Paleoearthquakes of the Düzce fault (North Anatolian Fault Zone): implications for earthquake recurrence Pantosti, D. a, *, S. Pucci a,b, N. Palyvos a, P.M. De Martini a, G. D'Addezio a, P. E.F. Collins c, C. Zabcid a Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Rome, Italy b Dipartimento di Scienze della Terra, Università degli studi di Perugia,Piazza Università, 06123 Perugia, Italy c Geography & Earth Sciences, Brunel University, Uxbridge UB8 3PH, UK d ITU Istanbul, Turkey * Corresponding author. Current address: Istituto Nazionale di Geofisica e Vulcanologia, Via di Vigna Murata 605, 00143 Rome, Italy , Tel.: +39-6-51860483, Fax: +39-6-51860507. E-mail address: [email protected] (D. Pantosti). Abstract The November 12, 1999, Mw 7.1 earthquake, ruptured the Düzce segment of the North Anatolian Fault Zone and produced ca. 40 km-long surface ruptures. To learn about recurrence of large surface faulting earthquakes on this fault, we undertook paleoseismological trench investigations. We found evidence for repeated surface faulting paleoearthquakes pre-dating the 1999 event. Dating was based on radiocarbon and 210Pb analyses as well as on archaeological considerations. By merging information obtained from all the trenches we reconstructed the seismic history of the Düzce fault for the past millennium. We correlated coeval events between different trench sites under the assumption that, similarly to the 1999 event, paleoearthquakes ruptured the whole Düzce fault. Besides the 1999 earthquake, prior surface faulting earthquakes are dated as follows: AD1685-1900 (possibly end of 19th century); AD1685-1900 (possibly close to AD 1700); AD1185-1640; AD685-1220 (possibly AD800-1000). Thus, the AD1719, AD1878 and AD1894 historical earthquakes, may have ruptured the Düzce fault and not the faults they are usually associated to or, alternatively, a cascade of events occurred on the Düzce and nearby faults (similarly to the Izmit and Düzce 1999 earthquakes). Five events since AD 685-1220 (possibly AD800-1000), would yield an average recurrence time for the Düzce fault, of 200-325 yr (possibly 250-300 yr). The three most recent earthquakes, including 1999, occurred within 300 yr and may be suggestive of clustering. Assuming that the average 1999 slip is characteristic for this fault, the above recurrence times yield slip rates of 6.7-13.5 mm/yr. 2 1. Introduction The November, 12, 1999, Mw 7.1 earthquake ruptured the Düzce fault segment of the North Anatolian Fault Zone (NAFZ, fig. 1A) producing a ca. 40 km long surface rupture with up to 5 m right-lateral offset (2.7 m average) and up to 2.5 m vertical throws (Akyuz et al., 2000 and 2002; Pucci et al., submitted). This earthquake is considered to have been triggered by the Mw 7.4 Izmit earthquake that occurred 3 months earlier (August 17) on the fault segment to the W of the Düzce one. At a regional scale, the Düzce fault is located just west of the Bolu basin, where the NAFZ starts splaying into two main strands, the Düzce/Karadere to the north and the Mudurnu to the south, to splay again into three major strands in the Marmara Sea (NNAF, CNAF and SNAF in fig.1A, Wong et al., 1995; Armijo et al., 1999; Okay et al., 1999). The Düzce/Karadere strand, together with the Mudurnu fault to its S, accommodate most of the 2-3 cm/yr present-day strain of the NAFZ (Straub et al., 1997; Reilinger et al. 1997 and 2000; Ayhan et al., 1999 and 2001). The Mudurnu segment ruptured entirely during the 1957 and 1967 earthquakes, whereas high seismic potential of the Düzce fault was recognized well before 1999 by Barka and Erdik (1993) that considered this fault the possible source of a near-future earthquake. The Düzce fault has an average E-W trend and a clear geomorphic expression, being the boundary between the Quaternary Düzce and Kaynasli basins to the north and the Paleozoic- Eocene rocks of the Almacik block to the south (fig. 1B). The eastern and western boundaries of the 1999 earthquake rupture appear to be structurally controlled. To the west, the ca. E-W Düzce fault forms a releasing fault junction (sensu Christie-Blick and Biddle, 1985) with the NE-SW trending Karadere section (Pucci et al. submitted), which ruptured during the Izmit earthquake. To the east, the Düzce fault joins the eastern 3 single trace of the NAFZ via a ca 10-15 km-wide, right-releasing step-over involving