- Split View
-
Views
-
Cite
Cite
John Ristau, Caroline Holden, Anna Kaiser, Charles Williams, Stephen Bannister, Bill Fry, The Pegasus Bay aftershock sequence of the Mw 7.1 Darfield (Canterbury), New Zealand earthquake, Geophysical Journal International, Volume 195, Issue 1, October 2013, Pages 444–459, https://doi.org/10.1093/gji/ggt222
- Share Icon Share
Abstract
The Pegasus Bay aftershock sequence is the most recent aftershock sequence of the 2010 September 3 UTC moment magnitude (Mw) 7.1 Darfield earthquake in the Canterbury region of New Zealand. The Pegasus Bay aftershock sequence began on 2011 December 23 UTC with three events of Mw 5.4–5.9 located in the offshore region of Pegasus Bay, east of Christchurch city. We present a summary of key aspects of the sequence derived using various geophysical methods. Relocations carried out using double-difference tomography show a well-defined NNE–SSW to NE–SW series of aftershocks with most of the activity occurring at depths >5 km and an average depth of ∼10 km. Regional moment tensor solutions calculated for the Pegasus Bay sequence indicate that the vast majority (45 of 53 events) are reverse-faulting events with an average P-axis azimuth of 125°. Strong-motion data inversion favours a SE-dipping fault plane for the largest event (Mw 5.9) with a slip patch of 18 km × 15 km and a maximum slip of 0.8 m at 3.5 km depth. Peak ground accelerations ranging up to 0.98 g on the vertical component were recorded during the sequence, and the largest event produced horizontal accelerations of 0.2–0.4 g in the Christchurch central business district. Apparent stress estimates for the two largest events are 1.1 MPa (Mw 5.9) and 0.2 MPa (Mw 5.8), which are compatible with global averages, although lower than other large events in the Canterbury aftershock sequence. Coulomb stress analysis indicates that previous large earthquakes in the Canterbury sequence generate Coulomb stress increases for the two events only at relatively shallow depths (3–5 km). At greater depths, Coulomb stress decreases are predicted at the locations of the two events. The trend of the aftershocks is similar to mapped reverse faults north of Christchurch, and the high number of reverse-faulting mechanisms suggests that similar reverse-faulting structures are present in the offshore region east of Christchurch.
INTRODUCTION
The Pegasus Bay aftershock sequence is the most recent aftershock sequence resulting from the 2010 September 3 UTC moment magnitude (Mw) 7.1 Darfield earthquake in the Canterbury region of New Zealand (Gledhill et al.2011). The Pegasus Bay aftershock sequence began on 2011 December 23 UTC with three events of Mw 5.4–5.9 in Pegasus Bay east of Christchurch, New Zealand's second largest city (population ca. 377 000). The vast majority of the events occurred in the offshore region several kilometres east of Christchurch, which helped minimize their impact on populated areas.
A summary of the aftershock sequence is presented here with analysis using a variety of geophysical methods. Focal mechanisms derived from regional moment tensor (RMT) analysis and kinematic source models derived from strong-motion data are used to characterize the source processes for the largest events. Coulomb stress analysis and analysis of apparent stresses shows how previous large events in the Canterbury aftershock sequence influenced the Pegasus Bay sequence. Ground motions derived from strong-motion data for the two largest events are compared with the largest earthquakes in the Canterbury aftershock sequence. Aftershock relocations using double-difference tomography and RMT-derived focal mechanisms are used to examine how the aftershocks compare with previous aftershocks in the Canterbury sequence and with mapped faults in the region.
