Revised slip rates for the Alpine fault at Inchbonnie: Implications for plate boundary kinematics of South Island, New Zealand

Partitioning and transfer of strain at obliquely convergent collisional plate boundaries over millennial time scales are poorly documented worldwide. The Alpine fault, and its transition to the Marlborough Fault System in the northern South Island of New Zealand, offer such an opportunity using late Holocene geologic slip rates and vectors to assess an important on-land transitional plate boundary.

The Alpine fault is a major component of the collisional zone between the Australian and Pacific plates across the South Island (e.g., Cox and Sutherland, 2007). Through the southern half of the island, the northeast-striking and southeast-dipping Alpine fault exhibits primarily dextral-reverse slip and accommodates 50%–80% of the 37 ± 2 mm/yr of convergent motion across the plate boundary (DeMets et al., 1994; Sutherland et al., 2006; Berryman et al., 1992) (Fig. 1). This section of the fault, between Milford Sound and Toaroha River, is generally referred to as the Central segment, as it represents the ≥325-km-long, straight on-land part of the fault with a consistently high dextral slip rate averaging ∼27 ± 5 mm/yr (Norris and Cooper, 2001) and which is also responsible for the uplift of the Southern Alps. At its southern end, the Central segment of the Alpine fault evolves offshore into partitioned strike-slip faulting and oblique subduction in the Fiordland region, related to the Puysegur margin (Barnes, 2009; Sutherland and Norris, 1995; Barnes et al., 2005), whereas at its northern end the fault transitions into the Marlborough Fault System, a zone of distributed strike-slip deformation (Yeats and Berryman, 1987; Van Dissen and Yeats, 1991; Langridge and Berryman, 2005). The Alpine fault is a highly evolved fault with total dextral displacement of bedrock geology of ∼480 km (Wellman, 1955; Cox and Sutherland, 2007). Rupture of the Central segment represents a significant seismic hazard, capable of generating M_w 7.8–8.0 surface-rupturing earthquakes every few hundred years (Yetton, 1998, 2000; Rhoades and Van Dissen, 2003; Sutherland et al., 2007; Wells et al., 1999).

Late Quaternary slip rates have been estimated for the Alpine fault from offset geomorphic markers that range over more than one order of magnitude in age (generally from Last Glacial to late Holocene), and have been used to demonstrate the variability of strike-slip and dip-slip partitioning along the length of the fault (Norris and Cooper, 2001; Berryman et al., 1992). Dextral slip-rate measurements along the Central segment of the fault are high, e.g., 27 ± 5 mm/yr (Waikukupa River); 29 ± 6 mm/yr (Kakapotahi River) (Fig. 1). Geologic dip-slip rates vary along strike, ranging from >12 mm/yr (Gaunt Creek) to 0 mm/yr (Hokuri Creek). For such a major plate boundary structure, these rates show considerable variability and uncertainty, and importantly, owing to the rugged and vegetated nature of the West Coast terrain, they are derived from sites spaced tens of kilometers apart along the fault.

In this paper we used a geomorphic and structural approach to calculate revised slip rates for the Alpine fault at Inchbonnie in north Westland (Fig. 2) based on the offset of late Holocene features. This area is important because the dextral slip rate along this portion of the fault decreases by 50%–70% compared with those sites to the southwest (Norris and Cooper, 2001), and the site is present at a key location for understanding the kinematic transition from the Alpine fault to the Marlborough Fault System. Because the previously published slip rates at Inchbonnie of Berryman et al. (1992) (i.e., 10 ± 2 mm/yr dextral; 6 ± 2 mm/yr reverse) come from displacement of a very young surface (1 k.y.) dated using only a weathering rind technique, an essential part of this study has been to recognize and date alluvial surfaces using radiometric and relative dating techniques to estimate slip rates averaged over a longer time, i.e., over more earthquake cycles.

Geologic and geodetic data and their derivative models indicate that a large proportion of the tectonic plate motion is partitioned to the northeast from the Central segment of the Alpine fault onto the Hope and Kelly faults near Inchbonnie (Robinson, 2004; Berryman et al., 1992; Wallace et al., 2007; Stirling et al., 2002). In this paper we use our geologic slip-rate data to construct a kinematic model for the plate boundary transition from the Alpine fault to the Marlborough Fault System in central South Island.

The main tools used in this study were aerial photograph interpretation, geomorphic mapping, 2-D scarp profiling, topographic surveying using differential GPS, trench logging, soil chronology, and AMS (accelerator mass spectrometry) radiocarbon dating. Initial studies in the Inchbonnie area suggested that paleoseismic trenching would not yield a straightforward paleo-earthquake record (see Toy, 2007; Langridge et al., 2008). However, two significant outcomes of trenching included a need for (1) detailed geomorphic mapping of the fault scarps and Holocene alluvial surfaces in the area, and (2) precise age control on those surfaces, in order to reassess local slip rates for the Alpine fault.

In the following sections the geomorphic development of the Taramakau valley is described, along with a chronosequence of late Holocene alluvial surfaces (Figs. 3, 4). This is followed by a description of the location and height of the scarp of the Alpine fault between the Taramakau River and Lake Poerua (Fig. 5) along with presentation of scarp profiles used to assess the vertical displacement history (Fig. 6). This is followed by a description of the Harris trench site, where a microtopographic map is used to characterize the lateral slip at the site. Two of five trench exposures are discussed to provide background to the stratigraphy and age of surfaces and their deposits, and the tectonic structure there. In addition, to aid interpretation of surface and landscape ages, soil profiles from these trenches are also described. Last, in order to derive slip rates for the site, the ages of the key displaced alluvial surfaces are discussed.

