Distribution, microphysical properties, and tectonic control of deformation bands in the Miocene accretionary prism (Whakataki Formation) of the Hikurangi subduction zone

We analyse deformation bands related to both horizontal contraction and horizontal extension in Miocene 10 turbidites of the Whakataki Formation south of Castlepoint, Wairarapa, North Island, New Zealand. In the Whakataki Formation, four sets of cataclastic deformation bands are identified: [1] normal-sense Compactional Shear Bands (CSBs); [2] normal-sense Shear-Enhanced Compaction Bands (SECBs); [3] reverse-sense CSBs; and [4] reverse-sense SECBs. During extension, CSBs form most frequently with rare SECBs. Extensional CSBs are often, but not exclusively, associated with normal faults. During contraction, distributed SECBs are observed 15 most commonly, sometimes clustering around small reverse faults and thrusts. Contractional CSBs are primarily found in the damage zones of reverse faults. The quantitative spacing analysis shows that most outcrops are characterised by mixed spatial distributions of deformation bands, interpreted as a consequence of overprint due to progressive deformation or distinct multiple generations of deformation bands from different deformation phases. Since many deformation bands are parallel to adjacent juvenile normaland reverse-faults, bands are likely 20 precursors to faults. With progressive deformation, the linkage of distributed deformation bands across sedimentary beds occurs to form through-going faults. During this process, bands associated with the wall-, tip-, and interaction damage zones overprint earlier distributions resulting in complex spatial patterns. Regularly spaced bands are pervasively distributed when far away from faults. Microstructural analysis shows that all deformation bands form by inelastic pore collapse and grain crushing with an absolute reduction in porosity 25 relative to the host rock between 5 and 14%. Hence, deformation bands likely act as fluid flow barriers. Faults and their associated damage zones exhibit a spacing of order ten metres on the scale of 10 km and are more commonly observed in areas characterised by higher mudstone to sandstone ratios. As a result, extensive clay smear is common in these faults, enhancing the sealing capacity of faults. Therefore, the formation of deformation bands and faults leads to progressive flow compartmentalisation from the scale of ten metres down to about ten 30 centimetres, the typical spacing of distributed deformation bands.


105
The onset of subduction is expressed in the sedimentary record by the wide-spread deposition of olistostrome deposits in the earliest Miocene (Chanier and Ferrière, 1991;Bailleul et al., 2007;Bailleul et al., 2013), which also define the base of the studied Whakataki Formation. The basal olistostrome is overlain by deep-marine deposits to higher-energy flysch deposits (Neef, 1992a), consisting of a succession of laterally continuous finegrained sandstone and siltstone turbidite deposits with an estimated thickness of 900-1500 m (Neef, 1992a;Field, 110 2005;Bailleul et al., 2013).
The present study is based on field observations and samples collected from coastal exposures of the Miocene Whakataki Formation 5 km southwest of Castlepoint, in the Wairarapa region of the North Island, New Zealand  (Neef, 1992a;Morgans, 1997;Field, 2005;Bailleul et al., 2013)) was deposited within 115 tectonically controlled, confined basins on the lower trench-slope of the subduction margin (Bailleul et al., 2007;Bailleul et al., 2013). The sediments preserve a record of not only the onset of subduction ca. 25-19 Ma (D1), but also a period of extension ca. 15-5 Ma (D2) and renewed contraction from the Pliocene to recent (D3) (Chanier and Ferrière, 1991;Chanier et al., 1999;Bailleul et al., 2007;Bailleul et al., 2013). Each deformation phase induced corresponding deformation structures in the field area ( Fig. 1, Fig. 2): 120 D1: margin-perpendicular contraction expressed by both landward and seaward emplacement of thrust-sheets, gentle folding, and reverse faulting along the Hikurangi margin, with the development of trench-slope basins bound by structural highs (Bailleul et al., 2013;Chanier and Ferrière, 1991;Chanier et al., 1999;Maison et al., 2018;Rait et al., 1991). The main D1 structure in the field area is the NE-SW trending Whakataki thrust fault that has emplaced the pre-Miocene Whangai Formation onto the Miocene Whakataki Formation and acts as the 125 western boundary to the Whakataki Formation in the field site (Fig. 2).
D3: A renewal of dominant margin-perpendicular contraction of the wedge with associated folding, thrusting, and strike-slip faulting (Chanier et al., 1999;Nicol et al., 2002;Bailleul et al., 2007;Nicol et al., 2007;Bailleul et al., 2013). Also associated with this deformation phase is the uplift and exposure of the Coastal Ranges (Nicol et al., 2002). In the field site, this deformation phase is expressed by pervasive deformation bands, small thrust faults, 135 and intense folding. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. 145 the shortening direction, estimated from the pole to the average fold axial plane. i) Poles to D1 deformation bands. j) Poles to D2 deformation bands. k) Poles to D3 fault-damage-zone-associated deformation bands. l) Poles to D3 deformation bands that are not located in fault damage-zones. The similarity of the fault and fault-associated deformation band orientations indicates that they formed in the same stress field. Also, axial plane data from D3 aligns with D3 non-fault damage-zone associated bands as shown in (l). All deformation band and fault data are back tilted 150 as there is evidence of folding occurring coeval to, or later than all events. Stereoplots were produced using Stereonet 10 software (Allmendinger et al., 2011). Paleostress analysis completed using WinTensor (Delvaux and Sperner, 2003).

2.2.
Classification and microstructures of deformation bands Deformation bands are the most common strain localisation structure observed in deformed porous rocks (Fossen et al., 2007). The bands typically have widths in the range of micrometres to centimetres and lengths most 155 commonly > 100 m (Aydin, 1977;Antonellini et al., 1994;Aydin et al., 2006;Fossen et al., 2007). Associated displacement, if any, is often on the order of mmcm. The displacement associated with deformation bands is less than that associated with faults/slip surfaces, permitting a distinction between the two. Deformation bands can be broadly classified in two ways; [1] by deformation mechanism and [2] by kinematics (Fossen et al., 2007).
[1] The dominant deformation mechanisms observed in deformation bands are [a] granular flow; [b] cataclasis; 160 [c] phyllosilicate smearing; and [d] dissolution and cementation (see Fossen et al. 2007 for review). The deformation mechanism depends on the grain size, sorting, mineralogy, diagenetic history, porosity and stress state (Fossen et al., 2007) and controls petrophysical properties (Fossen and Bale, 2007;Ballas et al., 2015;Fossen et al., 2018). Disaggregation bands are associated with minor grain fracture, resulting in bands with porosities and permeabilities comparable to the host rock. However, bands formed dominantly by cataclastic deformation exhibit 165 considerably lower porosities and permeabilities compared to the host rock. Dependent upon the host rock mineralogy, porosity can be reduced by one to two orders of magnitude, and permeability can be reduced by up to six orders of magnitude with phyllosilicate-bearing sandstones generally characterised by larger reductions (Fossen et al., 2007). Field and laboratory observations show that deformation bands are most commonly characterised by a reduction in both porosity and permeability compared to the host rock (Aydin, 1978;Aydin 170 and Johnson, 1978;Underhill and Woodcock, 1987;Antonellini et al., 1994;Antonellini and Aydin, 1995;Fossen and Bale, 2007;Fossen et al., 2007;Torabi et al., 2013;Ballas et al., 2015;Fossen et al., 2018).
[2] Kinematically, deformation bands can be classified as compaction, dilation, shear, and hybrid bands, combining shear with compaction or dilation (Aydin et al., 2006;Fossen et al., 2007). The most common type in nature is Compactional Shear Bands (CSBs) (Fossen et al., 2007). Here, we use the kinematics-based classification 175 system because it is most readily applied in the field. Deformation mechanisms are best assessed with microscopic methods, also done in this paper. Microstructural characteristics of deformation bands are intimately related to their kinematics and are summarised in Table 1. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

