Is the electrical conductivity of the northwestern Pacific upper mantle normal?

Authors


Abstract

[1] The Normal Oceanic Mantle project, based on ocean bottom geophysical observations, has been underway since 2010 to investigate the physical state of the oceanic lithosphere and asthenosphere. We have conducted electromagnetic surveys on old (∼130 Ma) seafloor in the northwestern Pacific region, where no active tectonic processes have been identified, in order to image the electrical conductivity structure beneath the region. So far, data have been collected at four sites through a pilot survey conducted from June 2010 to August 2012. A one-dimensional electrical conductivity model was obtained by preliminary analysis of the data by using the magnetotelluric method. The model shows that the resistive (<0.01 S m−1) lithospheric mantle is as thick as ∼80 km, and that the asthenospheric mantle below has a conductivity of ∼0.03 S m−1. The resistive layer is slightly thicker than that beneath the Philippine Sea but significantly thinner than that beneath the area off the Bonin Trench in the Pacific Ocean. There is a greater difference in age between the survey area and the Philippine Sea (0–60 Ma) than between the survey area and the area off the Bonin Trench (140–155 Ma). This comparison suggests that the relation between age and lithospheric thickness is not as simple as that predicted by the concept of lithospheric cooling. It also suggests that the lithosphere beneath the area off the Bonin Trench in the Pacific Ocean is abnormally thick if the mantle beneath the survey area in this study is “normal,” as expected from the plate cooling model.

1. Introduction

[2] The horizontal flow zone between the upwelling and downwelling of mantle convection, which occupies a large portion of the ocean floor on its surface, is thought to represent a “normal” mantle that is located away from tectonic activities. Investigation of the “normal” mantle is thus expected to reveal the fundamental nature of mantle dynamics. The Normal Oceanic Mantle (NOMan) project (http://www.eri.u-tokyo.ac.jp/yesman/) has been underway since 2010 for the investigation of the “normal” oceanic mantle through years-long marine seismic and electromagnetic (EM) observations in the northwestern Pacific region. One of the goals of the project is to elucidate the physical condition of the lithosphere-asthenosphere boundary (LAB). There are various views regarding the fluidity of the asthenosphere, such as the effects of partial melting, grain size reduction, and/or hydration [e.g., Karato, 2012, and references therein]. The analysis of the seismic and EM data sets collected by the project is expected to distinguish these effects. The EM component contributes to the success of the project through imaging electrical conductivity structure of the mantle beneath the observation areas.

[3] Two survey areas in the northwestern Pacific Ocean were selected for the research: northwest (Area A) and southeast (Area B) of the Shatsky Rise (Figure 1). Both areas have seafloor as old as 130–145 Ma, which has never been investigated. The relatively flat seafloor in these areas indicates that the mantle underneath is likely “normal.” The criteria given by Korenaga and Korenaga [2008] support the fact that the area is classified as Normal Ocean, mostly away from bathymetric anomalies.

Figure 1.

Location of the observation sites of the NOMan project superimposed on a bathymetry map. Crosses, squares, and a star indicate the sites of OBEMs, EFOSs, and the Kakioka geomagnetic observatory (KAK), respectively. Yellow crosses and squares indicate the sites of the pilot survey. Contour lines indicate seafloor age [Müller et al., 2008]. Inset shows the close-up of the pilot survey area, but the bathymetry map is based on the collection of multinarrow beam echo sounding data (no available data in the white area). Labels are the site names.

[4] The main observation phase was initiated in 2011 and will end in 2014 (the observation is ongoing, replacing instruments every year). For the EM observations, a total of 34 ocean bottom electromagnetometers (OBEMs) have been deployed at 17 sites in Area A and 8 sites in Area B. In addition, newly developed electric field observation systems (EFOSs) [Utada et al., 2013], have also been deployed in Area A. An EFOS has a 10 km long cable at the ocean bottom that is installed by a remotely operated underwater vehicle (ROV). Much longer electric dipole of EFOS than OBEM enables us to measure an electric field with higher signal to noise ratio, which is applied to probe deeper in the mantle. Prior to the initiation of the main observation phase, a pilot survey was conducted at five sites in Area A from June 2010 to August 2012. The EM observation array consisted of five OBEM sites and one EFOS site (Figure 1 and Table 1). The instruments were successfully recovered except for the OBEM at site NM03.

