The subduction interface beneath the northern part of the Hikurangi subduction margin on the east coast of New Zealand's North Island is exceptionally shallow and cool compared with other margins where slow slip has been observed. Here we use magnetotelluric data to show that a marked decrease in the conductivity of the fore-arc sediments coincides with the onset of seismicity at ∼10 km depth. Below the sediments, a dipping band of seismicity and intermediate conductivity at the subduction interface connects to a deeper more conductive zone above the down-going plate. This deeper conductive zone is interpreted to be under-plated sediments. These results and results from previous seismic tomography in the area suggest that the intermediate resistivity zone represents a region of upward fluid transport near the plate interface followed by fluid escape into the upper-plate.
If you can't find a tool you're looking for, please click the link at the top of the page to "Go to old article view". Alternatively, view our Knowledge Base articles for additional help. Your feedback is important to us, so please let us know if you have comments or ideas for improvement.
 The Raukumara Peninsula, the eastern most part of New Zealand's North Island (Figure 1), exposes part of the northern Hikurangi subduction margin fore-arc, where the Australian and Pacific plates converge at ∼60 mm/a [e.g.,Wallace and Beavan, 2010]. Here, the Hikurangi Plateau, a 120 Ma old oceanic plateau with numerous seamounts, is being subducted beneath the Australian Plate. This has resulted in a shallow subduction interface (∼15 km deep at the eastern coast of the peninsula) and an uplifted fore-arc [Davy and Wood, 1994]. GPS data from the Raukumara Peninsula suggest that the upper and lower plates are weakly coupled [Wallace et al., 2004]. Shallow (<15 km) slow-slip events also occur just off-shore of the southern part of the peninsula [Wallace and Beavan, 2010]. The mechanism controlling the occurrence of slow slip initiation is poorly understood, but is thought to occur on the subduction interface between regions where aseismic-slip and stick–slip behaviors dominate, which may in turn be controlled by the fluid pressure on the interface [e.g.,Liu and Rice, 2007; Schwartz and Rokosky, 2007]. Fluids at near-lithostatic pressure on the subduction interface may also provide an explanation for the weak coupling of this part of the Hikurangi subduction margin. On the Raukumara Peninsula, a rich source of fluids is available from both the thickened oceanic crust of the subducting Hikurangi Plateau and sediments carried into the subduction zone [e.g.,Bell et al., 2010]. Sediment under-plating has also been suggested to cause the uplift of the Raukumara Ranges [Walcott, 1987].
 At mid and lower crustal depths, the presence of interconnected fluid in crustal rocks will increase the electrical conductivity detected at the surface by magnetotelluric (MT) instruments. Conductive zones have been observed in MT surveys at several subduction margins and have been interpreted in terms of the de-hydration processes occurring during subduction [Kurtz et al., 1986; Wannamaker et al., 1989; Jödicke et al., 2006; Soyer and Unsworth, 2006; Brasse et al., 2009; Wannamaker et al., 2009]. Thus MT surveys may contribute to the debate about the role of fluids in determining the frictional properties of the subduction interface. Identifying the location of the fluid sources and areas of fluid escape using MT will also contribute to the debate on the role of fluid pressure in slow slip events. Here we present results from a pilot MT survey on the Raukumara Peninsula along two profiles approximately parallel and perpendicular to the strike of the Hikurangi subduction interface (Figure 1).
2. Magnetotelluric Data and Analysis
 MT measurements consist of surface recordings of natural variations in the earth's surface magnetic and electric fields. These recordings are transformed into the frequency domain and the relationships between the measured electric- and magnetic-vector field components form the MT data used to determine the electrical conductivity structure. These relationships are expressed by the impedance tensorZ (units Ω), defined as E = ZH, relating the electric field vector E (V/m) to the horizontal magnetic field vector H (A/m) and the induction vector K (dimensionless), defined as Hz = −K·H, relating the vertical component of the magnetic field (Hz) to its horizontal components (H) [Parkinson, 1962]. Essentially these two relations normalize the observed electric field and its horizontal curl (Hz) by the strength of the inducing field H. By analyzing these data as a function of period and location, the subsurface distribution of conductivity can be determined.
