Structure and paleoearthquake records of active submarine faults, Cook Strait, New Zealand: Implications for fault interactions, stress loading, and seismic hazard
Abstract
[1] A new interpretation of active faulting in central Cook Strait, New Zealand, reveals tectonic structures associated with the spatial transition from subduction to continental transform faulting. Marine seismic reflection profiles and multibeam bathymetric data indicate that there are no throughgoing crustal faults connecting the North Island Dextral Fault Belt and the Marlborough Fault System in South Island. The major faults terminate offshore, associated with 5–20 km wide step-overs and a change in regional fault strike. This structure implies that propagation of strike-slip earthquake ruptures across the strait is not probable. Faulted sedimentary sequences in the Wairau Basin (Marlborough shelf), correlated to glacioeustatic sea level cycles, provide a stratigraphic framework for fault analysis. A high-resolution study of the postglacial (<20 ka) vertical displacement history of the Cloudy and Vernon faults reveals up to six and five paleoearthquakes since 18 ka, respectively. These long-timescale records indicate variable recurrence intervals and possibly variable stress drop, thus conforming to the variable slip model of earthquake behavior. Integration of these data with other submarine and terrestrial paleoearthquake records indicates the presence of clustered earthquake sequences involving multiple faults. Different sequences do not always involve the same faults. It appears that earthquake clustering is driven by fault interactions that lead to specific loading conditions favoring the triggering of earthquakes on major faults in relatively short time intervals. Present-day regional Coulomb stress distribution has been calculated in two scenarios considered to represent minimum and maximum loading conditions. The models, incorporating secular tectonic loading and stress changes associated with major crustal earthquakes, indicate high stress loading in a large part of central Cook Strait. These conditions may favor the triggering of future damaging earthquakes in this region.
1. Introduction
[2] Fault structure and earthquake records spanning multiple seismic cycles are critical for evaluation of fault behavior, fault interactions, stress distribution, and seismic hazard. Recurrence intervals between surface-rupturing earthquakes (M > 6.5) on any one fault are generally much larger than instrumental observation periods. Over the last few decades, paleoearthquake data have commonly been derived on land from geological and geomorphological studies of active faults [Wallace, 1981; Van Dissen and Berryman, 1996; Rockwell et al., 2001; Hubert-Ferrari et al., 2002; Van Der Woerd et al., 2002]. For active submarine plate boundary faults, however, the development of paleoseismological methodologies and data sets are still very much in their infancy [e.g., Armijo et al., 2005; Cormier et al., 2006; Henstock et al., 2006; Bull et al., 2006], and most marine records have been derived from indirect off-fault observations, such as turbidite records of strong ground shaking [e.g., Goldfinger et al., 2003; McHugh et al., 2006] or uplifted coseismic marine terraces associated with offshore faults [e.g., Berryman, 1993; Barnes et al., 2002]. In a separate paper developed concurrently from our study [Barnes and Pondard, 2010], we demonstrated that direct on-fault submarine paleoearthquake records can be derived from high-resolution seismic reflection profiles of active fault growth sequences. We recognized up to eight surface-rupturing earthquakes since 18 ka on the strike-slip Wairau Fault, Cook Strait, New Zealand, and were able to determine the magnitude of coseismic vertical displacements and earthquake recurrence. These data enabled us to evaluate the slip behavior of the fault over multiple seismic cycles. Wider application of these methodologies to other submarine faults provide tremendous potential for development of submarine paleoearthquake data sets, and for improving understanding of regional seismic hazard where major plate boundary faults lie offshore.
[3] In central New Zealand the major strike-slip faults of the Pacific-Australian plate boundary zone extend offshore beneath the continental shelf of Cook Strait (Figure 1). This region straddles the transition from oblique subduction along the Hikurangi margin of North Island, to continental transform tectonics in South Island. Prior to this study, the location and geometry of the active submarine faults in central Cook Strait, where the Marlborough Fault System (MFS) and the North Island Dextral Fault Belt (NIDFB) approach each other, were poorly mapped due to a lack of well-resolved seismic reflection and bathymetry data. In this paper, we analyze new multibeam bathymetry data, together with boomer, 3.5 kHz, and multichannel seismic reflection profiles from central Cook Strait, to identify and map the major strike-slip faults between North and South islands at a level of precision not previously achievable. We confirm there is a significant structural discontinuity between the islands [Carter et al., 1988], which has important implications for seismic hazard assessment of New Zealand's capital city, Wellington. We document the late Quaternary seismic stratigraphic framework for the continental shelf sediments of northern South Island, and apply the submarine paleoseismic methodologies developed by Barnes and Pondard [2010] to derive long-timescale (<18 ka) paleoearthquake records of the Vernon and Cloudy fault. These data, combined with other paleoearthquake records and historical seismicity published elsewhere, enable us to evaluate regional fault interactions over multiple seismic cycles, and to calculate present-day Coulomb stress distribution [e.g., King et al., 1994]. Our results provide important insights into the cyclic loading conditions that may favor the triggering of future damaging earthquakes in this region.
2. Tectonic Setting: The Australian-Pacific Plate Boundary in Central New Zealand
[4] In the Cook Strait region the Pacific-Australian plate boundary is characterized by oblique convergence at a rate of ∼42 mm/yr [Beavan et al., 2002]. Cook Strait lies broadly at the junction of the Alpine Fault, the North Island Dextral Fault Belt, and the Hikurangi subduction zone (Figure 1a). The area corresponds to the southern termination of subduction, where strike-slip faulting dominates, and relative plate motion is transferred to continental faulting in South Island.
[5] In the North Island, oblique convergence is partitioned largely between the North Island Dextral Fault Belt (NIDFB) and the Hikurangi subduction margin [e.g., Beanland and Haines, 1998; Barnes et al., 1998; Nicol et al., 2007]. The convergent component of plate motion is mainly accommodated on the interplate thrust [Barnes and Mercier de Lépinay, 1997; Nicol and Wallace, 2007], while most of the right-lateral motion is accommodated in the upper plate on dextral faults and by clockwise rotation of crustal blocks [Beanland and Haines, 1998; Wallace et al., 2004]. At Cook Strait the most prominent strike-slip faults of the NIDFB are the Wairarapa, Wellington, and Ohariu faults, which have dextral slip rates estimated at 8–12 mm/yr, 6–7 mm/yr, and 1–2 mm/yr, respectively (Figure 1b) [Van Dissen and Berryman, 1996; Heron et al., 1998; Langridge et al., 2005; Little et al., 2009].
[6] Although the Pacific plate has been subducted beneath northern South Island (Marlborough) to depths of more than 200 km, since 10 Ma [Furlong and Kamp, 2009], geological studies indicate that most of the present plate motion is now being accommodated by strike-slip faulting in the Marlborough Fault System (MFS) [e.g., Holt and Haines, 1995]. These faults, including the Wairau, Awatere, Clarence, Kekerengu, and Hope faults, transfer plate motion from the Hikurangi margin to the Alpine Fault and other structures in central South Island (Figure 1) [Berryman et al., 1992; Norris and Cooper, 2001]. In northern Marlborough, the Wairau and Awatere faults have slip rates of 3–5 mm/yr [Lensen, 1976; Knuepfer, 1992; Zachariasen et al., 2006] and 4–8 mm/yr [Knuepfer, 1992; Little et al., 1998; Benson et al., 2001], respectively. A reduction in the slip rate to <1.5 mm/yr on the coastal part of the Awatere Fault [Little et al., 1998], indicates that up to 4.5 mm/yr of dextral slip may be transferred northward to the Vernon Fault [Barnes et al., 2008], which branches off the Awatere Fault about 15 km inland of the coast [Benson, 2000].
[7] The Wairau, Vernon, Awatere, Kekerengu, Wairarapa, Wellington, Ohariu, and Shepards Gully faults all extend offshore into Cook Strait (Figure 1b). The submarine fault structure was already relatively well constrained in the northern [Lamarche et al., 2005; Nodder et al., 2007] and southern [Barnes and Audru, 1999a, 1999b; Barnes et al., 1998; Mountjoy et al., 2009] reaches of the strait prior to this study. In southern Cook Strait and off eastern Marlborough the shelf is underlain by large strike-slip faults including the Hope, Needles, Chancet, and Boo Boo faults, whereas the continental slope is underlain by predominantly seaward verging (i.e., NW dipping) thrust faults (Figure 1b). Previous studies of central Cook Strait proposed various submarine fault geometry and relationships, including continuity of structures across the strait, and various discontinuities [Carter et al., 1988, and references therein].
