Rupture direction of paleoearthquakes on the Alpine Fault, New Zealand, as recorded by curved slickenlines

The Alpine Fault is an 850-km-long dextral continental transform fault that accommodates up to 80% of current motion between the Pacific and Australian plates in South Island, New Zealand (e.g., DeMets et al., 1994; Wallace et al., 2007; Fig. 1). The fault has one of the most spatiotemporally complete paleoearthquake records on Earth, with the timing and extents of the last 20 surface-rupturing earthquakes constrained over a 300 km length of the fault over the past 4 k.y. (Howarth et al., 2021). The spatial resolution of this record reveals that 55% of the events have stopped at or near the boundary between the Central and South Westland sections of the fault to be “single-section” ∼M_w 7.5 earthquakes, whereas 45% (including the three most recent paleoearthquakes) have passed through to become “multi-section” ∼M_w 8.0 earthquakes. While Howarth et al. (2021) demonstrate that fault geometry could lead to this conditional “earthquake gate,” little is known about its underlying causes or behavior, including the gate's current state and whether throughgoing ruptures can initiate on either side.

The fast slip rate (∼30 mm/yr), regular recurrence interval (249 ± 58 yr, with a coefficient of variation of 0.22), and time of last earthquake (1717 CE) yield a 75% probability of Alpine Fault surface rupture in the next 50 years, making this one of the world's most anticipated earthquakes and distinct national-level hazards (Wells et al., 1999; Berryman et al., 2012; Barth et al., 2014; Howarth et al., 2021). Much science and policy interest has focused on anticipating the properties of the next Alpine Fault earthquake, including its magnitude, shaking intensity, rupture direction, rupture extent, slip distribution, and landscape impacts, such as landsliding and aggradation (e.g., Robinson and Davies, 2013; Orchiston et al., 2016). The more that can be uncovered about the properties of the last several paleoearthquakes, such as their surface slip distributions and paleoearthquake rupture propagation directions, the greater the potential to anticipate realistic earthquake scenarios for the Alpine Fault's twenty-first century earthquake(s).

Rapid fault slip during earthquakes causes the two bodies of rock on either side of a natural (i.e., imperfectly planar) fault to abrade against each other, forming striations or grooves known as slickenlines. Slickenlines have long been used by geologists to infer the past orientation (and in some cases also sense) of slip on faults (e.g., Petit, 1987; Doblas, 1998). Field observations of curved slickenlines on exposed fault free-faces of the Kekerengu Fault generated during the 2016 Kaikōura earthquake in New Zealand (Kearse et al., 2019)—and analysis of eight other published accounts of curved striations for historic earthquakes with known hypocentral locations, focal mechanisms, and propagation directions (Kearse and Kaneko, 2020)—has led to a theoretical framework of curved slickenlines that relates slickenline curvature to direction of rupture propagation. Dynamic stresses in the process zones of ruptures cause curved slip-paths, and the curvature of slip (e.g., convex-up versus convex-down, or clockwise versus anticlockwise) can be used to interpret the local direction of propagation of an earthquake rupture. The curved shape also depends on the fault kinematics, such as reverse, normal, sinistral, dextral, or oblique-slip, and the view direction (Kearse and Kaneko, 2020; Aoki et al., 2023; Fig. 1B). This framework for interpreting curved slip has led to the powerful paradigm that curved slickenlines, as observed either on natural free-faces or on excavated fault planes, can be used to infer prehistoric rupture directions during large surface-rupturing earthquakes.

In this study, we visited three sites on the Alpine Fault that span the Central–South Westland earthquake gate. Southwest to northeast, these are: Hokuri Creek, Martyr River, and Chasm Creek (Fig. 1C). Our goals are to: (1) develop best practice methods to exhume and document slickenlines produced by prehistoric surface ruptures; (2) interpret paleoearthquake rupture directions of Alpine Fault paleoearthquakes, including the most recent event in 1717 CE; and (3) reevaluate the earthquake gate model in light of these additional paleoseismic data. Our study is the first to use observations of curved slickenlines on a major plate boundary fault for which there has been no historic or instrumentally observed surface rupture and the first to document the preservation of curved slickenlines resulting from multiple earthquakes on the same fault outcrop.