 The phase relationships contained in Z also take the form of a tensor [Caldwell et al., 2004] and we will use the phase tensor Φ, defined as Φ = X−1Y, where X and Y are the real and imaginary parts of Z, to visualize the impedance data. Unlike the impedance magnitude or apparent resistivity information, at longer periods Φis independent of near- surface small-scale heterogeneities [Caldwell et al., 2004]. Graphically, Φ can be represented as an ellipse with the principal axes (Φmax and Φmin) defined by the ellipse axes. Phase tensor ellipse maps at different periods provide a method of visualizing spatial conductivity gradients at different depths; longer periods sensing more deeply. Similarly, gradients in the conductivity structure are indicated by the magnitude and direction of the real part of the induction vectors; with the induction vectors pointing towards regions of greater conductance.
 Thirty four broadband (0.003 s–1000 s) MT soundings were recorded along two profiles approximately orthogonal (Figure 1). (See auxiliary material for details of the instrumentation and data collection). Phase tensor ellipses and induction vectors at period T = 42.6 s are shown in Figure 1. The colours used to fill the phase tensor ellipses in these and later figures show the geometric mean Φ2 of the principal phase values Φmax and Φmin, i.e., Φ2 = √(ΦmaxΦmin). Values of Φ2 > 45° indicate decreasing resistivity with depth. The phase tensor and induction vector data along each of the two profiles are also shown in Figure 2as period-distance pseudo-sections. Apparent resistivity sounding curves are shown inFigure S1.
 In the south-east, Φ2 values are >45° and are almost circular at periods of ∼<100 s (Figure 2) indicating an almost 1-D resistivity structure of considerable thickness. Induction arrows in the south–east are also small, consistent with a 1-D structure. MT apparent resistivity values near the surface are ∼10 Ωm, suggesting that the conductive clay-rich sediments that crop out along the eastern side of the peninsula are very thick (∼8 km) at the south-eastern end of Profile-1 (seeFigure 3). In the west of Profile-1 where the Raukumara Ranges (Figure 1) expose resistive meta-sediments, Φ2 values > 45° (shaded zone in Figure 2) indicate the presence of higher conductivities at depth. In the northern part of Profile-2, the induction vectors increase in magnitude and rotate with increasing period until at the longest periods they point north–eastwards towards the deepest part of the ocean (Figure 2). The rotation of induction vectors and phase tensor skew angle (see Figure S2) indicate that at deeper levels the conductivity structure is 3-D.
3. Three-Dimensional Inverse Modelling
 To transform the MT data into resistivity-depth sections we used the 3-D non-linear inverse modeling code ofSiripunvaraporn et al. [2005a] to simultaneously invert the impedance tensor data along both profiles. Siripunvaraporn et al. [2005b]show that 3-D inversion modeling gives better results for profile data in the presence of 3-D effects than a 2-D inversion. Noisy sections of data in the ‘dead-band’ (1–10 s) where the MT signal is weak were omitted from the data inverted. To reduce computation time and memory requirements, 3 data points per (period) decade were used in the inversion in the period range from 0.1 to 680 s. Outside the dead-band, data quality is generally good and a 5% error floor was used. Details of the 3-D model mesh and inversion parameters are given inFigure S3.
 To test the possibility of model convergence to a localized minimum 3 different starting models for the inversion were tried. The major features are similar in all the models (see Figure S4for details). To further assess the model fits, we compared the distribution of misfit between the observed and calculated phase-tensors using a tensor measure of the misfit which provides a way of visualizing the spatial and period dependence of the model fit. Pseudo-sections of the misfit tensor are shown inFigure S5. Figure 3shows slices through the 3-D inversion model along the two measurement profiles for the model where the misfit tensors are smallest in the period range where the phase response to the conductor C1 occurs (shaded area inFigure 2). The normalized RMS misfit for this model is 1.26.