[8] Two surface rupturing earthquakes, both of which are considered to have extended into Cook Strait, occurred since European settlement (∼A.D. 1840). These include the 1848 M 7.5 Marlborough earthquake on the Awatere Fault [Grapes et al., 1998; Mason and Little, 2006a], and the 1855 M 8.2 Wairarapa earthquake on the Wairarapa Fault [Darby and Beanland, 1992; Rodgers and Little, 2006]. Substantial paleoseismic data from widely across the MFS and NIDFB attest to repeated large magnitude earthquakes (M > 7) prior to the historical (and instrumental) period [Cowan and McGlone, 1991; Van Dissen and Berryman, 1996; Heron et al., 1998; Benson et al., 2001; Langridge et al., 2003, 2005, 2007, 2009; Mason and Little, 2006b; Zachariasen et al., 2006; Litchfield et al., 2006; McSaveney et al., 2006; Van Dissen and Nicol, 2009; Little et al., 2009; Barnes and Pondard, 2010]. While there is ample evidence of historical and paleoearthquakes, there is no evidence that any of the major faults creep.
[9] The Cook Strait shelf and canyons are underlain by three major sedimentary basins, each about 15–20 km wide. These include the onshore-offshore Wairarapa and Wairau basins, and the entirely offshore Flaxbourne Basin (Figure 1b) [Carter et al., 1988; Field et al., 1997; Barnes and Audru, 1999a]. The seabed forms part of the New Zealand continental shelf at water depths of less than 150 m, except in the Narrows Basin (∼300 m) and Cook Strait canyons (Figure 1b). The canyon system incises the continental slope and shelf with a three-branched canyon head, and intersects the Hikurangi Trough in 2500 m water depth. Cook Strait is characterized by very strong (maximum >2 m s−1) tidal currents [Stanton et al., 2001], which have locally scoured the seafloor, particularly along the south Wellington coast (Figure 2) [Lewis et al., 1994; Carter, 1992]. This area is also subject to southern ocean swells and regular storm waves, which move sediment [Lewis, 1979].
3. Seismic Reflection and Bathymetric Data
[10] In this study we use KONSBERG SIMRAD EM300 multibeam bathymetry data and a variety of seismic reflection profiles (e.g., Figure 1c). The multibeam echo sounder data were acquired by NIWA between 2001 and 2005 on board R/V Tangaroa. This system operates with 135 beams at 30 kHz frequency, and with approximately 1° beam width. Vessel navigation comprised a POS/MV system with differential GPS. These bathymetry data were collected from water depths typically >100 m and gridded at 10 m using ESRI topogrid functions to develop a high-resolution digital terrain model (DTM). They provide unprecedented geomorphic details of fault scarps, sediment transport bed forms, erosional scours, submarine canyons and landslides (Figures 2 and 3) [Mackay et al., 2009; Barnes et al., 2009; Mountjoy et al., 2009]. DTMs were illuminated with low grazing angles to reveal subdued fault traces in areas of strong seafloor erosion. These data are augmented by contoured bathymetry from shallower areas of shelf, derived from Royal New Zealand Navy single-beam echo soundings [Mitchell, 1996].
[11] The primary seismic data used in this study include 400 km of high-resolution seismic profiles (penetration ∼ 200 m) acquired from the Wairau Basin shelf in 2005 during R/V Tangaroa voyage TAN0510 with a 500–2000 Hz Applied Acoustics CSP-300 boomer system. These profiles have variable line spacing typically of about 1.5–2.0 km and multiple ties (Figure 1c). The boomer source has a nominal frequency of about 1200 Hz, and was sampled at 24 kHz. Band-pass (350–500–1750–1950 Hz), swell, and coherency filters were applied, and the data displayed with an 80 ms AGC window. The data provide exceptional imaging (maximum theoretical resolution ∼30 cm) of the postglacial (<20 ka) sedimentary sequence and the active faults. We interpreted the profiles on large hard copy displays with vertical scale of 1.5 m/s, and with Geographix Seisvision software enabling zoom to more detail. In practice, the smallest fault vertical displacement detectable is about 0.4 m. This is the magnitude of the smallest detectable difference in the converted depth to the midpoint (wavelet crest) of any particular reflector that can be traced from footwall to hanging wall locations. It is close to the scale of a typical seismic wavelet. For depth conversion, we used a P-wave velocity of 1500 m/s (two-way travel, TWT) within the water column and 1550 m/s (TWT) within shallow sediments, based on measurements from similar sequences elsewhere on the New Zealand shelf [e.g., Taylor et al., 2004; Gerber et al., 2010].
[12] The TAN0510 boomer profiles are complemented by 3.5 kHz profiles acquired concurrently with EM300 and multichannel seismic data. The 3.5 kHz profiles have subsurface penetration of ∼5–40 m, and are positioned ∼3–5 km apart on the Wairau shelf, and 300–1000 m apart in the area of EM300 multibeam data (lines not plotted on Figure 1c for clarity). In addition to the high-resolution seismic data, we interpreted an excellent quality, high-fold deep-penetration (>4 km) seismic section acquired in 2005 by Discovery Geo Corporation. These data are also complemented by archived, typically low (3–12) fold, seismic sections recorded by oil companies between 1969 and 1973 [Holdgate and Shaw, 2001], by NIWA and the former New Zealand Oceanographic Institute (DSIR) in the 1980s and 1990s [Carter et al., 1988; Lewis et al., 1994; Barnes and Audru, 1999b; Lamarche and Harrison, 2000], and by GNS Science in the mid-1990s [Field et al., 1997].
4. Continental Shelf Stratigraphy
4.1. Late Quaternary Sediments and Stratigraphic Markers (<150 ka)
[13] The TAN0510 boomer seismic profiles image clearly a succession of late Pleistocene and Holocene sedimentary deposits and unconformities beneath the Wairau continental shelf (Figures 4 and 5). The stacking pattern of these deposits and the characteristics of their seismic facies are similar to those observed widely on the New Zealand shelf [e.g., Lewis, 1973; Browne and Naish, 2003; Lamarche et al., 2006; Paquet et al., 2009]. Three clearly imaged erosional unconformities are referred to here as PGS2/1, SB1, and PGS6/5. These unconformities and the intervening sedimentary units can be correlated to glacioeustatic sea level cycles and climate induced changes in the nature of sediment delivered to the shelf (Figure 4c). The upper two sedimentary units and the PGS2/1 surface can also be correlated to an 80 m deep coastal borehole (Figure 4b, P28/W1773) [Ota et al., 1995] by geometrically projecting two tie lines about 3.5 km to the coast (Figures 1c).
[14] The youngest sedimentary unit, shaded green in Figures 4 and 5, is part of the postglacial (<20 ka) sediment wedge which is developed in several connected depocenters on the Cook Strait shelf [Carter, 1992; Barnes and Audru, 1999a, 1999b]. This unit is thickest (up to about 45 m) off the eastern Marlborough coast and in the Cloudy Bay, and is essentially absent from a 25 km wide, current swept and scoured zone off the south Wellington coast (Figures 2b, 2c, and 3). Surficial sediments are predominantly silty sand beneath the middle shelf, grading to sandy silt nearshore and sand beneath the outer shelf [Lewis and Mitchell, 1980; Carter, 1992]. Seismic reflectors within the postglacial wedge typically down-lap onto the PGS2/1 surface beneath the outer shelf (Figure 4b), and onlap it toward the coast. These reflectors are associated with transgressive and highstand sediments that accumulated during marine oxygen isotope stage (MIS) 1 [Imbrie et al., 1984]. It is from detailed fault growth analysis of this unit that submarine paleoearthquake records can be derived (see section 6). Unconformity PGS2/1 is a widespread, diachronous (18–7 ka) erosion surface representing the post–last glacial (MIS 2/1) marine transgressive ravinement (Figures 4 and 5) [Gibb, 1986], which formed nearshore within the migrating wave-abrasion zone.
[15] Beneath PGS2/1 are fluvial outwash gravels (shaded blue in Figures 4 and 5) that are currently exposed, and are being actively reworked by strong currents, on the outer continental shelf beyond the postglacial sediment wedge (Figures 2a, 3, and 4b) [Carter, 1992]. These deposits are interpreted to have been deposited primarily during the last glaciation (MIS 2), ∼28–18 ka, when sea level reached some 120 m below present (Figure 4c). They correlate with the Wairau, Speargrass and Rapaura gravel units in borehole (P28/W1773) (Figure 4b) [Ota et al., 1995]. Near the present shelf break, the PGS2/1 surface converges with unconformity SB1, which is interpreted as the last sea level cycle sequence boundary at the base of the gravels. During the glacial maximum lowstand of MIS 2, marine sediments were deposited directly into the Cook Strait canyons and Narrows Basin (Figure 3) [Lewis et al., 1994; Mountjoy et al., 2009].
[16] Beneath the SB1 erosion surface are marine deposits that we infer are related to the episodic lowering of sea level between MIS 5e and MIS 2 (120–30 ka) (Figure 4). These deposits, highlighted in shades of yellow on Figures 4 and 5, preferentially accumulated in basins associated with the offshore Cloudy and Wairau faults (Figures 5b and 6), and locally are >150 m thick. They include seaward prograding clinoforms that reflect the general downward shift in sea level (regression) and sediment depocenter that followed the last interglacial (MIS 5e). The clinoforms down-lap onto a prominent erosion surface (PGS6/5) (e.g., Figure 4b), which we interpret to have developed largely during the penultimate postglacial marine transgression (MIS 6/5, ∼140–125 ka). While the down-lapping clinoforms and PGS6/5 surface are clearly imaged beneath the outer shelf, and can be tied between intersecting seismic lines, they are not readily identified beneath the inner to middle shelf where the multiple is shallow and gas masking in the post glacial sequence is patchy.
