figs. 4 and 7; Table 1). As shown in Fig. 6b, the offset of event 3 is about 1.5 m, which is equivalent to that of event 2. Differences in vertical displacements of stratigraphic units that straddle the principal fault zone also indicate the number and timing of the past events. Vertical displacement was estimated from the elevation difference of correlative units between the hanging wall and footwall outside the fault zone, under the assumption that these units were horizontal at the time of deposition. Vertical displacement of the base of unit 10 and both the base and top of unit 9 is m (shown as green dashed lines in Fig. 4), whereas that of the tops of units 6 and 5 is m (shown as purple dashed lines in Fig. 4). The vertical displacement of the base of unit 1 is only 10 cm (Figs. 4 and 5). These significant differences in vertical displacement suggest that slip events must have occurred between units 9 and 6 and between units 5 and 1, which coincide with the events inferred from the dip-slip retrodeformation and the stratigraphic features described above. Vertical displacement of the three individual events can be calculated as 10 cm for event 1, m for event 2, and m for event 3. The individual vertical displacements of events 2 and 3 are thus nearly identical. Evidence of additional previous event(s) is manifested by the presence of (i) an angular unconformity between units 12 and 11 and (ii) faults that cut unit 12 but are truncated by unit 11 (Fig. 4). Along the principal fault zone, fault rock, including a fault breccia-gouge zone, is developed only within the weakly consolidated unit 12, which also attests to previous seismic episodes, prior to the deposition of unit 11. We carefully observed and logged the trench walls, searching for small 2004 earthquake-type rupture events. Although a number of faults with a small slip were observed in the trench walls, we could not find compelling evidence of prior earthquakes with such a small rupture. For example, the faults that cut unit 9 but are covered by unit 8 might suggest a past 2004 earthquake-type rupture event (Fig. 4). However, because the faults are restricted to around the axial surface of the footwall syncline and occur only within unit 9 (they do not extend to the underlying units), we cannot preclude the possibility that the formation of these faults was related to growth of the fold during the large earthquake events (events 2 and 3). Past small ruptures are difficult to identify, mainly because (i) small displacements are within the observation uncertainty and (ii) they may be overprinted by subsequent larger events, and (iii) because it is difficult to distinguish secondary ruptures related to larger events from primary ruptures associated with smaller events. It is well known that ductile deformation such as drag folding and stratal thickening accommodates a large portion of the net fault displacement in reverse and thrust faults (e.g., Yeats et al., 1997). Estimating how much of the net displacement of the faults is taken up by such ductile deformation is fundamental for better understanding the near-surface deformational mechanism in a contractional deformation zone. As reported above, the amount of each individual vertical

13 T. Maruyama et al. / Tectonophysics 429 (2007) displacement of events 2 and 3 is similar to the amount of the slip along the principal fault zone. This vertical displacement:dip-slip ratio contradicts that inferred from the shallow fault dip (20 30 ) exposed in the trench. If we assume that all of the fault slip was accommodated by brittle faulting along the dipping principal fault zone (Fig. 4c), then net displacement of the two individual pre-2004 events (events 2 and 3) can be calculated using simple trigonometry as m, which is more than twice that of the slip along the principal fault zone obtained directly by the retrodeformation technique (about 1.5 m). As shown in Fig. 4, the strata along the principal fault zone exhibit distinct drag folding. Furthermore, units 10 to 4 thicken toward the principal fault zone because of stacking of the strata by reverse faulting and/or ductile