4. Discussion and Conclusions
 The upper part (<10 km) of the resistivity model shown in Figure 3is in good agreement with the surface geology. The high resistivity area on the northwestern side of Profile-1 coincides with the area where the old (85–160 Ma) meta-sediments that make up the Raukumara Ranges are exposed. Along Profile-2, the surface geology consists mainly of young (2–25 Ma) marine sediments that are clay rich and highly conductive. These sediments reach thicknesses of ∼5 km at the north and south ends of Profile-2, separated by an area where older (25–115 Ma) allochtonous meta-sediments are exposed [Mazengarb and Speden, 2000].
 In Figure 3a, hypocenters of earthquakes [Reyners et al., 2011] occurring within 25 km of Profile-1 are shown superimposed on the resistivity cross-section. A zone of intermediate resistivity (∼50–100 Ωm) coincides with the uppermost part of the dipping band of seismicity that marks the subduction in the south–eastern part of Profile-1. To the northwest, below ∼20 km, the zone of intermediate resistivity becomes more highly conductive (∼15–30 Ωm) forming a wedge shaped region (C1) above the band of seismicity marking the top of the underlying subducting plate. Below these two regions, the underlying plate is resistive (>200 Ωm) in good agreement with the high resistivity of the subducting slab seen ∼100 km to the southwest [Heise et al., 2007].
 A conductive region similar to C1 is seen under the northern part of Profile-2 in approximately the same depth range as C1. This conductor, C2 inFigure 3b, dips to the north in the direction of Profile-2 consistent with the oblique direction of the profile with respect to the down-dip direction of the plate interface (Figure 1). Thus C1 and C2 may be different sections of a single conductive zone that exists above the subduction interface. At its southern end, C2 underlies a tabular block of high resistivity (R1 in Figure 3b) that separates these areas of thick sediment to the north and south. Gravity values [Woodward and Boyle, 1972] are higher in this area (R1), suggesting that the high resistivity material is a block of denser basement rock similar to the meta-sediments exposed in the Raukumara ranges.
 The most striking feature of the inverse model is the conductor (C1) in Profile-1 (Figure 3a). This feature is also present in all the inversions tried (Figure S4). To assess the sensitivity of the MT data to C1, a forward model was computed without the conductor (Figure S6). Removing C1 produces a 10–30% phase tensor change in the period range 10–60 s showing that C1 is strongly constrained by the data. We also inverted the data omitting site 919 (Figure 3) which lies in an isolated position in the centre of Profile-1. Again, the resulting inversion model showed a conductive zone similar to C1.
 In Figure 4awe have superimposed p-wave velocity (Vp) contours [Reyners et al., 1999] onto a slice through the 3-D resistivity model along the tomographic cross-section that is closest to Profile-1. The contours shown (6.5–8 km/s) correspond to the transition inVp expected between crustal rocks and the oceanic mantle in the down going plate. The oceanic crust on the Hikurangi Plateau is thought to be 10–15 km thick [Bassett et al., 2010] corresponding approximately with the 6.5 km/s contour in Figure 4a. Although there is an inherent trade-off between thickness and conductivity in MT models and C1 could be thinner provided it were also more conductive, theVp values suggest that C1 is located entirely in crustal rocks above the plate interface.
Reyners et al.  and others [Eberhart-Phillips and Chadwick, 2002; Bassett et al., 2010 ] interpret the zone of low Vpvalues above the plate corresponding to C1 as a zone of under-plated sediment first proposed byWalcott to explain the uplift of the Raukumara Ranges. The MT data show that this zone is electrically-conductive and therefore must contain an interconnected conductive phase. Given the low temperature of the subducted plate and the lowVpvalues, the most plausible cause of the low resistivity is the presence of aqueous fluids within a zone of under-plated sediment. Supporting this interpretation are the highVp/Vs values seen in the seismic tomography (Figure 4b), which are indicative of over-pressured fluid [Reyners et al., 1999].