Table 1. Information on the EM Sites of the NOMan Project Pilot Survey
SiteInstrumentLatitudeLongitudeDepth (m)Available Data Length
  1. a

    Failed to recover.

NM01OBEM39°12.01′N154°47.09′E575719 Jun 2010 to 8 Mar 2012
NM02OBEM39°42.09′N153°21.17′E573524 Jun 2010 to 27 Feb 2012
NM03OBEM38°45.84′N155°54.73′E5766N/Aa
EFOS38°45.81′N155°54.69′E576621 Jun 2010 to 24 Aug 2012
NM04OBEM38°12.67′N154°11.40′E594719 Jun 2010 to 27 Nov 2011
NM05OBEM40°15.00′N155°24.40′E561724 Jun 2010 to 27 Feb 2012

[5] This paper presents a preliminary result from the EM component of the pilot survey. We introduce the acquired data set and a one-dimensional (1-D) electrical conductivity structure model that best approximates the upper mantle (down to ∼400 km depth) beneath the pilot survey area. Although we aim to elucidate not only the vertical but also the lateral heterogeneity, we reserve three-dimensional (3-D) analysis until the complete data set is acquired through the main observation phase. The obtained 1-D model should be used as a background or a starting model of the future 3-D inversion analysis. Here, we briefly compare the model obtained from the NOMan pilot survey with the 1-D conductivity models obtained similarly in our previous study on the mantle structure beneath the Philippine Sea and the area off the Bonin Trench in the Pacific Ocean [Baba et al., 2010]. Then, we discuss the differences in the models in terms of the lithosphere thickness.

2. Data Analysis

[6] We analyzed the data collected by the OBEMs at the four sites of the pilot survey, based on the magnetotelluric (MT) method. The time series of the observed EM field was first edited to remove abnormal fluctuations such as spikes and rectangular steps. They were detected by eyes comparing different field components. Then, the spikes were linearly interpolated and the steps were shifted to reduce the discontinuity. Further, the instrumental clock shift and tilt were corrected. The coordinate system of the EM data was adjusted to a geographical one in which the x, y, and z directions are northward, eastward, and vertical downward, respectively. The International Geomagnetic Reference Field (IGRF) [IAGA Working Group V-MOD, 2010] was used for declination correction. For the electric field data, drifts longer than 5 days were removed by polynomial fitting. Quasiperiodic solar quiet daily geomagnetic variations (Sq), ocean tides, and their harmonics for each time series component were removed by fitting sinusoids for these periods, as reported by Lizarralde et al. [1995]. The cleaned time series data were processed into the period-dependent and site-dependent MT impedance tensor relating the horizontal magnetic field variations to the horizontal electric field variations and geomagnetic transfer functions relating the horizontal magnetic field variations to the vertical magnetic field variations, using a bounded influence algorithm [Chave and Thomson, 2004]. The magnetic field at the Kakioka observatory was used as a remote reference to reduce the effect of site-dependent noise.