The glacial history of the Taramakau valley is well recorded for the Last Glacial Maximum (LGM) by extensive moraines, one of which has impounded Lake Brunner (Figs. 1, 3) (Suggate, 1965; Suggate and Waight, 1999; Nathan et al., 2002). It was recognized that during the LGM the former Taramakau valley glacier branched into three lobes at the range front of the Southern Alps near Inchbonnie. These lobes excavated distinct, moraine-bounded troughs, referred to here as the Poerua, Orangipuku, and Taramakau lobes (Fig. 2). On the northwest side of the Alpine fault the glacial troughs are separated by a series of ice-sculpted, dome-shaped Cretaceous to Paleozoic granitoid hills of the Hohonu Group, e.g., Mount Te Kinga (Figs. 2–4,³⁴) (Suggate and Waight, 1999; Nathan et al., 2002). For more detail of the bedrock geology, see Figure 3.

Figure 4 shows a series of fanning alluvial surfaces related to the Taramakau River near Inchbonnie. During the Holocene, and following retreat of the valley glacier, the glacial troughs have been partly filled by alluvium from the Taramakau River system. In this paper we recognize a series of inset, overlapping fanning alluvial surfaces in the Inchbonnie area. Based on their geomorphology, surface texture (i.e., “smoothness”), and channel flow directions, it has been possible to distinguish several distinct “Inchbonnie” alluvial surfaces locally described as I-Ø to I-4 (Fig. 4). Radiocarbon samples from exposures, and from in situ and buried tree stumps, are used to infer ages for these surfaces (Table 1). In general, I-1 to I-4 are identified on the upthrown side of the Alpine fault by risers that separate them. On the downthrown side of the fault the differentiation of surfaces based on topography is difficult, as older surfaces are partially reoccupied or buried by large floods from the active younger floodplain.

On the West Coast of the South Island, graywacke alluvium is characteristic of larger rivers with headwaters in the lower grade graywacke rocks of the Torlesse Supergroup that crop out near the Main Divide of South Island (Nathan et al., 2002). The modern river course (referred to as I-Ø) exclusively follows the Taramakau lobe flowing southwest to the Tasman Sea (Figs. 2, 3). Surface I-1 corresponds to the most recently abandoned and faulted surface within the Taramakau valley. I-1 also grades geomorphically to the Taramakau lobe of the system but may have at times spilled into the Orangipuku lobe (Fig. 4). Buried forests and remnant “snag” trees are useful in estimating the times of construction and abandonment of surfaces I-Ø to I-2. At site TR1 a number of in situ tree stumps are exposed in the bed of a secondary channel of the Taramakau River, southwest of Inchbonnie (Figs. 3, 4) and in the adjacent edge of an eroding, sandy fill terrace that corresponds to I-1. The outer rings from stump TR1 were dated and yielded an AMS radiocarbon age of 416 ± 20 yr B.P. (330–498 calibrated yr B.P. at 2σ) (Table 1). This date suggests that the Taramakau River recently shifted to its current course and also probably provides a maximum age for I-1. That is, a forest was formerly established in a place which has been both inundated by sediment and reoccupied by the river during the past 500 yr.

The I-2 surface is a lobate-shaped alluvial surface that extends across the Alpine fault from the Taramakau River northward in the direction of the Orangipuku lobe and Lake Brunner. At least three subdivisions of I-2 have been made using old aerial photographs combined with field reconnaissance. Despite its lack of topographic separation and modification by farming practice, surfaces I-2a to I-2c are mapped from west to east across this floodplain. The spring-fed and underfit Orangipuku River is characterized by a braided network of channels that were beheaded by the recent switch of the Taramakau River (I-Ø and I-1). These channels are pinned on the west side of the Orangipuku lobe and form surface I-2a.

Surface I-2b follows the central axis of the Orangipuku lobe and is characterized by a variegated network of northward-flowing underfit braided channels that merge to form Bruce Stream (Fig. 4). A tree stump is exposed in the channel of Bruce Stream at its lower end where the I-2c surface has been superseded by I-2b drainage. A wood sample from the outside of stump B3 yielded an age of 332 ± 25 yr B.P. (302–447 cal. yr B.P. at 2σ) (Table 1). On aerial photographs, surface I-2c is characterized by a smooth geomorphic texture lacking in distinct channels. This texture is typical of surfaces covered by a veneer of overbank flood deposits. Exposures of coarse, cobbly alluvial graywacke in the Bruce Stream catchment are typically overlain by ∼0.5 m of sandy to silty deposits that resemble overbank deposits. A previous maximum age estimate for this surface of 1100–1300 yr comes from measurements of weathering rinds on graywacke clasts in a gravel pit within the scarp of the Alpine fault (Fig. 5) (Berryman et al., 1992).