deformation. These observations imply that a large portion ( 50%) of the fault slip was consumed by drag folding and thickening of the strata along the principal fault zone. 4. Discussion and conclusions Reconstruction of past slip behavior of faults in seismically active fold-and-thrust belts is fundamental for forecasting future seismic activity and identifying seismic hazard accurately in a contractional deformation zone. Such reconstruction also helps us to better understand how seismic strain is accumulated and released by the reverse faults in the belt. Paleoseismological trenching provided direct constraints on the past slip behavior of a thrust fault associated with a small surface rupture of the 2004 Mid-Niigata Prefecture earthquake sequence (Figs. 4 6). Evidence for past earthquakes is clearly displayed by tectonic scarps with progressive displacements (Fig. 2) and by the structural and stratigraphic relationships of the deposits exposed in the trench walls (Figs. 4 and 6). By retrodeformation of the trench logs, we recognized two large paleoearthquakes in the past ca years prior to the 2004 event. The slip amounts of two individual pre-2004 events were nearly identical (at minimum 1.5 m) (Figs. 4 and 6), and 15 times that of the 2004 event ( 10 cm), indicating that the surface slip behavior of the 2004 event was different from that of the two previous events at the trench site. On the basis of the vertical displacement of the ca years old stratigraphic unit (unit 10), which records three well-defined recent surface-rupturing episodes, including the 2004 event, we estimated a Holocene vertical slip rate for the fault of, at minimum, 0.4 mm/yr, which corresponds to a net slip rate of mm/yr taking into account the of fault dip exposed in the trench walls (Fig. 4c). This slip rate is comparable to those of major active reverse faults in active fold-andthrust belts of northern central Japan (e.g., Research Group for Active Faults of Japan, 1991), although it is a little smaller than those in the main portion of the MBSFZ (Kim, 2001). Based on the similarity of the nearby geological and geomorphological characteristics described in Section 2, together with the slip rate, we infer that the fault segment studied in this paper, which was unmapped prior to the 2004 event, belongs to the northernmost portion of the MBSFZ (Fig. 1). Large slips (N1.5 m) of two pre-2004 events in the northernmost portion of the MBSFZ suggest that past earthquakes ruptured greater portions of the fault zone. According to the empirical relationship between average surface displacement and moment magnitude (Wells and Coppersmith, 1994), the vertical displacements of the two pre-2004 events imply earthquakes with Mw 7.1 (using the regression for all slip types). These seismic moments are N5 times that of the 2004 event (Mw 6.6). To generate earthquakes with such magnitudes, surface rupture lengths of 40 km are required, corresponding to the entire length of the MBSFZ (Kim, 2001; Yoshioka et al., 2005) (Fig. 1). Our paleoseismological data thus indicate that the two events prior to the 2004 event might have had Mw 7.1, which is a substantially larger magnitude than that of any historical earthquakes recorded in the epicentral region (e.g., Research Group for Active Faults of Japan, 1991; Usami, 2003) (Fig. 1). Seismological studies show that the 2004 earthquake sequence ruptured mainly the deeper portion of the northern MBSFZ and its sub-parallel faults, although the aftershock distribution indicates that the dynamic shear rupture extended, with a low dip angle, to the shallower fault (Kato et al., 2005a). Because the two pre-2004 events show nearly identical slip amounts that apparently ruptured greater portions of the fault zone than the 2004 event, they may represent maximum slip events. Patterns of fault behavior have major implications for seismic hazard evaluation and determination of the mechanism of earthquakes based on geological and geomorphological studies. It has been proposed that some active faults are represented by characteristic earthquakes (e.g., Schwartz and Coppersmith, 1984). Characteristic earthquakes are the result of characteristic slip that defined as displacement (slip) at a specific location along a fault does not vary greatly from event to event (Grant, 2002). This implies that characteristic earthquakes display approximately the same amount and distribution of slip during each successive event. If a fault does indeed display characteristic earthquakes, knowledge of only a single faulting event would provide