 At the southeastern end of Profile-1, a dipping zone of intermediate resistivity (30–100 Ωm) underlies the thick conductive layer of sediments linking these sediments to C1. Although forward modeling (Figure S7) shows that the resistivity of this zone is only weakly resolved by the data due to the screening effect of the overlying conductive sediments [e.g., Bedrosian, 2007], a dipping zone of intermediate resistivity appears in the inversions that use different starting models (Figure S4). This gives confidence that this zone is not an artifact produced by the smoothing used in the 3-D inversion. We further tested this feature by inverting the data with less horizontal and vertical smoothing. The resulting model again contained a dipping zone of intermediate resistivity above the plate (Figure S8).
 The region of high Vp/Vs values (Figure 4b) overlaps with the zone of intermediate resistivity suggesting that over-pressured fluid may be also present in this zone. Here, the seismicity extends well above the oceanic material into the overlying upper plate (Figure 4a). Taken together these observations suggest that the intermediate resistivity zone may show a zone of upward fluid transport and fluid escape into the overlying-plate.
 A question that arises from this interpretation is the cause of the enhanced conductivity of C1 compared to the dipping zone of intermediate resistivity. In situations where a conductive zone is sandwiched between more resistive material, only the conductance (i.e., the thickness resistivity ratio) of the conductive layer can be reliably resolved by MT measurements. A reduction in the thickness of the intermediate zone by a factor of about two would reduce the resistivity to a value similar that of C1. Alternatively the enhancement of the conductivity in C1 may be the result of on-going low-grade metamorphic reactions within the under-plated sediment [Jödicke et al., 2006; Saffer and Tobin, 2011; Fagereng and Cooper, 2010] that are not present to the same degree in the zone of intermediate resistivity.
 There is also remarkable agreement between the base of the conductive sediments and the shallow aseismic-seismic transition in the south–east of Profile-1 where the sediments deepen slightly (Figure 3a). The depth of the conductive sediments is a well resolved feature of the inversion models as illustrated by the inversion model computed with reduced smoothing shown in Figure S8. Several authors [e.g., Hyndman et al., 1995] have proposed that this transition coincides with the transformation of smectite to illite known to cause an increase in resistivity [Pellerin et al., 1996]. However, Moore et al. argue that cementation and the associated increase in cohesive strength is more likely to control the depth of the shallow aseismic-seismic transition. Both mechanisms are consistent with the increase in resistivity seen at the onset of seismicity.
 Slow slip episodes have been observed just off-shore the end of Profile-1. Several studies [e.g.,Schwartz and Rokosky, 2007; Wallace and Beavan, 2010] have invoked aqueous-fluids at the subduction interface as a key factor in the occurrence of slow slip.Wech and Creager interpret seismic tremor observed down dip of the slow slip region as slip too small to be detected geodetically. Tremor has recently been detected on the Hikurangi subduction at a location ∼10 km west of the southern end of Profile-2 [Kim et al., 2011]. Assuming that the tremor occurs on or near the subduction interface, the tremor is occurring within the region of intermediate resistivity that we interpret to be a zone of upward fluid transport. This is consistent with the view that the intermediate resistivity is a zone of slip enhanced fluid interconnectivity.
 Our thanks go to Weerachi Siripunvaraporn and his co-workers who have made their 3-D MT inversion code available to the research community. Without their generosity much of this work would not have been possible. Our thanks also go to Nick Cozens who assisted with the field work and to Susan Ellis, Martin Reyners, Laura Wallace, Mark Rattenbury and Nick Mortimer who helped our understanding of the subduction process and geological context. Finally the MT measurements would not have been possible without the cooperation of the many farm and forest owners in the area. In particular, EW would like to thank Pia Pohatu from the He Oranga mo Nga uri Tuku Iho Trust for her help.
 The Editor wishes to thank an anonymous reviewer for assistance in evaluating this paper.