Notice: Wiley Online Library will be unavailable on Saturday 27th February from 09:00-14:00 GMT / 04:00-09:00 EST / 17:00-22:00 SGT for essential maintenance. Apologies for the inconvenience.
 This paper presents a comparison between the pattern of surface ruptures produced by a single earthquake and patterns of cumulative deformation. We performed a detailed study of the 1999 earthquake coseismic ruptures and of the long-term tectonic landforms in a key area of the Düzce fault segment of the North Anatolian fault. We observed a scale-independent en echelon arrangement of the coseismic surface ruptures. As a whole, the long-term geomorphic expression of the Düzce Fault near the 1999 ruptures is evidence of the principal slip zone at depth that accommodates the bulk of the displacement during an individual rupture event. This may stay localized through many rupture episodes with persistent geometry and kinematics. The long-term tectonic and geomorphic expression of the fault in a broader area around the 1999 ruptures defines a wider deformation zone. In fact, an old and complex fault arrangement has been mapped, partially coinciding with the 1999 rupturing fault, suggesting that the 1999 ruptures are an incomplete expression of the long-term Düzce fault system. The relationships between the coseismic and the old fault systems suggest an evolution of the fault pattern trough time, with a tendency to simplify a geometric complexity into a straighter, mature trace. The integrated investigation of long-term tectonic morphologies and structural pattern offers a noteworthy frame to interpret the coseismic rupture kinematics and clarifies their complexities. Moreover, to fully understand the principal slip zone at depth, this work shows the importance of the study of strain distribution pattern and evolution of surface rupturing faults.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 Large earthquakes are a unique opportunity to shed light to seismogenic sources and the related tectonic processes. They provide a snapshot on the long-term tectonic processes that are controlling the evolution of the tectonic structures of a specific region. Combined observations of coseismic surface ruptures, tectonic geomorphology and earthquake geology can identify the principal deformation zones at the surface (PDZ) that, as expression of a localized slip plane at depth (PSZ, principal slip zone), can improve the understanding and characterization of seismogenic structures and of their evolution in space and time. These disciplines have developed significantly during the past 2 decades, but there is still a need for detailed case studies to cover the wide variety of tectonic/geodynamic settings, deformation rates, geological and geomorphic environments that seismogenic zones are associated with. Freshly collected observations on recent earthquake ruptures and on their geomorphic setting are therefore a critical input to improve and to develop seismogenic source models. Under this light, recent large earthquakes with sizable coseismic effects at the surface represent excellent case studies.
 We investigated the area struck by the 12 November 1999 Mw 7.1 earthquake that ruptured the Düzce segment of the North Anatolian Fault Zone (NAFZ) (Figures 1 and 2) . This area is favorable for this type of studies because of the occurrence and preservation of the spectacular 1999 surface ruptures coupled with impressive tectonically driven landforms. Thus the Düzce fault is a natural laboratory where coseismic deformation, tectonic landforms and geological structures can be investigated and compared in order to reconstruct the fault zone history and to understand if forecasting some of the characteristics of the future earthquake rupture could be possible.
 In this paper we discuss the results of an investigation performed in a key area of about 70 km2 along the central part of the Düzce fault segment (Figure 2). This area was selected because of the extensive presence of faulted Pleistocene and Holocene continental deposits that constrain timing and amount of ongoing deformation. We describe and analyze the 1999 coseismic rupture patterns along with the long-term (Late Holocene to Late Pleistocene) landforms and structures as result of cumulative fault-related deformation. Rupture style and complexity were investigated by means of field survey, structural geological analysis, and aerial photo interpretation. We carried out a detailed tracing and measurement of the 1999 ruptures, coupled with geological and geomorphological mapping at a 1:25,000 scale. Geomorphological observations were made on 1:18,000 and 1:35,000 scale aerial photographs, a 20-m-resolution Digital Elevation Model (interpolated from 1:25,000-scale topography, 10-m digital contours) and standard morphometric derivatives (hill-shaded and slope angle maps), as well as during the field survey.
 In the following, after a brief introduction to the tectonic framework of the North Anatolian Fault Zone, we describe (1) the patterns of the coseismic ruptures at different scales and (2) the long-term fault patterns that are derived from cumulative tectonic landforms, along the coseismic ruptures and at a broader area around them. Subsequently, we make a comparison between these two data sets, in order to understand whether their relationships reveal any space and time evolution of the fault pattern at the surface.
2. Overview on the North Anatolian Fault Zone (NAFZ)
 To the east of the town of Bolu (Figure 1), the NAFZ is formed by a main single trace but, west of it, it splays into two main strands, the Düzce and the Mudurnu fault segments, where GPS data indicate that the former accommodates up to 10 mm/yr [Ayhan et al., 1999]. Farther west (Figure 2a), the NAFZ splays again into three major strands and the northernmost one is considered to accommodate most of the present-day strain [Barka and Kadinski-Cade, 1988; Straub et al., 1997].
