Using displaced river terraces to determine Late Quaternary slip rate for the central Wairarapa Fault at Waiohine River, New Zealand

Abstract

We analyse a flight of post-Last Glacial Maximum terraces at Waiohine River, New Zealand that are progressively displaced by the dextral-slip Wairarapa Fault. The Waiohine River is shown to have smoothed its faulted river banks after each earthquake so that riser displacements are only recorded after abandonment of their lower bounding terrace tread. Digital elevation models produced from newly collected high-precision topographic data allowed us to precisely measure the cumulative dextral displacement of five risers and two palaeochannels, and the vertical displacement of six treads. The ratio of horizontal to vertical slip at the terraces ranges from 5.0 to 10.2 (average of 6.9±3.5 (2σ)). Combining our new displacements for the ‘Waiohine’ aggradation terrace and next younger terrace with new and previous optically stimulated luminescence (OSL) ages for Waiohine terrace silts, we calculate Late Quaternary dextral slip rates for this central part of the Wairarapa Fault of 11.9±2.9 mm a⁻¹ (2σ) and >9.2±1.3 mm a⁻¹ (2σ). Based on the smallest and next smallest observed offsets, the magnitudes of the inferred 1855 (smallest) and penultimate (next-smallest) single-event displacements are 12.4±0.8 m (2σ) and 9.7±1.7 m (2σ), respectively.

Keywords:

• fluvial terrace
• slip rate
• strike-slip fault
• tectonic geomorphology
• Wairarapa Fault

View correction statement:

Erratum

Introduction

The Wairarapa Fault is a major active dextral strike-slip fault in the North Island of New Zealand (Fig. 1). It most recently ruptured in an M_w >8.1 earthquake on 23 January 1855, over an onshore distance of c. 120 km (Rodgers & Little 2006). Despite this, the magnitude and distribution of its single-event displacement or its Late Quaternary slip rate remain poorly constrained. Estimates of 1855 single-event dextral displacements on the Wairarapa Fault range from as high as c. 18.5 m at Pigeon Bush on the southern part of the fault (Rodgers & Little 2006) to as low as 4–7 m on the Alfredton Fault, discontinuously linked to the northern end of the Wairarapa Fault (Schermer et al. 2004). Estimates of the Late Quaternary dextral slip rate along the central Wairarapa Fault range from 6.5 mm a⁻¹ (Van Dissen & Berryman 1996) to 11.5±0.5 mm a⁻¹ (Grapes & Wellman 1988; Wang & Grapes 2008). The uncertainties in slip rate are largely due to uncertainties in the age of the chief geomorphic marker transected by the fault, the widely post-distributed Last post-Glacial Maximum (LGM) aggradation terrace termed the ‘Waiohine’ surface.

Figure 1  Major active faults in the lower North Island, New Zealand (modified from Little et al. 2009). Small open circles are main towns. Large open circle is area of this study (Waiohine River). Different sections of the Wairarapa Fault referred to in the text are labelled. Inset: tectonic setting of New Zealand, where Pacific plate motions are shown relative to a fixed Australian plate according to Nuvel-1a plate motion model of De Mets et al. (1990)Mets et al. (1994).

Flights of fluvial degradation terraces can provide valuable geomorphic markers of progressive fault displacements and can be used to study rates and directions of fault slip through time. To use displaced terrace risers to measure slip rates on strike-slip faults requires knowledge of how those alluvial landforms interact with the river and fault scarp, and when they begin to record fault slip. A riser is the incisional slope between two adjacent terrace surfaces. The age of a riser represents the time at which fluvial lateral erosion of that palaeo-river bank ended. Temporally, this must have taken place some time between the abandonment ages of the upper and lower terrace treads bounding that riser. However, the exact relationship between the timing and magnitude of a terrace riser displacement is commonly uncertain (e.g. Suggate 1960; Lensen 1964; Avouac 1993; Cowgill 2007; Harkins & Kirby 2008). The cessation of lateral erosion of a riser is most commonly assumed to occur closely in time to the abandonment of the lower bounding terrace tread (e.g. Lensen & Vella 1971; Little et al. 1998; Van der Woerd et al. 2002; Meriaux et al. 2004, 2005; Mason et al. 2006). Less often, the cessation of lateral erosion of a riser is assumed to occur more closely in time to the initial incision of the upper bounding terrace tread of that riser (e.g. Van Dissen et al. 2005; Zhang et al. 2007). Theoretical uncertainty in a riser's age results in uncertainty in a slip rate estimate. Cowgill (2007) illustrates that the minimum and maximum slip rates derived from terrace riser offsets and the ages of the upper and lower bounding terraces, respectively, can vary by as little as a factor of 1.2 but commonly vary by a considerably larger factor (as large as a factor of 5).

In this study we examined a classic flight of terraces to the northeast of the Waiohine River in the central North Island, New Zealand, that have been variably displaced by the Wairarapa Fault (Fig. 1). The Waiohine terraces were formed by sequential downcutting of the river into previously deposited aggradation gravels of post-LGM age (Lensen & Vella 1971). The Waiohine terraces provide an exceptional setting in which we can study the interaction between an active fault and a flight of degradation terraces, as well as being a key site for measuring and dating rates of slip on the central section of the Wairarapa Fault. We quantify displacements of terraces, risers and palaeochannels using digital elevation models (DEMs) produced from topographic data points collected using real-time kinematic (RTK) global positioning system (GPS) and electronic distance meter (EDM) techniques. We attempt to date terraces through optically stimulated luminescence (OSL) dating of terrace sands and silts. We use our new displacement and age data together with previous OSL ages to: (1) draw conclusions about the nature of the interaction between flights of degradation terraces and active faults and determine the most valid model of progressive lateral displacements of flights of degradation terraces at the Waiohine River, allowing us to accurately assign terrace ages to the measured terrace riser displacements; (2) determine single-event displacements for the 1855 and penultimate earthquakes; and (3) derive the ratio of horizontal to vertical slip and the Late Quaternary dextral slip rate on this central part of the Wairarapa Fault.

This paper is structured to first provide the reader with background information on the tectonic setting, models of progressive fault displacement and previous geomorphologic and dating studies at the Waiohine terraces. Secondly, we describe the data collection methodology in this study and our new terrace interpretation. We follow by describing our method for determining fault displacements and a summary of our new displacement measurements. The terrace ages determined in this study are then described, followed by our discussion and conclusions.

Tectonic setting

New Zealand straddles the boundary between the Pacific and Australian plates (Fig. 1). Oblique subduction of oceanic crust occurs along the northeast-trending Hikurangi margin in the North Island and oblique continental collision takes place in the South Island. The southern North Island lies on the Australian plate to the west of the margin. In central New Zealand, the relative convergence rate of the Pacific plate relative to the Australian plate is c. 39 mm a⁻¹ along an azimuth of c. 261° (NUVEL-1A of De Mets et al. 1990, 1994).