the WNW-ESE trending Bakacak and Elmalik faults (fig. 1A) (Barka and Erdik, 1993; Altunel et al., 2000; Barka et al., 2001; Hitchcock et al., 2003). Both these step-overs appear unfavourable to rupture propagation and possibly represent persistent barriers to earthquake ruptures. Although Turkey has one of the richer records of historical seismicity in the Mediterranean, no clear evidence for historical earthquakes produced by the Düzce segment of the NAFZ during the past centuries has been found. This is probably due to the scarce density of population and lack of cultural settlements in historical times in the Düzce region. The only historical earthquakes that are known to be close enough to be potentially associated to the Düzce fault are: AD967, AD1719, AD1754, AD1878, AD1894, and (Ambraseys and Finkel, 1995; Ambraseys, 2002) (fig.1A). Interestingly, local people living near the eastern part of the Düzce fault recall their grandparents telling them about an earthquake at the end of the 19th century, producing ground ruptures exactly where these occurred in 1999. Because historical information is very limited, knowledge about recurrence of large earthquakes on the Düzce fault can be derived only from paleoseismology. Soon after the 1999 earthquakes several paleoseismological investigations were carried out at different locations along the fault (fig 1B). On the basis of trenching Hitchcock et al (2003) find evidence for three to five paleoearthquakes in the past 2100 years, with a recurrence interval ranging from 300 to 800 years and the penultimate event occurred about 300 years ago. Komut (2005) recognized 6 paleoevents since BC1750 with the one prior to 1999 occurring during the past 500 yr. Emre et al. (2001, 2003a and 2003b) found evidence for three paleoearthquakes since AD665, the oldest and the youngest of which are dated AD665-1050 and AD1650-1750, respectively. Finally, by 4 paleoseismological geo-slicing and coring investigations, Sugai et al. (2001) developed a surface faulting history for the past 2 millennia at a site in the western part of the fault. Here, these authors recognize four possible/probable paleoearthquakes preceding 1999 and suggest an average recurrence time of 4-500 yr. In this paper we present the results of trenching at five sites performed during the E.U. project RELIEF, we compare them with results from previous studies, and discuss their implications for the seismic behaviour of the Düzce segment of the NAFZ. 2. Trenching the Düzce fault We excavated a total of 10 trenches, 7 across the fault and 3 fault-parallel ones, at five different sites along the Düzce fault (Fig. 1B). High water table and the lack of sites with slow and continuous sedimentation made the site selection and trench interpretation problematic. Because of the type of sediments and sedimentary structures crossed, no piercing points to measure individual or cumulative horizontal coseismic offset were found. Dating of paleoearthquakes was based both on radiocarbon and 210Pb analyses. Both dating methods contain large uncertainties. Samples for radiocarbon dating were quite small with high possibility of reworking. Moreover, most of the trenched deposits appear to be younger than 300 yr. The past 300 yr are a very problematic time interval for radiocarbon dating because of the "radiocarbon plateau" produced by fossil fuel combustion (Suess effect, Bradley, 1985) and increasing solar activity following the Maunder minimum (Stuiver and Quay, 1980). As a consequence of this, a precise age cannot be determined because measured radiocarbon ages in the 5 plateau calibrate with almost equal probability to any age within it. For this reason we limited the number of samples for 14C dating In this work we also experimented with the use of 210Pb analyses for dating colluvial and marsh deposits, following the sample preparation method used in Cundy et al. (1998). 