 The number of large, historical earthquakes attributed to the NAFZ testifies that the present-day high strain rate is seismically accommodated and places this fault among the most active strike-slip faults worldwide [Ambraseys, 1970; Ambraseys and Finkel, 1995; Ambraseys, 2002]. Seismicity along the active segments of the NAFZ is characterized by frequent moderate to large earthquakes (M > 7) with focal mechanisms that show essentially pure right-lateral strike-slip solutions [Canitez and Ucer, 1967; McKenzie, 1972] (see also Reinecker et al., 2004 release of the World Stress Map, available at http://www.world-stress-map.org).
 The 1999 earthquakes are the most recent occurring on the NAFZ in the 20th century. The first of these earthquakes occurred on 17 August and struck the Izmit region, west of the Marmara Sea. This Mw 7.4 (USGS) earthquake nucleated at a depth of 13 km and its focal mechanism (Harvard CMT) is consistent with right-lateral movement along an E-W strike-slip fault (Figure 2a). Maximum surface dextral offsets exceeded 5 m and surface rupture extended along the fault for a total rupture length of ∼110 km [Barka, 1999; Barka et al., 2000, 2002]. Three months later, on 12 November [Muller et al., 2003; Utkucu et al., 2003], the Mw 7.1 Düzce earthquake (USGS, KOERI) occurred. The focal mechanism solution shows almost pure, dextral strike-slip movement on an E-W nodal plane, dipping 54° to 64° to the north, with rake between 177° and 167° (USGS, Harvard CMT [Tibi et al., 2001]) (Figure 2a). The pattern of aftershocks, recorded between 12 November 1999 and 20 November 1999 by the TÜBİTAK-MAM local network, is organized in clusters that are all located north of the surface trace of the Düzce fault [Özalaybey et al., 2000; Milkereit et al., 2000; Cakir et al., 2003]. This earthquake produced right-lateral surface ruptures for a total length of ∼40 km and a maximum dextral offset of 5 m [Akyüz et al., 2000, 2002].
3. Düzce Fault Segment of the NAFZ
 The Düzce fault segment has an average E-W trend and a clear geomorphic expression for about 35 km. It separates the Quaternary Düzce and Kaynasli basins to the north from the Paleozoic-Eocene formations of the Almacik block to the south (Figure 2b) and, overall, produces uplift of the range front to the south, with respect to the plain to the north. In the western part of the basin, the E-W Düzce fault splays out from the NE-SW trending Karadere section, that ruptured during the 1999 August earthquake. The eastern termination of the Düzce fault is not clearly defined and, according to A. A. Barka and M. Erdik (unpublished report, 1993), it may join the eastern single trace of the NAFZ via a right-releasing step-over, involving the WNW-ESE trending Bakacak and Elmalik faults (Figure 2a).
 In general, the 40-km-long coseismic rupture is simple and narrow with a deformation zone 0.5 to 5 m wide with local exceptions where the width reaches 50 m. The coseismic ruptures do not exactly run at the mountain-piedmont interface, but affect both basin infill deposits and bedrock with a slightly northward convex trajectory. Owing to the strike variability of the rupture, transpressional and transtensional structures formed along the fault trace [Akyüz et al., 2002; Hartleb et al., 2002] (Figure 2b). On the basis of geometry and the coseismic slip distribution, Akyüz et al.  divided the entire Düzce rupture into four sections (Figure 2b), each of them showing a symmetric slip tapering off toward both ends. The westernmost sections (1 and 2, Figure 2b) are separated by a releasing bend with a significant normal component of displacement, whereas the eastern sections (2, 3, and 4, Figure 2b) are separated by two 1-km wide, transpressional left step-overs. The area investigated in this paper includes the two westernmost sections of the fault, where the 1999 coseismic ruptures enter the Düzce basin and affect Quaternary continental deposits of key importance to record the cumulative fault-related deformations (Figure 3).
4. Expression of the Düzce Fault
4.1. Coseismic Expression of the Düzce Fault
 We mapped in detail the 1999 earthquake ruptures between Efteni Lake and Beykoy (Figure 3) and collected data on: (1) structural patterns of fractures and faults; (2) geomorphic modifications induced by the earthquake; and (3) coseismic offset of items such as roads, fences, tree lines, channels, streams, buildings, etc. (∼80 new measurements, additional to those made by Akyüz et al. [2000, 2002] soon after the earthquake, see Tables 1 and 2). The offset measurements were collected putting attention in separating single from multiple slip events. The 1999 offsets were mostly measured using recent manufacts that were crosscut by the still visible ruptures. These manufacts provide reliable piercing points for both horizontal and vertical offsets. We measured vertical displacements on the 1999 free face at locations where we could exclude the possibility of apparent vertical displacements due to juxtaposition of topographic features (i.e., locations where the ground surface was flat or regularly sloping, without undulations). Landslides and gravitational movements occurring where the coseismic rupture crossed steep, instable slopes (Figure 3) were also carefully mapped to avoid the inclusion of surface ruptures caused by gravitational collapse in the coseismic slip measurements.