A small proportion of the margin-perpendicular component of plate motion is accommodated by shortening of the upper (Australian) plate through thrust faulting and related folding in the onshore and offshore parts of the plate (Barnes & Mercier de Lepinay 1997; Barnes et al. 1998; Nicol et al. 2007). The majority (>80%) of the margin-perpendicular motion is inferred to be accommodated by contractional slip on the strongly coupled subduction megathrust, which is currently locked and accumulating elastic strain at 20–25 km depth beneath Wellington (Reyners 1998; Darby & Beavan 2001; Wallace et al. 2004, 2009; Nicol et al. 2007).

The margin-parallel component of plate motion is largely accommodated by dextral slip on the NNE-striking faults of the North Island Dextral Fault Belt (NIDFB) (Beanland 1995; Mouslopoulou et al. 2007). In the onshore region near Wellington, dextral slip on the Shepards Gulley, Ohariu, Wellington and Wairarapa Faults (Fig. 1) accommodates up to c. 18 mm a⁻¹ (Beanland 1995) of the margin-parallel motion (e.g. Berryman 1990; Van Dissen et al. 1992; Van Dissen & Berryman 1996; Rodgers & Little 2006; Wang & Grapes 2008). At the surface, these faults typically dip steeply to the northwest and have a small component of reverse-slip. Seismicity data suggest that these faults intersect the subduction megathrust at depths of 20–30 km beneath the southern North Island (e.g. Reyners 1998). Clockwise vertical-axis rotation of eastern parts of the North Island (e.g. Wallace et al. 2004; Nicol et al. 2007; Rowen & Roberts 2008), strike-slip on active ENE-striking faults in Cook Strait such as the Boo Boo Fault (e.g. Barnes & Audru 1999; Barnes 2005) and oblique slip on NE-trending offshore faults including the subduction megathrust (Barnes & Audru 1999; Nicol et al. 2007) also accommodate some of the margin-parallel component of plate motion.

The Wairarapa Fault

The Wairarapa Fault is the easternmost dextral-reverse fault in the southern NIDFB, bounding the eastern side of the Rimutaka Range (Fig. 1). It is interpreted to have been initiated in the Pliocene as a reverse fault, and to have reactivated as a strike-slip fault at c. 1–2 Ma in response to a clockwise vertical axis rotation of the forearc relative to the Pacific Plate (Beanland 1995; Kelsey et al. 1995; Beanland & Haines 1998). Based on the geometry of its trace, the Wairarapa Fault can be divided into southern, central and northern sections (Fig. 1).

The central section, when observed at scales more detailed than 1:250,000, consists of an en echelon array of mostly left-stepping faults. These discontinuous dextral-oblique faults are separated by contractional bulges or folds in the area of their overlap that cause warping of alluvial terrace surfaces (Grapes & Wellman 1988; Rodgers & Little 2006). Rodgers & Little (2006) mapped 15 individual fault strands 1200±700 m long on the southern part of the central Wairarapa Fault, separated by stepovers 400–600 m long and 20–200 m wide. Along its central section, the Wairarapa Fault forms the boundary between the greywacke of the Rimutaka Ranges and the Holocene and latest Pleistocene river gravels of the Wairarapa Valley (Grapes 1991). North of the Waiohine River, the northern section of the Wairarapa Fault bifurcates eastward into a series of east-northeast striking dextral splay faults (Fig. 1) including the Carterton, Masterton and Mokonui Faults (Langridge et al. 2005).

Previous models for progressive fault displacement of flights of degradational terraces

In this section, we describe three models for progressive lateral fault displacement during the formation of degradational river terraces (see also Carne 2009: 20–27). Comparisons between the models and our observations allow us to draw conclusions about the nature of the interaction between flights of degradation terraces and active faults, and to accurately assign terrace ages to our measured riser displacements. The first two models are end-member models that were first described by Suggate (1960). The third model is a hybrid model that falls between those two end-members and was described by Lensen (1964). We use Lensen's (1964) terrace nomenclature (Fig. 2). For example, terrace tread is the nearly horizontal surface of fluvially deposited terrace gravels that represents the top surface of a now abandoned, former river bed (Fig. 2), and terrace riser is the steep incisional slope between two adjacent river terraces that represents a former river bank (Fig. 2).

Figure 2  Selected river terrace terminology for a typical section through the Waiohine River terraces. Aggradation fill/gravel is deposited by the river during times of high sediment supply, and commonly contains sand beds between thicker gravel units. The top surface of the aggradation fill is an aggradation terrace. Degradation terraces are formed by erosion/incision during subsequent river downcutting into the aggradation fill. Both aggradation and degradation terraces can be mantled by coverbeds of silt overbank deposits or loess. The tread of a terrace is the top of the terrace gravels beneath the younger coverbed.

The first end-member model (Fig. 3A) assumes that, after a lateral displacement on a fault, the river is unable to laterally erode or trim any of the displacement of its river banks. In this model the lateral displacement of a riser is equal to the displacement of any palaeotopographic features incised into that riser's upper bounding terrace (e.g. palaeochannels). That riser's lateral displacement is ‘fossilised’ into the landscape from the initiation of incision of that riser into its upper bounding terrace, and so the lateral displacement of that riser is most precisely dated by the age of abandonment of its upper bounding terrace tread.

Figure 3  Two-stage (slip events D1 and D2) plan view models of a degradational terrace sequence undergoing the same amount of river downcutting and lateral faulting but different amount of lateral riser trimming (modified from Lensen 1964). A, No lateral riser trimming; B, complete lateral riser trimming and C, partial lateral riser trimming. Bold line is the fault and half arrows indicate dextral sense of motion; toothed lines are terrace risers; thin grey line is palaeochannel; dashed line indicates displacement of riser before lateral erosion; D1 and D2 are the displacements associated with the first and second faulting events, respectively; W1 and W2 are the terrace widths on the left and right sides of the fault, respectively.

The second end-member model (Fig. 3B) assumes that any lateral fault displacement is followed by rapid and complete lateral erosion of the displaced riverbanks by the still-active river occupying the lower surface of that riser. Lateral erosion can result in irregularities in the terrace sequence such as uncomplemented terraces, where a displaced terrace is completely removed by erosion on one side of the fault (Suggate 1960). In this model, the lateral displacement of the riser (e.g. lower riser) is equal to the displacement of any palaeotopographic features incised into that riser's lower bounding terrace (e.g. palaeochannels). That riser's lateral displacement is ‘fossilised’ into the landscape from the time of abandonment of the riser's lower bounding terrace, when the river that occupied that terrace can no longer completely trim any displacement of its former river bank. The lateral displacement of the riser is most precisely dated by the age of abandonment of that riser's lower bounding terrace tread.

The third model (Fig. 3C), a hybrid that falls between the two end-member models, assumes that the lateral fault displacement of the river bank is later only partially eroded or trimmed by the active river. In this model, the magnitude of the lateral displacement of a riser will fall between that of the displacements of any palaeotopographic features on that riser's upper or lower bounding terraces. The age of the apparent lateral displacement of a riser is between the abandonment ages of its upper and lower bounding terrace treads.