The I-3 alluvial surface traverses the area that includes the Harris trench site, and therefore an essential part of this study was to understand the extent and age of this surface. On the upthrown side of the fault near the trench site, I-3 is a typically smooth geomorphic surface, as evidenced by fine-grained overbank deposits exposed within trenches there. On the downthrown side of the fault the I-3 alluvial surface is characterized by a network of north- to northeast-directed channels that trend northeastward into the Poerua lobe of the valley (Fig. 4). These underfit channels grade to Lake Poerua and have dissected the smooth upper surface of I-3, and therefore probably represent a reoccupation of the I-3 surface by overbank flow, shown as I-2r in Figure 4.

The I-4 surface refers to the texturally smoothest and highest alluvial surface in the study area. I-4 is preserved only on the upthrown side of the Alpine fault to the northeast of the Harris site. I-4 also characterizes the surface that bounds the southeastern shore of Lake Poerua (Figs. 4, 5). Engineering soil pits excavated into I-4 near the lake show a variable thickness (0.2–1.2 m) of moderately weathered, fine-grained deposits that cover a paleotopography on alluvial gravel (Golder Associates, 2007). Northeast of Lake Poerua, I-4 is overlain by schist-bearing alluvial fans derived from the range front of the Alpine fault (e.g., Dry Creek; DCF in Figs. 3, 4). The southwest edge of I-4 is marked by a prominent riser cut at the edge of the large push-up adjacent to the Harris trench site (Figs. 4, 5). I-4 is buried on the downthrown side of the fault by I-3 and subsequent deposition. Although there is no direct age control for I-4, it is inferred from the predominance of graywacke cobbles that it corresponds to an older course of the Taramakau River through the Poerua lobe, which has been uplifted along the trace of the Alpine fault.

In summary, during the late Holocene the Taramakau River has occupied three main courses that correspond to former glacial troughs. Surfaces I-Ø to I-4 correspond to a series of inset alluvial fan lobes formed as the river migrated through avulsion from the northeast (I-4) to its current position (I-Ø).

The stretch of the Alpine fault between the Taramakau River and Lake Poerua has received much attention because of its ease of access and exposure (Figs. 4, 5) (Berryman, 1975; Berryman et al., 1992). The fault has a clear, fresh trace along this stretch and also forms the southeastern shorefront of the lake (Langridge and McSaveney, 2008). The fault displaces alluvial surfaces I-1 to I-4 along this stretch, each of which is typically recognized by a stepwise increase in scarp height to the northeast on the upthrown side of the fault. Along this stretch the fault is also characterized by a series of left-stepping fault traces with approximately kilometer-length sections separated by 80–100-m-wide stepover zones between sections (Fig. 5). Northeast of Lake Poerua the trace of the Alpine fault is buried by young alluvial fans that emanate from the range front of the Southern Alps and is not exposed (Fig. 3).

Figure 6 shows 10 topographic profiles of the scarp of the Alpine fault that are used to demonstrate the progressive growth in the vertical component of deformation along strike (profile locations shown in Fig. 5). Many of these profiles probably document minimum scarp heights owing to reoccupation and burial of older surfaces on the downthrown side of the fault. A clear fault trace is first observed on the I-1 surface near the Taramakau River. Based on the projection of I-1 across the fault, profile ZØ yields a scarp height of ∼2.3 m. Profiles across the I-2a and I-2b (surface profiles ZX1 and Zb) yield scarp heights of ∼3.1 and ∼3.3 m, respectively. A small shoulder (riser) on the upthrown side of the fault scarp marks the boundary between I-2c and I-3 southwest of Inchbonnie. Across the I-3 alluvial surface, profiles Za, S1, and HR yield typically larger scarp heights of ∼6.4 ± 0.2, ∼6.6 ± 0.2, and ∼4.8 ± 0.5 m, respectively (Figs. 5, 6).

Three further profiles across the I-4 surface between the trench site and Lake Poerua (S2, S3, and S4) yield scarp heights of 8.1 ± 0.6 m, 6.5 ± 0.4, and 9.3 ± 0.3 m, respectively. Finally, profile L1 was measured from tree stump WP3 (Fig. 5; Table 1) on the floor of Lake Poerua to the lakeshore and across the raised edge of the lake. This profile confirmed that the steep lake edge was the scarp of the Alpine fault, which has an overall scarp height of ∼6.9 ± 0.2 m there (Langridge and McSaveney, 2008).

At the Harris farm, detailed topographic surveying and paleoseismic trenching were undertaken to estimate the lateral slip and to determine the paleo-earthquake record of the Alpine fault. The site lies at the leading edge of a prominent left stepover in the fault. Directly northeast of the site the stepover is characterized by a compressional bulge with a range-facing back scarp (see profile S2 in Fig. 6). At the trench site the stepover zone comprises two main traces (f1, f7) and a number of subparallel, transpressive fault traces (f2–f6) that displace the alluvial surface and locally a surface stream channel (Figs. 5, 7).

Figure 8 shows trench 4 and an abbreviated stratigraphic legend and brief description of units at the trench site. Overall, the stratigraphy exposed in the Harris trenches is demonstrably coarse scaled and can be divided into five major late Holocene packages of units: Taramakau Gravels, Taramakau Overbank deposits, local channel deposits, colluvium, and soil units. Fault zone and anthropogenic deposits have also been identified in the stratigraphy. The major packages show the evolution from an aggrading gravel surface (I-3) to overbank deposition, followed by abandonment, local deposition (related to reoccupation of I-3), and soil formation. The surficial channel form exposed in trench 4 is a shallow (∼10 cm) erosional feature cut on fine-grained cover materials (silt-sand and soils). Trench 4 was excavated outside of the fault zone in order to map the undeformed stratigraphy of the I-3 surface and this channel.