4.2. Age of the PGS2/1 Erosion Surface and Post–Last Glacial (<20 ka) Sediments
[18] We selected five reflectors within the thickest, middle shelf part of the postglacial sediment wedge above PGS2/1 that can be traced confidently across the Wairau Basin, and tied between boomer seismic profiles (colored reflectors Figure 5b). In the absence of dated borehole material, we estimate the age of these reflectors as the means of random distributions of age estimates determined at six sites across the basin (labeled Sites 1 to 6, on Figure 5b). The estimates of reflector ages at each individual site were extrapolated assuming constant sedimentation rate above the PGS2/1 surface at that site, and a large error of ±40% was modeled to allow for potential variability in sedimentation rate through time. These uncertainties appear reasonable considering the middle shelf depositional environment, and the maximum inferred age for PGS2/1. The mean ages of the five colored reflectors, with errors expressed at 95% confidence, are estimated to be 2.7 ± 0.6 ka (dark blue), 6.2 ± 1.1 ka (orange), 8.4 ± 1.6 ka (red), 10.3 ± 2.1 ka (light blue), and 14.0 ± 3.8 ka (yellow), were checked for consistency across the basin considering the diachronous age of the underlying PGS2/1 surface. For detailed analysis of fault displacements at specific profile locations (see section 6), the ages of intervening reflectors between those referred to above were determined by simple extrapolation on each profile.
5. Active Faulting in Central Cook Strait: Structure and Slip Rates
[19] We describe below the submarine components of the active Wairau, Awatere, Vernon, Wellington, Ohariu, and Shepards Gully faults, as well as the newly identified, entirely marine Cloudy Fault (Figure 3).
5.1. Wairau Fault
[20] The Wairau Fault extends offshore as a continuous structure for about 40 km (Figure 3) [Barnes and Pondard, 2010]. The fault is well imaged in seismic sections (e.g., Figure 5b), and as a prominent (15–20 m) bathymetric scarp on the outer shelf (Figure 2a). The strike of the fault is about 75° over the northern 100 km, and 65° locally near the coast. There is a left bend in the surface trace about 6–12 km from shore, north of which the strike swings to be more easterly (80°), and is NW–SE (110°) near its northern tip. North of the main trace, beneath the outer shelf and west of the Terawhiti Sill, a number of fault scarps evident in multibeam bathymetry data are distributed across an 8 km wide zone (Figures 2a and 3). These are between 2 km and 5 km in length, have normal displacement, and scarp heights of up to 15 m.
[21] The offshore part of the Wairau Fault is down-thrown to the south, and bounds a tilted basin (Figure 5b) [Field et al., 1997]. Beneath the inner to middle shelf the fault displaces the PGS2/1 surface vertically by ∼8–20 m, but there is no seafloor scarp there because the fault tip is buried by late Holocene (<2 ka) sediments of the postglacial sediment wedge [Barnes and Pondard, 2010]. The fault scarp on the outer shelf (Figure 2a) reflects largely the absence of sediment fill beyond the postglacial wedge. The vertical separation rate, based on displacement of the PGS2/1 surface, varies typically from 0.8 to 1.5 mm/yr (Figure 7a). There are no direct submarine observations of strike-slip displacement, but a local area of folding and reverse faulting associated with the 3 km wide left bend in the surface trace, 6–12 km from shore, is consistent with dextral displacement [Barnes and Pondard, 2010]. On land, lateral offsets indicate that late Quaternary dextral displacement occurs at 3–5 mm/yr [Lensen, 1976; Knuepfer, 1992; Zachariasen et al., 2006].
5.2. Cloudy, Vernon, and Awatere Faults
[22] The newly identified Cloudy Fault is entirely submarine, about 23 km in length, and has a curved trace (Figure 3). The eastern section lies 7 km north of the Vernon Fault and strikes 105°, whereas the western section strikes 50° and possibly connects with the Vernon Fault. The eastern end approaches the southern end the Wellington Fault at an oblique angle in central Cook Strait, but the tips are separated by 7 km and profiles between them reveal no active deformation. In seismic profiles the Cloudy Fault dips moderately (∼60°) to the south, bounds a tilted subbasin with half-graben geometry, and has antithetic faults in its hanging wall section (Figure 6). The fault clearly has a significant extensional component of displacement, and displaces vertically the PGS2/1 surface by up to 25 m (e.g., Figure 5b), indicating a vertical displacement rate of up to 1.4 mm/yr (Figure 7b). Given the regional tectonic environment, fault relationships, and demonstrated strike-slip displacement on the Awatere and Vernon faults [Little et al., 1998], it is possible that at least the western section of the Cloudy Fault also has a component of strike slip. While we have no direct observations of dextral displacement, such displacement might be expected to be concentrated on the western portion of the fault, in contrast to the predominantly extensional eastern section where the largest vertical displacement rates occur. There is no seafloor scarp because the fault tip is buried by late Holocene (<2 ka) sediments of the postglacial sediment wedge.
[23] The Vernon Fault extends offshore for 28 km, bringing the total fault length to about 40 km (Figure 3). The fault has an average strike of 73° near the coast, but swings E–W about 7 km from shore. The eastern tip of the fault underlies the outer shelf, and is well constrained by the seismic profile coverage. In seismic sections the Vernon Fault dips steeply (>70°) to the south and is down thrown to the north (e.g., Figures 5 and 6), indicating it has a component of reverse displacement. This is consistent with elevated topography on land south of Big Lagoon, and with uplift of the contiguous nearshore bathymetric reef between the Vernon and Awatere faults (Figures 3 and 5c). Along the E–W striking section of the fault beneath the middle to outer shelf, >12 km offshore, the Vernon Fault displaces vertically the PGS2/1 surface typically by 2–6 m (e.g., Figure 5b), indicating a vertical displacement rate of 0.2 to 0.4 mm/yr (Figure 7c). However, across the converging overlap with the northern part of the Awatere Fault, about 9–12 km from shore, there may be an abrupt increase in the vertical displacement on the Vernon Fault. Nearshore profiles reveal vertical displacement of up to 20 m on a prominent erosion surface along the NW side of the reef (e.g., Figure 5c), but it is not certain if the surface is PGS2/1 (10–14 ka) or older. The vertical separation rate therefore could potentially increase rapidly inshore to >1–2 mm/yr (Figure 7c). There is no direct submarine evidence of strike-slip displacement on the Vernon Fault, but dextral displacements have been reported on the fault onland [Benson, 2000], and a dextral rate of up to 4–5 mm/yr inferred from a comparable slip rate reduction on the coastal part of the Awatere Fault [Little et al., 1998].
[24] The Awatere Fault extends offshore for at least 11 km, across the southern part of a bathymetric reef off White Bluffs (Figures 3 and 5c). The fault has an average strike of about 63°, and converges with the Vernon Fault beneath the middle shelf. A separate active trace about 10 km in length is mapped east of the main trace, parallel to the Vernon Fault. Nearshore, the Awatere Fault displaces vertically the PGS2/1 surface, where it is about 10 ka, by at least 5 m, indicating a vertical displacement rate of ∼0.5 mm/yr. The eastern trace south of the Vernon Fault has up to 4 m of vertical displacement of the PGS2/1 surface.
5.3. Wellington, Ohariu, Shepards Gully, and Terawhiti Faults
[25] The Wellington Fault extends offshore for about 20 km, and terminates in central Cook Strait (Figure 3). The strike of the fault nearshore is about 35°, similar to its contiguous trace on land, and changes to 55°–60° toward the southwest. The submarine trace is recognized as a discontinuous lineament in EM300 multibeam bathymetric data (Figure 2b). At the shelf edge, seaward of a submerged rocky reef, a series of sinuous bathymetric ridges, possibly beach ridges or sand/gravel bars associated with the glacial lowstand shoreline, are truncated by the fault but not observed southeast of the fault scarp. Off the shelf, the seafloor in this region is characterized by widespread erosional scour and large sedimentary dunes associated with tidal currents and active sediment transport. The postglacial muddy sediment wedge that is well developed on the Marlborough shelf is absent. The fault appears to be down-thrown to the southeast, but the surface scarp off the shelf is generally not well preserved. Although the fault trace is crossed by numerous 3.5 kHz seismic profiles, it is typically not recognizable in these data due to prolonged acoustic echos resulting from gravel and sandy gravel substrates [Lewis and Mitchell, 1980; Carter, 1992].
[26] The Ohariu Fault appears to extend offshore for about 25 km, across the heavily scoured seafloor about 5–7 km north of the Wellington Fault (Figure 3). Off the south Wellington coast, the fault may be associated with a bathymetric ridge spur at the shelf break south of the fault trace on land (Figure 2b). To the southwest of this spur are a series of diverging lineaments that define what appears to be a graben. These inferred fault traces strike approximately parallel to the Wellington Fault, and terminate on the Marlborough shelf between the northeastern end of the Wairau and Cloudy faults. The two parallel, southwestern traces have northward facing fault scarps, <3 m in height, vertical displacement rates of the order of 0.1 mm/yr, and no evidence of dextral displacement of current scour depressions (Figure 2c).