[7] The obtained MT impedance tensors and geomagnetic transfer functions are plotted as the sounding curves of the apparent resistivity and phase in Figure 2 and as the phase tensor ellipses [Caldwell et al., 2004] and the induction vectors with Parkinson's convention [Parkinson, 1959] in Figure 3, respectively. The overall feature of the MT responses is similar among the sites. The apparent resistivity of the xy element is the largest, that of the yx element is secondary, and the apparent resistivity of the diagonal elements is much smaller than that of the off-diagonal elements. The major apparent resistivity has a peak at around 500 s, suggesting a relatively conductive uppermost lithosphere or crust, a resistive lithospheric mantle, and a conductive asthenospheric mantle below. The splitting of the xy and yx elements tends to be larger in the western site (NM02) and smaller in the eastern site (NM05). However, the principal axes of the MT impedance detected from the phase tensor ellipses are different from the x (north) and y (east) directions. The major axes of the ellipses in the middle period range are around N60°E, which is approximately parallel to the strikes of the past seafloor spreading axis in the area and the Kuril Trench (Figure 1). The phase tensor skews indicate that the three-dimensionality is small but not negligible compared to the confidence limit. The features of the induction vector are slightly more complicated. The vectors of the real part of the periods shorter than ∼104 s point to the south to SSE for NM01 and NM05 and the north to NNW for NM02, which are roughly perpendicular to the major axis of the phase tensors. For NM04, they rotate from the northeast to southeast as the period increases. These apparently non-1-D features are explained qualitatively by the topographic effect, demonstrated later. Abrupt changes in the responses appear at ∼104 s, especially in the xy and yy elements of the MT sounding curves and the induction vectors. The three-dimensionality indicated by the phase tensor skew also abruptly increases at the period. These are probably nonuniform source effects because of Sq and tides, which were not perfectly removed from the raw time series data [Shimizu et al., 2011].

Figure 2.

MT responses in terms of (top) the apparent resistivity and (bottom) the phase. Symbols with error bars and lines are the ones observed and calculated from a conductivity model consisting of the 3-D topographic heterogeneity over the 1-D mantle (shown in Figure 5a), respectively. Colors correspond to the different elements, as indicated in the top right plot. The error bars indicate 95% confidence limits.

Figure 3.

(left) MT phase tensor ellipses and (right) the induction vectors of the geomagnetic transfer functions. North is pointing up. Each column shows those at different site indicated at the top. The period increases from the top to the bottom as shown in the center. Black ellipses with gray shades and light red and light blue arrows with ellipses are the observed ones with 95% confidence limits. Green ellipses and deep red and deep blue arrows are the ones calculated from a conductivity model consisting of the 3-D topographic heterogeneity over the 1-D mantle (shown in Figure 5a), respectively.

[8] We estimated a 1-D conductivity structure model that fits the averaged MT responses of all sites, applying the topographic effect correction iteratively. This procedure is critical to obtain a reliable mantle conductivity structure model because the large contrast in the conductivity between seawater and crustal rocks can severely distort the observed EM field. We used the square root of the determinant of the MT impedance tensor, which is invariant for the horizontal rotation of the coordinate system, as a representative 1-D response for each site, and averaged the values calculated at the sites. The averaged response was then inverted using Occam's inversion method [Constable et al., 1987] to obtain a 1-D model. Next, the topographic effect was simulated with 3-D forward modeling with a model that contains 3-D surface heterogeneity representative of topography/bathymetry, overlying a 1-D structure which is solved for by inversion. The topographic effect was then removed from the observed MT responses [Baba and Chave, 2005] and the 1-D inversion was run again using the corrected MT responses in the same manner as above. These procedures were iterated until the root-mean-square (RMS) normalized misfit between the observed (noncorrected) and simulated MT impedance tensors of all sites changed less significantly (less than 1%). For the final model, the uncertainty of conductivity for each layer was estimated by running a number of inversions with different a priori constraints. This iterative topographic effect correction and inversion method is the same as that demonstrated in the previous study [Baba et al., 2010], except for the 3-D forward modeling, to which we applied a new approach proposed by Baba et al. [2013]; the details of the method can be obtained from these papers.

[9] The method was applied to the data under the following conditions. The 3-D surface heterogeneous model was constructed using ETOPO1 data [Amante and Eakins, 2009] for regional large-scale topography and a collection of multinarrow beam echo sounding data for local small-scale topography. The lateral dimension of the model was 5500 km × 5100 km. The conductivities of the seawater and land crust were assumed to be 3.2 and 0.01 S m−1, respectively. In the 1-D inversion, the conductivity at ∼20 km depth was constrained to be close to 3.2 × 10−4 S m−1. These parameters also followed Baba et al. [2010]. The iteration converged at the third cycle.