The Taramakau Gravels were deposited when the active floodplain of that river occupied the I-3 surface northeast of Inchbonnie (Fig. 4). The Taramakau Gravels comprise fluvial gravel and sand beds derived from graywacke source material. The coarse, permeable nature of the gravels makes the preservation of organic material within them unlikely. Consequently, no wood or charcoal was available from these units for dating the I-3 surface.

The Taramakau Overbank deposits comprise a series of sheetlike silt to sand units that blanket the top of the gravels and fill in topographic lows on that surface. These overbank deposits are extensive at the trench site and are considered to be flood deposits from the Taramakau River, laid down subsequent to aggradation of the I-3 surface. In trench 4 the overbank deposits cover and fill an old paleochannel formed on Taramakau Gravels (Fig. 8).

Widespread alluvial sedimentation ceased at the Harris site following deposition of the Taramakau Overbank deposits. Local channel deposits refer to the units found underlying the displaced channel mapped through the site (see unit 1b in Fig. 8). Local channel deposits typically consist of reworked sand and silt derived from the Taramakau Overbank sequence. These are associated with a reoccupation of the I-3 surface that formed the surficial channels ascribed to surface I-2r (Fig. 4). Following broad abandonment of the trench site area, soils began to develop within this sequence of units. Apart from large earthquake faulting events, which acted to expose the section to colluviation, the next major event to occur at the site was clearing of native forest for grazing during the mid–twentieth century.

A microtopographic map of the Harris site is shown in Figure 7. We used a Leica 500 RTK-GPS unit to survey the site. Almost 15,000 points were collected in transects with a 1-m point spacing and a measurement accuracy of ±5 cm. The scattered data points were gridded through a Triangular Irregular Network interpolation method (Akima, 1978) using a quintic polynomial surface that honors the data values and predicts some degree of over- and underestimation above and below local high and low values. This technique allowed us to obtain a 1-m-resolution digital elevation model that had minimal errors compared with the meter-scale displacements across the site.

Continuous, multi-meter-high scarps occur on either side of the stepover zone at the trench site (e.g., f1, f7) but diminish in vertical expression as they pass into the stepover zone (Figs. 6, 7). The main trace to the southwest (f7) steps to the left by ∼80 m, to the main trace toward the northeast (f1). The microtopographic map also shows that fault displacement is partitioned among north-northeast–striking structures within the stepover, which manifest themselves as oblique-slip faults and folds on the surface and in exposure; e.g., fault trace f6 was exposed at the upper end of trench 3 (Fig. 9).

At the trench site a shallow channel crosses the fault zone at a high angle and was used as a piercing line to estimate a cumulative dextral separation of 22.5 ± 2.0 m in the field across the major faults of the stepover zone. This channel is clearly younger than the I-3 surface and probably originated as part of an overflow episode related to I-2 surface construction. At the southwest edge of the stepover zone the channel crosses fault trace f7 and is dextrally displaced by ∼11.4 ± 1.3 m (estimated by projecting the channel thalweg into and across the fault). The channel is also dextrally displaced 10 ± 1 m across a second main trace of the fault at the northwest edge of the stepover zone (f1 in Fig. 7). Within the stepover zone the channel is further displaced by ∼1.1 m across a small, broad scarp (fault trace f6). This scarp was confirmed as a rupture trace in both trenches 5 and 3. Within the stepover zone the channel is guided by (runs parallel to) the oblique structures there (e.g., f3–f5) (Fig. 7). The amount of internal strike-slip deformation within the stepover zone is difficult to characterize. Based on their strike and the relatively shallow dips observed for fault traces f3 and f4, exposed in trenches 2 and 3, these faults are inferred to have a dominantly reverse sense of movement. However, as this internal component of dextral displacement is unknown, the total of 22.5 ± 2.0 m must be considered a minimum value. The individual (a–c in Fig. 7) and total displacements presented here are equivalent to those shown by Berryman (1975).

In this study the main purpose of the trenches was to document the presence and style of faulting related to individual fault traces within the stepover and the correspondence between surface fault traces and their subsurface expression. The faults and folds logged in the Harris trenches are consistent with strike-slip to reverse-slip faulting. Each linear trace (f1–f8) identified in Figure 7 that was intercepted in a trench exposure was confirmed as an active fault or fold (see Fig. 9).

Figure 9 presents an example of the expression of faulting observed at the Harris site. Fault f6 is exposed in trench 3 as a steep, southeast-dipping zone of oblique-slip faulting, identified by the juxtaposition of a section of Taramakau Overbank deposits against a shear zone characterized by graywacke cobbles rotated toward vertical. The juxtaposition of sands and silts against gravel is clear evidence for faulting, whereas, in addition, the Taramakau Overbank deposits are drag folded beyond vertical adjacent to the fault zone (Fig. 9). An additional trench log from trench 1, displaying steep, oblique-slip faulting, can be viewed in Toy (2007).