[27] Northwest of the Ohariu Fault, in Oteranga Bay, the Shepards Gully and Terawhiti faults are recognized only within 2 km of shore, where they are inferred to be associated with a graben 500–1000 m wide in the wave-cut rocky platform (Figure 3).
6. Submarine Paleoearthquake Records
6.1. Submarine Paleoearthquake Records From High-Resolution Fault Growth Sequences
[28] Tectonic fault growth sequences such as those illustrated in Figures 5 and 6 develop when sedimentation rate is sufficient to fill the accommodation space created by vertical deformation [e.g., Childs et al., 2003]. Sediments accumulate preferentially in tectonically depressed areas when sedimentation regime is sufficiently dynamic, leading to an increase in stratigraphic thickness on the relatively down-thrown side of faults. As vertical displacement accumulates on a fault, the strata forming the sedimentary sequence are associated with increasing vertical offsets with depth. For tectonic growth faults, for which displacement occurs only during earthquakes (and any potential period of postseismic slip that may last weeks to years), this indicates that the progressively older layers have been displaced by a larger number of earthquakes compared to the younger layers (e.g., Figure 8a). The precise characteristics of the growth sequence depend largely on the vertical displacement rate, the recurrence and variability of coseismic displacements, and the rate and style of sedimentation over a number of seismic cycles [Nicol et al., 2009; Pochat et al., 2009; Barnes and Pondard, 2010].
[29] In a separate detailed seismic reflection study of the late Quaternary (<20 ka) submarine growth sequence associated with the Wairau Fault, Barnes and Pondard [2010] demonstrated that at least eight seabed-rupturing paleoearthquakes have occurred on the fault since 18 ka. Here, we first summarize their methodology (Figures 8a–8c), and illustrate the analytical procedure with an example from the Wairau Fault (Figures 8d and 8e). We then apply this method to develop new paleoearthquake records of the Cloudy and Vernon faults (sections 6.2 and 6.3).
[30] Coseismic vertical increments of displacement are likely to be best preserved and recognizable in the architecture of the growth sequence when the long-term rate of postglacial sedimentation exceeds the rate of fault vertical displacement (Figure 8a). At such sites, the interseismic period of sedimentation may comprise a postseismic growth interval (shaded yellow, PS1–PS4) that developed in response to the earthquake-generated tectonic relief, covered by a uniformly thick sedimentation interval (clear layered sequences, IUS1/0 to IUS4/3) that developed after complete burial of the fault tip. This hypothetical scenario may apply to any sedimentation regime that is sufficiently dynamic to result in sedimentation preferentially developing in the hanging wall where there is tectonic topography, a situation we believe applies to the silty and sandy sediments of the northern Marlborough shelf. In contrast to the gravity-driven fault growth model of Castelltort et al. [2004] and Pochat et al. [2009], in which fault displacement may be constant and continuous over various intervals, on crustal faults that accumulate elastic strain and do not creep there is no displacement accrual between earthquakes (and any associated short-term postseismic slip). Provided no long-term interseismic creep occurs, all vertical throw is the sum of coseismic displacements that each led to tectonic topography. In this context, each cycle of the earthquake growth model in Figure 8a is equivalent to the fill-to-the-top model of Cartwright et al. [1998] and Castelltort et al. [2004], with the overall displacement history indicative of alternative periods of fault activity and inactivity.
[31] In the hypothetical scenario in Figure 8a, the postseismic growth intervals and intervening uniformly thick sedimentation intervals can be recognized on strongly stepped vertical displacement history curves, from which the earthquake timing and coseismic vertical displacements can be inferred. The earthquakes are recognized as scarp forming events at the base of each postseismic growth interval (PS1–PS4). These occur at the top of each inclined section of the vertical displacement history curve. The postseismic growth intervals are differentiated from the intervening uniformly thick sequences, which correspond with flats on the displacement history curve. Provided that sedimentation is sufficient to completely bury the fault scarp between successive earthquakes, the coseismic vertical displacements, shown in red (Figure 8a; CSD1–CSD4), are equivalent to the height of the inclined risers on the displacement history curve. As with conventional terrestrial paleoseismology, the coseismic vertical displacement recorded in the displacement profile includes the coseismic surface rupture and any potential postseismic slip such as that which may occur over durations of weeks to years. The general sequence architecture in Figure 8a applies even if sedimentation rates are not strictly constant. Sedimentation variability could contribute to nonlinear inclined sections of the displacement history curve, but never to a flat in the curve provided the strata are part of the same growth sequence. The inclined risers of the displacement curve, however, could be expected to progressively flatten at their youngest (lower) ends, representing the late stages of the postseismic growth sequence. Barnes and Pondard [2010] demonstrate that in the situation where the average fault vertical displacement rate exceeds the sedimentation rate, successive postseismic growth sequences may be stacked, and the absence of intervening uniformly thick intervals may impair recognition of individual coseismic vertical displacements.
[32] In the specific Wairau Fault example (Figures 8d and 8e), from the inner shelf (Figure 3, where the post glacial sequence is inferred to be <12.1 kyr, six of the total eight paleoearthquakes recognized on the fault are illustrated [Barnes and Pondard, 2010]. In this example, the above model (Figure 8a) applies to five paleoearthquakes (EQ8/6 to EQ8/2), but it does not apply to EQ8/1. The youngest earthquake in this profile is stacked on top of the postseismic growth sequence following EQ8/2, which had not completely developed at the time of EQ8/1, and is therefore not separated from it by a uniformily thick interval. EQ8/1 therefore lies on the inclined riser of the displacement history curve (Figure 8e). Earthquake EQ8/1 is recognized by a renewed phase of growth faulting indicated by a prominent wedge in the middle of the composite growth sequence. Barnes and Pondard [2010] identified this event on seven other seismic profiles of the Wairau Fault. On several of those profiles the same earthquake is recognized by down-lapping postseismic reflectors at the base of the youngest growth sequence (e.g., Figure 8c; type 2 postseismic sequence; see below). Given the possibility that additional coseismic vertical displacements with magnitude below the seismic resolution could have occurred, or that events close in time could be stacked in a postseismic sequence and not recognized (especially in type 1 sequences; see below), the record represents a minimum number of earthquakes on the fault.
[33] On the basis of sequence architecture, Barnes and Pondard [2010] identified two types of postseismic growth sequence, characteristic of different continental shelf environments (Figures 8b and 8c). Type 1 postseismic growth sequences (Figure 8b) are (1) representative of relatively quiet-water, silt-dominated environments; (2) commonly wedge or lens shaped in their lower parts and more tabular in their upper parts; and (3) typically characterized by subhorizontal, drape-type internal reflection configurations that, in some cases, converge locally across the fault where the sedimentary growth strata wedge against the scarp topography. Examples of this type of sequence occur in Figures 8d, 9b, and 9c, where they are interpreted according to the criteria in Figure 8a. They are not equivalent to the pelagic clay suspensive deposition of Pochat et al. [2009], in which there is zero current influence on sedimentation, and an absence of displacement growth. Type 2 postseismic growth sequences (Figure 8c) develop in relatively higher-energy sand-dominated sedimentary environments associated with stronger bottom currents and bed load sediment transport. They are characterized by distinctive wedge geometry with inclined, down-lapping reflections on the relatively down-thrown side of the fault, typically limited to within a few hundreds of meters to 1 km from the fault trace (e.g., Figure 9d, EQ8/4, and Figure 9e, EQ9/4 and EQ 9/1). These intervals are analogous to the bed load growth deposits of Pochat et al. [2009, Figure 13]. In the case of a type 2 postseismic growth sequence, the height of the former fault scarp (the coseismic vertical displacement) can be estimated from the maximum thickness of the inclined, down-lapping reflection package developed against the fault. If a second displacement event had occurred within such a sequence, it may be identifiable as a second, stacked down-lap surface. On the Cloudy Bay shelf, type 1 postseismic growth sequences are observed to be dominant between about 25 m and 60 m water depth (e.g., Figures 8d and 9b), whereas type 2 growth sequences dominate within the outer part of the postglacial sediment wedge in water depths of 60–100 m. The favorable occurrence of type 2 growth sequences may enable recognition of paleoearthquakes in the otherwise unfavorable situation where the long-term vertical displacement rate exceeds the sedimentation rate, and successive postseismic growth intervals are largely stacked [Barnes and Pondard, 2010].