Table 1. Raw Data of the Coseismic Horizontal Offsets
 On the basis of sharp changes in the strike direction of the coseismic ruptures, the fault trace, in the study area, can be divided into four 3- to 4-km-long subsections. These are named from west to east: the Lake Efteni, Yenikoy, Cinarli, and Beykoy subsections that trend WNW-ESE, WSW-ENE, E-W and WSW-ENE, respectively (Figure 4a). Even though the kinematics of the Düzce fault is dominantly strike-slip, field observations indicate the presence of widespread vertical motion, which was also observed in InSAR data [Cakir et al., 2003]. The ruptures along the first two subsections produced relative subsidence of the block to the north. Conversely, the Cinarli and Beykoy subsections produced topographic offset with relative subsidence of the block to their south (Figure 4). This had impact on local north-flowing drainages, which were dammed at several points by coseismic scarps across them.
 Overall, the slip measurements (Figure 4b and Tables 1 and 2) exhibit a great variability along strike, with 4.5-m maximum dextral offsets (2.7-m average) and 2.5-m maximum vertical offset. The characteristics of the slip distribution are: (1) the maximum dextral offset of the Efteni Lake is higher than that of the other sections, (2) the average dextral offset of the Efteni Lake and Cinarli subsections is higher than that of the other sections, (3) the vertical component reaches its maximum along the Efteni subsection, and (4) the dip-slip direction changes at the boundary between Yenikoy and Cinarli subsections (Figure 4b). Modeled slip distribution at depth, obtained by joint inversion of GPS and InSar data, indicates a maximum right-lateral slip of more than 4 m, in the study area [Cakir et al., 2003; Burgmann et al., 2002]. The great variability of the strike-slip distribution with lower values points could denote losses of slip because of the anastomosing of the rupture, possibly due to scarce connection between the ruptured fault at depth and the surface ruptures (mechanical problems), or because of the plastic deformation in the unconsolidated deposits that overlay the bedrock. If we consider that the coseismic strike-slip distribution at the surface is representative of the movement along the seismogenic fault (see focal mechanism), the maximum dextral offsets, measured at the surface, would better represent the fault slip at depth. This is not true for the coseismic dip-slip at the surface, which includes local transtensional or transpressional components and may also contain an unknown component related to possible gravitational effects (i.e., sediment compaction).
 The expression of the 1999 coseismic ruptures is predominantly mole-track type. Each mole-track deformation zone, 5 to 15 m wide and 30 to 100 m long, is referred to here as a single displacement zone (SDZ) and is the morphological expression of a complex dextral fault zone, mainly consisting of synthetic, 1- to 10-m-long, strike-slip Riedel shears (Figure 5). These features are typical of poorly consolidated sediments (i.e., continental Quaternary deposits) that overlay a rigid bedrock (i.e., Paleozoic-Eocene bedrock), as observed by Riedel  in analogical deformation experiments. The main characteristics of the system are an over-stepping, en echelon array of left-stepping synthetic Riedel shears (R-shears) and a right-stepping array of compressional fold axes (Figures 5b and 5c). Antithetic Riedel shears (R'-shears) are rarely observed even in the transfer zones at the R-shears step-overs. At places, where the width of the deformation zone increases, Riedel systems develop a second array of right-stepping synthetic shears (P-shears or Thrust-shears, Figure 6) that locally join or end up to the R-shears [e.g., Tchalenko, 1970; Wilcox et al., 1973; Naylor et al., 1986]. At a few sites, the zone of deformation consists not only of a set of en echelon R-shears, but also P-shears together with compressional fold axes. Here P-shears have grown and linked to R-shears, forming shear lenses, and afterward Y-shears have also developed, in a typical anastomosing pattern of a through-going wrench fault, parallel to the SDZ [e.g., Woodcock and Schubert, 1994; Naylor et al., 1986] (Figure 6). Locally, arrays of fractures form as well, with a prevalent extensional kinematics (T-fractures).