14 58 T. Maruyama et al. / Tectonophysics 429 (2007) a remarkable understanding of both previous and future rupture patterns. This model has been widely applied to assess seismic hazard and understand fault mechanics. Examination of historical and paleoseismological slip measurements of well-studied faults shows that individual fault patches do fail characteristically, that is, increments of slip are constant at a particular locality (e.g., Sieh, 1996; Klinger et al., 2003). In contrast, some paleoseismic data along major active faults have demonstrated complexity in rupture behavior (e.g., Berryman and Beanland, 1991; Grant, 1996; Marco et al., 2005). Although the data obtained from a single paleoseismic site provide limited information about the behavior of the fault zone, our trenching suggests the possibility that the MBSFZ has generated surface ruptures with not only characteristic slips during large earthquakes such as events 2 and 3, but also small noncharacteristic slips during moderate earthquakes such as the 2004 event (event 1) in the same part of the fault zone, as discussed above. This suggests that the surficial rupture behavior is more complex than that expected from a simple characteristic model and that the rupture properties of the fault zone that affect the amount and extent of the surface rupture are not necessarily same during successive events. Therefore, we conclude that (i) the 2004 surface rupture was a non-characteristic slip of the fault, which has the potential to generate meter-scale (characteristic) surface ruptures over tens of kilometers, and (ii) seismic hazard and risk assessment in the epicentral region should be reconsidered in light of the possibility of large surface displacements and strong ground motions such as result from Mw 7 earthquakes. Unfortunately, because the timing of the two pre-2004 events is only loosely constrained owing to poor age controls and cultural disturbance of the strata, the present seismic potential cannot be well evaluated (Fig. 7). In order to assess seismic potential quantitatively, further comprehensive investigations, including the detailed mapping of tectonic landforms, seismic profiling, and archaeo- and historical seismology, as well as paleoseismological trenching along the fault zone and nearby faults, are required. We could not find firm evidence of past 2004-type non-characteristic slip events in the trench exposure, but the possibility that the MBSFZ has generated such events between the characteristic slip events cannot be precluded. Thus, it may be difficult to estimate the timing and recurrence of non-characteristic events, even though they can cause severe damage and pose considerable hazard, as the 2004 event did (N50 dead and N4800 injured), by using only conventional trenching. Several major active faults and fault-related folds are developed in and around the epicentral region (e.g., Research Group for Active Faults of Japan, 1991; Watanabe et al., 2001; Nakata and Imaizumi, 2002) (Fig. 1). Thus, even if events with small slips had been identified in the trench exposure, slips produced by noncharacteristic events of the MBSFZ would be hard to distinguish from triggered slips caused by earthquakes on nearby faults, unless reliable earthquake information (such as focal mechanisms, aftershock distribution, and damage distribution) were available. Recurrence intervals of large earthquakes generated by an active fault in inland Japan are, generally, longer than 1000 years (e.g., Research Group for Active Faults of Japan, 1991), and no faults are known to have ruptured twice in historical time, except for the Tanna fault, which ruptured in the AD 1930 and AD 