6.2. Cloudy Fault Submarine Paleoearthquakes
[34] We analyzed in detail four boomer seismic profiles across the central section of the Cloudy Fault to determine its postglacial (<20 ka) vertical displacement history (Figures 3 and 9). The postglacial growth sequence is exceptionally well imaged and is up to 45 m thick (e.g., Figures 9a and 9b). In these profiles the long-term rate of sedimentation exceeds the rate of fault vertical displacement by a factor of 3, and the fault tip is currently buried by late Holocene (<1 ka) sediment that has accumulated since the last earthquake. Displacement history curves for each profile are strongly stepped (Figure 9f), with inclined sections of the curves coinciding with postseismic growth intervals, and horizontal sections with intervening uniformly thick sequences (compare Figure 8a). The postseismic growth intervals have predominantly type 1 characteristics (e.g., compare Figures 8b and 9c, EQ8/1 and EQ8/2), but superb examples of type 2 growth sequences are present also (e.g., Figure 9d, earthquake EQ8/4, and Figure 9e, EQ9/1 and EQ9/4). The displacement curves, combined with identification of type 2 sequence architectures, indicate that repeated earthquakes developed a bathymetric fault scarp, leading to the onset of growth sedimentation in the immediate postseismic period (yellow shaded intervals on Figures 9b–9d). The presence of intervening, uniformly thick sedimentation intervals with blanket-type horizontal reflections indicates that the fault scarp was buried and completely leveled by growth sedimentation during several, but not all, of the interseismic periods.
[35] Between three and six Cloudy Fault paleoearthquakes are recognized on individual profiles (Figures 9f and 10a). We correlated displacement events between profiles, considering (1) their relative stratigraphic positions on individual profiles and displacement history curves, (2) their age estimates, (3) their coseismic vertical displacements, and (4) the stratigraphic characteristics of their postseismic growth sequences [Barnes and Pondard, 2010]. A composite record of at least six paleoearthquakes is recognized (Figure 10a), including a tentative interpretation of event EQ3. We determined arithmetic mean ages of the offshore paleoearthquakes, with errors expressed at 95% confidence, derived from the sum of random distributions around the age estimates (±30%) of each paleoearthquake on individual seismic profiles. The recurrence intervals between earthquakes range from 1.0 to 5.7 ka, with a mean of 2.7 ka. The mean coseismic vertical displacement per event ranges from 1.0 to 6.4 m, with an overall mean of all events of 2.7 m. The largest vertical displacement events coincide with the steepest part of the displacement history curves, between about 8 ka and 13 ka (Figure 9f). The earthquake history of the fault has been determine only from profiles across the central part of the fault trace where there is an ideal relationship between postglacial sedimentation and vertical displacement rate (Figures 3 and 7b). Further east and offshore, a long-timescale earthquake record is not preserved where the postglacial sequence becomes condensed, while the vertical displacement rate increases. While the coseismic vertical displacements recorded (Figure 9) may be close to the net slip per event at these sites, a strike slip component could be present. Irrespective of the true net slip, the largest coseismic vertical displacement (Figure 10a, EQ4; mean 6.4 m at 9.7 ± 1.7 ka) equates to a seismic moment too large for the length of the Cloudy Fault alone (Table 1), yet we find no evidence for an additional event between EQ3 and EQ4, and therefore infer that on occasions the fault may have coruptured with the Vernon Fault. For example, this largest Cloudy Fault displacement may be coeval with EQ3 on the Vernon Fault (see below), which occurred at 9.2 ± 2.5 ka (compare Figure 10b).
| Fault | sr (mm/yr) | Age (years B.P.) | M | L (km) | u (m) | utot (m) |
|---|---|---|---|---|---|---|
| Wellington | 6.6 | ∼400; ∼800 | 7.5 | 130 | 5 | 10 |
| Wairarapa | 11.3 | 155 (A.D. 1855); ∼800 | 7.9 | 140 | 10 | 20 |
| Ohariu | 1.5 | ∼1000 | 7.4 | 90 | 4 | 4 |
| Wairau | 4 | no event < 2000 years B.P. | – | 150 | – | – |
| Cloudy | 1.5 | ∼1800 | 7.0 | 25 | 3 | 3 |
| Vernon | 4.5 | no event < 2000 years B.P. | – | 40 | – | – |
| Awatere | 6.0 | 162 (A.D. 1848); ∼1000; ∼1500 | 7.6 | 110 | 6.0 | 18 |
| Clarence | 4.5 | ∼1800 | 7.7 | 130 | 7 | 7 |
- a For events <2000 years B.P.; sr is modeled slip rate calculated from active fault source parameter and Australian-Pacific relative plate motion rates derived from GPS velocities and tectonic observations [Beavan et al., 2002; Barnes et al., 2008; Robinson et al., submitted manuscript, 2010]. Age is age of modeled historical events (see also Figure 13). M, magnitude; L, rupture length; u, average slip relieved per event; utot, cumulated slip relieved over the past 2000 years. See Figure 1 for fault locations.
6.3. Vernon Fault Submarine Paleoearthquakes
[36] We analyzed in detail two boomer seismic profiles of the Vernon Fault to determine its postglacial (<20 ka) vertical displacement history (Figures 3 and 11). At these sites the long-term rate of sedimentation exceeds the rate of fault vertical displacement by a factor of 5–12, leading to relatively condensed type 1 postseismic growth intervals (yellow shaded intervals on Figure 11b), separated by relatively thick intervening sequences. The displacement history curves are strongly stepped, particularly for profile L7 (Figure 11c), enabling recognition of coseismic vertical displacements. The flatter displacement history curve for profile L8 for the interval 5–13 ka, reflects reduced vertical deformation at this site (compare L7), which we infer to result from complexities in the surface trace of the fault combined with predominantly dextral strike-slip displacement. The fault tip is currently buried by late Holocene (<3 ka) sediment that has accumulated since the last earthquake (Figure 11b enlargement).
[37] At least five paleoearthquakes are recognized on the Vernon Fault since 18 ka. This includes a tentative interpretation for earthquake EQ4, for which the vertical displacement measured (0.4 m) is at the maximum resolution (Figures 10b and 11c). The recurrence intervals between earthquakes range from 2.0 to 4.0 ka, with a mean of 3.1 ka. The mean coseismic vertical displacement per event ranges from 0.4 to 2.6 m, with an overall mean of all events of 1.2 m. Because we interpret the offshore section of the Vernon Fault to be predominantly strike slip, the coseismic vertical displacements are considered to be minimum estimates of the net slip per event.
7. Discussion
7.1. Regional Fault Structure and Earthquake Source Implications
[38] Our new mapping in central Cook Strait supports the interpretation of Carter et al. [1988] that major active strike-slip faults of the North Island Dextral Fault Belt (Wellington, Ohariu, Shepards Gully and Terawhiti faults) are not connected with those of the northern Marlborough Fault System (Wairau, Vernon, and Awatere faults) (Figure 3). Furthermore, we find that whereas the strike of these faults on land is generally SW–NE, there is a notable swing (∼20°–30°) in their average strike offshore to a more E–W orientation across the Wairau Basin (Figure 1). Their strike offshore is similar to the most active structure in the Marlborough Fault System, the Hope Fault (Figure 1a), and some other entirely submarine strike-slip faults in southern Cook Strait, including the Boo Boo and Chancet faults [Barnes and Audru, 1999a, 1999b; Mountjoy et al., 2009]. Their strike is typically within ±20° of the azimuth of the Pacific-Australian plate motion vector in this region, and reflects the transition from Hikurangi subduction beneath southern North Island to continental transform tectonics in southeastern Marlborough.
[39] The current deformation of the Wairau Basin is characterized by a combination of dextral strike-slip and extensional faulting (Figure 3). The late Quaternary subsidence of the basin is consistent with (1) the geometry of the faults relative to the azimuth of the Pacific-Australian plate motion vector derived from GPS data [Beavan et al., 2002], (2) an extensional component of deformation associated with the Wairau and Cloudy faults, (3) the easterly swing in the strike of the northern part of the Wairau Fault, and (4) the 10–15 km wide right step-over between the Wairau and Wellington faults. The inferred minor component of reverse vertical displacement on the offshore part of the Vernon Fault (Figure 6), however, attests to significant kinematic complexities at the boundaries of crustal blocks. The predominance of dextral strike slip and extensional faulting in the Wairau Basin contrasts with the combination of strike slip and contractional faulting in southern Cook Strait (Figure 1) [Barnes and Audru, 1999a, 1999b; Mountjoy et al., 2009].