In a similar way to the strike-slip displacements, the three models also make similar predictions about the progressive accumulation of vertical displacement on faulted river terraces (e.g. Lensen 1964). The models (depicted in Fig. 4A–C) predict only small changes in the gradient of the river across the fault due to co-seismic throw and assume that, away from the fault, the regional river gradient remains the same before and after faulting. Figs. 4A–C only consider cases where the scarp is downhill-facing.

Figure 4  Cross-sectional models of a river bed and bank undergoing the same amount of vertical faulting but different amounts of river downcutting (modified from Lensen 1964). Here we consider only cases where the scarp is downhill facing. A, Immediately after vertical faulting with no downcutting of the fault scarp in the river bed; B, complete downcutting of the fault scarp in the river bed by erosion and C, some downcutting of the fault scarp in the river bed with erosion on the upthrown side and deposition on the downthrown side of the fault. Heavy black line is the active river bed with direction of river flow; thin black line is the top of the river bank; bold vertical line is the fault; dashed line indicates location of river bed prior to faulting; small arrows indicate height of riser on either side of the fault; light grey indicates areas of downcutting; dark grey indicates areas of deposition.

Cowgill (2007) presents six geomorphic indices (summarised in Table 1) that are useful in determining which model of progressive lateral displacement of flights of degradation terraces is most valid for a particular terrace sequence, and the most accurate age to assign to a terrace riser displacement.

Table 1  Geomorphic indices of Cowgill (2007) for determining the most valid model of progressive lateral displacement of flights of degradation terraces for a particular terrace sequence.

	Indices	Description
1	Comparison of riser offset with width and/or displacement of main channel incised into lower terrace tread	Riser displacements greater than the offset or width of the channel incised into the lower terrace tread must have occurred prior to incision of the new channel into the lower terrace tread, indicating incomplete trimming
2	Similarity of riser and tread displacements	The displacement of a terrace riser will be equal to the displacement of a feature incised into that riser's lower or upper bounding terrace tread, under a complete or no riser trimming model, respectively
3	Morphological analysis of scarp profile	The age of the scarp is estimated using the diffusion equation that models the risers topographic evolution as it degrades
4	Riser deflection in plan view	A riser that is curved in a sense consistent with the fault slip sense suggests that the riser was incompletely trimmed prior to the lower terrace's abandonment
5	Diachroneity of terrace abandonment	Isochronous terrace abandonment suggests complete riser trimming as the river migrates quickly across the flood plain, while diachronous terrace abandonment suggests incomplete or no riser trimming as the river migrates away from the riser for extended periods of time allowing riser displacements to be preserved
6	Whether riser crests or bases yield the slip vector	The intersection between the active tread and the riser crest or base makes a piercing line that records fault motion occurring after the abandonment of the upper and lower terrace, respectively. The slip vectors recorded by these piercing lines can be compared to known slip vectors, or slip vectors from other terrace-riser pairs in the sequence to identify any parallelism, if the slip direction is uniform over time.

Previous geomorphologic and dating studies at the Waiohine terraces

The Waiohine terraces northeast of the Waiohine River have been variably displaced by the Wairarapa Fault (Fig. 5). During the Holocene, these degradation terraces were sequentially downcut into the post-LGM Waiohine aggradation terrace (Lensen & Vella 1971). This is the most extensive aggradation terrace in the Wairarapa Valley, which formed by the coalescence of alluvial fans along the Tauwharenikau, Waiohine, Waingawa and Ruamahanga Rivers (Grapes 1991; Begg & Johnston 2000). In this section we describe the results of previous geomorphologic and dating studies at the Waiohine terraces.

Figure 5  Overview map of Waiohine River terraces. Wairarapa Fault trace is bold line. Unit ‘Wai’ depicts post- Last Glacial Maximum Waiohine aggradation surface. Unit ‘Po’ depicts older Porewan aggradation surface. Coordinates are in New Zealand Map Grid (NZMG).

Age estimates for the Waiohine surface or similar post-LGM aggradation sequences elsewhere in New Zealand range from c. 5 ka to 30 ka (see Carne 2009: 151–152 for ages and references). Debate over the Waiohine surface age is largely based on the identification and/or correlation of the Waiohine (or equivalent) surface across different parts of the North Island, and the different methods used to date the surface. Three recent studies have estimated the timing of Waiohine surface abandonment from OSL and ¹⁴C dating of samples collected in the Wairarapa Valley (Formento-Trigilio et al. 2003; Wang & Grapes 2008; Little et al. 2009). Formento-Trigilio et al. (2003) report four OSL ages between 5.12±0.37 ka and 10.5±1.2 ka for loess coverbeds on Waiohine terrace surfaces in the Huangarua River Valley, and suggest that the Waiohine terrace tread was abandoned at c. 10–12 ka. Little et al. (2009) bracket the abandonment age of the Waiohine terrace between 12,400±300 cal. yr BP and 5,440±140 cal. yr BP using ¹⁴C samples collected above and below the terrace tread at Cross Creek on the southern Wairarapa Fault. Because the c. 12 ka sample was collected only 1–2 m below the terrace tread, they suggest that terrace abandonment probably occurred soon after this time. Wang & Grapes (2008) present eight OSL ages between 16.1±1.6 ka and 10.0±0.9 ka for overbank silts/loess coverbeds overlying Waiohine terrace gravels from six sites on the central-southern Wairarapa Fault. Three of those samples (MR1, WH1 and WH2) were collected near the Waiohine River terraces (Fig. 5) and yielded ages of 10.0±0.8 ka, 13.0±0.9 ka and 10.2±1.2 ka, respectively.

Two important previous geomorphologic interpretations of the faulted Waiohine terrace sequence are those of Lensen & Vella (1971) and Grapes & Wellman (1988). Lensen & Vella (1971) describe a flight of eight terraces (Fig. 6A) referred to as A–H (A being the oldest and highest and H being the youngest and lowest of these terraces). They report seven vertical displacements, measured with a level and rod with an estimated accuracy of c. 0.15 m, and five horizontal displacements measured with a measuring tape with an estimated accuracy of c. 0.9 m. They did not identify any landforms on the Waiohine surface from which a horizontal displacement could be measured (Lensen & Vella 1971). Grapes & Wellman (1988) mapped the Waiohine terraces using vertical aerial photos on an approximate scale of 1:17,000. They describe a flight of seven terraces – Ag3 and terraces A–F – where Ag3 is the oldest and highest Waiohine aggradation surface and terraces A–F are degradational terraces, terrace F being the youngest and lowest (Fig. 6B). They report seven vertical displacements and six horizontal displacements. Because their interpretive maps lack topographic contours and their displacement measurements are not accompanied by methodological explanations, survey data or assessments of uncertainty, these results are difficult to evaluate. The uncertainties for all horizontal and vertical displacement measurements were later uniformly quoted as±1 m by Grapes (1999). Grapes & Wellman (1988) also describe a dextral displacement for the Waiohine surface of 130 m, based on their recognition of four closely spaced channels incised into that surface (Fig. 6B).