Tectonic setting and historical seismicity
New Zealand straddles the boundary of the Pacific and the Australian plates and its active tectonics are dominated by three main features (Fig. 1). Beneath the North Island and northern-most South Island the Pacific Plate subducts obliquely beneath the Australian Plate at the Hikurangi Trough. In contrast, in the Fiordland region in the southwest area of the South Island the Australian Plate subducts obliquely beneath the Pacific Plate at the Puysegur Trench. Linking the two subduction zones is the Alpine Fault which delineates the plate boundary and runs across the South Island for approximately 650 km. In the central South Island Australian/Pacific plate convergence occurs at 38 mm yr–1 (DeMets et al.2010), with about three-quarters of the convergence taken up in a narrow zone along the Alpine Fault with dextral and reverse slip rates of up to 25 mm yr–1 and 10 mm yr–1, respectively (Sutherland et al.2006; Norris & Cooper 2007). Palaeoseismic evidence suggests that the Alpine Fault ruptures in major earthquakes (M > 7.5) with recurrence intervals of ∼200–300 yr, with the most recent event in 1717 (e.g. Cooper & Norris 1990; Yetton et al.1998; Rhoades & Van Dissen 2003; Sutherland et al.2007; Berryman et al.2012). In the northern South Island plate motion is largely accommodated by a series of strike-slip faults known as the Marlborough Fault System (MFS); the slip rate on the Hope Fault at the southern end of the MFS is approximately 20 mm yr–1 (Cowan 1991; Van Dissen & Yeats 1991; Langridge & Berryman 2005). A balancing of the plate motion budget across the central South Island using geodetic, seismological and geological data suggests that up to 5 mm yr–1 of active deformation is possible on faults distributed within the Southern Alps and up to 100 km to the east of the Alpine Fault (Wallace et al.2007). The Canterbury region is located >100 km from the plate boundary with substantially lower slip and strain rates.
A number of M 6–7 earthquakes have occurred in the Southern Alps and their eastern foothills in the past 150 yr. These include 1888 North Canterbury Mw 7.1 (Cowan 1991), 1929 Arthur's Pass Mw 7.0 (Doser et al.1999), 1994 Arthur's Pass Mw 6.7 (Abercrombie et al.2000) and 1995 Cass Mw 6.2 (Gledhill et al.2000) earthquakes. There are also a number of mapped active faults in the eastern foothills of the Southern Alps (see Stirling et al.2008 for a summary). Faults are inferred to extend beyond the range front as buried structures beneath the flat-lying topography, as revealed in high-resolution seismic reflection surveys in the western Canterbury Plains (Dorn et al.2010). The Canterbury earthquake sequence was located east of the foothills where no active surface faults had previously been mapped, and demonstrates that the zone of active deformation in the eastern South Island extends beyond the visible range front. Crustal structure in the region of the Canterbury earthquake sequence is consistent with that of the Chatham Rise on the converging Pacific Plate (Reyners & Cowan 1993). The Chatham Rise is a buoyant marine plateau formed by fragmentation of the Gondwana continental margin in the mid-to-late Cretaceous. It is cut by major east–west trending faults established sometime in the late Early Cretaceous (Wood & Herzer 1993; Eberhart-Phillips & Bannister 2002; Browne et al.2012; Jongens et al.2012).
The 2010–2012 Canterbury earthquake sequence (Bannister & Gledhill 2012) occurred within the 30 km thick continental crust belonging to the Chatham Rise within the Pacific Plate. Local geology comprises a basement of highly deformed metagraywackes and metamorphosed equivalents, overlain by a Late Cretaceous–Neogene cover sequence up to ∼2.5 km thick (Forsyth et al.2008). The cover sequence consists of Late Cretaceous–Palaeogene sedimentary units overlain by a Miocene–Pliocene clastic sequence that contains the predominantly basaltic Late Miocene Banks Peninsula volcanics. In the Canterbury Plains the older units are largely obscured by Pliocene and Quaternary alluvial gravels up to a few hundred metres thick (Forsyth et al. 2008). These post-glacial gravel deposits tend to mask evidence of tectonic structures east of the Southern Alps range front in the Canterbury Plains.
Prior to 2010 September seismicity in the Canterbury Plains had been very low with no large earthquakes occurring since written records began ∼170 yr ago (Fig. 1). The strongest shaking experienced historically from local events was MM 7, both from a shallow M 4.7–4.9 earthquake near Christchurch (ca. 43.6°S 172.6°E) on 1869 June 4, and from a lower crustal M 5.6–5.8 earthquake near Lake Ellesmere (ca. 43.8°S 172.5°E) on 1870 August 31 (Pettinga et al.2001).
Darfield earthquake
The Pegasus Bay aftershock sequence is part of the 2010–2012 Canterbury sequence (Bannister & Gledhill 2012), which began with the 2010 September 3 Mw 7.1 Darfield earthquake, with an epicentre approximately 40 km west of Christchurch. The Darfield earthquake produced a 29.5-km long E–W striking surface rupture, which has been named the Greendale Fault. Beavan et al. (2010), Quigley et al. (2010) and Gledhill et al. (2011) provide detailed overviews of the Darfield earthquake and its associated surface rupture. Since the Darfield earthquake, GeoNet has located more than 10 000 aftershocks in the Canterbury Plains and the offshore region east of Christchurch with local magnitude (ML) as low as ML 1.1 (Fig. 2).