210Pb derives from the decay of 222Rn, dating is based on the assumption that all the main sources for 210Pb (in situ, from the atmosphere, and from eroded material in the catchment) can be considered constant through time. If this is true, a near- exponential decline of activity with depth would be expected. Previous work by Cundy and Stewart (2004) highlighted the difficulty of developing a precise geochronology based on short-lived radionuclides in depositional settings where sediment texture is very variable and deposition has occurred in pulses. It is possible, however, to derive an outline chronology, and to use variations in the 210Pb activity to help identify episodes during which older material was being remobilised in the local sedimentary environment and delivered to the sampling site. 2.1 The Kaynasli trench (KAY) This trench was excavated across the 1999 ruptures in the floodplain of the Asarsu river (KAY, fig. 1B) at the western edge of a sag pond that is artificially drained by a nearby man-made channel. In this area the dextral and vertical offsets of the 1999 ruptures were 0.7-1.7m and 0.3m, respectively (Akyüz et al., 2000 and 2002; Pucci et al., submitted). The trench was about 18 m long and 2 m deep and exposed a sequence of predominantly fine sediments (silt and clay), with intercalated layers of sand and pebbles. Fluvial gravel was exposed at the bottom of the central part of the trench. A description of all stratigraphic units is given in fig. 2. Four charcoal samples were AMS 6 dated from units d, e, f, and g (samples KW-45, KW-02, KW-20, KE-08, see table 1 and fig. 2). They yielded ages ranging from modern to AD 685-890. Two main fault zones were exposed on both walls of the trench (1 and 2 in Figure 2A and B). They were composed of several splays, most of which ruptured in 1999. A third fault zone (3 in Figure 2A), that was not activated in 1999 was located in the southern part of the trench and it is also highlighted by a sharp change in color of units h and g, possibly because of preferential water circulation in the fracture zone.. In 1999 the rupture reached the surface along at least one of the main branches fault zones 1 and 2 and was subsequently sealed by the post-event unit b in fault zone 1. During this event, clay and fine sand were injected along the rupture (fault zone 1, west wall, unit z) and at the c/e and d/e contacts. On the basis of stratigraphic and structural relations, we find evidence for three surface faulting paleoearthquakes before 1999 (Kay2 to Kay4 in the following and in fig. 2). Evidence for the penultimate earthquake, Kay2, are fault terminations below unit d as well as the presence in both fault zones of large cobbles and small boulders (unit k) completely unrelated to the surrounding stratigraphy and buried by unit d (see also figs. 2C and D). These cobbles are interpreted as man-made fill of ground fissures formed during an earthquake rupture along which water is flowing, as indicated by alteration coatings on the cobbles and boulders of the unit k and by the presence of laminated fine sediment at the base of channel-like features. The pre-penultimate event, Kay3, was recognized only in fault zone 1 of the western wall, where two fault splays deformed the trenched sequence up to unit f and the base of unit e. The top part of unit e has not been affected. Thus, we place the event horizon somewhere near the base of unit e. Evidence for an older event, Kay4, was found at fault zones 2 and 3. On the western wall, the northern strand of fault zone 2 offset only g and older units. 7 Similar features define Kay4 also in fault zone 3. On the basis of the dated samples, the timing of the above paleoearthquakes can be constrained as follows: Kay2 is younger than AD1475, Kay3 occurred between AD1035 and 1640, and Kay4 between AD685 and 1220. 2.2 The Mengencik trench site A total of six trenches were excavated at this site (fig.3) across the 1999 ruptures that attained ca. 3.7m dextral and 0.4m vertical offset (Akyuz et al., 2000 and 2002; Pucci et al., submitted). Five of the trenches were located in the western part of the site and one in the eastern. In the following we present results only from the across-fault trenches (Men1, Men5 and Men6, fig. 3B) because observations in the fault-parallel ones were inconclusive. 