[10] The final RMS misfit was 8.47, while the initial value was 11.92. The reduction was 29%. One may be impressed that the RMS misfits are large. But, note that we did not apply error floor for the RMS misfit calculation because we did not invert the observed MT responses directly. The relative errors in the middle period range are around 1% for the off-diagonal elements and several to ∼10% for the diagonal elements (Figure 4), which are quite good for seafloor MT data. Applying the same procedure assuming a uniform half space as the mantle structure, we obtained the mantle conductivity of 0.026 S m−1 and the RMS misfit of 23.00. The reduction of the RMS misfit from this simplest model to the best 1-D model is as large as 63%. This fact suggests that the vertical change in conductivity is primary feature of the mantle beneath the study area although the lateral heterogeneity corresponding to the major part of the misfit of 8.47 should not be ruled out.

Figure 4.

(left) Relative errors of the observed MT impedance and (right) the residuals (the difference in log apparent resistivity (circles) and phase (diamonds) between observed and calculated from a conductivity model consisting of the 3-D topographic heterogeneity over the 1-D mantle shown in Figure 5a) normalized by the data error. Colors correspond to the different sites, as indicated in the top left plot.

[11] The MT responses and induction vectors calculated from the final model are also plotted in Figures 2 and 3. In the periods shorter than ∼104 s, the calculated MT responses (the sounding curves and phase tensors) reconstructed the overall features of the observations. However, the splitting between the xy and yx elements is smaller than the observations, which yields relatively larger misfits (Figure 4). It is also demonstrated by the phase tensor ellipses as their shape and direction of the major axis diverse from the observations at around 103 s. The features of the observed induction vectors were also roughly reconstructed although they were not used for the 1-D model estimation. The major discrepancies are seen in smaller dimensions of the calculated induction arrows than the observed ones. These results suggest that non-1-D features of the MT responses and geomagnetic transfer functions are roughly explained by the topography taken into account and that the discrepancies may be explained by lateral heterogeneity and/or anisotropy in the conductivity of the mantle, which needs to be examined in future studies.

3. Results

[12] The obtained 1-D model shows the resistive upper layer and the underlying conductive zone, as anticipated from the apparent resistivity curves (Figure 5a). The resistive (<0.01 S m−1) layer extends down to ∼80 km depth. The conductivity changes gradually at ∼0.03 S m−1 below ∼120 km depth. This high-conductivity zone is better constrained by the data. Kawakatsu et al. [2009] found a sharp boundary in the seismic velocity—considered to be the LAB—at ∼80 km depth beneath the borehole seismic observatory, WP2, which is ∼400 km to the east of the survey area in this study. The depth range at which the conductivity increases rather steeply is comparable to the estimation of the seismic LAB depth. However, the inversion adopted in the present study includes a priori constraints so that the model changes smoothly with depth; therefore, it is originally hard to resolve a discontinuous change in the conductivity at a depth corresponding to the LAB. We tested if the MT data have any sensitivity to a sharp increase in conductivity. We run inversions relaxing the smoothness constrain at various depths between 50 and 120 km and obtained models that have conductivity jump (Figure 5b). All models fit the data as well as the smooth model does. These results indicate that the sharp conductivity increase at similar depth with the seismic LAB may exist but it is not required by the current MT data.

Figure 5.

(a) 1-D electrical conductivity models. Black line is the one obtained from the averaged MT response. Gray shades represent, respectively, 70% and 95% limits of the uncertainty of the conductivity estimates at each depth. Color lines are the models obtained through the site by site analysis. (b) 1-D conductivity models obtained by the inversions relaxing the smoothness constraint at various depths. Gray shades represent the uncertainties for the model which has a jump at ∼75 km depth.

[13] The electrical conductivity structure in the mantle is possibly three-dimensional. Although we reserve detailed 3-D inversion analysis until the ongoing observation is complete, here we attempted a simple test to estimate possible lateral heterogeneity beneath the pilot survey area. We applied the iterative topographic effect correction and inversion of the determinant average MT response site by site to obtain a 1-D model beneath each site. The RMS misfits were 6.06, 8.83, 9.04, and 8.10 for NM01, NM02, NM04, and NM05, respectively. They are slightly better than the partial RMS misfit of the average 1-D model for each site; 6.25, 9.41, 9.22, and 8.62. The resultant 1-D models are shown in Figure 5a. The difference among the models can be seen at the depths between ∼30 and ∼130 km. The model for NM02 is slightly more conductive and those for NM04 and NM05 are more resistive than the average model. The model for NM01 is very close to the average model. These differences are as small as all models fall within the 70% uncertainty of the average model.