841 earthquakes, although the older event is not well documented (Tanna Fault Trenching Research Group, 1983). Thus, it is difficult to evaluate whether the extent and behavior of surface ruptures associated with a certain event are representative of a fault without paleoseismological data. The results of our paleoseismological investigation suggest that the fault that slipped during the 2004 event has the potential to generate significantly larger earthquakes than the 2004 event. Historically, surface ruptures with possible noncharacteristic behavior similar to that of the 2004 event, defined by exceptionally smaller surface displacements and shorter rupture lengths than expected based on fault dimensions, have been reported in Japan (e.g., the Yamada fault in the 1927 Kita-Tango earthquake, southwest Japan, M JMA 7.3, Okada and Matsuda, 1997; the Nishine fault group in the 1998 Iwate-ken Nairiku Hokubu earthquake, northeast Japan, M JMA 6.1, Miyauchi et al., 1998; Azuma et al., 1999). To evaluate the maximum magnitude of earthquakes that can occur on the faults associated with these earthquakes and their behavioral characteristics, which is essential information for seismic hazard assessment, paleoseismological investigations coupled with geologic and geomorphological mapping as used in this study are indispensable. Acknowledgements We thank Y. Sugiyama, Y. Fusejima, Y. Okamura, Y. Awata, T. Komatsubara and H. Kaneda for helpful discussions and A. Miyawaki for technical assistance. We are indebted to the landowners who graciously allowed us access to their properties to conduct the paleoseismic investigation. We thank Tectonophysics editor M. Sandiford and two anonymous reviewers for their careful

15 T. Maruyama et al. / Tectonophysics 429 (2007) reviews and suggestions that greatly improved the manuscript. References Azuma, T., Awata, Y., Yoshioka, T., Fusejima, Y., rupturing and paleoseismicity of northern Shizukuishi-bonchi Seien fault. Interim Report on Active Fault and Paleoearthquake Researches in the 1998 Fiscal Year, Geological Survey of Japan Interim Report no. EQ/99/3, pp (in Japanese with English abstract). Berryman, K., Beanland, S., Variation in fault behaviour in different tectonic provinces of New Zealand. J. Struct. Geol. 13, Carver, G.A., McCalpin, J.P., Paleoseismology of compressional tectonic environments. In: McCalpin, J.P. (Ed.), Paleoseismology. Academic Press, California, pp Chapter 5. Grant, L.B., Uncharacteristic earthquakes on the San Andreas fault. Science 272, Grant, L.B., Paleoseismology. In: Lee, W.H.K., Kanamori, H., Jennings, P.C., Kisslinger, C. (Eds.), International Handbook of Earthquake and Engineering Seismology Part A. Academic Press, California, pp Hikima, K., Koketsu, K., Rupture processes of the 2004 Chuetsu (mid-niigata prefecture) earthquake, Japan: a series of events in a complex fault system. Geophys. Res. Lett. 32, L doi: / 2005GL Hirata,N.,Sato,H.,Sakai,S.,Kato,A.,Kurashimo,E.,2005.Faultsystem of the 2004 Mid Niigata Prefecture Earthquake and its aftershocks. Landslides 2, doi: /s Kato, H., Yamazaki, H., Neotectonic map of Shin-etsu district in Japan. Tectonic Map series, 5. Geological Survey of Japan. Kato, A., Kurashimo, E., Hirata, N., Sakai, S., Iwasaki, T., Kanazawa, T., 2005a. Imaging the source region of the 2004 mid-niigata prefecture earthquake and the evolution of a seismogenic thrustrelated fold. Geophys. Res. Lett. 32, L doi: / 2005GL Kato, N., Echigo, T., Sato, H., Tateishi, M., Ogino, S., Sakai, S., Toda, S., Koshiya, S., Ito, T., Toyoshima, T., Imaizumi, T., Kato, H., Abe, S., 2005b. Geologic fault model based on the high-resolution seismic reflection profile and aftershock distribution associated with the 2004 Mid-Niigata Prefecture earthquake (M6.8), central Japan. Earth Planets Space 57, Kim, H.