Figure 6  Previous interpretations of the terraces at Waiohine River shown at the same scale. A, Interpretation by Lensen & Vella (1971) and B, interpretation by Grapes & Wellman (1988) including palaeochannels incised into Ag3 (Waiohine surface) displaying c. 130 m dextral displacement. Wairarapa Fault is bold line. Terrace edge/top of fault scarp is fine-toothed line. Coordinates are in NZMG.

Collection of new displacement and age data at the Waiohine terraces

In this study, we reinterpret the displacements of geomorphic features across the Wairarapa Fault at the Waiohine River terraces. We mapped the terrace sequence using c. 1:3,000 scale colour vertical aerial photographs and modern geospatial techniques. A Trimble R8 GNSS RTK GPS and a Sokkia 3030R EDM laser ranging theodolite were used to collect over 18,000 topographic data points (± 1 cm (2σ) horizontal and±18 cm (2σ) vertical precision) across the terrace sequence. The uncertainty in the vertical precision was calculated from the standard deviation of the differences in the elevation of points within 1 m of each other, collected during two different RTK GPS surveys at the Waiohine terraces. The data were krigged (i.e. gridded at a grid spacing of 1–4 m; 4 m for larger scale maps and 1 m for smaller scale individual offsets) and variously contoured using ArcMap and ArcGIS software to produce microtopographic maps or DEMs for the terrace sequence and for each specific terrace and riser displacement. From these maps, precise estimates of vertical and horizontal displacements were made from each terrace and riser, respectively.

We also attempted to construct a detailed map of the previously mentioned channels incised into the Waoihine surface that were recognised by Grapes & Wellman (1988). We did this by means of an RTK GPS-based survey across the proposed site of the channels on both sides of the fault (Fig. 7). In addition, in hope of verifying the presence of any infilled palaeochannels, an auger survey was made across the terrace surface on the downthrown side of the fault along the base of the fault scarp, to measure the depth to the top of the Waiohine gravels in the area where Grapes & Wellman (1988) mapped the four channels. Eleven auger holes were made in a fault-parallel profile (Fig. 7).

Figure 7  Geomorphologic interpretation and sample locality map of the terraces at Waiohine River based on this study. Six terraces (A–F) were identified; where terrace A corresponds to the oldest and highest aggradation terrace (Waiohine surface) and terraces B–F are degradation terraces of which terrace F is the youngest. Terraces C’ and D’ may represent locally developed palaeotopography on terraces C and D, respectively, and are not repeated on the southeast side of the fault. Distinct palaeochannels include Channel 1 (Ch1) and Channel 2 (Ch2). Wairarapa Fault is depicted by a bold line. Coordinates are in NZMG.

We collected a suite of nine OSL and two radiocarbon samples from above and below the treads of two terraces at four different sites at and near the Waiohine terraces (Fig. 5). The location, lithology and stratigraphic context of each sample are described later in the paper, after presentation of our new terrace map interpretation. The OSL samples were collected by inserting steel tubes into cleaned vertical cliff or pit faces. OSL samples were dated by the Victoria University of Wellington OSL Laboratory, and radiocarbon samples were dated by the University of Waikato Radiocarbon Laboratory.

New map interpretation of the Waiohine terraces

In this study, we identified six main terrace surfaces (A–F, where A is the Waiohine surface) and two indistinct or locally developed terraces (C’ and D’) (Fig. 7). In addition, we mapped two palaeochannels; Ch1, incised into terrace C, and Ch2, incised into terrace F (Fig. 7). We did not observe any horizontal displacement across the Wairarapa Fault in the modern river bank, or any vertical displacement of the modern river bed as a result of the 1855 earthquake. The risers are typically not curved in plan view in a sense that is consistent with a dextral sense of slip on the fault. The locations and trends of terrace riser segments near the fault (risers AB, BC and CD) determined in this study are different to previous studies by Lensen & Vella (1971) and Grapes & Wellman (1988) (Fig. 6A–B & 7). Neither contouring and profiling of the DEM produced from RTK GPS data collected on the Waiohine surface (our terrace A) nor the augering survey revealed any topographic evidence for the channels described by Grapes & Wellman (1988) (Fig. 8). We found no evidence for channels in the topography of the top of the terrace gravels, as would be expected if channels had been incised into this surface. The depths to the top of the gravels ranged between just 6 cm and 15 cm for the eleven auger holes.

Figure 8  Detailed site investigation to search for evidence of palaeochannels on the Waiohine surface described by Grapes & Wellman (1988). Central part of the figure shows topographic contours derived from a GPS survey of terrace A (Waiohine surface). Light grey polygons and solid black lines indicate location of palaeochannels according to Grapes & Wellman (1988). Fine grey toothed line shows edge of terrace on upthrown side of Wairarapa Fault. Black toothed lines show terrace risers. Inserts show topographic profiles across the Terrace A surface on upthrown (A–A’) and downthrown (B–B’) sides of the fault. Black circles are locations of auger holes in a targeted auger survey on the downthrown side of the fault to determine the depth to the top of the Terrace A gravels (B–B’ insert in grey).

Method for determining horizontal and vertical displacements of terraces and risers

At the Waiohine River, vertical and horizontal displacements of each terrace, riser and palaeochannel across the Wairarapa Fault were measured using the method outlined below. This method assumes a complete riser trimming model in which any vertical and lateral fault displacements are followed by rapid and complete erosion of the scarp in the river bed and displaced river bank (model 2 in Fig. 3B & 4B), until the river finally downcuts to abandon that fully trimmed terrace and riser. A complete riser trimming model is supported at the Waiohine terraces by the lack of any horizontal or vertical displacement in the modern river bank and bed, respectively, and the observation that none of the risers are curved in a sense that is consistent with a dextral sense of slip on the fault, as might be expected under an incomplete trimming model according to the fourth index of Cowgill (2007). Because the Wairarapa Fault plane is not exposed in the Waiohine River gorge, we assume a near-vertical fault plane at the Waiohine River on the basis of magnetotelluric and gravity surveys of the Wairarapa Fault (e.g. Hicks & Woodward 1978; Von Borstel & Ingham 2004).