All three sites are remote and required hand tools to excavate the fault (e.g., shovel, pickaxe, trowel, knife). At each site, we first documented all naturally exposed parts of the fault planes and their slickenline tracks (structural attitudes, track lengths, photographs, field sketches). We then identified places to clean the exposed fault surfaces or exhume them further to make additional slickenline observations. When exhuming a fault surface, we took care not to disturb or modify existing slickenlines or create false ones. Tools and techniques became increasingly precise as we approached the principal slip surface. Carefully poured plastic bags of water were effective at cleaning loose detritus from gravel-hosted gouge surfaces. Key outcrops were documented with field sketches that placed all measurements in context and with overlapping photos in flat light conditions to use for structure-from-motion (SfM) photogrammetry. None of the exposures included materials that could be used to date the slickenlines. Our structural data include the host rock type, attitude of fault surfaces and pitch of striations, and—for curved striation tracks—the range of pitch and sense of convexity (up or down; see Supplemental Material¹. Tracks were classified by type (e.g., grooves scoured by cobbles or gouge-hosted striations), and we noted their length, incision depth, width, and density. Critical for rupture direction interpretation, we specified the direction of view (reference frame) of all striation observations and noted any apparent cross-cutting relationships between individual striae.

We made 233 slickenline measurements on 146 individual slickenline tracks, of which 30 (21%) were curved. The shortest curved track was 10 cm in length, and the longest two were greater than 3.5 m. The average pitch deflection of the curved subset of slickenlines was ∼15°, and the largest deflection for a single track was 42°. We identified several types of slickenlines on or within 10 cm of the inferred principal slip surface of the Alpine Fault. These include striations or fine scratches (both wallrock-on-gouge and entirely gouge-hosted), centimeter-scale-amplitude ridges and troughs attributed to the passage of protruding cobbles and boulders (on either wallrock or gouge), and polished striations on clasts embedded in the fault plane (resembling glacial polish). Over half of the slickenline tracks that we observed were artificially exposed by hand tools. In many cases, slickenlines that initially appeared to be short and linear were found to be curved after further excavation, a result that highlights the ease with which these curved features can be overlooked.

At Hokuri Creek, the Alpine Fault is expressed as a 10-m-wide slab of saponitic fault gouge with entrained clasts that is bounded by slickenlined fault planes of much harder fluvioglacial gravel on both sides (Fig. 2; Barth et al., 2013). Fault slip and gouge erosion have exposed these planes. Field and microstructural observations and friction experiments suggest that most surface ruptures have occurred on these bounding fault planes rather than within the rate-strengthening slab of gouge between them (Barth et al., 2013; Boulton et al., 2018). On the subvertical fault outcrop that we call HC#1, which faces to the northwest, we exhumed four curved slickenlines, ∼20 cm-long, that deflect in pitch by 20° to 28°; these were all convex-up, indicating rupture from the southwest. On the subvertical HC#2 fault outcrop, which faces to the southeast, we observed 17 curved slickenline tracks in two different exposures (HC#2A and HC#2B). The sense of convexity (up or down) and the morphologies of the slickenlines were different between these two spatially separated exposures. At HC#2A, the 13-m-long fault plane exposure contains curved grooves (wallrock-on-gouge) that are up to 3.5-m-long and thin gouge-hosted curved striations that are up to 20 cm long; the total angular deflection on these convex-down tracks ranged from 4° to 12° (Fig. 2). The two longest slickenlines occur as deep, subparallel tracks and are inferred to have been inscribed by boulders embedded in the adjacent gouge. The convex-down sense of curvature of all 10 tracks on this outcrop indicates fault rupture from the southwest. At HC#2B, the 10-m-long exposure archives gouge-hosted striations with 15° to 42° of total angular deflection and lengths to 20 cm. The convex-up curvature of seven slickenline tracks on this outcrop indicates fault rupture from the northeast during a different earthquake than that recorded at HC#2A.

In summary, at Hokuri Creek, after taking account of the opposite directions of view on the two exposed fault planes, we found curved slickenlines indicative of both northward and southward rupture propagation, including examples preserved on different parts of the same fault plane. We infer that slight undulations in the fault surface between outcrops HC#2A and HC#2B may have preferentially preserved patches of slickenlines that were inscribed during different previous ruptures (i.e., in abrasion shadows). We conclude from this that all slickenlines on a fault cannot be assumed to have formed during the same or most recent rupture. At this site, we do not know which of the two planes slipped in the most recent rupture (1717 CE). The paleo-propagation directions indicated above are thus undated. The three longest curved slickenline tracks on the HC#2A outcrop exceed 1 m in length. Their shape indicates that early in the slip-path, motion pitched at ∼21°, after which it progressively shallowed to sub-horizontal (∼4° pitch on average). This later, near-pure dextral-slip increment was much longer and matched the long-term fault kinematics. An interpreted Alpine Fault single event displacement of ∼7.5 m dextral and ∼1 m northwest-side-up is recorded 1 km to the northeast of the HC outcrops (Berryman et al., 2012); this displacement, resolved onto our average fault plane at Hokuri Creek of 052°/82°SE, yields a calculated pitch of 8° southwest, the same as the average of all 171 Hokuri slickenline measurements.