To estimate slip rates for the faulted channel at the Harris site it is necessary to determine a tractable age for either the channel or the surface that it cuts into. Dates for the I-3 surface from the Taramakau Gravels and Overbank deposits would likely provide maximum ages for the displacement of the channel. No deposits or materials were located to directly determine the age of the channel itself; however, local channel deposits probably provide a reasonable age for the formation of the channel. Nevertheless, owing to the paucity of datable organic material at the trench site, it was necessary to constrain the age of I-3 by other means. Geomorphic mapping concluded that the I-3 surface is continuous between Lake Poerua and the Harris site. For a fault as active as the Alpine fault, which causes punctuated landscape change every few hundred years (Berryman et al., 2009; Wells et al., 2001), dendrochronology is a useful technique for determining surface and paleo-earthquake ages.

At the southern shore of Lake Poerua a cut slab was extracted from the stump of an in situ, large (2.9 m circumference) podocarp tree (probably Matai; Prumnopitys taxifolia) (WP13 in Figs. 4, 5). We infer that this tree began growing on soft sediments following the full abandonment of I-3 (i.e., post–Taramakau Overbank deposits) so that WP13 probably provides a reasonable maximum age with respect to the offsets at the trench site. Podocarps are intolerant of saturated soil, such that when the local groundwater table rose, this tree effectively drowned and was preserved in place.

The age analysis of stump WP13 indicated when this tree both began growing and died, and hence provides a minimum age for the full abandonment of I-3 (Fig. 10). Three AMS radiocarbon samples were submitted from counted ring sections from the slab, i.e., rings 22, 46, and 222 (Table 1; each sample was 3 rings wide). The sample from near the outside of stump WP13 yielded an age of 1203 ± 35 yr B.P. The second sample (ring 46) yielded an age of 1217 ± 25 yr B.P. The third ring section (222 ± 1) was selected for dating on the basis of its postulated position on the radiocarbon calibration curve. That is, the rings were counted to intercept a portion of the calibration curve that was steep and unimodal (1329 ± 20 yr B.P.; see Table 1 for 2σ calibrated ages). Using a Bayesian statistical technique for ordering radiocarbon samples and events (i.e., the OxCal program; https://c14.arch.ox.ac.uk/oxcal/OxCal.html), it is possible to order the three dates and place time constraints on each dated ring section by inserting gaps in the OxCal program that correspond to counted annular ring gaps (Fig. 11). In this way it is possible to “shave” the probability density functions that represent each calibrated age (e.g., Biasi and Weldon, 1994) and ultimately to refine the distribution for the death age of tree WP13. Based on this analysis and an estimate of 22 rings to the exterior of the tree, tree WP13 stopped growing at ca. 947–1064 cal. yr B.P. (Fig. 11).

The seedling age, i.e., the time when tree WP13 began growing, is considered to represent a minimum age for the abandonment of surface I-3. The seedling age can be estimated by counting the ring sequence back to the center of the tree. Figure 10 provides an attempt to estimate when WP13 began to grow. A tree ring count of ≥480 rings was determined from the cut slab of stump WP13, which implies that this tree was alive at least 1500 yr ago or more. Because it was not possible to extract a full slab from this large stump in the lake, we mapped out the visible ring structure to estimate the time (rings) preserved in the core of the tree. Thus, it was possible to estimate that the podocarp WP13 was a further 80–100 yr older than could be counted from the incomplete slab section. Therefore, this tree was at least 570 ± 10 yr old when it died. In addition, it has been shown that Matai are relatively rapid colonizers of alluvial surfaces and have an average colonization time to corer height of 28 yr (5–40 yr; Wells et al., 1999). Based on the tree ring countback from the OxCal distribution for the death of WP13 (Figs. 10, 11), the tree probably began growing on abandoned surface I-3 at ca. 1530–1670 cal. yr B.P. (1590–1730 yr ago).

If the rate of soil development can be adequately calibrated, then soil morphology can be used to estimate surface exposure age. The rate of soil development is usually calibrated through studies of soil chronosequences, and a number of chronosequence studies exist for the Westland region, in which the study site is located. Tonkin and Basher (1990) review three of these chronosequences and provide a general pathway of soil development. Soils develop from entisols to inceptisols to spodosols (U.S. Natural Resources Conservation Service, 1999) but at differing rates, depending on mean annual rainfall.

In Trench 3 at the Harris site, a sequence of Taramakau Overbank deposits on gravel is preserved in fault contact with Taramakau Gravels (Fig. 9). A moderately developed inceptisol (brown soil; see Hewitt, 1998) has formed in these fine deposits, adjacent to the fault zone (f6). This soil profile is described in Appendix 1. A similar inceptisol was logged in trench 2. In trench 1, Taramakau Gravels with a thin silt cover show oxidation and reduction features that are indicative of a similar amount of relative soil development to those soils described above (see logs in Toy, 2007). The Wanganui River chronosequence in South Westland, reviewed by Tonkin and Basher (1990), is probably the most appropriate for comparison with the Inchbonnie area, although the former has a mean annual rainfall in the order of 6500 mm in contrast to Inchbonnie's 5000 mm. At Wanganui River the transition from entisol to inceptisol would have taken more than ∼400 yr but less than ∼600 yr, whereas the transition to a spodsol would have taken at least ∼1500 yr but no more than 3000 yr. Assuming that soil development was not as rapid at Inchbonnie as at Wanganui River, we conclude that the inceptisol in trench 3 is at least 400–600 yr old but younger than 1500–3000 yr old.

