Focused fluid seepage related to variations in accretionary wedge structure, Hikurangi margin, New Zealand

Hydrogeological processes influence the morphology, mechanical behavior, and evolution of subduction margins. Fluid supply, release, migration, and drainage control fluid pressure and collectively govern the stress state, which varies between accretionary and nonaccretionary systems. We compiled over a decade of published and unpublished acoustic data sets and seafloor observations to analyze the distribution of focused fluid expulsion along the Hikurangi margin, New Zealand. The spatial coverage and quality of our data are exceptional for subduction margins globally. We found that focused fluid seepage is widespread and varies south to north with changes in subduction setting, including: wedge morphology, convergence rate, seafloor roughness, and sediment thickness on the incoming Pacific plate. Overall, focused seepage manifests most commonly above the deforming backstop, is common on thrust ridges, and is largely absent from the frontal wedge despite ubiquitous hydrate occurrences. Focused seepage distribution may reflect spatial differences in shallow permeability architecture, while diffusive fluid flow and seepage at scales below detection limits are also likely. From the spatial coincidence of fluids with major thrust faults that disrupt gas hydrate stability, we surmise that focused seepage distribution may also reflect deeper drainage of the forearc, with implications for pore-pressure regime, fault mechanics, and critical wedge stability and morphology. Because a range of subduction styles is represented by 800 km of along-strike variability, our results may have implications for understanding subduction fluid flow and seepage globally.


INTRODUCTION
Fluids-in their gaseous or liquid phaseplay a critical role in controlling the dynamics of accretionary wedges by influencing pore pressure (Dahlen, 1990;Saffer and Bekins, 2006) and the mechanical behavior of faults (Sibson, 1992;Kodaira et al., 2004;Liu and Rice, 2007;Bangs et al., 2015). Fluids are released from both the subducting plate and the accretionary wedge, due to compaction and thermally controlled mineral-phase transformations. Efficient fluid migration relies on the occurrence of high permeability, i.e., flow pathways through faults, fractures, and high-porosity lithologies (Saffer and Tobin, 2011;Plaza-Faverola et al., 2012;Bangs et al., 2015).
The Hikurangi margin, offshore the North Island of New Zealand, exhibits evidence for widespread fluid seepage (Lewis and Marshall, 1996;Barnes et al., 2010;Greinert et al., 2010), together with variations in tectonic structure and processes capturing a range of subduction styles (Fig. 1). These variations include changes in the rate and obliquity of plate convergence (Wallace et al., 2004), occurrence of subducting seamounts, accretionary versus erosional subduction behavior, and wedge morphology (Lewis and Pettinga, 1993;Barnes et al., 2010). From south to north, the deformation setting changes from subduction to transform transition, to classical frontal accretion, and finally to widespread frontal tectonic erosion with seamount subduction. These changes accompany a northward reduction in sediment thickness on the subducting Pacific plate, from ∼9 km ( Fig. 1; Plaza-Faverola et al., 2012) to ∼1-2 km (Barker et al., 2009(Barker et al., , 2018. The margin attains a width of 150 km, characterized by a deforming Late Cretaceous-Paleogene presubduction foundation inboard of a late Cenozoic accretionary wedge (Lewis and Pettinga, 1993;Barnes et al., 2010;Ghisetti et al., 2016). Estimates of the average volume of fluids subducted and accreted along strike are high (∼57 m 3 yr -1 m -1 ; Pecher et al., 2010), with a significant reduction expected northward, and the vast majority drained from the wedge (Ellis et al., 2015).
We investigated the distribution of focused fluid seepage using an exceptional high-quality regional data set from the Hikurangi margin, including: (1) hydro-acoustic data that image active gas bubbles; (2) seafloor camera observations of chemosynthetic seep ecosystems; and (3) geomorphological and seafloor acoustic backscatter data indicative of substrates associated with focused fluid seepage ( Fig. 2; see the GSA Data Repository 1 for individual presentation of seepage features, grid sizes, and observation thresholds). We used observed spatial patterns of seep distribution to map shallowwedge permeability, and then we considered the implications for hydrological framework and subduction processes.