 At smaller scale, the 30- to 100-m-long SDZs themselves are also organized in slightly overlapping en echelon arrays. At this scale the SDZs constitute a left-stepping synthetic Riedel system spaced 3 to 20 m, which shows a clear continuity for a length of 300 to 800 m. The envelope of this en echelon array defines a principal displacement zone (PDZ) that has a strike direction diverging 5°–10° from the SDZ and produces a consistent topographic offset with the uplift of one of the fault blocks (Figure 7). The alignment of the PDZs defines the average location, strike and relative movement of the 1999 surface fault trace of each subsection of Figure 4a in which it is located.
4.2. Long-Term Expression of the Düzce Fault
 As already mentioned, one of the reasons for selecting the study area was the fact that the 1999 ruptures affected Holocene and Pleistocene deposits. Because of this, the Quaternary deformational history of the Düzce fault can be reconstructed by investigating the setting of their depositional and tectonic landforms. In fact, the generally loose and soft recent deposits can develop well-expressed tectonic morphologies that record persistent landform modification due to faulting. Following this perspective and with the aim of highlighting the long-term tectonic setting and evolution of this part of the Düzce fault, we conducted geomorphic analysis, first along the 1999 rupture zone (hereinafter referred as near-fault), then in a broader area around it (hereinafter referred as far-field).
 The contribution of gravitational movements to the landform evolution and to the observed relief was considered with particular attention. Frequently, landslides are responsible for the degradation of the tectonic landforms (e.g., interrupting the continuity of fault escarpments). Most of the landslides, both old and generated by the 1999 earthquake, are located along the fault-related topographical anomalies testifying that, in the study area, the slope evolution is strongly linked to tectonics.
4.2.1. Tectonic Landforms in the Near-Fault
 In the following section, we discuss the geomorphological expression of the four subsections inferred from the coseismic ruptures (see previous section). This subdivision appears, in fact, to fit also the long-term fault trace, since each subsection is associated with homogeneous tectonic landform assemblages that are clearly distinct from those of the adjacent subsections.
 The 1999 ruptures of the Efteni Lake subsection (Figure 3) coincide with the base of a steep stretch of the Almacik rangefront, and is characterized by the lack of well-developed piedmont deposits. NNE-facing triangular facets (Figure 8) are present along the range, deeply incised by canyons that exhibit hanging terraces at their outlet (x and y in Figure 8a). This setting is clearly indicative of a narrow fault zone (PDZ) located at the base of the range front that is characterized by an important and persistent NNE-trending dip-slip component.
 The trace of the Yenikoy subsection (Figure 3) coincides with the northward limit of a composite Pleistocene bajada (i.e., alluvial fans merging into a single apron of sediments that covers the foot of the range), but also active Holocene fans, deposited in subsequent incisions of the Pleistocene fan surface (Figure 3). This bajada has a topographic profile with a steep fanhead characterized by an abrupt transition to the fluvio-lacustrine deposits of the Düzce plain, indicating that the bajada is no longer related to the present dynamics of the latter. The generally linear contact between the bajada and the plain, is a c.a. SW-NE degraded and indented escarpment. The presence of this escarpment suggests relative subsidence north of the Düzce fault, a fact further indicated by the stream incisions across the bajada. The morphology of the bajada-bounding escarpment is not simple in its details owing to dissection by north-flowing drainages, but in general it can be subdivided into two main right-stepping elements. The westernmost is a fairly continuous, about 10-m-high, escarpment (from y to z in Figure 9) that is up to 20 m high in its central part. The eastern section is a substantially higher scarp averaging 15 m that is more degraded to the west where is perceptible only as a linear slope gradient variation. This becomes very prominent to the east (from k to w in Figure 9) where it rises up to 50 m above the Düzce plain (x in Figure 9). The right step-over, recognized in the escarpment, is also illustrated vividly by morphologic ridges along the two distinct stretches. These ridges, 300 to 1000 m long and 10 to 40 m high, have elongated shapes with long axes paralleling the escarpment elements they are associated with. The ridges have asymmetric cross profiles, with steep northern flanks and gentle south flanks, that express a smooth (ridges 1 and 2 in Figure 9) or abrupt (ridges 3 and 4 in Figure 9 and see profile A-A′ in inset) transition to the bajada surface. These observations suggest that the Yenikoy subsection is formed by two main, ∼500 m spaced, WSW-ENE striking, ∼2-km-long right-stepping fault traces (y-z and k-w in Figure 9) that have been producing uplift of the block to their south.