To develop viable slip rates for the Alpine fault at Inchbonnie, a comprehensive understanding of surface ages was required. While this has been challenging in this area, the most promising techniques for dating the alluvial surfaces have been dendrochronology combined with AMS radiocarbon dating, and relative soil chronosequence dating. These two techniques provide bounding age constraints of differing precision for the I-3 surface and displacements upon it. The preferred age of I-3 and the offset features, which comes from geomorphic and dendrochonologic data (1590–1730 yr ago), lies within the middle of the broad age range described from soil chronosequence dating. In addition, both the younger (400–600 yr) and older (3000 yr) limits of the soil chronosequence range can be reasonably ruled out owing to the expectation that 22 m of dextral displacement is unlikely to have occurred over such a short time frame (i.e., some 400–600 yr), and that up to 5–6 m of scarp height is unlikely to account for the vertical deformation over 3000 yr. Therefore, the preferred age of abandonment for the I-3 surface (i.e., effectively a maximum age of 1590–1730 yr), which grades toward Lake Poerua, fits with a general understanding of the sequence of late Holocene fanning surfaces and soils, from northeast to southwest across the Inchbonnie area. Similarly, young dates on exposed tree stumps in the valley and on the floor of Lake Poerua (e.g., B3, TR1, WP3; all less than 500 cal. yr B.P.) (Table 1) probably relate to young avulsions of the Taramakau River that have occurred following the last one or two surface ruptures on the Alpine fault.

A third dating technique, optically stimulated luminescence (OSL) dating, was attempted on fine-grained deposits of the Taramakau aggradation sequence within the Harris trenches (see Rieser and Wang, 2009). OSL age results have not been presented in detail, as they all yielded feldspar OSL ages from the silt fraction that are consistent with the early Holocene and Last Glacial period for the West Coast of South Island (Suggate, 1965). Two of these OSL dates (14.1 ± 1.6 and 26.3 ± 4.2 k.y.) are consistent with a time when the Inchbonnie area was under ice and not depositing alluvial gravels at this location. Two other OSL dates from fine-grained deposits within the Harris trenches (7.5 ± 1.5 and 8.15 ± 0.81 k.y.) are also significantly older than the soil chronosequence ages and would imply very low rates of deformation for the Alpine fault. Therefore, we recognize that there is a problem with using OSL ages from this site, i.e., poor or non-resetting of glacial age materials redeposited in late Holocene deposits, and consequently we have rejected these dates from our results.

Finally, as the former slip rates for the Alpine fault at Inchbonnie are derived from weathering rind ages of 1100 ± 100 yr on graywacke clasts from the I-2 surface, it is important to test the validity of these ages (Berryman et al., 1992). We accept that this was the only dating technique, and therefore best available data, for this area at the time. However, we infer that the sample location cannot be used in association with current New Zealand weathering rind calibration curves (McSaveney, 1992; M.J. McSaveney, 2009, personal commun.). That is, the sample site on surface I-2 was formerly a forested alluvial surface where the sandstone clasts were within the soil-forming zone; c.f. dry alpine environments, where chemical weathering processes are minimal. Therefore, we conclude that the Inchbonnie weathering rind data are invalid and cannot be used to determine geologic slip rates from these surfaces. The implications of this will be discussed in the following sections.

A new minimum estimate for the vertical slip rate for the Alpine fault at the Harris site is 2.9 ± 0.4 mm/yr. This result combines the vertical scarp height from profile HR (4.8 ± 0.4 m; Fig. 6) with the minimum age of the I-3 (maximum age of I-2r) surface estimated from stump WP13 (1590–1730 yr). Two other profiles across the Alpine fault (Za, S1) near the Harris site across surface I-3 yield larger scarp heights of 6.4 ± 0.2 and 6.6 ± 0.2 m, respectively, and subsequently higher vertical slip rates using the available dates. We use profile HR for the vertical slip rate because it comes from the same locality (Harris site–stepover) as the dextral displacements (Fig. 7). The reasons for a lower vertical slip rate at the stepover are probably due to the rejuvenation (erosion) of the scarp post–I-3 abandonment by flooding across the I-3 surface that ultimately formed the I-2r channels. These data indicate the importance of understanding the geomorphic data inputs that go into deriving slip rates.

In order to calculate the reverse-slip movement and shortening, fault dip is required. Locally, measurements of hanging-wall bedrock attitudes imply that the dip of the Alpine fault is 58° ± 5° SE (Nathan et al., 2002), whereas measurements from the Harris trenches show that the faulting is typically high angle near the ground surface (Fig. 9). In this study a dip of 60° ± 5° SE is used with a vertical slip rate of 2.9 ± 0.4 mm/yr to derive an estimate for the minimum dip-slip rate of 3.4 ± 0.6 mm/yr. Similarly, shortening rates of 1.2–2.3 mm/yr can be calculated from profile HR using the fault dip and dip-slip rate above.

Our new values of vertical and dip-slip rates represent a significant drop compared with the published values, i.e., dip slip 6 ± 1–2 mm/yr (Norris and Cooper, 2001; Berryman et al., 1992). Earlier, we discounted the validity of the slip rates calculated at Inchbonnie using the weathering rind dating technique. Therefore, our lower estimates are valid and represent a more robust analysis of both the age and displacement components of the slip rate.