SEEP INDICATORS AND THEIR DISTRIBUTION ON THE HIKURANGI MARGIN
There is extensive evidence for both active and relict (or dormant) focused fluid seepage in water depths of 50-2400 m along the margin (Figs. 2A-2C). Pockmarks are widespread on the outer shelf and upper slope (28% of the database); however, very few pockmarks have concomitant indicators of seep activity and are not considered herein.
Active focused seepage was identified from hydro-acoustic flares and/or direct observations of live seep fauna (Figs. 2A-2C), commonly with characteristic rough seafloor geomorphology, high seafloor backscatter intensity, and mounds. Approximately 75% of authigenic carbonate observations were corroborated by evidence for active seepage ( Fig. 2; see also the Data Repository).
Focused seepage occurs across the inboard margin, at sites located on the outer continental shelf, slope, and thrust ridges. Generally, seepage indicators are rare across the outer margin (Figs. 2A-2C and 3A-3C).
The northern margin is characterized by active, spatially concentrated fluid expulsion on the outer shelf and upper slope (including active pockmarks with coincident flares; seep site spatial density: ∼1/17 km 2 ; Figs. 2A and 3A). For example, the Tuaheni Seep Field covers ∼90 km 2 and includes >1700 seep indicators (including flares, mounds, and pockmarks) in water depths of 58-532 m, making it the most spatially concentrated and shallowest seep area observed anywhere on the margin ( Fig. 2A; Higgs et al., 2019). Farther offshore, active seepage is scattered mainly along midslope ridges (Figs. 2A and 3A).
Across the southern margin, active, sustained, and focused seepage is widespread but concentrated mainly on thrust ridges (Figs. 2C and 3C). The Glendhu and Honeycomb Ridges (∼2100-2400 m water depths) are the only sites with evidence of active seepage at seafloor depths >2100 m and proximal to the deformation front (Fig. 2C). At the southern extent of the margin, seepage is widespread along the crest of Kekerengū Bank, while the southernmost seep site lies in the Kōwhai Sea Valleys (Fig. 2C).

RELATIONSHIP BETWEEN FLUID SEEPAGE AND MARGIN CHARACTERISTICS
The comprehensive spatial coverage and quality of our data are exceptional for subduction margins globally (e.g., Ranero et al., 2008;Geersen et al., 2018;Riedel et al., 2018). Hydroacoustic flares are the most common seepage indicator (∼46% of the database), indicative of active focused gas expulsion (Fig. 2).
The distribution of active focused seepage on the Hikurangi margin may reflect variations in the shallow permeability architecture of the wedge. Shallow fluid pathways may in turn reflect higher permeability at depth, governed by the regional tectonic framework (Figs. 2A-2C and 4A-4C). In the north, where the convergence rate is ∼5 cm/yr, the Tuaheni Seep Field is spatially correlated with localized and possibly shallow extensional faulting (Böttner et al., 2018), immediately east of major thrust faults that intersect with the weakly coupled plate interface inland of the coast (Barnes et al., 2002;Mountjoy and Barnes, 2011). Active focused seepage between Rock Garden and Tolaga Knoll ( Figs. 2A and 2B) lies landward of the frontal accretionary wedge and coincides with active thrust faulting and upper-plate fracture networks above and landward of subducting seamounts Pedley et al., 2010;Bell et al., 2010Bell et al., , 2014Plaza-Faverola et al., 2014;Barker et al., 2018).
Zones of active focused seepage at North Hikurangi overlie regions characterized by microseismicity and tectonic tremor (Todd and Schwartz, 2016;Todd et al., 2018), shallow slow slip events Wallace et al., 2016), and a highly reflective subducting unit that has been inferred to be fluid rich (Barker et al., 2009;Bell et al., 2010), moderately overpressured, and associated with seamount subduction ( Fig. 1; Bassett et al., 2014;Ellis et al., 2015). Hydrofracturing at the innerplate interface facilitates fluid drainage from the subducting sequence by providing secondary permeability in the overriding plate (Ellis et al., 2015), as supported by mantle/slab He isotope signatures in onshore fluid seeps (Reyes et al., 2010). Similarly, active seepage off the coast of Costa Rica is closely associated with subducted seamounts on the Cocos plate (Ranero et al., Figure 1 and/or the Data Repository (see footnote 1) for mapped coverage of the multibeam bathymetry and water-column data on the margin. Features described in the text are labeled. Examples and implications of each fluid expulsion indicator presented in this study are provided in the legend. Seep site spatial density was calculated by dividing the total mapped area within the zone (km 2 ) by the corresponding number of seepage indicators for that zone, presented as one seep indicator per unit area. Figure 2 for the broad-scale margin morphology associated with these graphs. Seep site spatial density for each zone is annotated.