2.3.
Spatial distribution of deformation bands: controls of tectonic stress regime? Field studies demonstrate that deformation bands exhibit two spatial distributions; [1] networks of bands concentrated into clusters or zones in the vicinity of faults (Antonellini and Aydin, 1995;Cowie, 2001, 2003;Solum et al., 2010;Soliva et al., 2013;Ballas et al., 2014) and [2] a pervasive periodic and/or clustered 185 distribution across a deformed region, unrelated to faults (Saillet and Wibberley, 2010;Solum et al., 2010;Ballas et al., 2013;Soliva et al., 2013;Soliva et al., 2016). The distribution and type of deformation band that form have been linked to the tectonic stress regime (Saillet and Wibberley, 2010;Soliva et al., 2013;Ballas et al., 2014;Ballas et al., 2015;Soliva et al., 2016). In a thrust regime, periodically spaced bands with a compaction component ± shear were observed. In horizontal extension, clusters of bands in fault damage zones characterised by a decay 190 of density away from the fault plane and a large shear component ± dilation/compaction (volumetric sign dependent upon the effective mean stress) are observed. The Whakataki Formation contains deformation bands associated with horizontal extension and horizontal contraction. By analysing the distribution of deformation bands associated with each regime at Castlepoint, we test if the tectonic stress regime controls the spatial distribution of the observed bands. 195

A conceptual mechanical model for deformation bands
The kinematics and orientation of deformation bands are commonly explained through rate-independent plasticity theory. The broadly accepted approach is called the cam-cap model of yielding and band formation ( Fig. 3) (Wong et al., 1992;Schultz and Siddharthan, 2005;Fossen et al., 2007). The cam-cap yield surface often matches the failure response of brittle granular media (e.g., Wong and Zhu, 1999). Triaxial deformation experiments on porous 200 media show that their yield is a non-linear surface in Q-P-space, with a positive slope at low mean stresses and a negative slope at higher mean stresses (Fig. 3) (Wong and Zhu, 1999;Rudnicki, 2004;Karner et al., 2005). Q and P signify differential stress and effective mean stress, respectively: where σ1, σ2 and σ3 denote the principal stresses of the Cauchy stress tensor, and Pf is pore-fluid pressure.
Kinematically, the cam-cap yield surface can explain the continuous transition from pure dilation bands to pure compaction bands, via shear bands, with its change in slope ( Fig. 3) (Bésuelle, 2001). The Q-P diagram links the state of stress at the point of inelastic yielding to the kinematic type of deformation band that will form and, broadly, the orientation of that band in relation to σ1 (Fig. 3) (Fossen et al., 2007;Schultz, 2019). Dilation and 210 compaction band orientation can be adequately predicted in Q-P space. However, the orientation of bands with https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. associated shear is less adequately predicted because the Coulomb criterion, used for predicting orientation, does not consider volumetric changes across a shearing surface (Schultz, 2019). If the orientation and band kinematics are known, the intersection of the yield envelope can be predicted. Current literature hypothesises that the point of intersection is controlled by the tectonic regime, and therefore, the orientation and band kinematics are 215 controlled by the tectonic regime (Soliva et al., 2013;Ballas et al., 2014;Soliva et al., 2016;Fossen et al., 2018).

Field Data
Field mapping took place in a 17 km 2 area south of Castlepoint (coordinates: NZGD 2000, UTM Zone 60S, E: 0430670, N:5468211), in the Wairarapa region of the North Island, New Zealand ( Fig. 1). While significant hinterland mapping took place, coastal outcrops were the focus of the fieldwork because exposure is poor in the 225 hinterland. A detailed sedimentological and facies analysis was conducted on sediments of the uplifted Hikurangi accretionary prism to contextualise the structural data. It will not be included in this paper and will be presented elsewhere. Orientations of structural elements including bedding planes (S0), faults, and deformation bands were taken across the area and plotted in lower-hemisphere, equal-area stereograms (Allmendinger et al., 2011). All data are shown with bedding restored to horizontal. The restoration was completed by rotating back from the 230 associated S0 measurement of the bed hosting the structures using Stereonet 10 (Allmendinger et al., 2011).
Individual S0 measurements were used rather than fold axes because the folds are non-cylindrical and plunge gently. All data are restored as there is evidence of rotation in almost all features (Fig. 5). Paleostress analysis of back-rotated faults was completed using Win-Tensor (Delvaux and Sperner, 2003). For this analysis, only faults https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. with unambiguous slip vectors and shear sense were used. The PBT-kinematic-axes method was used to find the 235 orientation of the principal stresses. An assessment of whether faults and deformation bands form a conjugate (bimodal) or polymodal pattern was conducted using a statistical test developed by Healy and Jupp (2018). The test analyses the orientation distribution of the poles to fault planes to distinguish between bimodal and quadrimodal patterns. Differentiation between bimodal and polymodal is established by first finding the ratio between the eigenvalues of the second-rank orientation tensor of the datasets and secondly by calculating the p-240 value. P values close to 0 describe a polymodal pattern with higher values reflecting bimodal patterns. The dihedral angle, the angle bisected by σ1 (Chemenda et al., 2012), between conjugate sets of deformation bands was estimated from a cylindrical best fit to orientation data. The ratio of net shear/compaction (Ds(net)/Dc) was calculated using the methods from Soliva et al. (2013) and Ballas et al. (2014). These values are used to define the kinematics of the deformation bands. However, these authors assumed that they cut their samples in the plane 245 spanned by the smallest and largest principal strains, the orientation of which they determined from the rather tightly bimodal orientation patterns of the studied bands. In our case, this approach is more difficult because the orientation distributions are less focussed (Fig. 2). Therefore, typical geometric section errors in terms of true bandwidth and true shear displacement can be expected.