The dextral displacement of 22.5 ± 2 m on the shallow surficial channel measured across three fault strands at the Harris site is used together with the minimum age for the abandonment of the I-3 surface to calculate a minimum dextral slip rate of 13.6 ± 1.8 mm/yr for the Alpine fault at Inchbonnie. Again, we infer that the previously published value of 10 ± 2 mm/yr cannot be used as a valid slip rate for the fault in this area, based on the use of weathering rind ages. Nonetheless, this new value represents a significant increase in the accepted slip rate for the Alpine fault in this area, and such an increase (and decrease in the case of fault normal rates) has a significant impact on hazard and tectonic models for the area.

Additional dextrally displaced geomorphic features have been difficult to identify across the fault trace. However, Berryman et al. (1992) recognized an area on the I-2 surface near the Taramakau River where four channels had been displaced 11–13 m dextrally (Fig. 5). One of these former channels was later abandoned, leaving an active channel that displays a dextral displacement of ∼6 m. Although the slip rate based on this 11–13 m displacement is not upheld in this paper, these observations suggest that there have been single to multiple co-seismic dextral displacements across the Alpine fault at Inchbonnie of ∼6 ± 1 m. In this way the stepwise growth of the fault scarp between the Taramakau River and Lake Poerua (Figs. 5, 6) is mirrored by a similar increase in dextral displacement from the river to older surfaces to the northeast. This gives us further confidence that our geomorphic assessments and slip rates derived from dating the abandonment of the I-3 surface using AMS radiocarbon dating are more reliable and more time-averaged slip rates compared to previously published data.

The slip vector at Inchbonnie is estimated from the combination of the dextral and shortening rates at the Harris site, which are used in the next section as vectors with direction (trend) and magnitude (rate) in a kinematic circuit. The strike-parallel component of the total vector is 13.6 ± 1.8 mm/yr at 052°. There is a small change in the strike of the fault in the Inchbonnie area (effective to the NE of the Alpine-Hope fault junction) compared with the Central segment of the Alpine fault to the southwest of Inchbonnie (Fig. 2). To compare the resultant geologic slip rate with a geodetic rate, we have used the dextral slip rate in combination with the horizontal shortening rate (1.2–2.3 mm/yr at 142°). These components yield a combined minimum rate of 13.7 ± 1.9 mm/yr at an orientation of 061° ± 3° for the Inchbonnie slip vector. This value is in close agreement with the geodetic slip rate estimated for this locality (14.7 ± 1.1 mm/yr at 062.4° ± 3°) (see Wallace et al., 2007) and implies that the differences between short-term strain accumulation from GPS and medium-term strain release from geology are minimal relative to their differing time frames.

Geophysical, geodetic, and seismic hazard models indicate that a large proportion of tectonic motion is partitioned from the Alpine fault southwest of Inchbonnie onto the Hope and Kelly faults (Fig. 2) (Berryman et al., 1992; Wallace et al., 2007; Stirling et al., 2002). The revision of slip rates at the Inchbonnie site, in combination with recently published slip rates on nearby structures such as the Hope and Kakapo faults (Yang, 1991; Langridge and Berryman, 2005), allow for a geologic analysis of the partitioning of strain about this major transition in the Australian-Pacific plate boundary.

First- and second-order tectonic changes can be assessed by using geologic fault slip rates and vectors in a kinematic circuit model, following a similar technique to that of Humphreys and Weldon (1994). Figure 12 documents a kinematic vector sum constructed using the rates, vectors, and uncertainties of the major regional faults. The circuit passes from Inchbonnie, between the Clarence and Hope faults, crossing the western end of the Hope fault before returning southwest to the Kakapotahi River site on the Alpine fault via the Kakapo fault (Fig. 12A). The slip vectors for the Kakapotahi River are shown as positive vectors, whereas all other “returning vectors” have a negative sense. The Hope and Kakapo faults are both primarily dextral-slip faults and have small to negligible shortening components. Slip vector data for the Hope fault comes from the McKenzie fan site (9.6 ± 1.5 mm/yr at 067°) and for the Kakapo fault from the Kakapo Brook site (6.4 ± 0.4 mm/yr at 037°) (Yang, 1991; Langridge and Berryman, 2005). Dip-slip rates have been converted to horizontal shortening rates in order to be in accordance with GPS data, which are expressed in a horizontal framework (Fig. 12).

To complete the kinematic circuit, valid slip rates for the Alpine fault are required from southwest of the Kelly fault. The Kakapotahi River locality is the closest to the northeast end of the Central segment, but it is situated within 65 km of the Inchbonnie site, which is wholly northeast of the Alpine-Kelly-Hope fault junction. From the Kakapotahi River to Inchbonnie, the dextral rate decreases from 29 ± 6 to 13.6 ± 1.8 mm/yr, with the shortening rate decrease from 3 ± 1 to 1.7 ± 0.6 mm/yr, assuming a near surface dip of 60° SE (Norris and Cooper, 2001; this study). There is an associated strike change from 048° to 052° between these two sites, respectively. The dextral slip rates estimated in this study represent an ∼50% decrease from the Kakapotahi River site, and there is a similar relative drop in the calculated reverse-slip rates.