The largest aftershock of the Darfield earthquake was Mw 6.2 on 2011 February 21 UTC approximately 6 km SE of the Christchurch city centre on the Port Hills Fault (e.g. Bannister et al.2011; Holden 2011; Atzori et al.2012; Elliot et al.2012; Kaiser et al.2012). The Christchurch aftershock resulted in more than 180 fatalities and caused significant widespread building damage and liquefaction, particularly to the Christchurch central business district (CBD; e.g. Cubrinovski et al.2011; Iizuka et al.2011). A Mw 6.0 aftershock occurred on 2011 June 13 UTC ∼4 km east of the February earthquake, which also produced significant damage and liquefaction in the Christchurch area (Kaiser et al.2012).
PEGASUS BAY AFTERSHOCK SEQUENCE
Overview
The Pegasus Bay aftershock sequence began with a Mw 5.8 earthquake on 2011 December 23 at 00:58 UTC in the offshore region 20 km east of Christchurch, followed by a Mw 5.4 at 01:06 UTC and a Mw 5.9 at 02:18 UTC. The earthquakes were widely felt in Christchurch and the Canterbury region with 1481 felt reports for the Mw 5.8 event and 1042 felt reports for the Mw 5.9 event. Nearly 100 aftershocks were located in the offshore region east of Christchurch in the first 48 h and 1075 events have been located through 2012 September 2.
Ground motions
Peak ground accelerations in the two largest earthquakes of the Pegasus Bay sequence (Mw 5.8; Mw 5.9) are shown in Fig. 3. Particularly large vertical accelerations were recorded during the first Mw 5.8 event (up to 0.98 g at New Brighton Library, NBLC; Fig. 3a). However, during the larger event (Mw 5.9), strong motion records were not returned from NBLC, the closest station to the epicentre. Recorded ground motions in the larger event (Mw 5.9) ranged up to 0.67 g (horizontal) at Heathcote Valley (HVSC) and 0.39 g (vertical) at Hulverstone Drive Pumping Station (HPSC; which recorded 0.36 g in the earlier Mw 5.8 event). In the CBD, peak ground acceleration ranged from 0.2–0.4 g (horizontal) and 0.1–0.2 g (vertical). These values are comparable to accelerations observed in the CBD during the Mw 6.0 June, and Mw 7.1 September earthquakes for spectral periods up to ∼1.2 s, but significantly weaker than those recorded during the Mw 6.2 February earthquake (Fig. 4). In general, the observed horizontal spectral accelerations at 0.5 s during the larger (Mw 5.9) earthquake are reasonably well encompassed by the range of Next Generation Attenuation (NGA) and New Zealand models (Fig. 5; shown for deep soil stations).
The spatial variability in observed ground motions is also likely to be influenced by site effects, including high-frequency amplification at HVSC as well as both basin and near-surface effects at stations located on the plains (outside the Port Hills areas shown in Fig. 3) (Bradley & Cubrinovski 2011; Fry et al.2011; Kaiser et al.2013). At stations on the plains, significant high-frequency energy on the vertical component is inferred to arise from reverberations above the shallow water table, whereas a corresponding lack of high-frequency energy on the horizontal components arises from attenuation of shear wave energy in shallow saturated (or liquefied) soils (e.g. Fry et al.2011).
Moment tensor solutions
53 RMT solutions have been calculated for the Pegasus Bay aftershock sequence with Mw 3.4–5.9, including five with Mw ≥ 5, using the RMT method developed at the University of California, Berkeley Seismological Laboratory (Dreger & Helmberger 1993; Pasyanos et al.1996; Dreger 2003) (Fig. 6). Fig. 7 shows the waveform fits and changes in variance reduction and focal mechanism with depth for the largest event in the sequence (Mw 5.9). The observed waveforms and Green's functions were bandpass filtered at 0.01–0.04 Hz and the best-fit depth was calculated at 6 km. For RMT solutions the observed waveforms and Green's functions are typically filtered in a subrange of 0.01–0.1 Hz (e.g. 0.02–0.05 Hz; 0.03–0.1 Hz) depending on the signal-to-noise ratio. For all events the epicentres (latitude/longitude) are fixed to the relocated epicentre when available, otherwise the original catalogue epicentre is used. Depth is solved for by calculating solutions over a range of depths to find the depth with the largest variance reduction between the observed and synthetic waveforms.