Figure 3. Bar graphs showing distribution of each fluid expulsion indicator as a function of geomorphic zones for (A) frontal tectonic erosion and seamount subduction zone, (B) subduction accretion-low-taper zone, and (C) subduction accretion-transpressional zone at the Hikurangi margin, offshore the North Island of New Zealand. Note the different yaxis scales in A-C. Refer to
2008; Sahling et al., 2008). Intriguingly, we observed no evidence for active focused seepage in the Poverty re-entrant, across the ∼7000 km 2 Ruatoria re-entrant and associated debris-avalanche deposits, and in the Poverty Sea Valleys and canyons. These features are considered to have evolved from major impact scars (wake avalanches) associated with subducted seamounts ( Fig. 2A; Lewis et al., 1998;Collot et al., 2001;Pedley et al., 2010). The chaotic nature of the debris-avalanche deposits may not favor localized fluid expulsion. We also observed a distinct lack of focused seepage on the relatively narrow (15-25 km), highly tapered (10°-15°) frontal accretionary wedge in the northern margin.
The wide (150 km), low-taper (4°) central margin wedge, characterized by a decreased orthogonal convergence rate (3-4 cm/yr) and increased sedimentary thickness on the subducting plate (3-4 km), has the lowest density of seepage observed on the margin (Figs. 2B and 3B). The absence of active focused seepage in this region, compared to the northern and southern areas, could be related to more distributed, diffuse and/or smaller-scale fluid expulsion, which may not be detectable in acoustic water-column backscatter data or at the scale measured in this study. Diffuse fluid expulsion has been documented to represent the majority of fluids expelled at some accretionary margins (e.g., Nankai and Barbados; Bekins, 1998, 1999), and thus may also play a significant role in wedge dewatering of the central Hikurangi margin wedge. Hydroacoustic flares occur predominantly on thrust ridges deforming the presubduction foundation of the inner margin Pecher et al., 2010). These fluids may reflect microbial methanogenesis coupled with shallow permeable pathways and/or fluids released from compaction of subducting sediments, and clay dehydration reactions (e.g., Bekins et al., 1994;Hyndman et al., 1997;Reyes et al., 2010;Plaza-Faverola et al., 2016), with upward flow along structure-induced permeability ( Fig. 4B; Barnes et al., 2010).
To the south, the margin-normal convergence rate decreases nearly fourfold, from ∼3 to 0.6 cm/yr, and the wedge narrows from ∼100 km to ∼40 km wide (Fig. 1). Focused seepage extends to the deformation front at Honeycomb Ridge, and up to 130 km farther south than previously recognized along the midslope Kekerengū Bank thrust ridge (Fig. 2C). While seep sites at Urutī Ridge are proximal to the northern strike-slip section of the Palliser-Kaiwhata fault, no seepage was observed directly along the other major strike-slip faults of the southern margin (Fig. 2C). A correlation was found, however, between the distribution of fluid seepage on thrust ridges and the spatial distribution of concentrated gas hydrate indicators in regional seismic data (cf. Crutchley et al., 2019). We surmise that deforming thrust ridges in this context are more efficient wedge dewatering structures than strike-slip faults or are more effective structural traps for migrating fluids, enabling the development of productive seep environments. Although strike-slip faults have been previously considered to be effective dewatering conduits in other accretionary settings (e.g., Wecoma fault, Oregon; Tobin et al., 1993), the lack of observed focused seepage along such faults on the southern margin supports the notion that fluid expulsion along strike-slip margins is relatively sparse compared to subduction settings ( Fig. 4C; Maloney et al., 2015).
We observed an overall lack of evidence for focused fluid seepage across the Hikurangi margin frontal wedge, despite rapid accretion of thick trench sediments ( Fig. 2B; Ghisetti et al., 2016), and the expectation of compaction-driven dewatering dominating fluid sources (Saffer and Bekins, 2006;Ellis et al., 2015). At least 90% of observed seepage occurs above the deforming Late Cretaceous-Paleogene foundation, and it is most prevalent ∼20-70 km landward of the deformation front ( Figs. 2A-2C). The absence of focused seepage across the central margin frontal wedge may reflect poor drainage, resulting in elevated fluid pressures, a relatively weak plate interface, and low-taper (2°-4°) morphology (Fig. 4B). Alternatively, dewatering of the central Hikurangi margin and frontal wedge manifests predominantly as diffuse seepage, below our detection threshold (cf. Nankai accretionary margin; Saffer and Bekins, 1998). Low-taper, stable accretionary wedges typical of the central Hikurangi margin are thought to reflect poorly drained systems with a basal fault zone that is either fluid overpressured or frictionally weak, relative to the wedge (Dahlen, 1990;Sibson and Rowland, 2003;Ellis et al., 2019). We note that the wide low-taper Washington sector of the Cascadia margin has a similar distribution of seepage (Riedel et al., 2018) and strong interseismic coupling (Schmalzle et al., 2014). Active focused seepage on the Hikurangi margin does not show any obvious spatial relationship with the slow slip events and/or interseismic coupling (Figs. 1 and 2A-2C).