The resultant kinematic circuit shows a number of interesting features. First, the slip rates for these four faults (or segments) do not account for the total convergence rate required across the Australian-Pacific plate boundary in this area; i.e., 37 ± 2 mm/yr at 250° (DeMets et al., 1994). The average strain release from the “outgoing” (Kakapotahi River) and “incoming” paths of the kinematic circuit is 29 ± 2 mm/yr at 236°, or ∼69%–89% of the plate rate. This suggests that there must be further deformation on other structures and/or within crustal blocks to the south and east of our circuit that account for the remaining ∼11%–31% of plate motion. Other fault systems that could account for this shortfall in slip rate include the Main Divide Fault Zone (Cox and Findlay, 1995), the Poulter fault (Berryman and Villamor, 2004), the Porters Pass Fault Zone (Cowan et al., 1996; Howard et al., 2005), and structures within the Australian plate (Nathan et al., 2002) (Figs. 1, 12).

Second, though the kinematic circuit accounts for only 69%–89% of the plate convergence rate, the circuit itself is virtually closed, i.e., the net vector sum is close to zero. The net difference between the outgoing and incoming paths is only ∼1.2 mm/yr at an azimuth of 348°. At first order, this implies that these major faults account for the greater proportion of strain accumulation and release in this part of the plate boundary. The resultant vector difference can be completely accounted for within the uncertainties of the slip-rate estimates presented here (e.g., the large errors for Alpine fault rates).

Third, excluding the Kakapo fault, the azimuths of each of the slip vectors and the plate convergence direction are remarkably similar, i.e., generally at 062°–071° compared with 070° (250°), respectively (Fig. 12). This implies that the faults operate in accordance with the regional stress field at the plate boundary scale. Similarly, the principal horizontal stress (PHS) directions for most faults throughout South Island are subparallel to each other and the regional stress field (Berryman, 1979). For the Alpine fault, this effectively means that owing to the step-down in dextral slip rate that occurs from Kakapotahi to Inchbonnie, there must be a local strike change (i.e., as observed from 228° to 232°), step-down in the rate of dip-slip motion (6 ± 1 to 4 ± 1 mm/yr) consistent with the regional stress direction across the plate boundary, and/or changes in the partitioning of strain release. From a kinematic perspective, this helps to explain why the highest parts of the Southern Alps in South Island (Fig. 1) are limited to the length of the Central segment of the Alpine fault, southwest of the transition to the Marlborough Fault System.

Finally, slip vectors used here are estimated from dated surfaces that range from late (Inchbonnie–western Hope) to mid-Holocene (Kakapo fault) and represent medium- to long-term estimates of strain release. Despite their differing time scales, the magnitude and direction of resultant slip vectors derived from the results of this study and contemporary GPS rates and vectors are remarkably similar at first order (e.g., Wallace et al., 2007). The geologic slip-rate and kinematic data presented in this study confirm a major change (junction and bend) in the Australian-Pacific plate boundary through New Zealand in central South Island. In this area, approximately half of the Alpine fault motion (∼52%) is partitioned onto the Hope, Kelly, and Kakapo faults at the southeastern edge of the Marlborough Fault System. These results have significant implications for the treatment of seismic hazard at junctions along major continental plate boundary systems. The new, higher dextral slip rate presented from Inchbonnie is also more consistent with the overall strain budget that is partitioned farther to the northeast among faults of the Marlborough Fault System (Van Dissen and Yeats, 1991; Holt and Haines, 1995).

1. Geomorphic and geochronologic studies in the Inchbonnie area have resulted in refined local dextral and vertical slip rates for the Alpine fault. A series of late Holocene (0–2 k.y.) alluvial surfaces (I-Ø to I-4) have been mapped in this part of the Taramakau River valley. The scarp of the Alpine fault crosses these surfaces, growing in height stepwise from I-1 to I-4.

2. Detailed mapping and surveying at the leading edge of a push-up structure along the fault at the Harris site have resulted in new estimates of vertical (4.8 ± 0.5 m) and dextral (22.5 ± 2 m) displacement there. Based on precise dating of a remnant tree stump, the minimum age for the abandonment of the I-3 surface at the Harris trench site is ∼1590–1730 yr.

3. The resultant dextral, vertical, and reverse-slip rates for the Harris site are 13.6 ± 1.8 mm/yr, 2.9 ± 0.4 mm/yr, and 3.4 ± 0.6 mm/yr, respectively. These values are somewhat larger (dextral) and smaller (dip slip) than previous estimates for this area, which have been rendered invalid by improvements in the dating of late Holocene surfaces locally.

4. A kinematic circuit using the new slip vector in conjunction with other fault slip vectors in the transition area between the Alpine fault and the western end of the Marlborough Fault System is almost closed (i.e., net ∼0). The analysis shows that 69%–89% of the plate boundary strain in the area can be accommodated across the major faults (Alpine-Hope-Kakapo), with approximately half of the strain release being transferred from the Alpine fault to the Kelly-Hope-Kakapo system.

5. The higher estimated strike-slip rate for Inchbonnie is consistent with the amount and partitioning of geologic strain between the Alpine fault and faults of the Marlborough Fault System in this part of the Australian-Pacific plate boundary.

The authors wish to thank Mark Hemphill-Haley, William Ries, and Zion Klos for assistance in the field. We also thank Peter Harris for access and permission to dig trenches and to the Department of Conservation for access to Lake Poerua. Fabian Hurter provided several of the scarp profiles. The authors thank Laura Wallace, Mauri McSaveney, Kelvin Berryman, and Liz Schermer for discussions and insightful reviews that helped the content of this paper. This research was funded by the New Zealand Foundation for Research, Science and Technology project PLT Alpine Fault earthquake geology (PGST Contract CO5X0702).