The focal mechanisms derived from the RMT solutions show a high degree of consistency with the vast majority (45 out of 53 events) being reverse or oblique-reverse faulting mechanisms (Fig. 6; Fig. 8, left). The seismic consistency (Cs) is a measure of the consistency of a group of moment tensor solutions which has a value of 0 if the individual moment tensor solutions cancel one another and 1.0 if all the moment tensors are the same except for their scalar moment (Frolich & Apperson 1992). For the RMT solutions in the Pegasus Bay sequence Cs = 0.82, which suggests a high degree of similarity among the moment tensor solutions. A small number of strike-slip mechanisms are present at the southern end of the aftershock zone, which could possibly be related to ENE–WSW structures.
Fig. 6 (bottom left) shows a WNW–ESE cross-section using depths derived from double-difference relocations when available (see ‘Aftershock relocations’), otherwise from RMT solutions. The RMT-derived depths are shallower than those determined from double-difference relocations; however, the uncertainties in the RMT depths is around ±3–5 km. Uncertainties in depths from double-difference relocations are typically on the order of several hundred metres (e.g. Walhauser & Ellsworth 2000). Most events in the Pegasus Bay sequence are located offshore and the depth uncertainties will be greater than several hundred metres; however, they are still likely to be more reliable than the RMT depths. Even with the improved double-difference depths, the cross-section does not clearly show a group of aftershocks consistent with either fault plane of the two largest events (Fig. 6, bottom left).
The P-axes show a high degree of consistency with most events having a NW–SE azimuth and an average azimuth of 125° (305°) (Fig. 8, right). A stress tensor inversion of the focal mechanisms using the method of Michael (1984, 1987) gives a well-constrained maximum compressive stress (Φ1) of 123° (Fig. 9), which is similar to the regional stress field across the South Island where the Φ1 is ∼115° ± 5° (e.g. Sibson et al.2011). Elliot et al. (2012) found the P-axes for the various fault segments of the Darfield and Christchurch earthquakes lie in the range of 110°–140°. This is within the error of the principal contraction of the Canterbury region reported by geodetic results of 110°–120° (Wallace et al.2007).
Kinematic source inversion
Kinematic source models have been calculated for the Mw 5.9 event following Holden et al. (2011) and Holden (2011). Three-component strong-motion data are inverted using a fixed fault geometry and synthetic waveforms are computed from 1-D velocity models (Fig. 11) using the discrete wavenumber approach of Bouchon (1981). 13 stations were used in the inversion and the data were filtered from 0.1 to 0.5 Hz and integrated into velocity. The fault planes were constrained by the location of the relocated hypocentre and the fault geometry is determined by the RMT solution. Both fault planes are tested and data is inverted for one elliptical patch of slip with various sizes, locations, amplitudes, directions of slip and rupture times. To search for a minimum waveform misfit, we used the non-linear neighbourhood algorithm (Sambridge 1999). The misfit function is a least-squares scheme applied to the first 70 s of all three components of the recordings. The parameter ranges are 1–7 m s–1 for the rupture velocity, 0.5–5.5 m for maximum slip and 90–180 for the rake angle.
Of the two fault planes calculated for the Mw 5.9 event the southeast-dipping fault plane (strike/dip/rake: 57/51/123) gave the better waveform fits (Fig. 10). The misfit for the SE-dipping plane is 0.00294, and for the NW-dipping plane 0.00365. The rupture is characterized by an elliptical slip patch of 18 × 15 km2 and maximum slip of 0.8 m at 3.5 km depth. This is similar to the 2011 February Christchurch earthquake where the aftershocks happened at depth and the region above the slip area had very few aftershocks. The fault area is large for a Mw 5.9 earthquake compared with empirical relations of ∼85 km2 (e.g. Somerville et al.1999), and the reason for this requires further study. The slip direction is 127° and the rupture propagated up the fault plane. The calculated moment (Mo) was 1.42 × 1018 Nm (Mw 6.0), which is a factor of two larger than the moment derived from the RMT solution (Mo 7.01 × 1017 Nm; Mw 5.9). This model is similar to the model proposed by Beavan et al. (2012) using InSAR and GPS; however, the geodetic model is only well constrained at its western end due to the offshore hypocentre. Beavan et al. (2012) estimate a moment of 1.7 × 1018 Nm; however, this includes aftershocks as well as some potential post-seismic slip. The fits between the observed and synthetic waveforms are good (Fig. 11). A resonant period of ∼3 s is notable on the horizontal components of soft soil sites (e.g. CBGS; SHLC), similar to that noted by Cousins & McVerry (2010) and Bradley & Cubrinovski (2011) during other large Canterbury events. The majority of the slip is concentrated at shallow depths, less than ∼7 km depth, which is above much of the relocated aftershock activity. The RMT-derived depths are typically shallower; however, the depths derived from double-difference relocations are preferred.