CONCLUSIONS
Active focused seepage along the Hikurangi margin manifests mostly above the deforming backstop, likely exploits tectonically generated conduits that disrupt regional gas hydrate stability, and is largely absent from the frontal wedge. Patterns of seepage vary in accordance with changes in subduction setting (from the subduction accretion-transpressional zone, to subduction accretion-low-taper zone, and finally to seamount subduction and frontal tectonic erosion zone), convergence rate, and subducting plate roughness and sediment thickness. North-to-south variations in focused seepage observed in the current data set support previous suggestions that changes in subduction processes influence structural permeability and drainage pathways. We expect that large-scale seafloor seepage patterns may provide important insights into the deeper mechanics of accretionary and nonaccretionary systems.

ACKNOWLEDGMENTS
Data used for this study came from a large number of marine surveys using several different vessels, but primarily RV Tangaroa. We are grateful to the science officers and ship officers and crew staffing those surveys for the provision of the accumulated survey data that underpin this study. Some of the data collection and analysis was funded by New Zealand Petroleum and Minerals (NZP&M) for two petroleum basin screening reports (Bland et al., 2014;Crutchley et al., 2016). We are grateful to OMV New Zealand (Wellington) for providing access to multibeam data from the inner central margin. Funding for National Institute of Water and Atmospheric Research (NIWA) staff came from the Ministry of Business, Innovation, and Employment (MBIE) Strategic Science Investment Fund (SSIF) Marine Geological Processes Programme of the NIWA Coasts and Oceans Centre. Funding for Crutchley and Hillman, and partially for Watson and Mountjoy, was provided by the New Zealand MBIE Endeavor Fund Programme Gas Hydrates: Economic Opportunities and Environmental Impact (HYDEE; contract CO5X1708). We thank D. Saffer and two anonymous reviewers for their constructive feedback that greatly improved this manuscript.