1.	The mean elevations of equivalent terrace treads were calculated from GPS points (usually at least 50) collected within c. 20 m of the fault scarp (Fig. 9). The difference in these mean elevations gives a precise vertical displacement or throw estimate, assuming negligible tilt of the terrace surfaces within c. 20 m of the fault scarp. The uncertainties in the vertical displacement were calculated as the standard error of the difference between the means of the two terrace elevations. The vertical displacements have standard errors of <9 cm at 2σ level. The uncertainties in the elevation of each data point are insignificant for this calculation because it is the relative elevations between points that are important in this study.
2.	The top edges of the terrace risers have probably retreated as a result of erosion at the top of each riser, and the bottom edges or toes have probably advanced by deposition of younger colluvial sediment at the base. For each riser, a contour was arbitrarily selected on the downthrown side of the fault at the midpoint of the riser, which is most likely to have remained unmodified. On the upthrown side of the fault, a matching contour was chosen that differed in elevation by exactly the throw of the riser's lower terrace surface (Fig. 9). These two contours are assumed to have been continuous with one another at the time of abandonment of that riser's lower bounding terrace and prior to subsequent faulting of the riser. For palaeochannels incised into terrace surfaces, the thalwegs of the channels were selected on either side of the fault.
3.	A polygon was drawn along the fault trace to represent the topography near the fault trace that has been modified by erosion, deposition and/or anthropogenic processes (dark grey polygon on Fig. 9). The edge of the polygon was generally taken to coincide with the points in the topography where the terrace riser topographic contours deflect (often subtly) away from the main riser trend and begin to curve inwards towards the trend of the nearby fault trace. This polygon also included areas near the fault scarp where the density of GPS and EDM points was particularly low.
4.	We assume that the mid-riser contours (or palaeochannel thalwegs) once had an identical trend on either side of the fault where they met. This follows directly from the ‘complete trimming’ model which infers an originally linear and well-trimmed riser shape prior to faulting. Where the risers have different trends today on either side of the scarp, one or the other (or both) of the risers must have curved in such a way that they once met at the fault with the same trend (site of future fault scarp).
5.	The most fault-proximal segment of each preserved riser (i.e. that part nearest the fault and immediately outside of the polygon, typically within the last c. 5–10 m of it) was linearly projected to the fault trace with a trend that was parallel to the mean riser trend near the fault (Fig. 9). For both faulted risers and channels, the strike-slip displacement is measured as the distance between the two matched projection lines, along the trace of the fault.
6.	Triangular uncertainty regions were produced by drawing projection uncertainty lines on one side of the projected line segments to encompass the extreme range of curvature variations that could satisfy the assumption that the risers or thalwegs once met at the fault with the exact same trend (see step 4, above). The projection uncertainty line on one side of the fault was drawn with the same trend as the projected riser line segment for the equivalent riser on the other side of the fault (Fig. 9). These asymmetric triangular uncertainty regions are inferred to approximate 2σ confidence regions in the location of each projected riser line segment. The root-mean-square (RMS) of the two projection uncertainties (one for each riser segment on either side of the fault) approximates a 2σ confidence uncertainty (generally <3 m) for that displacement (Fig. 9). Any uncertainty in dextral displacement resulting from the uncertainty in the throw measurement is considered to be negligible here.

Figure 9  Method of calculating displacements as based on a complete lateral riser trimming model. (1) Mean elevations of equivalent terraces across the fault are found from GPS points close to the fault, from which the vertical displacement can be estimated (Throw = 1–2). (2) A mid-point contour is selected on the downthrown side of the fault and the equivalent contour is found on the upthrown side that differs by the vertical displacement magnitude. (3) A dark grey polygon is drawn to represent topography near the scarp that has been modified by erosion or deposition. (4 & 5) The trends of the fault-proximal parts of the risers are projected linearly across this grey area to hit the fault, giving an estimate of horizontal displacement. (6) Triangular uncertainty regions (assumed to approximate 2σ uncertainties) are determined by projecting uncertainty lines that satisfy the assumption that riser trends were once identical at the fault.

New displacement estimates for the Waiohine terraces

Cumulative displacements

Using the method outlined above, we measured cumulative vertical and dextral displacements of the Waiohine surface (terrace A) and five degradational terraces (terraces B–F) at the Waiohine River (Fig. 10A–F). The vertical and dextral displacements range from 1.30±0.02 m (2σ) to 19.67±0.09 m (2σ) and 11.4±1.7 m (2σ) to 101.1±3.4 m (2σ), respectively, and are listed in Table 2 from the oldest and highest terrace (terrace A) to the youngest and lowest terrace (terrace F) (see Carne 2009: 144–145 for detailed displacement calculations). For each of the terraces in the sequence, the sense of throw was up-to-the-northwest. Our measured displacements differ from those reported by Lensen & Vella (1971) and Grapes & Wellman (1988) (Table 3). For all six terraces, our vertical displacements are greater than those reported in either of the previous studies by up to 1.5 m (Table 3). Our dextral displacements are within error of those from the previous studies for four of the displacements; however, our dextral displacements for risers DE and BC are c. 12 m and c. 10 m smaller, respectively, than in the previous two studies.

Figure 10  Microtopographic maps of terrace, riser and palaeochannel displacements across the Wairarapa Fault. See Fig. 3 for locations. All co-ordinates are in NZMG. A, Terrace F, riser EF and Ch2 displacements, based on c. 4900 survey points, which are omitted for clarity. B, Terrace E and riser DE displacements, based on c. 1850 survey points. C, Terrace D and riser CD displacements based on c. 2920 survey points. D, Terrace C and Ch1 displacements based on c. 3620 survey points. E, Terrace C and riser BC displacements based on c. 3540 points. F, Terrace A, terrace B and riser AB displacements based on c. 2790 points.

Table 2  Cumulative and incremental vertical and horizontal displacements and differences in riser heights and terrace widths across the Wairarapa Fault.

Terrace	Mean cumulative throw across fault (m)	Standard error of differences in cumulative throw 95%	Incremental vertical offset relative to next younger terrace (m)	Standard error of differences in incremental throw 95%
A	19.67	0.09	3.93	0.10
B	15.74	0.04	2.75	0.05
C	12.99	0.03	1.15	0.05
D	11.85	0.03	9.18	0.06
E	2.67	0.05	1.36	0.05
F	1.30	0.02	1.30	0.02
Riser	Cumulative horizontal offset (m)	Projection uncertainty (c. 95%)	Incremental horizontal offset relative to next younger terrace (m)	Projection uncertainty (c. 95%)
AB	101.12	3.43	27.97	4.13
BC	73.15	2.30	−8.66^3	21.00
Ch 1 NE edge	83.71	12.06^4	–^2	–
Ch 1 thalweg	85.84	9.88^5	–^2	–
CD	81.81	20.87^4	59.71^3	20.93
DE	22.10	1.52	9.74^1	1.72
EF	11.43	1.71	–	–
Ch 2	12.36	0.80	12.36	0.80
^1Incremental horizontal offset of riser DE relative to riser EF was calculated using Ch 2 offset only, as EF riser has larger uncertainty
^2Dextral displacements of Ch 1 are within error of riser BC offset, no incremental horizontal displacement of riser BC relative to Ch1 is calculated
^3Because of the large uncertainty in riser CD displacement, we calculate an incremental displacement for riser BC relative to riser DE of 51.05±2.76 m
^4The large uncertainties for these measurements chiefly arise due to the difference in trend of the riser segments on either side of the fault, which yield large uncertainty triangles when the risers are projected across the grey polygon
^5The large uncertainty for this measurement chiefly arises from the uncertainty in the palaeochannel thalweg location on the downthrown side of the fault, made more difficult by the wide, shallow morphology of the channel

Table 3  Horizontal and vertical displacements at Waiohine River compared to results of previous studies (Lensen & Vella 1971; Grapes & Wellman 1988).