Apparent stress estimates
Preliminary analyses of apparent stress derived from teleseismic data suggest that the apparent stress drops associated with the Pegasus Bay earthquake sources are compatible with global averages for crustal events (0.31 MPa for continental reverse-faulting earthquakes, and 3.62 MPa for continental strike-slip earthquakes; Choy & Boatwright 1995). Radiated energy (Es) is calculated using the method described in Fry & Gerstenberger (2011), which is based on Boatwright & Choy (1986) and Choy & Boatwright (1995). The method involves measuring the energy flux from high-frequency velocity records, and incorporating corrections for rupture processes and wave propagation. Apparent stress (τα) is defined as the product of the rigidity (μ) and Es per unit moment (| $ \tau _\alpha = {({\mu \times E_s})}/ {M_0}$|) (Wyss & Brune 1968). Fry & Gerstenberger (2011) used teleseismic data as regional data requires refined knowledge of local attenuation structure and site responses which are currently not well resolved.
Apparent stress (radiated energy per unit moment) of the two largest aftershocks is ∼1.1 MPa (Mw 5.9) and ∼0.2 MPa (Mw 5.8), respectively. The apparent stress estimates for the three other large events of the Canterbury sequence as determined by identical methodology are higher (∼16 MPa, 2010 September; ∼4 MPa, 2011 February; ∼6 MPa, 2011 June; Fry & Gerstenberger 2011). The high stress of the previous large earthquakes has been interpreted as indicative of strong regional faults (Fry & Gerstenberger 2011; Fry et al.2011), but this assumption may not apply to the Pegasus Bay region. However, we note that comparisons of source stress properties are notoriously challenging and dependent on the methodology. Ongoing work based on local recordings (B. Fry & R. Abercrombie 2012, personal communication; Oth & Kaiser 2013) suggests that in contrast the two largest Pegasus Bay earthquakes may still be associated with higher than average stress drop.
The interpretation of the differences in apparent stress of the largest Pegasus Bay events and the earlier large Canterbury events is further complicated by our limited understanding of the implications of Banks Peninsula, a 6 Ma volcano to the southeast of the city, on the regional stress regime. Townend et al. (2012) compared the orientation of the maximum horizontal compressive stress (SHmax) from focal mechanisms recorded before and after the 2011 February 21 Mw 6.2 Christchurch earthquake. They found no significant change in SHmax orientation suggesting that the high-stress drop Christchurch earthquake was incapable of substantially modifying the ambient stress field. Therefore, the lower apparent stress drops associated with the Pegasus Bay earthquakes would not be caused by any changes in the regional stress field related to the Christchurch earthquake.
Aftershock relocations
Events from the Pegasus Bay sequence have been relocated using the double-difference tomography approach of Zhang et al. (2009), which builds on earlier work by Zhang & Thurber (2003). The technique minimizes the residuals between observed and calculated arrival-time differences for pairs of closely located earthquakes, while also minimizing the residuals of absolute arrival times. The algorithm solves for the hypocentral parameters of the earthquakes, also allowing for some modification of the P- and S-wave velocity structure used for traveltime calculation.
The initial 3-D velocity model used was based on the most recent version of the 3-D New Zealand velocity model (Eberhart-Phillips et al.2010). The New Zealand-wide model is interpolated to a denser rectilinear grid using Delaunay triangulation. Then a series of inversion runs is carried out using the tomoDDPS approach of Zhang et al. (2009), slowly decreasing the inversion node spacing as the event and station density allow, with appropriate smoothing and weighting constraints (Zhang et al.2009). Bannister et al. (2011) contains a detailed description of the double-difference relocation technique used for earlier events in the Canterbury earthquake sequence.
Fig. 12 shows the original locations (top) and the relocated epicentres (middle) for 556 events in the Pegasus Bay sequence. It is difficult to quantitatively determine uncertainties in the relocated epicentres, but the main concern is the uncertainty in the absolute locations rather than relative location, due to the lack of constraint on the crustal velocity structure offshore, away from the seismometers, and the limited range of azimuthal coverage for these offshore events. As almost all of the events are located in the offshore region east of Christchurch the azimuthal station coverage is less than 180°, which limits improvement in the relocations.