Spacing Analysis 250
The spacing analysis was conducted on two scales in the mapping area: [1] on macroscopic faults with > 20 cm offset exposed along a 5 km stretch of coastline, and [2] on deformation bands at individual outcrops exposed in sandstone units. For macroscopic fault spacing, the location, dip and dip direction, heave and throw, where possible, and shear-sense were recorded, and faults plotted onto a map. The spacing of macrofaults was measured along scan-lines oriented perpendicular to the average fault strike using the ruler tool in ImageJ (Schindelin et al., 255 2012). The scan-lines were 200 m in length in an E-W orientation (orientation perpendicular to the average fault orientation) and had a 50 m N-S spacing. Due to the oblique relationship between the coastline and the fault strike, scan-lines were shifted N after 400 m, with the first scan-line located in the SW of the coastline. Twenty scanlines were used to measure the spacing, and the median is reported. The use of multiple scan-lines minimises, but does not eradicate, measurement errors arising from variation in fault strike and an average strike being used to 260 generate the scan-line orientation. Spacing greater than 20 m was removed from the data as these points represent artefacts of exposure conditions. Horizontally measured spacing was corrected for the average dip of the faults using the sine transformation (Eq. 3).
True spacing = spacing × sin (average dip angle) The correction procedure also introduces errors because the faults do not all dip the same way. Moreover, they 265 were not rotated to their original position and orientation in space. However, these are common shortcomings of studying fracture spacing in 2D (Soliva and Benedicto, 2005;Laubach et al., 2018). Here, we are mainly interested in the general shape of the spacing distribution and the order of magnitude of absolute spacing. Pearson correlation coefficient values were calculated for the correlation between the corrected spacing and the associated cumulative field overprinting criteria were used to identify the older set, which was analysed (Fig. 5). The older set was chosen as these bands formed in rock with properties closer to the undeformed state. Changes in rock properties that occur during deformation band initiation are more likely to influence the spatial position of a later set Regenauer-Lieb et al., 2013a). 280 For spacing measurements of deformation bands in damage zones, the fault was put into the origin of the digitised image. For non-FDZ bands, with apparent 'regular' spacing, the first band was treated as the origin of the image.
Deformation band clusters were treated as single bands . The resulting maps were exported as a binary image. Matlab (Mathworks, 2011) was used to obtain spacing statistics along scan-lines at a spacing of 1% of the image height. When the horizontal image dimension was not perpendicular to the traces of the 285 deformation bands in the map, spacing was again corrected with the sine transformation (Eq. 3). When normalised spacing is reported for damage zones, normalisation was done by division through damage-zone width.
In addition to the analysis of natural deformation band distributions, synthetic images were created to show ideal spacing distributions and to highlight how natural variation in heterogeneous rocks and data collection error can impact the measured spacing of natural samples. Six synthetic images were produced: 290 [1] Deformation bands with constant spacing; [2] Deformation bands with constant spacing and added noise to replicate measuring bias and outcrop conditions. The noise in image [2] was generated by adding an array of random values, between 0 and 0.8 of the constant spacing value, to the space. The random values are collectively characterised by a normal distribution and a median value of 0; 295 [3] Deformation bands spatially characterised by an exponential spatial decay away from a fault; [4] Deformation bands characterised by an exponential spatial decay away from a fault with integrated noise. Noise in the image [4] is generated by adding values up to 0.4 of the maxima spacing onto each spacing measurement, with added values collectively characterised by a normal distribution with a median value of 0; 300 [5] Deformation band spacing that reflects the overprint of an equally spaced distribution [1] by an exponentially decaying damage zone [3]; and [6] Deformation band spacing reflecting the overprint of an equal distribution with integrated noise [2] by an exponentially decaying damage zone also containing integrated noise [4].

Microstructural Analysis 305
Twenty polished thin-sections from samples of host rock and deformation band were examined by back-scattered electron (BSE) imaging using a VP Zeiss Sigma scanning-electron microscope (SEM). A 10 nm carbon-coat was applied to the samples. The instrument was run using a working distance of 8 mm, an acceleration voltage of 15 kV, an aperture size of 30 µm with an angle-selected backscattered detector. Images of host rock were taken at 400 x magnification and those of deformation bands at 800 x. Multiple images were taken with a 5-10% overlap 310 and stitched together to produce ca. 1 mm 2 image. Of the 20 samples, 5 were taken from zones showing a 'regular' spacing of D3 deformation bands with zero to minimal apparent offset. The remaining 15 samples are from fault https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. damage zones, 10 from D2 damage zones, and 5 from D3 damage zones (Fig. 1). For each sample, the host rock and deformation band were analysed in the same thin section. Samples were selected from different locations along the coastline in attempts to analyse a broad range of lithologies. Sample locations of bands with a 'regular' 315 spacing are more clustered as these were the most appropriate outcrops with good exposure of the facies hosting non-FDZ bands.
BSE maps were analysed for porosity and mineralogy of the samples. Grey-level slicing was used to extract four different phases from the images [1] porosity, [2] quartz, [3] feldspar, and [4] 'other' which includes reflective oxide components, micas, and clays. For porosity phase analysis, the images were eroded and dilated to generate 320 upper and lower bounds for the estimate (Fig. 4) (Liu and Regenauer-Lieb, 2011). Grain size estimates were obtained through manual tracing of grains from SEM images in ImageJ (Schindelin et al., 2012). An average of 100 intact grains were traced for the host rock and the deformation band of each sample, and the equivalent circle diameter was calculated for each polygon. Microfracture density was measured on BSE maps along 1-mm long scan-lines oriented normal to the deformation bands, as in Ballas et al. (2013). Along each scan-line, intercepted 325 microfractures were counted to produce a fracture/mm count. Twenty scan-lines were drawn with a spacing of 0.1 mm in images, and the results were averaged for deformation bands and for host rock. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

4.1.
Rock Descriptions and Structure 335 All three regional deformation phases discussed above (Setting 2.1) can be recognised in the Whakataki Formation. Expressions of the deformation in the Whakataki Formation include folds, faults, and deformation bands. Deformation bands are most common within sandstone units in areas characterised by equal sandstone to mudstone ratios, or areas where sandstone beds are dominant. In areas dominated by mudstone, bands are less common. 340 In the following, we describe structures associated with each tectonic phase. Throughout this description, we use Anderson's nomenclature (Anderson, 1951) where the normal-faulting regime reflects a (sub-)vertical σ1 (horizontal extension) and a thrust-faulting regime reflects a (sub-)horizontal σ1 (horizontal contraction). A summary of the results is shown in Table 2. 345

Faults and Folds
The main regional fault associated with D1 horizontal contraction is the lower Miocene NE-SW trending 350 Whakataki thrust fault ( Fig. 1) (Bailleul et al., 2013;Maison et al., 2018). Rare exposures of the thrust damage zone contain slip planes which indicate a shallow NW dip (30°). NE-trending km-scale folds in the hinterland define the topography (

Deformation Bands
D1 deformation bands are rare and are distinguished from D2 and D3 deformation bands through cross-cutting relationships (Fig. 5). The orientation distribution of D1 deformation bands can be considered bimodal with a pvalue of 0.5. D1 dihedral angle ranges from 65° to 84° with a mean of 73°. With bedding restored to horizontal, 365 deformation bands trend 037°. The set dipping SE has a dip angle of 40° and the set dipping NW dips at 26°.
When bands pass through beds with thin clay-rich layers, they have a dark colour without relief. In beds with lower clay content, bands show high to moderate relief and are lighter in colour (Fig. 5). The width of bands ranges from 0.2 to 0.5 mm. Bands can extend for metres along strike. Eye and ramp structures are present, however, single strands are most commonly observed (Antonellini and Aydin, 1995). Offset associated with the 370 bands is variable, with some characterised by minimal offset, yet others accommodate reverse shear offsets at the https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. millimetre scale. Conclusively, this data suggests that CSBs and Shear-Enhanced Compaction Bands (SECBs) formed during D1 horizontal contraction Fossen et al., 2018;Schultz, 2019). The rarity of D1 deformation bands does not permit for meaningful spacing analysis.