 The Cinarli subsection is associated with a prominent, 3 to 15 m high, south-facing escarpment across the bajada deposits (Figure 3 and Figure 10b). This ∼E-W trending escarpment is formed by left-stepping en echelon, 200- to 700-m-long and ∼100-m-spaced scarps and constitutes the northern boundary of a flat and depressed area. The escarpment has a topographic profile characterized by a sharp asymmetric warp: A steep and short southern slope contrasts with a gentle northern slope, whose gradient is coincident with the bajada depositional gradient (see profile B-B′ in inset of Figures 10a and 10b). In some cases, the scarps of the Cinarli subsection act as natural dams for northward flowing drainages, being thus responsible for the formation of ponds, flooding and alluvial deposition. The major dammed areas are drained by large artificial channels (Figure 10a). In the easternmost part of the subsection, two escarpment-bounded depressions also trap Holocene pond and fan deposits against the scarp (x in Figure 10a). North-flowing drainages are characteristically deflected in an apparent left-lateral sense, since they cross the bajada through the escarpment step-overs, where the damming is less effective (i.e., the scarps are less high where they overlap; Figure 10a). The complex setting described above is suggestive of a set of short, up to 1-km-long left-stepping fault strands forming the Cinarli subsection that have been producing subsidence of the southern block.
 Farther east, the Beykoy subsection contrasts with the Cinarli subsection because it is associated with well-expressed, 400- to 800-m-long, WSW-ENE trending elongated tectonic ridges, along the base of the range front (Figures 3 and 11). Most of these ridges are made of a Paleozoic-Jurassic granitoid unit, are ∼70 m high and have symmetric profiles and steep slopes (Figure 11b). Only one of the eastern ridge is made of recent deposits of a remnant of a Holocene bajada. Compared to the bedrock ridges, it is lower (15 m high) and asymmetric, with a steep south-facing slope. Narrow and hanging WSW-ENE linear valleys, up to 40 m deep, have developed between the elongated ridges and the base of the range front. The north-flowing drainages show clear anomalies. Right-hand stream deflections of about 100 m (x in Figure 11a) occurred because of: the growth and offset of the ridges that act as natural dams to their flow; incision of both linear valleys and ridges that forced the streams to maintain their bed; and low stream power not enough to erode and to straighten their course. Tectonic influence on drainage patterns is also expressed by stream piracy along the fault (creating apparent sinistral deflections) and perched ponding areas (y and z, respectively in Figure 11). The setting of the Beykoy section suggests the existence of a WSW-ENE striking, persistent fault trace, located along the linear valley between the ridges and the range-front slope. A secondary parallel fault trace possibly bounds the northern side of the linear ridges, causing its relative uplift with respect to the Düzce plain.
4.2.2. Tectonic Landforms in the Far Field
 In this section, we enlarge the area of observation, from the near-fault features to a zone a few kilometers wide (Figure 12). In this broader view the long-term geomorphic characteristics of the Düzce fault, east of Lake Efteni subsection, indicates that the zone of deformation that separates the Almacik block from the Düzce plain is composed, besides the PDZ, of a complex WNW-ESE and SW-NE range front fault system (hereinafter referred to as RFS) (Figure 13a).
 The Pleistocene bajada is certainly the most significant geomorphic feature highlighting this fault system (Figure 3). It developed within a ∼2.5-km-wide and 7-km-long (b-c-d-f in Figure 12) fault-controlled triangular embayment of the Almacik range front. In Figure 12, stretch b-f-g of the RFS is essentially the extension towards the ESE of the Lake Efteni fault subsection. In fact, the alignments of linear ridges, valleys (y in Figure 12) and bases of indented and fairly well preserved facets, up to 250 m high (x in Figure 12), eloquently depict a tectonic boundary. This tectonic boundary is exposed in one outcrop (z in Figure 12), where facet-parallel, right-lateral fault planes juxtapose Late Eocene against Paleozoic-Jurassic units (see Figure 3). The geomorphologic expression of stretch f-d of the RFS is equally clear-cut, as a SW-NE straight boundary, characterized by facet-parallel alignments and an abrupt slope gradient change, between the 500-m-high range front slopes and the lower relief in front of them (Figure 13a). This SW-NE escarpment cannot be attributed to litho-structural control, since foliation and lineation of the outcropping bedrock strike oblique to range front (Figure 12). In addition, we observed SW-NE trending fault planes in bedrock outcrops along the escarpment. Even though clear kinematic indicators could not be identified in these fault outcrops, on the basis of their orientation with respect to the Düzce fault stress field, we may infer a transpressive character for the stretch f-d of the RFS (Figure 13a). Smaller morphological discontinuities are suggestive of the presence of several minor faults inside this major embayment, bounding a sector of bedrock of intermediate elevation (1 in Figures 12 and 13a).
 We also found within the bedrock units evidence for a tectonic contact along the SW-NE trending strands of the RFS (g-h in Figure 13a). In general, the geomorphic expression of the g-h strand and the lack of fault-controlled recent deposition do not suggest its present activity, but we note that Akyüz et al.  reported up to 15-cm right-lateral offset ruptures along its easternmost part (Figure 13a).