The relocated epicentres show a more defined NNE–SSW to NE–SW trend than the original locations. In particular, many of the events furthest east were relocated west of their original epicentres. The trend of the relocated aftershocks differs significantly from the E–W strike of the Greendale Fault and the ENE–WSW strike of the Port Hills Fault (Fig. 2). Geodetic results from Beavan et al. (2012) show a rupture plane with a 60° strike for the two largest Pegasus Bay events. This is consistent with the Port Hills Fault but differs from the trend of the aftershocks. Mapped faults in the offshore region east of Christchurch typically strike E–W similar to the Greendale Fault; however, no offshore faults have been mapped in the immediate vicinity of the Pegasus Bay sequence. Overall there is no distinct pattern to the location of the aftershocks over time. On 2012 January 1 a Mw 5.1 aftershock occurred in the NE part of the aftershock zone and a number of events were located in that area over the next few days (yellow events in Fig. 12b). The aftershocks appear to migrate to the central part of the aftershock zone where a Mw 4.8 and a Mw 4.5 occurred on 2012 January 6 (green events in Fig. 12b). However, aside from the time period in early January the aftershocks are scattered throughout the aftershock zone.
The relocated aftershock activity is largely at depths greater than ∼5 km with an average depth of ∼10 km (Fig. 12, bottom). The colours correspond to the time periods in Figs 12(a) and (b). They show the events in the NE part of the aftershock zone, which occurred in early January, to be among the deepest events. The overall trend of the aftershocks does not show a consistent NW or SE dip direction. With the aftershocks being located offshore the hypocentres have greater uncertainties than onshore events and identifying a consistent dip is difficult.
Coulomb stress analysis
We have computed the Coulomb stress changes in the vicinity of the Pegasus Bay sequence due to slip from the Mw 7.1 Darfield event, the February Mw 6.2 Christchurch event and the June Mw 6.0 Christchurch event. We use slip distributions for these events from Beavan et al. (2012). The stresses from these source faults are resolved onto receiver faults consistent with the Pegasus Bay Mw 5.8 event (strike/dip/rake: 45/63/105) and the Mw 5.9 event (strike/dip/rake: 57/51/123). We have also resolved the stresses onto the conjugate fault planes for the Pegasus Bay events and find very little difference in our results. Our Coulomb stress calculations were performed using the Farfalle code (McCloskey et al.2003). For all of our calculations we assumed a coefficient of friction of 0.8 and a Skempton's coefficient of 0.5; however, the results are not strongly dependent on the assumed frictional parameters. There are two simple Coulomb models that are commonly used: the isotropic poroelastic model and the constant apparent friction model (e.g. Beeler et al.2000). The constant apparent friction model assumes that changes in pore fluid pressure are proportional to the normal stress change across the potential failure plane, allowing the definition of an ‘apparent coefficient of friction’. Although convenient, Beeler et al. (2000) have shown that this approximation can provide misleading results. For that reason, we use the isotropic poroelastic model for our calculations. We have also used the constant apparent friction model, but there is no qualitative difference in our results.
Coulomb stress change with the Mw 5.8 and Mw 5.9 Pegasus Bay aftershocks as the receiver faults have been calculated with the following scenarios:
Using only the September Mw 7.1 Darfield earthquake as the source fault;
Darfield and February Mw 6.2 Christchurch earthquakes as source faults;
Darfield, February Christchurch and June Mw 6.0 Christchurch earthquakes as source faults.
Since the inferred fault planes for the Pegasus Bay events cover a range of depths, we examine the Coulomb stresses at several depth slices. For both receiver faults, we examine the depth corresponding to the depth inferred from double-difference relocations (10.7 km for the Mw 5.8 event and 7.9 km for the Mw 5.9 event), and then examine depths 5 km above and below these depths. The initial Mw 7.1 Darfield earthquake produced large increases in Coulomb stress at all depths at the eastern and western ends of the Greendale Fault, but only a moderate increase in Coulomb stress in the western suburbs of Christchurch and a negligible increase in the vicinity of the Pegasus Bay sequence (less than 0.02 MPa). When the February and June Christchurch earthquakes are included as sources (Fig. 13), the Coulomb stress pattern in the Christchurch region changes considerably. Figs 13(a)–(c) show the Coulomb stresses resolved onto the inferred fault plane for the Mw 5.8 event, while Figs 13(d)–(f) show the stresses resolved onto the inferred fault plane for the Mw 5.9 event. For both receiver faults, we see that the inferred epicentres only fall within positive Coulomb stress regions at very shallow depths (Figs 13a and d). At greater depths, both events fall within regions of decreased Coulomb stress.