Lensen & Vella 1971			Grapes & Wellman 1988			This study
Terrace equivalent	Horizontal displacement (m)	Vertical displacement (m)	Terrace equivalent	Horizontal displacement (m)	Vertical displacement (m)	Terrace	Horizontal displacement with 95% projection error (m)	Vertical displacement with 95% standard error (m)
A	–	18.3	Ag3	130	18.5–19.9	A/Waiohine	129±18^3	19.67±0.09
B	99	14.7	A	97	14.9	B	101.1±3.4	15.74±0.04
C	85.5	12.3	B	84	11.8	C	73.2±2.3	12.99±0.03
D	66.9	12.0	No equivalent	–	–	Ch1	84.8±15.6^1	–
E	–	10.2	C	63	10.1	D	81.8±20.9	11.85±0.03
F	–	–	D	–	–	D'	–	–
G	32.1	3.6	E	32	1.7	E	22.1±1.5	2.67±0.05
H	12.0	0.9	F	12	0.9	F	12.4±0.8^2	1.3±0.02
^1Average from Ch1 NE edge and Ch1 thalweg
^2Displacement of Ch 2 thalweg with smaller error than riser EF
^3Horizontal displacement of terrace A inferred from average H/V ratio for Waiohine terraces and vertical displacement of terrace A

Incremental displacements

Under a model of complete riser trimming and riverbed downcutting, the incremental horizontal displacement that took place between the abandonment of two adjacent, differently displaced terraces is equal to the difference between the cumulative strike-slip of the respective risers at the back of each terrace (Fig. 3B). The corresponding incremental vertical displacement for two adjacent terraces is the difference between the cumulative throws of those terraces. The vertical and incremental displacements are listed in Table 2 for the oldest terrace pair (terrace A and terrace B) to the youngest terrace pair (terrace F and the modern river). The uncertainties in the incremental dextral and vertical displacements were calculated as the RMS of the uncertainties of the two cumulative measurements (upper and lower terrace/riser displacements). The cumulative displacement for terrace F is interpreted to represent the single-incremental displacement that occurred during the most recent Wairarapa Fault earthquake in 1855 (Table 2).

Terrace ages at Waiohine River

Samples collected for dating of Waiohine terraces

We collected nine OSL samples and two radiocarbon samples from terraces E and B at the Waiohine River to constrain the ages of their abandonment (see Fig. 5 & 7 for sample locations). Terrace E was dated to support our inference that palaeochannel Ch2 incised into terrace F displays a dextral displacement generated in the 1855 earthquake. Terrace B was dated to constrain the Late Quaternary slip rate on the Wairarapa Fault at Waiohine River. The location, lithology and stratigraphic context of each sample are outlined in Table 4 (see Carne 2009: 148 for stratigraphic logs). Samples were collected from coverbeds of fluvial overbank silts deposited on the gravel terrace treads, except for sample Wai-1b which was collected from a sand layer interbedded with the fluvial terrace gravels below the tread of terrace B.

¹⁴C and OSL-based terrace ages

In this section, we summarise and interpret the results of OSL and ¹⁴C dating of terraces B and E, as displayed in Table 4 (see Carne 2009: 39–41 for description of OSL dating technique and p. 149 for OSL age calculations). The ¹⁴C ages obtained from samples Whine 3 and Whine 6 were very young or modern suggesting that (1) the wood sample may be a root and (2) the charcoal sample collected from a charcoal-rich layer near the ground surface may have been deposited during flooding onto terrace E long after its abandonment. These ¹⁴C results are not discussed further in this paper (see Carne 2009: 150). The OSL ages obtained for silts (Whine 1, Whine 2, Whine 4, Whine 5) overlying the gravel tread of terrace E range from 10.1±0.9 ka (1σ) to 33.5±3.0 ka (1σ) (Table 4). The four OSL ages obtained for silts (Wai-10a, Wai-10b, Wai-11a and Wai-11b) overlying the gravel tread of terrace B range from 7.4±0.6 ka to 26.8±2.2 ka (Table 4). An OSL age of 37.9±2.4 ka was obtained for sand sample Wai-1b, below the gravel tread of terrace B (Table 4).

Table 4  Results of OSL and ¹⁴C dating of samples from the Waiohine River area.

Field No^1	NZMG Easting	NZMG Northing	Elevation (m asl)	Terrace^2	Depth (cm)	Distance above gravel tread (cm)	Material	Dating method	OSL age^3 (ka±1σ)	Conventional ^14C age^4 (yr BP±1σ)	Calibrated ^14C age range^5 (cal. yr BP at 95.4%)
Whine 1	2711942.6	6014689.2	91.4	E	70	15	clay rich silt	OSL	33.5±3.0
Whine 2	2711942.6	6014689.2	91.4	E	50	35	fine sandy silt	OSL	15.1±1.4
Whine 3	2711942.6	6014689.2	91.4	E	14	71	charcoal	^14C		377±99	0–550 (95.4%)
Whine 4	2711986.2	6014638.9	87.4	E	98	12	silty clay	OSL	10.1±0.9
Whine 5	2711986.2	6014638.9	87.4	E	68	42	well sorted sand	OSL	19.8±1.6
Whine 6	2711986.2	6014638.9	87.4	E	58	52	wood from log	^14C		173±33	0-280 (95.4%)
Wai-1b	2711421.3	6014716.3	∼120	B	160	−87	coarse sand	OSL	37.9±2.4
Wai-10a	2712228.2	6014600.1	∼100	B	46	10	well sorted silt	OSL	14.2±1.3
Wai-10b	2712228.2	6014600.1	∼100	B	31	25	silt	OSL	7.4±0.6
Wai-11a	2711421.3	6014716.3	∼120	B	73	7	silt	OSL	19.3±1.5
Wai-11b	2711421.3	6014716.3	∼120	B	51	29	silt	OSL	26.8±2.2
MR1	2713528.0	6014005.9	89.9	A	60	–	stony loess	OSL	10.0±0.8
WH1	2711952.1	6015012.2	122.6	A	65	–	stony loess	OSL	13.0±0.9
WH2	2711906.4	6015017.6	123.0	A	77	–	stony loess	OSL	10.2±1.2
^1The last three samples in this table are from Wang & Grapes (2007)
^2Terrace A is equivalent to the Waiohine surface
^3OSL technique used: Infrared stimulation of finegrain feldspars, 410 nm detection band, multiple aliqout additive method with latelight subtraction
^4Conventional age as per Stuiver & Polach (1977). This is based on the Libby half-life of 5568 yr with correction for isotopic fractionation applied. Quoted errors are 1 standard deviation due to counting statistics multiplied by an experimentally determined Laboratory Error Multiplier
^5Radiocarbon ages were calibrated to calendric years before 1950 using OxCal v. 3.10 of Bronk Ramsey (2005) incorporating Southern Hemisphere data from McCormac et al. (2004)