[40] Modeling of geodetic and geologic deformation rates, and seismicity data, indicates that interseismic coupling extends to a depth of about 40 km beneath the entire southern North Island [Wallace et al., 2004, 2009]. Present interseismic coupling on the gently dipping (∼10°–15°) Hikurangi subduction interface ends beneath Cook Strait, and the locking depth is inferred to decrease progressively southwestward to ∼15 km beneath northern South Island [e.g., Wallace et al., 2009]. That elastic strain will be released in future large magnitude earthquakes [Stirling et al., 2002, 2008] is supported by historical seismicity, an absence of evidence of fault creep, and a growing body of paleoseismological data [e.g., Litchfield et al., 2006; Mason and Little, 2006a, 2006b; Zachariasen et al., 2006; Langridge et al., 2009; Van Dissen and Nicol, 2009; Rhoades et al., 2010; Barnes and Pondard, 2010; this study]. The discontinuity of major crustal faults in central Cook Strait has important implications for seismic hazard, because earthquake ruptures are commonly arrested at significant geometrical complexities such as fault jogs, bends and step overs, limiting the size of the earthquakes [Das and Aki, 1977; Aki, 1979; King and Nabelek, 1985; Sibson, 1985]. Wesnousky [2006] demonstrated that there is a limiting dimension of fault step-over width of 3–4 km, above which historical strike-slip earthquake ruptures have not propagated. From the geometry of fault discontinuities in central Cook Strait (Figure 3), throughgoing strike-slip earthquake ruptures across the strait appear improbable. Individual upper-crustal faults in the region are typically expected to be associated with earthquakes of Mw 6.9 to Mw 7.8, and with recurrence intervals of ∼500–5000 years [Stirling et al., 2008]. Larger events are possible in composite scenarios involving corupture of connected faults or combined ruptures along strike-slip faults and part of the subduction interface, such as occurred in the 1855 Mw 8.2 Wairarapa earthquake [Darby and Beanland, 1992; Rodgers and Little, 2006]. Considering the region of strong interseismic coupling, Wallace et al. [2009] suggested that a potential southern Hikurangi subduction interface rupture, pinned beneath Cook Strait, with ∼150–185 km width, ∼230 km length, and 8–12 m of slip (equivalent to a seismic moment of M0 = 8.3 1021 N m to M0 = 1.5 1022 N m), could be associated in an Mw 8.6–8.8 earthquake.
7.2. Long-Timescale (18 ka) Earthquake Recurrence and Displacement Variability in Central Cook Strait: The Vernon, Cloudy, and Wairau Faults
[41] A variety of simple conceptual models have been proposed for the long-term earthquake recurrence on tectonic faults [Shimazaki and Nakata, 1980; Sieh, 1981; Schwartz and Coppersmith, 1984]. These include the uniform slip model in which earthquake size, displacement per event at a given site, and slip rate along the fault are constant. In contrast, the variable slip model invokes variable earthquake size, and nonuniform coseismic slip distribution along the same fault segment, while constant slip rate is maintained along strike. The characteristic model invokes constant earthquake size and uniform slip distribution, but variable slip rate along the fault. In the latter case, earthquakes are assumed to be perfectly periodic (i.e., constant recurrence intervals), to occur when the tectonic loading (stress accumulation) reaches a critical threshold, and to be associated with the same stress drop (coseismic slip). Data sets required to test these models, and improve knowledge of cyclic tectonic loading and relaxation processes, include long-timescale paleoearthquake records from multiple sites along major active faults [Scholz, 2002].
[42] Our long-timescale (18 ka) paleoearthquake records from the Cloudy and Vernon faults indicate significant variability in earthquake recurrence and coseismic vertical displacements from sites spanning ∼2–4 km of the fault traces (Figures 3 and 10). The recurrence intervals between the six Cloudy Fault paleoearthquakes recognized range from 1.0 to 5.7 ka, based on the best estimates, with a mean of 2.7 ka, standard deviation of 1.8 and coefficient of variation of 0.7. The five Vernon Fault earthquakes recognized were relatively more regular, with recurrence intervals ranging from 2.0 to 4.0 ka, a mean of 3.1 ka, standard deviation of 1.0 and coefficient of variation of 0.3. The mean coseismic vertical displacements per event on these faults range significantly from 1.0 to 6.4 m, and 0.4 to 2.6 m, respectively. A similar degree of variability has been observed on the offshore Wairau Fault (Figure 12). From their analysis of 10 seismic profiles spanning a 20 km section of the fault, Barnes and Pondard [2010] documented recurrence intervals over the eight seismic cycles recognized ranging from 0.9 to 3.8 ka, with a mean of 2.2 ka, standard deviation of 0.85 and coefficient of variation of 0.4. They found the mean coseismic vertical displacement of individual Wairau Fault paleoearthquakes ranges from 0.5 to 5.3 m (Figure 12b), while the mean of all events is ∼2.5 m. These offshore Wairau Fault data were considered to be consistent with long-term dextral slip rate of 3–5 mm/yr [Lensen, 1976; Zachariasen et al., 2006], paleoearthquakes >Mw 7.5, and coseismic displacements (net slip) of about 6–11 m.
[43] Although it is possible that we have not identified all postglacial earthquakes on each of the Wairau, Cloudy, and Vernon faults, acknowledging the possibility of surface ruptures with vertical displacement <0.4 m, and possible stacked events in type 1 growth sequences, the present recurrence data derived from the vertical displacement histories indicate that their seismic cycles are not perfectly periodic. Furthermore, on the whole, the observed vertical displacement variability between different seismic cycles is consistent with the variable slip model of earthquake behavior. Static stress relief along a fault is a function of seismic displacement [e.g., Scholz, 2002]. If the coseismic vertical displacements observed on each of the Wairau, Cloudy, and Vernon faults are an approximately constant proportion of their net slip over multiple earthquake cycles, their displacement variability is noncharacteristic and implies that the stress drop may vary between different earthquakes on each fault. Figure 12b illustrates that, if the vertical to horizontal displacement ratio at any site is approximately constant, not all events have a vertical coseismic displacement that would be predicted from the duration of the elapsed time preceding the event [Barnes and Pondard, 2010]. The longest periods of tectonic loading (quiescent periods) on the Cloudy and Wairau faults were not followed by the largest apparent vertical displacements. In particular, the Cloudy Fault paleoearthquake EQ2 (C2 in Figure 12), at 3.1 ± 0.6 ka, had an anomalously small coseismic vertical displacement of ∼1.0 m considering the elapsed time of almost 6 ka corresponds to a vertical slip deficit of ∼5–6 m at a secular rate of ∼1 mm/yr. In contrast, the youngest earthquake on the Wairau Fault (W1), at 2.0 ± 0.3 ka, appeared to have been the largest with mean coseismic vertical displacement of 5.3 m, yet it followed a relatively short interseismic interval of about 1 ka. If the vertical to horizontal displacement ratio at any site is approximately constant, the observations suggest there may be complex relationships between tectonic loading and relaxation processes at the scale of an individual seismic cycle, however, over longer periods of time corresponding to multiple seismic cycles we may expect such processes to be more uniform [e.g., Nicol et al., 2009; Barnes and Pondard, 2010]. An alternative explanation for the vertical slip variability observed in the Cook Strait data, is that the vertical to horizontal displacement ratio at any site is not constant, and some smaller vertical displacement events have been associated with larger than average strike-slip displacements at the same sites (which we cannot measure with the data). The Cook Strait data, however, do appear consistent with evidence from other active faults elsewhere, indicating that variability in recurrence intervals and coseismic displacements may be common, particularly in regions of complexly distributed deformation [e.g., Palumbo et al., 2004; Weldon et al., 2004; Canora-Catalan et al., 2008; Nicol et al., 2009].
7.3. Fault Interactions
[44] The observed variability in recurrence intervals and coseismic displacements on the Cloudy, Wairau and Vernon faults may reflect interactions between adjacent faults [e.g., Rice, 1980; Kasahara, 1981; Scholz, 2002; Robinson, 2004]. Modeling of Coulomb stress changes associated with large magnitude earthquakes in such fault systems has shown that earthquake triggering and certain earthquake sequences result from secular tectonic loading below the faults, combined with lateral loading due to stress transfer associated with specific earthquake ruptures [King et al., 1994; Harris and Simpson, 1998; Stein 1999; King and Cocco, 2000]. Examples of this process include the clustered earthquakes (M > 7) of the Landers sequence in California [e.g., Stein et al., 1992; Harris and Simpson, 1992], and the 20th-century propagating earthquake sequence along the North Anatolian Fault [e.g., Stein et al., 1997; Pondard et al., 2007]. Our own Coulomb static stress modeling of earthquake rupture scenarios and historical events in the Cook Strait region (Tables 1 and 2; see section 7.4) indicates that lateral stress loading is of the order of ∼100 bars within an area of ∼10 km from the tips of ruptured faults, decreasing rapidly to ∼1–10 bars >10 km away from the faults. Because of the density of faulting and the numerous fault terminations in Cook Strait, lateral stress loading associated with an earthquake rupture would be expected to enhance the triggering of earthquakes on nearby faults, particularly if the loading occurs on faults that have accumulated significant elastic strain. The development of long-timescale paleoearthquake records provide an opportunity to evaluate this process over multiple seismic cycles, to characterize the loading conditions that lead to triggering of earthquake sequences, and to test models of synthetic seismicity and earthquake clustering (e.g., R. Robinson et al., Using synthetic seismicity to evaluate seismic hazard in the Wellington region, New Zealand, submitted to Geophysical Journal International, 2010).
| Fault | sr (mm/yr) | Number of Modeled Events | M | L (km) | u (m) | utot (m) |
|---|---|---|---|---|---|---|
| Boo Boo | 11 | 3 | 7.6 | 90 | 7.3 | 22 |
| Needles | 16 | 8 | 7.3 | 50 | 4 | 32 |
| Chancet | 3 | 2 | 7.2 | 50 | 3 | 6 |
| Campbell Bank | 3 | 2 | 7.2 | 50 | 3 | 6 |
| Kekerengu | 19 | 5 | 7.7 | 90 | 8 | 40 |
| Wharekauhau | 2.5 | 2 | 7.2 | 50 | 2.5 | 5 |
| Palliser-Kaiwhata | 5 | 3 | 7.3 | 65 | 3 | 9 |
| Fisherman | 2 | 2 | 7.2 | 75 | 2 | 4 |
| Okupe | 1 | 1 | 7.2 | 65 | 2 | 2 |
| Mana-Otaheke | 0.5 | 1 | 7.1 | 90 | 1 | 1 |
| Subduction interface | 18 | 3 | 8.7 | 230 | 12 | 36 |
- a Hypothetical earthquakes modeled to relieve completely the slip deficit cumulated over the past 2000 years from secular loading in southern Cook Strait (Figure 14c). The modeled events involve ruptures of individual segments. Other earthquake scenarios relieving completely the slip deficit are possible (e.g., combined ruptures of nearby segments, different number of events), resulting in the same present-day loading.