Since the hypocentre locations inferred from double-difference tomography would be expected to represent the initiation point for the rupture, the Coulomb stress results would at first glance seem to indicate a failure of the Coulomb model to account for the occurrence of the Pegasus Bay events. There are a number of parameters with large uncertainties involved in Coulomb stress calculations, however, and variations in any of them could dramatically change the results. In addition to the uncertainties in the receiver fault parameters and frictional properties, there are large uncertainties in the source fault parameters. Other geodetic inversions provide significantly different fault geometries and slip distributions for the February Christchurch event (e.g. Atzori et al.2012; Elliot et al.2012), and it is possible that use of an alternate solution could alter the Coulomb stress results. It is equally possible, however, that neglecting material property variations could significantly alter the results. This is especially true for stresses computed near the Banks Peninsula volcanics, which have higher elastic shear strength than the surrounding material. An exhaustive search of parameter space is beyond the scope of this study. Using our current parameters for the source and receiver faults, a simple set of frictional properties, and an elastic half-space model, slip due to the three largest events does not appear to enhance the likelihood of slip for either the Mw 5.8 or Mw 5.9 Pegasus Bay events, except at shallow depths. Additional work will be needed to address the uncertainties in the source and receiver fault parameters, as well as the influence of material heterogeneities.
DISCUSSION
The Pegasus Bay aftershock sequence is the most recent aftershock sequence resulting from the 2010 September 3 Mw 7.1 Darfield earthquake. The largest event in the Pegasus Bay sequence was a Mw 5.9 reverse-faulting earthquake and the kinematic source model is consistent with the SE-dipping plane as the rupture plane with most of the slip occurring above the relocated aftershocks. The overall trend of the relocated aftershocks does not show a consistent NW or SE dip direction. However, the relocated hypocentres are offshore and have larger uncertainties than onshore events. One of the more interesting features of the aftershock sequence is the high degree of consistency among the focal mechanisms with the vast majority being reverse-faulting mechanisms with NW–SE oriented P-axes. The high percentage of reverse-faulting mechanisms differs significantly from the rest of the Canterbury aftershock sequence (Fig. 14). Excluding the Pegasus Bay aftershock sequence, 74 per cent of the events in the Canterbury aftershock sequence are strike-slip mechanisms and 21 per cent reverse or oblique-reverse faulting mechanisms. In contrast to the rest of the Canterbury aftershock sequence, only 7 out of 53 events (13 per cent) in the Pegasus Bay sequence are strike-slip mechanisms whereas 85 per cent are reverse or oblique-reverse faulting mechanisms.
Coulomb stress calculated using the Mw 7.1 Darfield, Mw 6.2 Christchurch and Mw 6.0 Christchurch earthquakes as source faults show the hypocentre of the December Mw 5.9 earthquake was located within a negative stress lobe, unless the stresses are computed at a shallow depth. Relocated epicentres shows a well-defined NNE–SSW or NE–SE trend, which differs significantly from the mapped faults in the offshore region which generally strike ENE–WSW; also there are no mapped faults within the aftershock zone (Fig. 15). Geodetic results for the Mw 6.2 Christchurch earthquake (Beavan et al.2012; Elliot et al.2012) show a SE-dipping fault plane with a strike of 58° and possibly extending a few kilometres offshore (e.g. Kaiser et al.2012). This fault plane extends offshore into the region of the Pegasus Bay sequence, but the 58° strike is quite different from the trend of the Pegasus Bay aftershocks. It is, however, consistent with the strike of the SW-dipping fault plane of the largest Pegasus Bay aftershock (57°).
The apparent stresses (∼4–16 MPa) for the three largest earthquakes in the Canterbury earthquake sequence (Mw 7.1, 6.2 and 6.0) are considerably higher than the global average for crustal earthquakes. In contrast, the apparent stresses for the Mw 5.8 and Mw 5.9 Pegasus Bay events (∼0.2–1 MPa) are lower, but compatible with the global average for crustal events. The apparent stress measurements suggest that offshore faults in Pegasus Bay may be weaker than those in the onshore Canterbury region, including the Port Hills Fault associated with the February Christchurch earthquake. However, we note that other ongoing source stress drop measurements based on local recordings suggest, in contrast, that the stress associated with the two largest Pegasus Bay events may still be higher than crustal averages.