-Y., Active faults along the western margin of the Muikamachi Basin, central Japan. Quat. Res. 40, (in Japanese with English abstract). Klinger, Y., Sieh, K., Altunel, E., Akoglu, A., Barka, A., Dawson, T., Gonzalez, T., Meltzner, A., Rockwell, T., Paleoseismic evidence of characteristic slip on the western segment of the North Anatolian fault, Turkey. Bull. Seismol. Soc. Am. 93, Marco, S., Rockwell, T.K., Heimann, A., Frieslander, U., Agnon, A., Late Holocene activity of the Dead Sea Transform revealed in 3D palaeoseismic trenches on the Jordan Gorge segment. Earth Planet. Sci. Lett. 234, Maruyama, T., Fusejima, Y., Yoshioka, T., Awata, Y., Matsu'ura, T., Characteristics of the surface rupture associated with the 2004 Mid Niigata Prefecture earthquake, central Japan and their seismotectonic implications. Earth Planets Space 57, Maruyama, T., Fusejima, Y., Kurosawa, H., Were location and style of surface rupture associated with the 2004 Mid Niigata Prefecture earthquake guided by pre-existing geologic structures? Zishin (J. Seismol. Soc. Jpn.) 58, (in Japanese). McCalpin, J.P., Field techniques in paleoseismology. In: McCalpin, J.P. (Ed.), Paleoseismology. Academic Press, California, pp Chapter 2. Miyauchi, T., Imaizumi, T., Watanabe, M., Yagi, H., Sawa, H., Hirano, S., Precise location map and late Quaternary activity of the western marginal active fault zone (Nishine faults group) of the Shizukuishi Basin, northeast Japan. Active Fault Res. 17, (in Japanese with English abstract). Nakata, T., Imaizumi, T. (Eds.), Digital Active Fault Map of Japan. 60 pp with 2 DVD-ROM and 1 map. University of Tokyo Press, Tokyo (in Japanese). Okada, A., Matsuda, T., Surface faults associated with the Kita- Tango Earthquake of 1927 in the northwestern part of Kinki district, central Japan. Active Fault Res. 16, (in Japanese with English abstract). Okamura, Y., Fault-related folds and an imbricate thrust system on the northwestern margin of the northern Fossa Magna region, central Japan. The Island Arc 12, Reimer, P.J., Baillie, M.G.L., Bard, E., Bayliss, A., Beck, J.W., Bertrand, C.J.H., Blackwell, P.G., Buck, C.E., Burr, G.S., Cutler, K.B., Damon, P.E., Edwards, R.L., Fairbanks, R.G., Friedrich, M., Guilderson, T.P., Hogg, A.G., Hughen, K.A., Kromer, B., McCormac, F.G., Manning, S.W., Ramsey, C.B., Reimer, R.W., Remmele, S., Southon, J.R., Stuiver, M., Talamo, S., Taylor, F.W., van der Plicht, J., Weyhenmeyer, C.E., IntCal04 Terrestrial radiocarbon age calibration, 26 0 ka BP. Radiocarbon 46, Research Group for Active Faults of Japan, Active Faults in Japan: Sheet Maps and Inventories (Revised edition). University of Tokyo Press, Tokyo. 437 pp. (in Japanese with English summary). Schwartz, D.P., Coppersmith, K.J., Fault behavior and characteristic earthquakes: examples from the Wasatch and San Andreas fault zones. J. Geophys. Res. 89 (B7), Sieh, K., The repetition of large-earthquake ruptures. Proc. Natl. Acad. Sci. U. S. A. 93, Tanna Fault Trenching Research Group, Trenching study for Tanna fault, Izu, at Myoga, Shizuoka Prefecture, Japan. Bull. Earthq. Res. Inst. Univ. Tokyo 58, (in Japanese with English abstract). Ueta, K., Inoue, D., Torigoe, Y., Deformational mechanism of upper Cenozoic system in the epicentral area of the 2004 Mid- Niigata Prefecture earthquake insights from physical modeling. Zishin (J. Seismol. Soc. Jpn.) 58, (in Japanese with English abstract). Usami, T., Materials for Comprehensive List of Destructive Earthquakes in Japan, [416]-2001 (Latest Edition). University of Tokyo Press, Tokyo. 605 pp. (in Japanese). Watanabe, M., Tsutsumi, H., Suzuki, Y., Kim, H.-Y., Sato, N., :25,000 Active Fault Map in Urban Area, Ojiya. D.1-No.388, Geographical Survey Institute, Tsukuba, Japan. Watanabe, M., Suzuki, Y., Ito, T., Fault model for the 2004 Mid- Niigata Prefecture earthquake built on the distinct tectonic landforms. Zishin (J. Seismol. Soc. Jpn.) 58, (in Japanese with English abstract). Wells, D.L., Coppersmith, K.J., New empirical relationships among magnitude, rupture length, rupture width, rupture area, and surface displacement. Bull. Seismol. Soc. Am. 84, Yanagisawa, Y., Kobayashi, I., Takeuchi, K., Tateishi, M., Chihara, K., Kato, H., Geology of the Ojiya district. With Geological Sheet Map at 1:50,000, Geological Survey of Japan, 177 pp. (in Japanese with English abstract 6 p). Yeats, R.S., Prentice, C.S., Introduction to special section: Paleoseismology. J. Geophys. Res. 101 (B3),