4. Discussion

[14] The average 1-D model is compared with those developed for relatively young (0–60 Ma) Philippine Sea mantle and very old (140–155 Ma) Pacific mantle beneath the area off the Bonin Trench obtained by our previous study [Baba et al., 2010]. The 1-D model in this study falls between these two 1-D models (Figure 6a). The thickness of the resistive layer is ∼50 km for the Philippine Sea mantle but ∼200 km for the Pacific mantle beneath the area off the Bonin Trench. The resistive layer beneath Area A is slightly thicker than that beneath the Philippine Sea, but the difference is less significant if the 95% uncertainty of the model is considered. The difference between this model and the one for the Pacific mantle beneath the area off the Bonin Trench is significant, even if the 95% uncertainty of the two models is considered. This is a surprising result for the small age difference (10–15 Ma) between the two old oceanic lithospheres.

Figure 6.

(a) 1-D electrical conductivity structure models for the three areas. Red, green, and blue lines with shades show the model in this study (same as the model in Figure 5a), and the models for the Philippine Sea mantle and the Pacific (off the Bonin Trench) mantle obtained by Baba et al. [2010], respectively. Dark and light shades represent, respectively, 70% and 95% limits of the uncertainty. Yellow line is the conductivity profile predicted from the plate cooling geothermal model for 130 Ma mantle shown in (b) and dry olivine conductivity by Constable [2006]. (b) Geothermal profiles predicted from the 125 km thick plate cooling model (solid lines) and the half-space cooling model (dashed lines), where the potential temperature of 1350°C and the adiabat of 0.3°C km−1 are involved for both models. Colors indicate different lithospheric ages: 130, 30, and 147 Ma, corresponding to the average ages of the three areas.

[15] The relationship between the electrical conductivity structure of the oceanic upper mantle and the lithospheric cooling with age has long been discussed since seafloor MT study started [e.g., Filloux, 1977] and it was recently reviewed by Baba [2005] and Ichiki et al. [2009]. Modern understanding on it given by these reviews can be summarized as follows. The age dependency of the resistive layer thickness is observed for older mantle where the thermal boundary layer become thicker than the depth of peridotite dry solidus (∼60 km), while the compositional control associated with partial melting is more significant for younger mantle. The transition of the dominant factor from the composition to the temperature should occur at ∼30 Ma for the mantle with the typical potential temperature of 1350°C [Baba, 2005].

[16] The 1-D conductivity model for the Philippine Sea is the average of the mantle with wide age range of 0–60 Ma so that it must include both effects but indistinctive. Actually, the transition depth to the high-conductive layer approximately matches the depth of the solidus and the depth of the transition from conductive to adiabatic geotherm. The other two models have thicker resistive layer so that the thermal structure should mainly control it. For very old mantle (>∼80 Ma), there is a discussion on the age-dependency of thermal structure. The bathymetry subsidence and heat flow change with age support cooling of a plate which has finite thickness [e.g., Parsons and Sclater, 1977]. However, Baba et al. [2010] discussed that the temperature estimated form the conductivity model for the Pacific mantle beneath off the Bonin Trench is more consistent with that predicted from a half-space cooling model rather than a plate cooling model. For the 1-D model in this study, the conductivity change with depth decreases at depths of 100–150 km, which is rather consistent with the plate cooling model. If the conductivity for both areas is dominantly controlled by lithospheric cooling, the relation between age and lithospheric thickness for the old mantle is not as simple as what was previously considered. Otherwise, the mantle beneath Area A is “normal,” as predicted from the plate cooling model, and that beneath the area off the Bonin Trench may have an abnormally thick lithosphere for unknown reasons. One possible reason for such lateral heterogeneity is the abrupt change in the conductivity structure toward the trench through, for example, plate deformation, which may also cause petit-spot volcanism [Hirano et al., 2006]. Another possibility is the presence of sub-oceanic small-scale convection [e.g., Richter, 1973], which produces variations perpendicular to the plate motion. The conductivity structure beneath Area B may be key to understand this issue because it is another sample of the old mantle whose lithospheric age is 135–145 Ma (Figure 1).