5.1. Coseismic Ruptures Versus Long-Term Tectonic Landforms in the Near Fault
 We identify a general coherence at the scale of principal displacement zone (PDZ), both in location and relative movement, between the 1999 coseismic fault trace and Holocene cumulative tectonic landforms along it (Figure 14a). The PDZ closely mimics the long-term tectono-morphologic pattern. Also, the location and distribution of the PDZ vertical component of displacement appear to correlate well with the cumulative tectonic landforms, emphasizing the preexisting relief (e.g., x in Figure 14a). On the contrary, at a larger scale (SDZ and larger, see Figure 7), the subordinate coseismic structures diverge from the long-term morphologies: the en echelon ruptures arrays generally “climb” diagonally across the landforms.
 The only relevant incoherences between the coseismic fault trace and cumulative tectonic landforms recognized at the scale of the PDZ are found at the Yenikoy and Cinarli subsections. In the first case, on the basis of cumulative tectonic landforms, we identify a right-step between long-term transpressive fault structure (y in Figure 14a) that was violated by two coseismic ruptures (q in Figure 14a), linking the right step-over and forming a duplex. The two, E-W trending, linkage ruptures are associated with a cumulative tectonic relief much lower than that of the two main fault stretches. This suggests that this linkage may be a secondary transfer structure, a pure strike-slip section with no vertical offset or be a late Y-shear-like feature. In the other case, along the Cinarli subsection, one of the cumulative major fault scarps was not entirely reactivated and the rupture crosscut the morphological warp linking two consecutive left stepping scarps (w in Figure 14a).
 The good correlation between 1999 ruptures at the PDZ scale and cumulative tectonic landforms in the near-fault indicates a persistent fault trace and style of deformation through several seismic cycles. However, apart from the Efteni Lake subsection, the 1999 ruptures represent only a part of the fault pattern inferred from cumulative tectonic landforms (and thus of the PDZ). The overall long-term deformation zone in the near-fault (Figure 14a) appears not only wider but also more complex than the coseismic one, including other fault strands to the south and/or to the north, which did not rupture in 1999. This raises the possibility that secondary structures have hosted relatively small, blind displacements or ruptures that only rarely reach the surface. In one case coseismic ruptures (near Yenikoy, z in Figure 14a) occurred along these secondary structures and described a PDZ up to 500 m wide. These secondary coseismic ruptures demostrates a superficial anastomosing of the fault system that contributes to the growth of tectonic landforms together with plastic deformation of loose deposits. In fact, as discussed by Sibson , the width of the deformation zones at the surface (corresponding to PDZ in this paper) is mainly related to the presence of young cover sequences of alluvium and to free surface effects. The PDZ appears to be the surface expression of a localized slip plane at depth (hereinafter PSZ, principal slip zone; see Figure 14b), as observed along other seismogenic structures (e.g., Landers [Sieh et al., 1993; Johnson et al., 1994; Lazarte et al., 1994]; Dast-e Bayaz [Tchalenko and Ambraseys, 1970; Tchalenko and Barberian, 1975]). This implies that measures along the 1999 surface ruptures could underestimate, although in a minor part, the total amount of the coseismic slip.
 Although at the outcrop scale the SDZs outline the fault location, they cannot unequivocally be indicative of the kinematics of the fault subsections. Only at a few locations, a prevalent presence of T-shears or push-ups is suggestive of transtension or transpression, respectively. In general, the mole tracks ruptures, that typify the 1999 rupture style, do not permit conclusions as to whether the relative uplift of the two fault blocks is due to a compressional or extensional component of displacement. In contrast, the cumulative tectonic landforms express clearly the local stress fields of the four fault subsections. The fault subsections reflect tectonic landforms according to strike-slip movement along an irregularly curved fault that produces transtension at releasing bends and transpression at restraining bends (Figure 14b). In fact, the E-W trend of the dextral Düzce shear couple tends to create a local transtensional strain field along Z-shaped fault trajectories and a local transpressional strain along S-shaped fault trajectories. The morphology of the Efteni lake fault subsection fits well with a normal component of the fault motion. The strike of the Yenikoy and Beykoy subsections suggests a local transpressive strain field, which is compliant with the observed right-stepping pressure ridges (the surface expression of cumulative positive flower structures).The coseismic rupture along the Beykoy subsection shows a lowering of the southern block. This displacement is most probably relative, only reflecting the growth of the pressure ridges of these subsection. This does not exclude the uplift of the Almacik block with respect to the basin as depicted by geodetic data. The Cinarli subsection shows a gentle negative flower structure, in accordance with the fact that it only slightly deviates from the average trend of the Düzce fault. This gentle flower structure affects the bajada forming local basins (Figure 14b).
 Summarizing, in the near-fault, the observed coseismic and long-term structures (PDZ) show vertical tectonic component that is related to local change of the fault trajectory. This geometrical complexity produces subsidiary folding and thrusting with small wavelength (less than 1 km) that suggests they are shallow-rooted to the same master fault (PSZ from Sibson ).