Structures and Relative Timing
Field overprinting criteria demonstrate that an extensional event followed D1 horizontal contraction (Fig. 5) (Chanier et al., 1999

Faults
Normal faults are the main structures associated with D2 ( Fig. 6). They trend NNW with an average dip angle of 69° (Fig. 2). Analysis of the fault pattern shows that it is polymodal with a p-value of 0.001 (Healy and Jupp, 2018   https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

Deformation Bands
D2 deformation bands are primarily observed in the damage zones of D2 faults. However, some single bands are observed between faults (Fig. 6). They are generally darker in colour compared to the host rock and show no or negative relief (Fig. 6). When suitable host-rock layers are present, clay smear is common (Fig. 6). When propagating through layers rich in shell hash, the bands show positive relief. The average trend of bands is 340°, 410 with two sets of poles spanning finite arcs dipping steeply NE and SW ( Fig. 2) (average dip angle is 75°). Band orientation ranges from ca. 40-60° to σ1 indicative of CSBs and SECBs. The bands most commonly occur as single bands; however, clusters restricted to sandstone beds are also observed. Clusters localise into single strands if they propagate into adjacent mudstone beds. Band thickness ranges from 0.1 cm to 0.35 cm, with an average thickness of 0.16 cm. No obvious variation in thickness is observed in the bands along their length, which typically 415 is on the order of tens of metres. Many D2 bands extend beyond the outcrop scale, making accurate estimates of length impossible. Bands hosted in areas dominated by sandstone, with only very thin mudstone intervals, extend beyond the outcrop scale through the mudstone. In areas characterised by thicker mudstone intervals, bands are restricted to sandstone intervals. Displacement associated with the bands generally ranges from 0 cm to 4 cm with the mean displacement ca. 1 cm. Bands without apparent offset are rare and may be an artefact of a sectioning 420 effect (Soliva et al., 2013). Pattern analysis of the band orientation distribution shows that bimodality can be rejected with a p-value of 0.01 (Healy and Jupp, 2018). The dihedral angle cannot be found. Therefore, the bands must be classified through microstructural analysis and characterisation. However, in many cases, band-parallel displacement is much larger than band width. So, the shear component is likely dominant kinematically indicating that most bands are CSBs, as confirmed microstructurally below. 425 https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. At the microscale, D2 bands are characterised by a reduction in grain size and porosity compared to the host rock which makes the bands easy to identify under the microscope (Fig. 7, Fig. 9). The bands show diffuse borders 440 with the surrounding host rock. The grain size distribution of deformation bands generally shows positively skewed distributions with a lower median value compared to the host rock indicating a reduction in grain size due to cataclasis (Fig. 8) (Fossen et al., 2007;Balsamo et al., 2010). The median grain equivalent circular diameter in deformation bands is 37 µm compared with 57 µm in the host rock, showing a 35% reduction (Fig. 8). Deformation bands also show a smaller range in grain size at 156 µm compared to 231 µm in the host rock, with the host rock 445 preserving larger grains. Deformation bands, compared to the host rock, are characterised by a high matrix content, due to grain size reduction, and a concentration of clay minerals permitting the distinction between the two (Fig.   7). The amount of matrix decreases from the centre to the outside of the bands and becomes almost non-existent in the host rock which is dominated by intact grains, with/without intragranular fractures, and pore space. There is an average absolute porosity reduction of 8% (from ca. 13% in the host rock to ca. 5%) in deformation bands 450 ( Fig. 9). This equates to 59% relative porosity reduction. Relict medium-sized pores (30-50 µm) are present within some bands, accounting for much of the remaining ca. 5% porosity. Quartz overgrowths are present in the host rock and deformation bands indicating active pressure-dissolution during the deformation history (Fig. 7).
Overgrowths are not abundant, yet they do contribute to the reduction in pore space. Using porosity reduction as a proxy for inelastic volumetric strain, we obtain a ratio of DS/DC of 65 on average indicative of CSBs (Soliva et 455 al., 2013;Ballas et al., 2014;Soliva et al., 2016;Fossen et al., 2018). Grain fracture is observed at grain contacts and within grains in both host rock and in deformation bands (Fig. 7). The microfracture density is greater in the host rock than in the deformation band. On average, 8.5 fractures/mm 2 are observed in host rock compared with 2.4 fractures/mm 2 in the deformation bands. Host rock microfracture density was measured within the same thin section as deformation bands, and therefore, is not representative of the entire unit. The presence of microfractures 460 in the host rock shows that the deformation is not solely concentrated within deformation bands. Due to a reduction in grain size, a significant reduction in porosity and a low density of microfractures, D2 deformation bands are classified as cataclastic CSBs (Antonellini et al., 1994;Mair et al., 2000;Fossen et al., 2007;Ballas et al., 2015).
The structures associated with D3 are: [1] folds; [2] reverse faults; [3] reverse-sense deformation bands located in 490 the damage zones of faults; and [4] deformation bands with constant spacing and minimal to no offset. Field overprinting criteria cannot be used to discern the relative timing of these structures during this contractional phase as all structures are not seen interacting in a single outcrop.