2.2.1 The western trenches Men1 and Men5 The 1999 earthquake surface fault at the western part of the Mengencik site crosses slope-wash deposits and small coalescent fans composed mainly of silt. Overall, the fault trace produces relative subsidence of the southern side, where ponding is observed against the scarp, especially where the rupture forms small grabens. Repetition of surface faulting events produced the formation of fault-parallel ridges of different size, which clearly control the drainage pattern (fig. 3). Trench Men1 was excavated across the 1999 ruptures (fig. 4 A, B) where they exhibit an apparent reverse component. The trench exposed a monotonous fan aggradation sequence of silt, fine sand and clay with rare gravel intercalations, derived from the marly deposits forming the range located to the south of the site (see fig. 4 for 8 description of stratigraphic units). Some layers were rich in organic matter, possibly due to periods of surface stability (soil formation) or ponding events related to pre-1999 surface ruptures (judging from the fact that ponding occurred at this site after the 1999 earthquake). Trench Men5 was excavated about 25 m east of trench Men1 (fig. 3), where the 1999 ruptures formed a graben 1 to 2 m-wide. Similarly to Men1 it exposed a monotonous fan aggradation sequence (fig. 5). Most of the stratigraphy can be directly correlated to that of trench Men1 and, in fact, the same labelling was used for the upper correlative layers. In both trenches the fault zone was about 2 m wide and consisted of several splays in an arrangement that reflects the type of structure observed at the surface: positive flower structure at Men1 and negative at Men5 (figs. 4 and 5). Most of the splays were reaching the surface, indicating that they ruptured also during the 1999 earthquake. Only fault splay B in Men 1 (fig.4 A and C) did not rupture in 1999 and thus it provides good evidence for a previous surface faulting event. To provide a timeframe to the trenched sediments, we collected samples for 14C, and 210Pb analyses. Radiocarbon dating was quite difficult because of the extremely small size of the available samples and important weathering and reworking. The dated samples (see Table 1 and Figs. 4 and 5) suggest that the upper 60-70 cm of sediments in the southern part of both trenches were deposited maximum during the past ca. 500 yr (samples Men1-W21 and Men5-W20) with the upper 40 cm being younger than AD1700 (possibly AD1880 - 60%probability, sample Men1-W23). Sample Men5-W16 from the bottom of the sediments that were trapped in the shear zone indicates that these are ca. 800 yr old. One further sample dated from unit d of trench Men5 (Men5-W13) yielded an out of sequence age indicating occurrence of reworking of material in this environment. 9 210Pb analysis was completed on the sequence to assess the validity of the radiocarbon chronology and to better constrain the age of the younger part of the stratigraphy in trench Men1. Sampling was performed in the southern part of the trench, from the surface to the bottom of the trench (fig. 4A); analyses were performed for the upper 1.24 m. 210Pb activity shows a general decline with depth (column 1 in fig. 4D). 210Pb activity is noticeably higher in the top 40 cm of the sediment sequence than lower in the sequence. This corresponds mainly to sediments that have buried a paleosol (unit e in fig. 4). Activity levels below this are low, with a slight increase with depth, suggesting in situ production. It is possible to generate 210Pb age estimates for the upper 40 cm using the constant rate of supply model (CRS; Appleby 2001). However, the variable activity levels mean that any derived chronology must be considered with caution. Low activity levels are associated with increased sand content, and probably reflect higher energy episodes on the fan surface. With this caveat, it is still possible to produce an outline chronology (column 2 in fig. 4D) which suggests that the upper 40cm accumulated since the mid 20th Century. Re-running the model with high and low values omitted suggests sediment from 40cm depth was deposited since AD1940-1960. This gives an average accumulation rate of around 0.75 cm a-1, with the highest rates (c. 1.5 cm a-1) possibly occurring since ca. AD1986. The stratigraphic unit at 40cm depth (unit e in fig. 4), which exhibits very low 210Pb activity, is characterised by a blocky ped stuctures and is interpreted as a probable paleosol.210Pb activity increases below the paleosol, probably reflecting in situ 210Pb production. A single 14C date of AD1450-1630 (sample Men1-W21) from below the palaeosol supports the hypothesis that this soil represented a phase of relative surface stability and pedogenesis between the 17th-19th Centuries. Similarly, a 14C date from immediately above the palaeosol yields a modern age, possibly younger than AD1880 10