5.2. Coseismic Ruptures Versus Long-Term Tectonic Landforms in the Far Field
 Examining the broader area around the coseismic ruptures we have seen that, with the exception of the Efteni Lake subsection, the RFS and the 1999 fault subsections (PDZ) form a complex fault zone (hereinafter Düzce Fault System, DFS), up to 2.5 km wide (Figures 12 and 13a) that could represent the Düzce fault damage zone [Sibson, 2003]. The northern strand of this fault system coincides with the 1999 ruptures and associated landforms, whereas branches of the RFS bound the Pleistocene bajada to the south. Most of the 1999 coseismic ruptures in the study area traversed the Pleistocene bajada (section b-c-d), instead of taking place along the RFS (Figure 13a). No significant ground ruptures were reported following the 1999 earthquake [Akyüz et al., 2002; Hartleb et al., 2002], or found during our own field reconnaissance, along the RFS sections. Besides the coseismic ruptures reported by Akyüz et al.  along the fault g-h of Figure 13a, only up to 20 cm right-lateral offset ruptures were decisively reported to us by 1999 earthquake local witnesses along 1 km of section f-g (Figures 12 and 13a). Although we cannot rule out shaking-related or sympathetic slip along preexisting fault traces that are embedded in and/or takes part to the damage zone, these observations may be suggestive also of some slip occurring in a wider area, that could contribute in the coseismic slip measurements being smaller along the main northern trace of the Yenikoy, Cinarli and Beykoy subsection (Figure 4b and strand b-c-d-e in Figure 13a).
 Long-term tectonic landforms along the Yenikoy and Cinarli sections suggest the persistence of the 1999 rupture complexity through several seismic cycles. However, these sections have hosted a smaller cumulative deformation and express younger morphologies with respect to the RFS. This suggests that: (1) the faults bounding the embayment to the south (b-f-d in Figure 13a) were the principal strands of the DFS in the past (sketches 1 and 2 in Figure 13b), (2) the Yenikoy and Cinarli subsections (b-c-d) started to be active during or after the bajada deposition (sketches 3 and 4 in Figure 13b) and now accommodate most of the Düzce fault slip.
 The northward migration of faulting, which the interpretation above implies, translates to a tendency of the fault to simplify a major geometric complexity towards a straighter trace. This is in agreement with the combination of displacement along Riedel and P-shears that leads to the formation of Y-shears, oriented parallel to the general direction of movement [Tchalenko, 1970; Woodcock and Schubert, 1994] that is presumably a mechanically more favorable setting. The presence of the intermediate faults between the 1999 ruptures and the RFS (x in Figure 13a), that look older than the long-term landforms along the coseismic ruptures, could be suggestive of a sequence of newer fault traces stepping out from the range front with time.
 In summary, the broader geologic and geomorphic setting shows a kilometer-scale wide fault system (DFS) that ruptured during the 1999 earthquake in correspondence of the PDZ (Figure 13a). The DFS at the surface has been evolving through time forming the younger Yenikoy and Cinarli subsections (Figure 13b) and contributing in the growth of the damage zone. Since there are no seismic lines available and the seismologic and the geodetic data of the study area can not resolve the complexities of the near-surface and of the far-field, is not possible to determine if parts of the RFS are directly connected to the principal fault (PSZ) at depth or not. However, even though the 1999 surface ruptures suggest that most of the strain is accommodated by the northern part of the DFS (PSZ), we cannot rule out the possibility that some amount of distributed permanent shear strain also develop coseismically at depth and that other structures composing the RFS, although secondarily, participate in the deformation.
5.3. Large-Scale Versus Small-Scale Patterns
 The patterns of 1999 ruptures and long-term deformation exhibit essentially the same en echelon arrangement of left-stepping transtensive structures and right-stepping transpressive structures, both at meter scale and kilometer scale. Thus it is noticeable that the faults arrangement is scale independent, persisting at the SDZ, at the PDZ and at the fault subsection scales (see Figures 5, 7 and 14). Looking at the coseismic ruptures at a meter scale it is possible to observe a reduced scale analogue model of the four rupture subsections in the study area: R-shears and T-shears at releasing bends (Efteni Lake subsection); P-shears and folds at restraining bends (Yenikoy and Beykoy) (Figure 15).
 A similar analogy can be proposed between the coseismic meter-scale rupture pattern to the broader scale long-term fault system (RFS) depicted in Figure 13a. The a-b-f and f-d-e range front faults of the latter can be interpreted as Riedel shears and P-shears or Thrust-shears, respectively, underlined in Figure 15.