[45] A compilation of paleoearthquakes and displacement history data for sites on the offshore Wairau [Barnes and Pondard, 2010], Cloudy, and Vernon faults confirms that over five to eight earthquake cycles during the last 18 ka, there have been similar periods of time when the faults have been relatively more active or quiescent, and that some earthquakes on each fault have occurred close in time (Figure 12a). Notably, all three faults reveal (1) significant elapsed time since their last earthquakes (Wairau Fault, 2.0 ± 0.3 ka; Cloudy Fault, 1.8 ± 0.3 ka; Vernon Fault, 3.2 ± 0.7 ka), (2) a significant active vertical growth period in which two to three earthquakes occurred on each fault between about 2 and 5.5 ka, and (3) a period of relative quiescence between about 5.5 and 8.5 ka. Furthermore, both the Wairau and Cloudy faults had a particularly significant growth period between about 8.5 and 13 ka, during which time large coseismic vertical displacements occurred. In Figure 13, we compare the timing of the Holocene submarine paleoearthquakes in Cook Strait with published paleoearthquake records from onshore studies. Notwithstanding the variable uncertainties, it appears notable that (1) four faults (Awatere, Ohariu, Wellington, and Wairarapa) experienced earthquakes between about 0.9 and 1.1 ka, (2) at least five (Wairau, Cloudy, Clarence, Wellington and Wairarapa faults) experienced earthquakes between about 1.8 and 2.3 ka, and (3) at least five (Wairau, Vernon, Awatere, Ohariu, and Wairarapa faults) ruptured between about 4.5 and 5.5 ka. Collectively, it appears that despite uncertainties in the precise earthquake ages, the available data indicate the occurrence of temporal earthquake clustering [e.g., Van Dissen et al., 2007]. This observation is consistent with the results of synthetic seismicity modeling of the same region by Robinson et al. (submitted manuscript, 2010).
[46] We interpret that fault interactions have enhanced earthquake triggering, that specific sequences of clustered events do not always involve the same faults, and that particular states of stress loading favor earthquake clustering. For clustering to occur, we infer that regional Coulomb stress loading must be relatively uniform and close to failure along several nearby fault segments, so the lateral loading associated with an earthquake rupture is capable of triggering earthquakes along other faults nearby. If the resulting coseismic slip (static stress drop) causes loading to be less uniform, then subsequent earthquake clusters may involve different faults. Furthermore, if a particular fault accrues a large slip deficit that is not fully relieved by an earthquake, a subsequent earthquake on that fault could potentially recur after a relatively short interval, depending on subsequent fault interactions.
7.4. Coulomb Stress Distribution: Implications for Seismic Hazard
[47] To gain insights into the stress conditions that might trigger future earthquakes in the Cook Strait region, we calculate the present Coulomb static stress distribution (Figure 14). The Coulomb stress distribution provides insight into which faults are relatively more loaded than others, but not the absolute state of stress along any particular fault.
[48] Changes of static stress associated with large historical earthquakes (M ≥ 7.0) are modeled using rectangular dislocations in an elastic half-space [e.g., King et al., 1994; Stein et al., 1997]. The model includes the effects of lateral stress changes imposed by the occurrence of previous earthquakes and secular tectonic loading from below the fault segments, described by Pondard et al. [2007]. The slip and length of rupture for the dislocation sources (Tables 1 and 2) are chosen to be consistent with seismic moments determined from morphologic and paleoseismic observations, and segment lengths [Van Dissen and Berryman, 1996; Litchfield et al., 2006; Zachariasen et al., 2006; McSaveney et al., 2006; Mason and Little, 2006a, 2006b; Cochran et al., 2007; Langridge et al., 2009; Van Dissen and Nicol, 2009; Little et al., 2009; Barnes and Pondard, 2010; this study]. Strike-slip earthquake ruptures are considered to extend from the surface to a depth varying progressively from 8 km in Southern Cook Strait to 22 km in Northern Cook Strait, and connecting to the Hikurangi subduction interface at depth. The coupling depth of the subduction interface is approximately 40 km across the Cook Strait region, according to Wallace et al. [2009]. A regional stress field is imposed in agreement with the kinematics of the Cook Strait region to determine optimum failure directions [Beavan et al., 2002], unless strike, dip and rake of the faults are specified.
[49] The secular tectonic loading model is calculated from active fault source parameters and Australian-Pacific relative plate motion rates derived from GPS velocities and tectonic observations (Figure 14a) [Beavan et al., 2002; Barnes et al., 2008; Robinson et al., submitted manuscript, 2010]. The elements for the secular loading extend from below the segments described above to great depth. We impose the secular tectonic loading following the cluster of large earthquakes that occurred about 2 ka. Therefore, all events older or equal than 2 ka have not been included in the modeling. The subsequent calculations of contemporary Coulomb stress depend critically on the stress relief associated with major paleohistorical and historical earthquakes since then (Tables 1 and 2 and Figure 13).
[50] Because some faults have not ruptured in historical times, and as yet have no paleoearthquake records, we test two regional scenarios. In Figure 14b we model the regional secular loading and only the stresses relieved from known paleohistorical and historical earthquakes (Table 1 and Figure 13). In this scenario no stress has been relieved on major faults in southern Cook Strait (Boo Boo, Needles, Campbell Bank, and Chancet faults) and the Narrows Basin (compare Figure 1), nor on the southern Hikurangi subduction interface, because there are no paleoearthquake records available (Figure 13) [Litchfield and Wilson, 2008; Wilson et al., 2008]. As a result, loading in southern North Island has been partially relieved during the 1855 M 8.2 Wairarapa earthquake and other paleoearthquakes associated with crustal (upper plate) faults, but all faults in that area remain in a high state of stress, principally as a result of stress transfer due to the secular tectonic loading along the subduction interface. Similarly, the region of very high Coulomb stress along the shelf of southern Cook Strait largely reflects the maximum secular loading on those faults, combined with the subduction interface. This model is clearly not realistic, as it is inconceivable that no earthquakes have occurred in the last 2000 years on such high-slip-rate (10–20 mm/yr) faults for which we have no paleoearthquake records. The model in Figure 14b therefore corresponds to the Coulomb stress distribution associated with the maximum possible regional cumulative slip deficit.
[51] In the second scenario (Figure 14c), we assume that the 2000 years of secular tectonic loading has been completely relieved along the above faults (for which we have no paleoearthquake records) by major earthquakes (Table 2). For example, 22 m of slip deficit accumulated along the Boo Boo Fault in 2000 years at a secular rate of 11 mm/yr has been relaxed by modeling 22 m of coseismic slip (the equivalent of three hypothetical M 7.6 earthquakes each associated with 7.3 m of coseismic slip), while 36 m of slip deficit on the Hikurangi subduction interface has been relieved (the equivalent of three M ∼ 8.7 earthquakes with 12 m of slip per event) [Wallace et al., 2009]. The present-day loading results from the total slip deficit cumulated from secular loading, competing with the total slip deficit relieved from earthquakes. Therefore, the precise timing of the earthquakes is not significant to present-day loading calculation in Figure 14c. This model corresponds to the Coulomb stress distribution associated with the minimum possible cumulative slip deficit (i.e., minimum present-day loading) in the Cook Strait region. In this scenario the tectonic loading along faults in southern North Island and Marlborough have been completely relaxed, as a result of stress relief associated with upper-crustal earthquakes and subduction interface earthquakes, but high Coulomb stress remains distributed throughout much of Cook Strait, where there are numerous fault terminations and step-overs.