Faults and Folds
Upright, NE-SW trending folds with 100 m-scale wavelengths and along-strike extents of several km's constitute 495 the most obvious large-scale D3 structures (Fig. 1). Fold axes are shallow and doubly plunging NE and SW (Fig.   2). Unlike D1 folds, D3 folds are asymmetrical, with steeper eastern limbs and non-cylindrical geometry (Fig. 1).
Synclines commonly have open hinges and resemble box folds in their geometry. Anticlines resemble kink folds with closed hinges and long straight limbs. Higher-order parasitic folds with matching geometry are observed throughout the mapping area and best exposed in road cuts in the hinterland. Within fold limbs, layer-parallel slip 500 and shear can be observed in interbedded mudstone layers, along with thinning-out of these layers. Slip surfaces with slickenfiber veins exhibit opposing shear sense on opposing fold limbs. These observations could be associated with both bending and buckling folding mechanisms (Donath and Parker, 1964;Chapple and Spang, https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. 1974). Given the fact that D3 axial planes trend parallel to the traces of the dominant thrust faults in the area, it is reasonable to assume, by assessing the poles to the axial planes, that D3 folds formed by SE shortening (Grujic 505 and Mancktelow, 1995). This is consistent with the shortening direction indicated by regional thrust faults (Chanier et al., 1999). While D1 fold geometry is inconsistent with D3 fold geometry, field observations cannot confirm whether D3 folds nucleated during D1 and tightened during D3 or whether they nucleated and tightened throughout D3.
D3 reverse faults can be observed along the coastline (Fig. 1, Fig. 6 Paleostress analysis of back tilted D3 reverse faults with slickenfiber veins and shear sense is consistent with the literature and is in alignment with the present-day movement of the Pacific plate (σ1: 01/092°) (Fig. 2) (Chanier et al., 1999;Bailleul et al., 2013). We have analysed back-rotated D3 faults because deformation bands in the damage zone of these faults mutually crosscut conjugate D3 bands with no apparent offset, and orientation analysis of the latter shows that they have been passively rotated during folding (see Supplement Section S.4). However, 520 we cannot rule out that faulting could have occurred anytime during D3, especially in steep fold limbs when deformation can no longer be accommodated by folding. We do not have evidence for this as most of the fieldwork took place at coastal outcrops that were in the gentle to moderately dipping back-limb of a syncline.

Deformation Bands
When comparing dihedral angle and DS/DC ratio, two types of deformation bands are associated with D3: reverse-525 sense CSBs and SECBs. SECBs are the most common.

D3 SECBs
SECBs are mainly hosted in regions dominated by sandstone, with less interbedded mudstone. SECBs primarily form single bands but also occur in narrow clusters in areas with higher band densities. On average, single bands are 5 mm in width with a range from 0.26 mm -1 cm. Clusters are on average 2 cm wide. Eye and ramp structures 530 can be recognised in the networks. The bands are lighter in colour than the host rock and show positive relief (Fig.   6). SECBs exhibit no observable shear displacement, except where clusters are present and millimetre offset can be observed. Some bands pass through multiple beds and can extend beyond the scope of observation. Most commonly, however, SECBs are confined to individual sand-and siltstone beds where they form two sets of bands trending NNE (Fig. 2, Fig. 6). Overprinting relationships show that commonly one of the two sets predates the 535 other. From 11 analysed outcrops back tilting of deformation band orientations results in tighter clustering in Schmidt nets, suggesting that SECBs formed before folding initiated. In addition, bands orientations mapped on the same bed across a fold hinge lie on the same great circle suggesting that these bands predate D3 folding (see Supplement Section S.4). Pattern analysis shows that most SECBs have a bimodal orientation distribution and can be considered as conjugate sets (Healy and Jupp, 2018). The average dihedral angle is 82° (after back-rotation), 540 ranging from 68-89°. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.
Microscopically, SECBs have diffuse borders with the surrounding host rock. The bands are characterised by a reduction in grain size and porosity (Fig. 7). Host rock median grain size is 39 µm and deformation band median grain size is 22 µm, representing a 44% reduction (Fig. 8). The range in grain diameters inside and outside of the bands is similar: 157 µm for deformation bands and 162 µm for host rock. However, the interquartile range is 545 much less for deformation bands: 49 µm for host rock and 27 µm for deformation bands. This small range for deformation bands results in many outliers indicating relict, less fractured grains are present within the bands (Fig.   8). Grain size within the bands has a unimodal, positively skewed distribution, indicating cataclasis with a larger number of smaller grains. Positive skew suggests that not all grains have been equally fractured. The porosity within the deformation bands, on average, reduced from 18.5% in the host to 8.8%, a relative reduction of 52% 550 ( Fig. 9). Quartz overgrowths are observed around grains in the host rock and deformation bands. Like D2 bands, microfracture density is high in the surrounding host rock, 6.3 fractures/mm 2 in the host rock and 3.8 fractures/mm 2 in the deformation bands. Within the bands there is a cataclastic fine-grained matrix, which is generally absent in the host rock. Analysis of compaction versus shear shows an average DS/DC value of 0.83 for SECBs, ranging from 0.2 -1.5. This value shows that compaction and shear magnitudes are similar, and the bands 555 are therefore microscopically classified as cataclastic SECBs. CSBs when clay smear is present (Fig. 6). They mainly occur as single bands but also in clusters. Single bands are on average 5 mm wide and range from 3 -7.5 mm. Clusters are ca. 2-3 cm in width (Fig. 6). CSBs accommodate mm-cm scale reverse offset. The bands trend NNE with one poorly clustered dominant set dipping ESE (Fig. 2). The dominant set dips ESE at 63° and the less dominant set dips WNW at 29°. This orientation data 565 is very similar to D3 faults indicating that they formed in the same stress state. SECB orientation does not match with fault orientation as strongly (Fig. 2). The bands generally propagate for multiple meters and extend out of outcrop observation, not unlike D2 normal sense deformation bands. However, some are truncated against thicker mudstone intervals. Pattern analysis rejects bimodality in CSBs associated with faults, suggesting a polymodal orientation distribution (Healy and Jupp, 2018). However, CSBs observed between faults, similarly to SECBs, are 570 bimodal with dihedral angles ranging from 51-80°.

D3 CSBs
Microscopically, CSBs are characterised by diffuse edges (Fig. 7). These bands show the largest reduction in grain size of all deformation bands observed at the field site. Median host rock grain size is 57 µm compared to 21 µm in the deformation bands, a reduction of 63%. The total range in grain size is similar for the host rock and deformation bands, 228 µm for the host rock and 224 µm for the deformation band. However, the interquartile 575 range is considerably different, 54 µm for host rock and 22 µm for the deformation band. This variation is similar to that seen in the D3 SECBs. As with the other bands, D3 CSBs have a lower porosity than the host rock, 20% outside of the band and 12.3% inside the banda reduction of 39% (Fig. 9). This is the smallest reduction in porosity observed across the varieties of deformation band. Pore space in deformation bands is filled with fragments of grains, clay, and quartz overgrowths. Quartz overgrowths have also reduced host rock porosity. 580 Grains within the bands and the host rock show intragranular fractures radiating from grain contacts (Fig. 7). https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.
There is a higher density of microfractures in the host rock surrounding the bands, as is observed in normal CSBs and SECBs. Microfracture density in the host rock is 6.3 fractures/mm 2 , while the density in the deformation bands is 3.1 fractures/mm 2 . Compaction and shear analysis shows that the Ds/Dc value is 63 ranging from 32 -106. This, as with D2 bands, indicates considerably more shear strain and characterises the bands as CSBs. 585

Proto-Deformation Bands
Host rock surrounding D2 normal-sense CSB bands and both types of D3 deformation bands, reverse-sense CSB and SECB, is characterised by pockets of cataclastic material that can be observed to originate from the crushing of individual grains. Throughout thin sections, these pockets can be seen to align with the same orientation as the deformation bands (Fig. 7). 590