The above OSL results are generally older than the expected abandonment ages of terraces B and E, given the most recent estimates for the abandonment age of terrace A (Waiohine surface) of c. 10–12 ka (e.g. Formento-Trigilio et al. 2003; Wang & Grapes 2008; Little et al. 2009). This age overestimation may have resulted from incomplete resetting of the OSL content of the grains during their final fluvial deposition, which is probably caused by a very short sediment transport distance (e.g. perhaps only a few hundred metres) or by turbid water or dim daylight conditions. Without having been exposed to sufficient light, the feldspar grains carry a remnant age lower or equal to their previous sedimentation age; this equates to the age of one of the surrounding older fluvial aggradation terraces where the grains are sourced by erosion. Sample Wai-11b (Table 4) was the only sample that significantly failed the plateau test, showing clear signs of partial bleaching. It must be noted that the plateau test can only detect insufficient bleaching if most grains in a sample have seen at least some light (although not enough to reset the OSL clock). In a grain mixture where some of the grains have seen plenty of light and others none, the plateau test must fail even though we are still dealing with an age overestimation due to partial bleaching (Duller 1994; Duller et al. 1995). We conclude that most of our samples fall into the latter category and that their ages are overestimates for the abandonment age of the terrace they were collected from. For example, sample Wai-11a yielded an age for terrace B that was almost twice the age of samples collected from terrace A at this site (Wang & Grapes 2008). The ages obtained from ¹⁴C samples collected from terrace E were modern, despite ages between 10.1±0.9 ka (1σ) and 33.5±3.0 ka (1σ) being obtained for OSL samples from that same terrace (Table 4). These examples support the likelihood that the OSL samples were incompletely bleached. The stratigraphically reversed OSL ages of samples Whine 4 and Whine 5 suggest that these older ages are not simply due to sampling of the older aggradation terraces underlying the degradation terrace deposits.

For these reasons, we interpret the OSL age of sample Wai-1b (37.9±2.4 ka (1σ)) to be a maximum age estimate for the end of the aggradation of the Waiohine terrace. Similarly, the OSL age of sample Wai-10a collected from terrace B (14.2±1.3 ka (1σ)) is thought to provide a maximum age for the abandonment of terrace B. This sample was taken from just above the gravel tread of terrace B, and was therefore deposited soon after the abandonment of the gravel tread.

On the other hand, sample Wai-10b with an OSL age of 7.4±0.6 ka (1σ), if completely bleached, is interpreted to provide a minimum estimate of the abandonment age of terrace B, as this sample was collected c. 50 cm above the gravel tread of terrace B. However, given that the other OSL samples from terrace B appear to be incompletely bleached, the abandonment age for terrace B from sample Wai-10b must be treated as provisional.

Discussion

Interpreting progressive fault displacements at the Waiohine terraces under a complete riser trimming model

Our study at the Waiohine terraces lends support to a model of complete riser trimming, both in map view and in vertical section (Fig. 3B & 4B). Support for this inference comes from the following features.

1.	The absence of any horizontal displacement in the modern river bank or of any vertical displacement in the modern river bed, indicating that the 1855 scarp in the Waiohine River has been completely removed by the river (map view and vertical section) in the c. 150 years since that earthquake.
2.	The equivalence of the dextral displacements of riser EF and palaeochannel Ch2 incised into terrace F immediately below riser EF (Table 2, Fig. 10A), in accordance with Cowgill's (2007) second index which compares the similarity of riser and terrace tread displacements (Table 1), which suggests that riser EF only began recording displacement after its lower flanking terrace F became abandoned.
3.	The absence of any risers that are curved in plan view in a sense that is consistent with the sense of slip on the fault, in accordance with the fourth index of Cowgill (2007). At the Waiohine terraces, none of the risers are sigmoidally curved in a dextral sense that would be consistent with the preservation of an incompletely trimmed dextral displacement.
4.	The differences in riser heights across the fault are equal to the vertical displacements of the riser's lower bounding terraces across the fault, as would be expected for a case of complete scarp removal by the active river after every earthquake (e.g. Fig. 4B).

A close examination of the models for progressive fault displacement of flights of degradation terraces shows that (1) the difference in terrace width across the fault will, in each model, be equal to the incremental displacement for that terrace and (2) a distinction between the different models cannot be made on the basis of a difference in terrace width alone. Following the second index of Cowgill (2007), the most appropriate trimming model for any flight of displaced river terraces should be determined from a comparison between the displacements of palaeochannels incised into the terrace surfaces and the terrace riser displacements.

Single-event displacements on the central Wairarapa Fault at Waiohine River

The displacements measured at the Waiohine terraces are used to interpret the number of earthquakes and the magnitude of co-seismic displacements occurring on the Wairarapa Fault since terrace formation. The vertical displacement of terrace F (1.30±0.02 m (2σ)) and the dextral displacements recorded by riser EF (11.4±1.7 m (2σ)) and Ch2 incised into terrace F (12.4±0.8 m (2σ)) are interpreted to represent co-seismic displacements that were generated during the 1855 M_w >8.1 earthquake on the Wairarapa Fault. The uncertainty in the dextral displacement of Ch2 is smaller than that of riser EF (Table 2), thus the displacement of Ch2 is assumed to more faithfully record the 1855 dextral displacement at the Waiohine terraces.

The incremental vertical displacement between terraces E and F of 1.36±0.06 m (2σ) and the incremental dextral displacement between riser DE and Ch2 of 9.7±1.7 m (2σ) are interpreted to represent displacements that were generated during the penultimate earthquake on the Wairarapa Fault. Twenty kilometres further to the south, the penultimate earthquake timing was most recently constrained as 920–800 cal. yr BP (Little et al. 2009). From the 1855 and penultimate single-event displacements, we infer a mean single-event dextral slip of 10.6±2.6 m (2σ) and a mean single-event throw of 1.35±0.07 m (2σ). The penultimate incremental displacements are similar to those attributed to the 1855 earthquake, suggesting that the last two events on the Wairarapa Fault had ruptures of comparable dimensions (>120 km) and magnitude (M_w c. 8) (Rodgers & Little 2006).

Given the above mean single-event dextral displacement, we infer that the dextral slip of riser BC (73.2±2.3 m (2σ)) resulted from an additional 4–6 earthquakes prior to the penultimate earthquake. Considering the mean single-event throw and the vertical displacement of terrace D (11.85±0.03 m (2σ)), we infer that it records slip from an additional 4–6 earthquakes that occurred after the abandonment of terrace D. Terrace C (vertically displaced by 12.99±0.03 m (2σ)) may record displacement as a result of one additional earthquake to those recorded by terrace D, occurring in the time interval between terrace C and terrace D's abandonment. Given that the dextral displacements of riser CD and riser BC are within error, it is possible that terraces C and D have experienced the same total number of earthquakes. In this case, the apparent difference in their vertical offset would have to be attributed to the lateral displacement of palaeotopography on terrace C to yield a locally higher scarp there. Considering the H/V ratio for the Wairarapa Fault (following section), it is unlikely that an earthquake on the Wairarapa Fault would result in c. 1.2 m of vertical displacement without any dextral displacement. We also infer that the vertical displacement of terrace B and the dextral displacement of riser AB occurred as a result of approximately two more earthquakes than were experienced by terrace C, and that terrace A experienced two or three more earthquakes than terrace B experienced (Table 2).