Browne et al. (2012) and Jongens et al. (2012) concluded that Mesozoic to Cenozoic basement faults in the Canterbury plains were reactivated during the Canterbury earthquake sequence. Campbell et al. (2012) noted there is sequential propagation of reverse-faulting systems from the west and progressively younger hybrid reverse and strike-slip assemblages from the north into the currently seismically active area. These processes are consistent with the pattern of strike-slip and reverse faulting seen during the Canterbury earthquake sequence. During the Canterbury earthquake sequence a number of faults have been active showing a high degree of consistency. Several parallel to subparallel reverse-faulting and strike-slip faulting segments have been identified relating to the Canterbury sequence (e.g. Bannister et al.2011; Sibson et al.2011; Atzori et al.2012; Beavan et al.2012; Elliot et al.2012). North of Christchurch the geology includes NE-trending reverse faults both in the onshore Canterbury region and offshore across the continental shelf and slope (Fig. 15; Pettinga et al.2001; Barnes et al.2011; Browne et al.2012; Campbell et al.2012). The reverse faults are evolving in response to oblique plate convergence and the transition from subduction in the north, to oblique continent–continent collision west of the Chatham Rise (Reyners & Cowan 1993; Reyners et al.2011; Campbell et al.2012).
In a 2011 seismic reflection survey in Pegasus Bay, Barnes et al. (2011) identified faults with evidence of late Pleistocene and late Pliocene activity north of the Pegasus Bay aftershock sequence. However, only sparse late Quaternary faulting was identified in southern Pegasus Bay in the vicinity of the aftershock zone. The Pegasus Bay aftershock sequence appears to be consistent with the reverse-faulting tectonic environment further north suggesting that active reverse-faulting offshore may extend further south than previously mapped. Reverse faulting is also present at similar latitude at the western end of the aftershock zone near the foothills of the Southern Alps.
Campbell et al. (2012) suggest that the Pegasus Bay aftershocks may be related to the Port Hills Fault extending offshore. However, the overall trend of the relocated aftershocks is rotated ∼25° counter-clockwise to the strike of the Port Hills Fault. Beavan et al. (2012), Elliot et al. (2012) and Atzori et al. (2012) model a NNE–SSW fault segment for the 2011 February Christchurch earthquake, in addition to one (Atzori et al.2012; Elliot et al.2012) or two (Beavan et al.2012) ENE–WSW fault segments. The trend of the Pegasus Bay aftershocks appears to be more similar to that of the NNE–SSW fault segment of the 2011 February Christchurch earthquake. The apparent stresses of the largest Pegasus Bay aftershocks are also much lower than the February Christchurch earthquake. Another possible interpretation could be the reactivation of old volcanic structures associated with Banks Peninsula, which is the site of an extinct volcano. The intrusion of the volcano may have produced highly segmented faults in the Christchurch region (e.g. Browne et al.2012; Davy et al.2012).
CONCLUSIONS
The Pegasus Bay aftershock sequence is the most recent series of aftershocks associated with the 2010–2012 Canterbury aftershock sequence. Relocations using double-difference tomography show a well-defined NNE–SSW to NE–SW trend in the aftershocks with depths ranging from 6 to 15 km. The two largest aftershocks occurred in a region with a predicted Coulomb stress decrease resulting from previous earthquakes in the Canterbury earthquake sequence. The kinematic source model for the largest earthquake (Mw 5.9) is consistent with reverse faulting on a southeast-dipping fault plane. 53 RMT solutions have been calculated for the aftershock sequence and they show remarkable consistency with 45 being reverse or oblique-reverse faulting mechanisms. This differs significantly from the rest of the Canterbury aftershock sequence where ∼74 per cent of the focal mechanisms derived from RMT solutions are strike-slip mechanisms. The Pegasus Bay aftershock sequence occurred in a region with no previously mapped active faults. The trend of the aftershocks is similar to mapped reverse faults north of Christchurch, and the high number of reverse-faulting mechanisms suggests that similar reverse-faulting structures are present in the offshore region east of Christchurch.
Martin Reyners, Sandra Bourguignon, Rick Sibson, Phil Barnes, Jan Burjanek and an anonymous reviewer provided valuable comments and discussion, which greatly improved this manuscript. Some of the figures were created using Generic Mapping Tools (Wessel & Smith 1998). Moment tensors were computed using the mtpackagev1.1 package developed by Doug Dreger of the Berkeley Seismological Laboratory, and Green's functions were computed using the FKRPROG software developed by Chandan Saikia of URS.