Synthetic Spacing
To explain the spacing analysis of natural data, six synthetic images representing different spatial distributions of deformation bands were constructed (Fig. 10). Bands with a strictly constant spacing show a zero Pearson correlation coefficient (Fig. 10). With the addition of normally distributed 'noise' to case [a], a distributed set of 595 spacing with Gaussian noise is obtained (Fig. 10). In this case [b], the Pearson correlation coefficient of spacing over distance is -0.08. The third synthetic image [c] represents a damage zone with exponential decay of deformation band density away from the fault plane. For this example, the Pearson correlation coefficient between spacing and distance to fault is exactly 1 (Fig. 10). A histogram of this spacing distribution shows positive skew.
The same results are expected for any other analytical expression in which spacing grows monotonously with 600 distance from the fault e.g.: the power-law relationship established by Savage and Brodsky (2011). With the addition of Gaussian noise to a damage zone [d], the Pearson correlation coefficient reduces to 0.56. However, its exact value depends on the amount of noise and can be larger or smaller. Synthetic image [e] represents an overprint of two distributions, equal spacing and variable spacing, simulating two different deformation events affecting the same bed subsequently. If the deformation bands resulting from the two events are morphologically 605 similar, they may be difficult to distinguish between in the field and can be mapped together as one. In this case, the Pearson value is 0.71, although it could be any value between 0 and 1 dependent upon the value of spacing in the background density compared with the damage zone spacing. In image [f], noise is integrated into both distributions and they are combined. This is the most realistic outcome if there is an overprint of two events with different distributions. In the case shown the Pearson value is 0.59, however, with different magnitudes of noise 610 and varied initial spacings, any value from -1 to 1 could be obtained. For the interpretation of our field data, we assume a positive correlation between band spacing and spatial location if the Pearson value > 0.5. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

Deformation Band Spacing
Qualitatively, deformation bands in the mapping area appear to exhibit two spacing distributions. Generally, those associated with faults (both normal-and reverse-sense) form fault damage zones with a variable distribution (  Non-FDZ Deformation Bands We examined D3 SECBs and CSBs with apparently constant spacing from twenty-eight outcrops where damage zones were not clearly detected, and fault planes were at least 3 m away. The spacing generally does not correlate with distance from the first deformation band (Fig. 13). Many peak Pearson correlation coefficients are close to 675 0, average of 0.02, with a normal distribution indicating no relationship between spacing and distance (Fig. 13).  Figure 13c where a scan-line along the top of the image would result in a negative Pearson value, as observed in the Pearson histogram (Fig. 13). This highlights the importance of using multiple scan-lines. Data for bands not associated with faults correlates with synthetic images [2] and [6], corresponding to constant spacing 685 with 'noise' and overprinted distributions, respectively. Discrete spacing values range from 1.4 cm to 14.8 cm.
Equivalent outcrops with D2 bands were not analysed for spacing statistics as suitable outcrops were proximal to faults and, therefore, did not meet the criteria.  However, variation can be seen with large ranges and a non-smooth curve. Bands not associated with faults are 705 expected to show a Gaussian distribution, representing a 'regular' spacing with noise (Fig. 10b). k shows this while l is characterised by a positive skew, as would be expected for a fault damage zone. Data for all analysed outcrops can be observed in Supplement Section S.8.

Figure 15. Boxplots showing the frequency distribution of skewness values for the spacing distribution of deformation
710 bands in fault damage zones and those located not adjacent to a fault (non-FDZ data). Periodic data shows less skew (0 being normally distributed). This correlated with synthetic images from Figure 10 showing regular spacing with noise. Fault data has a higher skew which also correlates with Figure 10 showing a damage zone with additional noise.

Discussion
In the following, we will discuss how our observations of deformation bands from Castlepoint, New Zealand 715 compare with previous studies and if similar relations to the tectonic regime are observed. We shall first discuss the association of tectonic regime with deformation band kinematics, and spatial distribution of bands before we conclude with a brief remark on implications for fluid flow in this deformed rock sequence. https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

Band Kinematics, Orientation, and Microstructure
Outcrop and microstructural data demonstrate that D2 bands, associated with a normal-fault regime, are 720 dominantly CSBs with rare SECBs. SECBs are identified by their lack of offset and higher angle to σ1. In other extensional regimes, only CSBs and Shear Bands (SBs) are observed (Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2015;Soliva et al., 2016;Fossen et al., 2018). SECBs are not predicted to form during normal faulting because smaller mean stresses and larger differential stresses are expected compared to a thrust-fault regime where SECBs are commonly observed (Soliva et al., 2013;Fossen et al., 2018). However, SECBs 725 associated with extension are observed in a carbonate host rock where bands formed during deposition as overburden was increasing (Lubiniecki et al., 2019). Since the Whakataki Formation was still compacting and lithifying during D2 (Chanier et al., 1999;Bailleul et al., 2013), D2 SECBs in our research area may also be explained in this way. Extension-associated bands from Utah and Provence formed post-lithification and were deformed in a tectonic stress state where the overburden, and hence, the vertical tectonic stress σv, are assumed to 730 be constant for the duration of deformation (Solum et al., 2010;Soliva et al., 2013;Ballas et al., 2014;Soliva et al., 2016). However, since normal faulting in our study area was accompanied by ongoing sedimentation, increasing mean stress and diagenetic expansion of the yield envelope could have caused the yield envelope to be intersected closer to the cap and thus explain the presence of SECBs in an extensional regime in the Whakataki Formation. In addition, if the initial stress path for D2 intersects the yield envelope close to the transition between 735 SECBs and CSBs, only small variations in the stress state are required to move from one structure to another.
D3 thrust-fault-associated bands are dominantly SECBs with some CSBs, which is consistent with observations from other contractional regimes (Ballas et al., 2013;Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2015;Fossen et al., 2018). D3 bands mainly show a bimodal orientation pattern. However, when CSBs are in the vicinity of faults, they are commonly unimodally oriented, with all bands sharing a similar dip to the main fault plane. 740 Whether the pattern is symmetric or asymmetric in the thrust-fault regime depends on the elastic properties of the layers and the friction coefficient between layers (Chemenda et al., 2014). Heterogeneity in the Whakataki Formation results in variable material properties which in some places may favour symmetry, and in others asymmetry. In addition, as a bed experiences progressive deformation, material properties change as well, and transitions from asymmetry to symmetry can also occur (Saillet and Wibberley, 2010;Klimczak et al., 2011;745 Chemenda et al., 2014).
Microstructurally, all bands are characterised by a reduction in porosity and grain size, which is consistent with literature for bands formed in normal-and thrust-fault regimes (Solum et al., 2010;Ballas et al., 2013;Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2015;Fossen et al., 2018). The absolute average porosity reduction of ca. 10% observed in all deformation bands from the Whakataki aligns with previous studies (Fig. 9). Permeability 750 was estimated from porosity values using an empirical relationship (Wu, 2005) (see Supplement Section S.6). A comparison of host rock and deformation band reveals a reduction of ca. two orders of magnitude for the normaland thrust-fault regimes. These values are within the bounds observed in previous studies Fossen et al., 2018).
Pattern analysis of orientation data reveals two distributions: [1] polymodal data associated with faults and CSBs 755 and [2] bimodal data associated with SECBs and CSBs not associated with fault damage zones. Polymodal patterns are consistent with 3D strain, while bimodal patterns are consistent with the 2D plane strain (Fossen et al., 2018;Healy and Jupp, 2018;Cai, 2019). The 3D strain is most commonly expected in nature (Healy et al., https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. 2015;Healy and Jupp, 2018). We propose that the presence of a bimodal distribution in our data is a consequence of the measurement scale. Orientations of faults and associated CSBs were measured over a large area at the 760 kilometre scale. At this scale, deformation is 3D in the research area, as seen qualitatively by the variability in the large-scale fold and fault geometry (Fig. 1). In contrast, conjugate SECBs and CSBs were documented in a much smaller area on the scale of hundreds of meters because of outcrop conditions. On this scale, structures can be fairly cylindrical (Fig. 1), and there is a higher chance to sample at a plane-strain scale. At the km-scale, there are many physical reasons for expecting 3D strain in D2 and D3 structures of the Whakataki Formation: [1] the layers 765 have already been folded and faulted in D1 horizontal contraction; [2] the Whakataki Formation has along-strike and stratigraphic thickness variations; [3] rough seafloor topography below the wedge, combined with oblique subduction can induce strong stress/strain heterogeneity (Jones et al., 2005;Wang and Bilek, 2014); [4] the stress field evolved gradually with the principal stress orientations rotating in the transition from D2 (vertical σ1) to D3 (horizontal σ1); [5] faults locally perturb the stress field and generate heterogeneous orientations (Maerten et al., 770 2016); and [6] the material properties in different layers can be anisotropic. All of these effects should result in spatial and temporal differences in stress state on a variety of length scales. Therefore, one can expect the contemporaneous formation of bands with quite different kinematics, possibly explaining the presence of CSBs and SECBs in both normal-and thrust-faulting regimes and complex patterns in outcrop.