At Martyr River, a single moderately SE-dipping dextral-reverse principal slip surface is visible at river level juxtaposing quartzofeldspathic fault rocks in the footwall against overlying chloritic protocataclasite in the hanging wall (Barth et al., 2013). Following this fault plane along strike and up slope, we observe a moderately SE-dipping frontal fault plane and a steeply SE-dipping fault plane that merge at the base of a prominent outcrop. Here we observed curved slickenlines on a freshly exposed, 12-m-wide by 4-m-high outcrop of the steeply SE-dipping fault plane (overhanging) cutting well-indurated boulder-gravel till (Fig. 3). The exposed fault plane extends to the ground surface, suggesting its rupture during the most recent event (1717 CE). The average orientation of all measurements of this fault plane is 042/79SE, and the average of 51 slickenline measurements is 11/051, suggesting dextral > reverse slip overall. We observed a dominance of slickenline tracks (six) with convex-up curvature (rupture from the southwest) primarily as gouge-hosted striations on the principal slip surface or striations in polish on boulders embedded within it. The convex-down curved slickenline tracks (three) are gouge-hosted features, two of which are 10 cm from the principal slip surface. We interpret the convex-up slickenlines on the main fault surface as likely recording the most recent rupture and the convex-down slickenline tracks on subsidiary slip planes as survivors from a previous rupture.

At our most northerly site of Chasm Creek, we documented a rare bedrock-on-bedrock exposure of the Alpine Fault principal slip zone. We were only able to uncover ∼200 cm² of the principal slip surface, and no curved slickenlines were observed.

We have demonstrated that curved slickenlines can be used to determine paleoearthquake rupture propagation directions from careful excavation, a new paleoseismic tool applicable to surface-rupturing faults globally. Curved slickenlines from multiple earthquakes can be preserved on a single principal slip surface. What such observations cannot easily provide, however, are dates for specific slickenline-inscribing events, for example by correlation of such marks to known historic or ancient earthquakes. While natural fault exposures may reveal curved tracks, we found that careful excavation of faults with hand tools approximately doubled the number of our slickenline observations and that some apparently short, straight slickenline tracks were revealed to be curved after exposing their buried extensions. The longest (>1 m) curved slickenlines we observed all indicated a short length of plunging early slip evolving to a length of subhorizontal slip that was more closely aligned with the mean striation direction and the inferred average finite-slip vector. The shape and scale of these curved slickenlines are consistent with dynamic simulations of 10 m of strike slip, which show the same transition from initially steep rake angles at the onset of slip toward sub-horizontal dextral slip in the immediate wake of the propagating rupture front (Kearse et al., 2019). We observed three fault surfaces >10 m in length and >3 m in height; despite these being sufficient in area to record an expected 8 m coseismic displacement, individual slickenline tracks never exceeded 4 m in length, and thus at best record less than half of expected slip.

We report the first curved slickenlines identified on the Alpine Fault; these indicate that past surface ruptures have initiated from both northeast and southwest of Martyr River and Hokuri Creek, and thus the fault does not rupture exclusively in a single direction. If the paleoseismically identified earthquake gate occurs at the major change in fault dip at the Martyr River, then our results indicate arrival of ruptures to the gate from both the southwest and northeast. Since the last three paleoearthquakes are known to be multi-section ruptures (Howarth et al., 2021), if we attribute the observed slickenlines to be no older than these, then multi-section ruptures could nucleate both northeast and southwest of the earthquake gate (i.e., the gate swings both ways). Based on the context and dominant curvature of the slickenlines on the surface-rupturing fault plane at Martyr River, we suggest that the most recent multi-section (∼M_w 8) 1717 CE rupture of the Alpine Fault initiated southwest of the earthquake gate at Martyr River and propagated northward, a scenario that leads to greater shaking to distal populated areas to the north, such as Nelson and Canterbury (Bradley et al., 2017).

We acknowledge support from the Marsden Fund Council from New Zealand Government funding, managed by Royal Society Te Apārangi. Fieldwork was authorized by the New Zealand Department of Conservation. We thank editor Rob Strachan, reviewer James Dolan, and two anonymous reviewers for constructive feedback.