 The central part of the Düzce segment of the North Anatolian fault was selected to explore the relationships between the pattern of the 1999 coseismic ruptures and long-term tectonic landforms, both along the ruptures and in a broad kilometer-scale zone around them. On the basis of geometrical considerations, the coseismic rupture in the study area is divided into four kilometer-scale subsections (Figure 4). Slip distribution in part shows a distinctive pattern: The maximum right-lateral and vertical slip (4.5 m and 2.5 m, respectively) are located in the western part of the study area (Efteni lake subsection, Figure 4b); a reversal of the dip-slip orientation is located at the boundary between Yenikoy and Cinarli subsections. The coseismic surface ruptures are organized in overlapping en echelon systems that can be recognized at different scales: (1) at the outcrop scale, where 1- to 10-m-long en echelon structures define single displacement zones (SDZ, Figures 5 and 6); (2) at the hundreds-of-meters scale, with 30- to 100-m-long en echelon SDZ, forming 300- to 800-m-long principal displacement zones (PDZ, Figure 7); (3) at the kilometer scale, where also the 1999 rupture subsections, both transpressional and transtensional, are arranged in an en echelon pattern. This scale invariance of the fault pattern of 1999 ruptures and long-term deformation suggests that it is meaningful and useful to compare natural analogues of deformation zones at different scales.
 On the basis of cumulative tectonic landforms we traced the long-term fault expression at the surface along the 1999 rupture zone (Figure 14). This long-term fault expression brings out that the bulk of the ongoing dextral displacement is localized mostly along the coseismic ruptures and, in addition, illustrates the existence of fault splays and a broad band of subsidiary deformation, both brittle or plastic. Because of this, the coseismic slip measured along the 1999 rupture trace could slightly underestimate the amount of seismic strain. The cumulative tectonic landforms along the 1999 rupture zone also indicate a fault pattern consistent with the coseismic one. Furthermore, they suggest different kinematics for the four coseismic subsections studied, which can be largely explained by the different orientation of each subsection with respect to the average strike of the Düzce fault. The transtensional and transpressional structures produce local vertical deformations that are characteristics of the subsections and that are related to shallow, near-surface features and do not represent the overall dip-slip component described by geodetic data or the almost pure strike-slip motion at seismogenic depth.
 On one hand, the geomorphic expression of the near-fault provides long-term evidence for slip localization, as observed along other active strike-slip fault zones. On the other hand, this case study shows that the width and the complexity of the surface deformation zone for the strike-slip faulting is affected by small irregularities of the fault trajectory transverse to the slip vectors, coupled with the presence and thickness of alluvial cover above the bedrock expression of the fault. Consequently, these observations draw attention to the strain partitioning between shallow transtensional and transpressional structures, above a main seismogenic structure.
 Cumulative tectonic landforms in the far-field zone of deformation (Figures 12 and 13a) suggest the Düzce Fault System (DFS) is complex long-term to the east of Efteni Lake subsection. This fault system is formed by: the range front fault system (RFS), WNW-ESE and SW-NE trending and bounding also a major embayment on the Almacik range front; and the PDZ represented by the WSW-ENE striking Yenikoy and Beykoy subsections and the E-W striking Cinarli subsection. Most of the coseismic ruptures and cumulative tectonic landforms (PDZ), only partially follow the RFS, but instead traverse the deposits of a Pleistocene bajada within the embayment, along the northernmost part of the long-term fault system. Overall, the DFS is formed by a PDZ as surface expression of a PSZ and by a RFS all embedded in the damage zone.
 The consistency between the geometry, location and style of the 1999 ruptures and the near-fault long-term tectonic landforms is a strong indication that the characteristics and complexities of the 1999 rupture in the study area have been persistent during several seismic cycles or more (Figure 14). The far-field landforms suggest that, during the Late Pleistocene, the Düzce fault developed structures forming the Yenikoy and Cinarli subsections, that abandoned the older rangefront-bounding faults. This evolution can be interpreted as a tendency of the deformation zone to simplify a complex geometry into a more straight fault trace (Figure 13b) and we expect the fault system to continue to straighten itself out in the future, until it becomes a smooth and narrow fault zone.
 One of the main conclusions of this study is to confirm that the comprehensive analysis of tectonic landforms and associated structures is a valuable tool to define the strain distribution pattern and its evolution in the near-surface. These data are the basis to understand how the surface deformation build up. Consequently, they provide a picture of the characteristics of the principal slip zone at depth and is certainly of value for the identification and characterization of faults with potential for surface rupturing earthquakes.
 This research is supported by the European Commission Project “Relief”: Large earthquake faulting and implications for seismic hazard assessment in Europe: The Izmit-Düzce earthquake sequence of 1999, Turkey, Mw 7.4, 7.1, EVG1-CT-2002-00069.