Deformation band spatial distribution 775
In the Whakataki Formation, deformation bands with a regular spacing, and deformation bands localised into zones and clusters are observed in normal-and thrust-faulting tectonic regimes. The observation of regularly spaced bands associated with a normal-fault regime contrasts with previous studies that identified a tendency of localisation, not pervasively distributed deformation bands (Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2018). Regularly spaced bands have only been observed in extensional settings when associated with; [1] large 780 relay ramps, [2] when deformed layers are above soft layers such as shale or salt, and [3] when forming in large rollover structures (Fossen et al., 2018). The Whakataki Formation most closely resembles case 2, with deformed strong sandstone beds embedded in soft mudstone layers. However, our observation is on a smaller scale than that suggested by Fossen et al. (2018). Conversely, in a thrust-fault regime, previous studies documented more evenly distributed bands (Soliva et al., 2013;Ballas et al., 2014;Soliva et al., 2016;Fossen et al., 2018). Variably spaced 785 CSBs and SBs in clusters are rarely observed in the contractional regime, and only when associated with faults Ballas et al., 2015;Soliva et al., 2016;Fossen et al., 2018). In the Whakataki Formation, pervasively distributed deformation bands are documented for normal-and thrust-fault regimes, with localisation occurring in the vicinity of faults in both cases. Of the twelve faults studied here, only 29% of damage zones associated with a normal fault regime and 40% associated with a thrust fault regime exhibit a positive correlation 790 between CSB spacing and distance from the fault plane. In the thrust fault regime, 90% of outcrops away from faults show no correlation between the band spacing and distance. In total, only seven of the forty studied outcrops are characterised by a clear positive correlation between spacing and distance. Therefore, both horizontal extension and contraction involve the pervasive formation of distributed CSBs and SECBs. Both stress regimes also come with localised damage-zone-type CSBs. In summary, our field observations show a less distinct 795 difference in the spatial distribution of deformation bands as documented previously for normal-and thrustfaulting regimes (Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2018). https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.
The recognition of deformation bands with regular spacing near faults is based on our interpretation of the statistics of Pearson correlations. Where we observe bimodal or very broad distributions of Pearson coefficients, the spacing distribution is interpreted to reflect progressive deformation and/or multiple tectonic events that have superposed 800 different generations of deformation bands with similar attributes (Fig. 14). This seems to be the most common case in our research area and can be explained with the following conceptual model.
We propose that in the early stages of horizontal extension and horizontal contraction, pervasively distributed CSBs and SECBs form (Fig. 16). The local orientation, spacing, and kinematics of individual bands in this early stage of distributed strain depend on the highly variable rheological properties of the sedimentary layers, layer 805 thickness, and the orientation of the layers relative to the far-field stress (Gross, 1993;Knott et al., 1996;Martel, 1999;Bai and Pollard, 2000a;Bai and Pollard, 2000b;Olsson and Holcomb, 2000;Ackermann et al., 2001;Rudnicki, 2003;Soliva and Benedicto, 2005;Soliva et al., 2006;Chemenda, 2009;Laubach et al., 2009;Chemenda et al., 2012;Regenauer-Lieb et al., 2013a;Regenauer-Lieb et al., 2013b;Chemenda et al., 2014;Zuza et al., 2017;Laubach et al., 2018). Our microstructural data demonstrate that there is no difference in the 810 deformation mechanisms of CSB and SECB formation. Therefore, only small changes in the local stress state and yield envelope are required to move from one structure to another because the kinematic transition is continuous and simply reflects different degrees of shear displacement and volume reduction (Fossen and Bale, 2007). With continued strain, favourably located deformation bands can link across sedimentary layers and cause the propagation and formation of throughgoing slip surfaces (Fig. 16) (Aydin and Johnson, 1983;Antonellini et al., 815 1994;Cowie, 2001, 2003;Schueller et al., 2013). Many CSBs associated with immature small faults are sub-parallel to the fault (Figs. 2,5,6), which supports the idea that faults nucleate through linkage of deformation bands across layers. Moreover, at the regional scale, normal-and reverse-faults in the Whakataki Formation are uniformly distributed (Fig. 11). This is consistent with the idea that fault spacing is strongly controlled by the mechanical stratigraphy of the turbidite stack (Knott et al., 1996;Martel, 1999;Ackermann et 820 al., 2001;Soliva and Benedicto, 2005;Soliva et al., 2006;Laubach et al., 2009;Zuza et al., 2017;Laubach et al., 2018). Once throughgoing faults have formed, progressive slip is expected to cause the formation of the wall-, interaction-, and tip-damage zones (Kim et al., 2004;Peacock et al., 2017) that then overprint CSBs with a variable spacing (Shipton and Cowie, 2003;Faulkner et al., 2011;Xu et al., 2012;Maerten et al., 2016). In this scenario of progressive deformation, one would expect a bimodal distribution of Pearson correlation coefficients where the 825 mode that is close to zero reflects the initial band generation with regular spacing, and a higher mode > 0.5 reflects that of the damage-zone overprint. Moreover, there are many outcrops with distinct generations of cross-cutting structures associated with the same deformation phase as well as those where D3 structures overprint D2 structures. Broad distributions of Pearson coefficients may reflect outcrops that experienced overprint by three or more distinct band-generating events (Fig. 14). 830 https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.