[17] The conductivity values below the resistive layer are similar among the three areas. They overlap at around 250 km depth. The values ∼0.03 S m−1 are higher than those predicted from the conductivity of dry olivine under possible mantle temperature (∼0.01 S m−1 for ∼1400°C) even if the 95% lower limit of the model uncertainty is considered (Figure 6a). Therefore, some mechanisms enhancing the conductivity (higher temperature, partial melting, and/or mantle hydration) are required to explain the differences between the observations and prediction. Baba et al. [2010] discussed that it is hard for such deep mantle to produce partial melt even if the reduction of solidus by mantle hydration is considered and estimated the maximum water content of ∼0.01 wt % based on laboratory measurements of hydrous olivine. The above discussion ignored the effect of carbonated melt which is thought to be stable in that depth and contribute to the enhancement of the bulk conductivity much more than basaltic melt [e.g., Gaillard et al., 2008]. Utada and Baba (U. Utada and K. Baba, 2013, Estimating the electrical conductivity of the melt phase of a partially molten asthenosphere from seafloor magnetotelluric sounding data, submitted to Physics of the Earth and Planetary Interiors) proposed a new approach to estimate the conductivity of the melt phase from observationally obtained conductivity models giving possible geotherm and melt fraction from mineral physics or seismological observations, under the assumption of partially molten asthenosphere. They applied it to the three 1-D models demonstrated in this study and indicated that the obtained melt conductivity fell between the values for silicate and carbonatite melts obtained by laboratory experiments [e.g., Gaillard et al., 2008]. It was emphasized the importance of consistency among different methods, such as electromagnetics, seismology, and mineral physics to test two major hypotheses for the origin of the asthenosphere, due either to partial melting or to hydration.

5. Conclusions

[18] Seafloor MT data collected from the four sites in the northwestern Pacific region were analyzed, and a 1-D electrical conductivity structure model of the upper mantle beneath the area was estimated. The model shows that the upper resistive (<0.01 S m−1) layer is as thick as ∼80 km and the highly conductive zone below has a conductivity of ∼0.03 S m−1. The resistive layer is slightly thicker than that beneath the Philippine Sea but significantly thinner than the one beneath the area off the Bonin Trench in the Pacific Ocean, although the age difference is larger between the Philippine Sea (0–60 Ma) and Area A (∼130 Ma) than between the area off the Bonin Trench (140–155 Ma) and Area A. These facts suggest that the relation between age and lithospheric thickness is not as simple as that predicted by the concept of lithospheric cooling, and that the lithosphere beneath off the Bonin Trench is abnormally thick if the mantle beneath the survey area in this study is considered “normal,” as expected from the plate cooling model. At least, large-scale lateral heterogeneity is most likely present beneath the oldest Pacific plate.

Acknowledgments

[19] The authors thank the captains, officers, and crew of R/V KAIREI of JAMSTEC for their dedication to the safety and success of the cruises KR10-08, KR11-10, and KR12–14. Takafumi Kasaya, Koji Miyakawa, Toyonobu Ota, Tsukasa Yoshida, Hitoshi Okinaga, Misumi Aoki, Kyoko Tanaka, Morifumi Takaesu, Satomi Minamizawa, Tatsuro Tanioka, and Hikari Hasegawa are also thanked for their technical assistance. Bathymetry data based on multinarrow beam echo sounding systems were provided by the JAMSTEC Data Site for Research Cruises. All figures were produced using GMT software [Wessel and Smith, 1998]. Comments by two anonymous reviewers yielded improvements in the manuscript. This study was partially supported by grant-in-Aid for Scientific Research 22000003, the Japan Society for the Promotion of Science (JSPS).