Horizontal to vertical slip ratios for the central Wairarapa Fault at Waiohine River

Based on the above-cited set of progressive (incremental) displacements, we infer that the ratio of horizontal to vertical slip (H/V) on the Wairarapa Fault has ranged from 5.0 to 10.2 for the Waiohine River terraces (Fig. 11) with a mean H/V of 6.9±3.5 (2σ). The slope of each graph segment indicates the incremental H/V slip ratio for each terrace (Fig. 11). An H/V slip ratio could not be calculated for terrace A (Waiohine surface) as no dextral displacement was measured for that terrace. It was also impossible to calculate an H/V ratio for terrace D/riser CD due to the large uncertainty in the dextral displacement of riser CD. Instead, a combined H/V ratio for terrace C/riser BC and terrace D/riser CD was calculated. The high H/V slip ratios demonstrate the dominantly strike-slip nature of the Wairarapa Fault. Our range of H/V slip ratios for the Waiohine terraces on the central part of the Wairarapa Fault is consistent with the H/V slip ratio of c. 6–9 reported for the southern part of the fault by Rodgers & Little (2006).

Although we did not identify any palaeochannel incised into terrace A, we can estimate the displacement of terrace A since its abandonment by assuming that the average H/V slip ratio for the Waiohine terraces is representative of the H/V ratio for post-terrace A displacements. Using the average H/V slip ratio for the Waiohine terraces of 6.9±3.5 (2σ) and the cumulative vertical displacement for terrace A of 19.67±0.09 m (2σ), we infer a finite dextral displacement of 129±18 m (2σ) for terrace A (Fig. 11).

Figure 11  Cumulative vertical versus dextral displacements across the Wairarapa Fault at the Waiohine terraces. The slope of each graph segment indicates the incremental H/V slip ratio for each terrace. Error bars are 2σ (95% confidence), error bars for vertical displacements are too small to be shown. The average H/V ratio for the Waiohine terraces is 6.9±3.5 (2σ). By projecting the vertical displacement of terrace A (Waiohine surface) of 19.67±0.09 m (2σ) across to intersect the average H/V ratio projection, a cumulative dextral displacement for terrace A of 129±18 m (2σ) is obtained.

Late Quaternary slip rate for the central Wairarapa Fault at Waiohine River

Two terraces at the Waiohine River provide displacement and age data that can be used to determine the Late Quaternary slip rate on this central part of the Wairarapa Fault. Under a complete riser trimming model, the displacement of a terrace riser is most precisely dated by the abandonment age of that riser's lower bounding terrace. The dextral displacement of riser AB is precisely known; however, the age of terrace B is only poorly constrained by results of OSL dating in this study. Conversely, the age of terrace A (Waiohine surface) is better constrained by OSL dates from Wang & Grapes (2008), while the displacement for this terrace is inferred from the vertical displacement of terrace B and the average H/V ratio for the terraces.

Using our inferred dextral displacement for terrace A, 129±18 m (2σ), together with an age for terrace A of 11.1±0.6 ka (1σ) calculated as the average of the three Wang & Grapes (2008) OSL ages for silts deposited on that surface in the Waiohine River area, we calculate a Late Quaternary dextral slip rate of 11.9±2.9 mm a⁻¹ (2σ) (Fig. 12). The vertical displacement of terrace A is 19.67±0.09 m (2σ); together with this average age for terrace A, a Late Quaternary vertical slip rate of 1.8±0.2 mm a⁻¹ (1σ) is implied (Fig. 12).

Our measured cumulative displacement of riser AB, 101.1±3.4 m (2σ), together with the above-mentioned average age for terrace A as a maximum age for the abandonment of terrace B's gravel tread, yields a new minimum Late Quaternary dextral slip rate estimate of 9.2±1.3 mm a⁻¹ (1σ) (Fig. 12). The vertical displacement of terrace B is 15.74±0.04 m (2σ) which, combined with this maximum age for terrace B, implies a minimum Late Quaternary vertical slip rate of 1.4±0.2 mm a⁻¹ (1σ) (Fig. 12). Our measured cumulative displacement of riser AB, together with the OSL age of sample Wai-10a of 7.4±0.6 ka (1σ) as a minimum age for the abandonment of terrace B (assuming this sample is completely bleached), yields a provisional maximum Late Quaternary dextral slip rate of 14.1±2.7 mm a⁻¹ (1σ), and a provisional maximum vertical slip rate of 2.2±0.4 mm a⁻¹ (1σ) (Fig. 12).

Figure 12  Cumulative dextral and vertical displacements across the Wairarapa Fault versus terrace age for terraces A and B at Waiohine River. Black dots are dextral displacements and grey dots are vertical displacements. Error bars are 2σ (95% confidence). Error bars for vertical cumulative displacements are too small to be shown.

A number of recent studies have quantified the single-event dextral displacements and placed constraints on the Late Quaternary slip rates for different sections of the Wairarapa Fault (e.g. Van Dissen & Berryman 1996; Schermer et al. 2004; Rodgers & Little 2006). Variations in slip rate and single-event displacement northwards along the Wairarapa Fault are the subject of another manuscript.

Conclusions

New fault displacement and age data collected in this study at the Wairarapa Fault in the Waiohine River area provides new insights into the geomorphic expression and rate of deformation along this active strike-slip fault. Our findings are summarised by the following conclusions.

1.	A model of complete riser trimming is most appropriate for the Waiohine River terraces (based on similar offset of palaeochannels and adjacent risers). From this, we infer that a riser begins recording displacement when its lower bounding terrace tread is abandoned. The most appropriate trimming model for any flight of displaced river terraces should be determined from a comparison between palaeochannel and riser displacements. A comparison between riser displacements and the difference in terrace widths across the fault cannot be used to distinguish between trimming models.
2.	The magnitudes of the inferred 1855 (smallest displacement) and penultimate (next-smallest displacement) single-event displacements at the Waiohine terraces are 12.4±0.8 m (2σ) and 9.7±1.7 m (2σ), respectively, considerably lower than the slip estimates of 15–18.5 m measured on the southern part of the Wairarapa Fault by Rodgers & Little (2006).
3.	Based on the age and inferred displacement of terrace A, the Late Quaternary dextral slip rate for this part of the central Wairarapa Fault is 11.9±2.9 mm a−1 (2σ). Moreover, from the displacement and maximum age for terrace B, we calculate a minimum slip rate of 9.2±1.3 mm a−1 (2σ). From the displacement and minimum age for terrace B, we calculate a provisional maximum slip rate of 14.1±2.7 mm a−1 (1σ). The ratio of horizontal to vertical slip ranges from 5.0 to 10.2 for the Waiohine River terraces, with an average of 6.9±3.5 (2σ).

Acknowledgements

We thank David Rhoades, Russ Van Dissen and Pilar Villamor for insightful discussions regarding statistical analysis of terrace ages and Wairarapa Fault tectonics, Ningsheng Wang for the preparation and processing of OSL samples, Mark Henderson for technical help with GPS data collection and processing and Hannu Seebeck, Dave Murphy, Dee Ninis, Keely Schreiber and the Colorado College undergraduate field trip students for their efforts surveying. We also thank Dave Rodgers and Pilar Villamor for helpful comments on an earlier draft, Nicola Litchfield and Mark Quigley for their extensive manuscript reviews and Simon Cox for his exceptional editorial efforts.