Tectonic control and the stress path
Stress path modelling was employed in previous field studies to explain the kinematics and orientation of the observed deformation bands as a function of the tectonic regime (Saillet and Wibberley, 2010;Solum et al., 2010;Soliva et al., 2013;Ballas et al., 2014;Fossen et al., 2015;Soliva et al., 2016). We do not use this approach here 840 because of the inherent complexity of the evolution of the state of stress within accretionary prisms. Critical wedge theory demonstrates that the state of stress in accretionary wedges is controlled by the geometry of the wedge (slope angle and dip angle of the subduction master fault), pore-fluid pressure, the frictional properties of the wedge, and the basal coefficient of the subduction thrust Wang and Hu, 2006). The spatial distribution of seismicity along subduction faults implies that the wedge can be separated into a velocity-845 strengthening section (outer wedge) and a seismogenic velocity-weakening section (inner wedge) . This causes the inner and outer wedges to exhibit significant differences in the mechanical state during failure. The Whakataki Formation has travelled from the top of the outer wedge into the inner wedge and back to the surface. During this journey, the formation has experienced significant changes in a stress state, as clearly indicated by the broad structural inventory (Fig. 2). It is difficult to constrain this stress history from field 850 observations. To emphasise this point, we recall that even within the inner wedge, one can obtain large vertical differences in stress regime, such as upper levels being under horizontal extension while lower levels of the wedge are simultaneously in horizontal contraction and vice versa . Stress regime in the accretionary wedge is largely controlled by the basal friction coefficient of the subduction thrust and the porefluid pressure. Both parameters cannot be constrained reliably throughout the Miocene from our field 855 observations. Therefore, to generate a comprehensive estimate of the stress path of the Whakataki Formation, at https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. least 2D stochastic mathematical inverse modelling complemented by time-resolved geophysical data and strong geochronological constraints of basin and deformation history is required. This is beyond the scope of the current study.

860
However, one generic idea proposed in the literature surrounding stress path modelling may apply to our case study: during extension, the mean stress is smaller than during contraction, resulting in an intersection of the yield envelope closer to the top of the cap and thus the formation of CSBs (Soliva et al., 2013). In D2 extension we also dominantly observe CSBs. Regional studies demonstrate that the deposition of the Whakataki Formation most likely lasted until the end of D2 (Bailleul et al., 2007). Therefore, the overburden stress probably increased 865 throughout D2. During D3, sediment thickness reached its maximum, and additional thickening through D3 folding and thrust-stacking would have added to the vertical stress. It is, therefore, likely that the mean stress increased from D2 to D3. In addition, a change of the mechanical properties of sandstones from D2 to D3 can be expected, due to porosity reduction associated with ongoing compaction and cementation. These mechanical changes are associated with hardening resulting in a bigger yield envelope during D3 (Fig. 17) (Fossen and Bale, 870 2007). The combination of higher mean stresses and a bigger yield envelope in D3 compared to D2 could explain why we also observe more CSBs during D2 extension and more D3 SECBs during D3 contraction (Fig. 17), as proposed by Soliva et al. (2013), Ballas et al. (2014) and Soliva et al. (2016). This idea is certainly appealing in its simplicity. However, it remains to be tested with an inversion of the effective stress path for our study area, which is a challenging problem. Finally, while a static yield model can explain the critical stress state required for 875 failure and the resulting failure angle and deformation band kinematics, it provides no information on the timing, rate of failure, spatial distribution or the number of deformation structures (Zhang et al., 1990;Wong et al., 1992;Wong et al., 1997;Schultz and Siddharthan, 2005;Wong and Baud, 2012). Therefore, one cannot predict the spatial distribution of localised instabilities based on the static far-field stress and static yield envelope alone. Our observations support the well-established notion that the mechanical stratigraphy plays a major role in determining 880 the spatial distribution of faults and deformation bands in layered rock sequences (Knott et al., 1996;Martel, 1999;Ackermann et al., 2001;Soliva and Benedicto, 2005;Soliva et al., 2006;Laubach et al., 2009;Laubach et al., 2018). In this case, predictive models for the time and spatial evolution of deformation structures must resort to at least 2D mathematical forward modelling (Chemenda et al., 2014). 885 https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License. Figure 17. Schematic diagram of the potential yield envelopes for the Whakataki Formation during D2 horizontal extension and D3 horizontal contraction. During D2, the unit is still lithifying and being buried, therefore, in D3 when the unit has lithified, the yield cap has expanded (red line). For a generic increase in effective stress (p) and differential stress (q), the yield cap may be intersected at different points due to the evolution of the yield envelope through time.

Implications for fluid flow 895
Our field observations demonstrate that the heterogeneous sedimentary architecture, in concert with folds, faults, and deformation bands, transforms the Whakataki Formation into a complex reservoir that is strongly compartmentalised over four orders of magnitude in length scale. At the kilometre scale, some large folds and thrusts juxtapose rock formations with different petrophysical properties. At the 100-to 10 m scale, normal-and reverse faults dissect the Whakataki Formation into monoclinic or triclinic blocks with a dominant thickness of 900 ca. 10 m. Faults can be both fluid conductors and inhibitors. However, D2 and D3 faults show lots of clay smear and gouge in our field area. Therefore, fluid flow inhibition is most likely (Fulljames et al., 1997;Yielding et al., 1997;Nicol and Childs, 2018). At the m-to dm-scale, interbedded mudstones can restrict fluid flow in sandstone layers in the vertical direction, resulting in an anisotropic hydraulic conductivity on the scale of metres to tens of centimetres. Finally, at the 10 cm-scale permeability-reducing deformation bands are observed in complex 3D 905 networks (Antonellini and Aydin, 1994;Antonellini and Aydin, 1995;Fossen and Bale, 2007;Ballas et al., 2013), especially in the sand-rich facies. Thus, individual sand-and siltstone beds are subdivided yet again into even smaller blocks. In an accretionary prism where fluid is readily available, such compartmentalisation can result in high pore fluid pressures that may result in catastrophic failure of the wedge (Audet et al., 2009). For petroleum extraction, one may expect that hydraulic properties determined at the core scale are not representative of those 910 at the reservoir scale. An integration of the multi-scale structural and sedimentary architecture with computational upscaling of flow properties is required for the prediction of reservoir properties (Field et al., 2006). https://doi.org/10.5194/se-2020-80 Preprint. Discussion started: 18 May 2020 c Author(s) 2020. CC BY 4.0 License.