[52] The true present-day Coulomb stress loading very likely corresponds to an intermediate scenario between those calculated in the two scenarios in Figures 14b and 14c, with elastic strain accumulated on some of the faults, but not all. Nevertheless, similar conditions are identified in both scenarios, in central Cook Strait and parts of northern South Island. The state of stress there has important consequences for seismic hazard because it likely corresponds closely to the true present-day loading in these areas where the paleohistorical and historical earthquakes have been accounted for [e.g., Rhoades et al., 2010]. Our study suggests that loading has been relieved in northern South Island, largely as a result of the stress shadow related to earthquakes on the Awatere and Clarence faults (Table 1) [e.g., Harris and Simpson, 1998], including the 1848 Marlborough earthquake [Grapes et al., 1998; Mason and Little, 2006a]. Interestingly, present Coulomb stress along the much of the northern Wairau Fault and its offshore extension appears to have been relaxed by the Awatere Fault ruptures, suggesting that the fault may not yet be close to failure. This occurs despite the Wairau Fault having not ruptured since 2.0 ± 0.3 ka, an elapsed time (1) close to the long-term average recurrence (2.2 ka) and more than 50% of the longest recurrence interval observed (3.8 ka) and (2) equivalent to ∼2 m of vertical slip deficit (compare 2.5 m mean coseismic vertical displacement, range 0.5–5.3 m) [Barnes and Pondard, 2010]. The present stress state along the Cloudy and eastern Vernon faults remains high as a result of slip deficit and stress transfer due to interactions with nearby earthquake ruptures (e.g., along the subduction interface; Figure 14). The high stress state on these faults is particularly noteworthy, as both could be in advanced stages of their respective seismic cycles (65% and 100% of their long-term average recurrence intervals, respectively; Figure 10). In southern Cook Strait the Needles and Campbell Bank faults, and the western part of the Boo Boo Fault, are also presently stressed as a result of stress transfer due to interactions with nearby earthquake ruptures, and possibly slip deficit. The current magnitude of slip deficit on of these faults is unknown. The modeling therefore indicates high present-day stress throughout much of central Cook Strait. However, apart from the offshore Wairau, Cloudy and Vernon faults, there is considerable uncertainty in the contemporary hazard in the absence of wider paleoearthquake data.
[53] Finally, our study indicates that high stress has accumulated where strike-slip faults terminate in central Cook Strait, due to stress loading associated with successive historical earthquakes. The likelihood of long-term high stress state in this area as a result of the fault step-overs may eventually lead to them propagating beyond their present-day termination to relieve stress accumulation at fault discontinuities.
8. Conclusions
[54] 1. New mapping of active submarine strike-slip faulting in Cook Strait, using extensive multibeam bathymetry and high-resolution seismic reflection data, provides a new understanding of the tectonic structure within the tripartite transition between the Hikurangi subduction zone, the strike-slip Marlborough Fault System, and the North Island Dextral Fault Belt. There are no throughgoing crustal faults connecting North and South islands of New Zealand. Rather, there is a general discontinuity of the major strike-slip faults and a change in fault strike in central Cook Strait. Whereas the predominant structural trend is SW–NE to WSW–ENE in South and North islands, faults in central and southern Cook Strait are predominantly E–W striking, dextral faults. Many are seaward extensions of strike-slip faults on land. The submarine fault system is characterized by a series of 5–20 km wide step-overs, which are associated with localized vertical deformation. The termination of several major crustal faults in central Cook Strait has important implications for seismic hazard, as throughgoing strike-slip earthquake ruptures across the strait now appear improbable. Active faults in the Wairau Basin accommodate strike slip and extension, in contrast to faults in southern Cook Strait which accommodate strike slip and compression.
[55] 2. Glacioeustatic sea level cycle sedimentary sequences and unconformities of 0–140 ka are identified beneath the continental shelf of northern South Island. Identification of the postglacial transgressive erosion surface (∼20–7 ka) and postglacial sediment cover in high-resolution seismic profiles provide a stratigraphic framework for detailed analysis of fault vertical displacement, growth faulting, and paleoearthquake history. In general the vertical separation rates represent minimum net slip rates, given the structures are predominantly strike slip. Vertical separation rates on faults in the Wairau Basin range from 0.1 to 1.5 mm/yr, which compare with onshore estimates of dextral slip rates of 3–5 mm/yr.
[56] 3. Submarine paleoearthquake records have been derived for the offshore Cloudy and Vernon faults, from identification of coseismic vertical displacements. Up to six paleoearthquake are recognized on the extensional Cloudy Fault, and up to five on the strike-slip Vernon Fault, since 18 ka. These long-timescale records indicate significant variability in earthquake recurrence and coseismic vertical displacements. The recurrence intervals between the Cloudy Fault paleoearthquakes range from 1.0 to 5.7 ka, with a mean of 2.7 ka, while those on Vernon Fault range from 2.0 to 4.0 ka, with a mean of 3.1 ka. The mean coseismic vertical displacements per event on these faults range from 1.0 to 6.4 m, and 0.4 to 2.6 m, respectively. The observed variability on these faults is similar to a long-timescale (18 ka) record of eight paleoearthquakes recognized by Barnes and Pondard [2010] on Wairau Fault. Collectively these data indicate that the faults conform to the variable slip model of earthquake behavior, that stress drop may possibly vary between earthquakes on the same fault, and that their seismic cycles are not perfectly periodic. If the vertical to horizontal displacement ratio is approximately constant at any given site, not all paleoearthquakes have a vertical coseismic displacement that would be predicted from the duration of the elapsed time preceding the event. Acknowledging the possibility that there are missing earthquakes not recognized, these observations may suggest there are poor relationships between tectonic loading and relaxation processes at the scale of individual seismic cycles in complex fault systems, but that these relationships may be more uniform over longer timescales, covering multiple seismic cycles. An alternative explanation for the vertical displacement variability is that the vertical to horizontal slip ratio is significantly nonconstant.
[57] 4. The submarine fault displacement histories and paleoearthquake records in Cook Strait, combined with paleoearthquake and historical seismicity records for faults on land, provide a valuable data set to evaluate possible fault interactions over many seismic cycles. Despite uncertainties in earthquake ages, earthquake clustering can be recognized from these combined records. These observations suggest that fault interactions associated with past earthquakes have contributed to loading conditions favoring the triggering of earthquake sequences, and that earthquake clusters do not always involve the same faults. A particular state of relatively uniform stress and advanced loading drives earthquake clustering on major faults in complex systems.
[58] 5. Present-day Coulomb stress distribution has been calculated in the Cook Strait and Wellington City region, including the effects of secular tectonic loading (slip deficit) on faults and stress transfer associated with earthquake ruptures since 2 ka. Two scenarios, considered to represent minimum and maximum present-day loading, both indicate that (1) stress has been relieved in parts of northern South Island as a result of earthquake ruptures on the Clarence and Awatere faults, (2) the above area of relaxed Coulomb stress envelopes much of the Wairau Fault, despite the fault having accumulated a vertical slip deficit of the order of 2 m and being advanced in its earthquake cycle, (3) Coulomb stress remains high on the submarine Cloudy and Vernon faults in central Cook Strait, both of which may be currently advanced in their seismic cycles (65% and 100% of their long-term average recurrence intervals, respectively), and (4) present Coulomb stress is high also in the vicinity of several faults in southern Cook Strait (Needles, Campbell Bank, western Boo Boo faults). However, there is considerable uncertainty in the contemporary hazard in the absence of wider paleoearthquake data for these faults.
[59] 6. Finally, high stress has accumulated where strike-slip faults terminate in central Cook Strait, due to stress loading associated with successive historical earthquakes. The likelihood of long-term high stress state in this area as a result of the fault step-overs may eventually lead to them propagating beyond their present-day termination to relieve stress accumulated at fault discontinuities.
Acknowledgments
[60] We thank John Mitchell, Steve Wilcox, and other participants of R/V Tangaroa voyages TAN0211 and TAN0215 (2002), during which the SIMRAD multibeam bathymetric data were acquired, and TAN0510 (2005), during which the boomer seismic reflection data were collected. Geoffroy Lamarche assisted with the early stages of TAN0510 seismic data planning, assisted with postvoyage seismic processing, and provided access to archived seismic data acquired on Tangaroa voyage CR3048. Discovery Geo Corporation is thanked for providing access to multichannel seismic line DG01. Stuart Henrys provided us the OGS Explora seismic line collected by GNS Science. Claire Castellazzi assisted with use of Geographix Seisvision software. We thank Scott Nodder, Geoffroy Lamarche, Russell Van Dissen, Tim Little, Rob Langridge, Kate Wilson, Nicola Litchfield, and Russell Robinson for various discussions concerning sequence stratigraphy, regional tectonics, earthquake sources, paleoearthquake histories, and Coulomb stress modeling. Laura Wallace provided the Hikurangi subduction interface source model for our Coulomb stress model. Geoffrey C. P. King and David Bowman developed the software Nutcracker Marzipan, used to model the Coulomb stress interactions. Geoffroy Lamarche provided a constructive review of a draft of the manuscript. Journal reviewers Jean-Yves Collot and Nathan Hayward provided useful critical reviews that improved this paper. Funding for this research was granted by the New Zealand Foundation for Research Science and Technology contracts CO1X0702 and CO1X0203, the NIWA Capability Fund, and the New Zealand Earthquake Commission research fund as part of the “It's Our Fault” project.
