Trans-Pacific temperature field in the mantle transition region from seismic and electromagnetic tomography Yoshio Fukao 1,3, Takao Koyama 2, Masayuki Obayashi 1 and Hisashi Utada 3 1 Research Program for Mantel Core Dynamics, Institute for Frontier Research on Earth Evolution (IFREE) 2 Center for Data and Sample Analyses, Institute for Frontier Research on Earth Evolution (IFREE) 3 Earthquake Research Institute, University of Tokyo This is the first attempt of synthesizing the images of temperature field at a semi-global scale from seismic and electromagnetic tomography. Introduction The first order structure of the Earth s mantle is vertically stratified. For example, seismic P and S velocities increase by 70% with increasing depth from the top to bottom of the mantle (Bullen, 1963), while their lateral variations remain within several percent at any depth (Su et al., 1994; Li and Romanowicz, 1996). This essentially vertically stratified feature, together with the vigor of plate motion at the Earth s surface, is diagnostic of solid-state convection within the mantle, which is most likely to be of thermal origin. Many of the recent geodynamic models attempted to relate, either explicitly or implicitly, the distribution of seismic velocity anomalies derived from global tomography to the temperature variations associated with the mantle convection (Yuen et al., 1993), which is then converted to thermally induced density anomalies by which mantle convection is driven (Hager et al., 1985; Forte and Woodward, 1997). A difficulty in such modeling is the nonuniqueness of converting seismic velocity anomalies to temperature anomalies with the possible significance of lateral heterogeneity of chemical composition in the mantle (Forte and Claire Perry, 2000). Seismic tomography versus geoelectromagnetic tomography Electric conductivity is a temperature-sensitive parameter, although it also depends strongly on the degree of partial melting and the dissolution of hydrogen or water (Karato, 1990). For example, for Mg-rich olivine (Mg,Fe) 2 SiO 4, the relative temperature dependences of P and S wave velocities, d(ln Vp)/dT and d(ln Vs)/dT, are -0.54x10-4 /K and -0.73x10-4 /K at room temperature (Anderson et al., 1968), while the relative temperature dependence of electric conductivity, d(ln σ)/dt, is 6.4x10-3 /K at 1700K (Xu et al., 1998). Because sensitivities of electric conductivity to temperature and other environmental parameters are very different from those of seismic velocities, simultaneous mapping of seismic velocity and electric conductivity distributions should more narrowly constrain the temperature field in the mantle than seismic mapping alone. As compared to seismic tomography, however, three-dimensional electromagnetic tomography is still at a premature stage, in part because the perturbation amplitudes in conductivity are too large to allow an iterative linearlized inversion. Tarits and Alexandnrescu (1998) and Schultz and Pritchard (1999) attempted to back-project the global magnetic field data to the three-dimensional conductivity distribution at depths down to the mid-mantle. More recently, Koyama (2001) inverted the electric and magnetic field data for the conductivity distribution in the mantle transition region, at depths from 350 to 850km under the Pacific Ocean. We compare this result with that of Fukao et al. (2002), who made a whole mantle P-wave travel time tomography by combining two million reported first arrival times with more than 7000 PP-P differential travel times read from broadband seismic records, which greatly improved the resolution in the upper to middle mantle beneath the central Pacific. Table 1 summarizes a comparison of the seismic tomography by Fukao et al. (2002) with the geoelectromagnetic tomography by Koyama (2001). Observations in seismic tomography measure the propagation speed of a change in the elastic field, while observations in geoelectromagnetic tomography measure the diffusive medium response to a change in Earth s electromagnetic field. With this difference the underdetermined nature of the inverse problem is more serious in geoelectromagnetic tomography. However, the lateral variation of electric conductivity is almost three orders of magnitude greater than that of seismic velocity, as demonstrated in Table 1, so that threedimensional geoelectromagnetic tomography should be still feasible. Koyama (2001) used the voltage data of 8 submarine cables and the magnetic field data of 17 geomagnetic observatories, in periods varying from 1 to 10 years (see Fig. 1 for their locations), from which the MT (Magneto-Telluric) and GDS (Geomagnetic Deep Sounding) responses were calculated. Spectral components of the signal range from 0.5 to 10 days in period. The depth range sensitive to this period range is roughly 350-850km. The starting model is a radially symmetric conductivity model with an ocean-land contrast at the top (Utada et al., 2002), that can explain all the relevant data in a consistent way such that their residuals may be attributed to the effect of lateral heterogeneity. Trans-Pacific tomographic images Fig. 1 shows the anomaly map of P-wave velocity at depths of 480-550km in comparison with that of electric conductivity at depths of 450-550km. The corresponding maps from the checkerboard resolution tests are shown in Fig. 2. This test examines to what extent the input pattern of alternate positive and negative anomalies can be recovered from a given synthetic dataset (Inoue et al., 1990). The test pattern for the seismic tomography (Fig. 2a) is reasonably well recovered, except for the eastern Pacific region. The area resolvable in the geoelectromagnetic tomography is, on the other hand, narrowly confined to the vicinity of the observational cables and stations, where 3
the most pronounced features are high conductivity anomalies under Hawaii and those beneath the south Philippine Sea, and low conductivity anomalies beneath the northwestern Pacific (Fig. 1b). The high conductivity anomalies under Hawaii and the low conductivity anomalies beneath the northwestern Pacific in Fig. 1b, correspond well to the low and high P-velocity anomalies in Fig. 1a, respectively. The high conductivity anomalies beneath the south Philippine Sea may be compared to the lower-than-surrounding velocities in the same region. Fig. 3 shows the cross-sections of the tomographic images along a profile across Hawaii, Midway and Japan (Figs. 3a and 3b), and along a profile subparallel to the Hawaii-Guam-Philippines cables (Figs. 3c and 3d). See Fig. 1a for the locations of the profiles. The corresponding cross-sections for the checkerboard resolution tests are shown in Fig. 4. The Hawaiian high conductivity anomaly (Figs. 3b and 3d) extends over a depth range of 400 to 700km, with poor resolution for its upper and lower extensions (Figs. 4b and 4d). The corresponding low velocity anomaly (Figs. 3a and 3c) extends upward well beyond this depth range, but its downward extension is limited to the uppermost part of the lower mantle (Fukao et al., 2002). The northwestern Pacific low conductivity anomaly (Fig. 3b) extends between the approximate depths of 350 to 650km, below which a zone of high conductivity anomaly exists at a marginal level of resolution (Fig. 4b). The seismic tomography (Fig. 3a) indicates that this is attributable to the cold slab of the Pacific plate downgoing from the Japan trench. This very strong slab anomaly is accompanied by a body of slow anomalies on the Pacific side. The association of a slow anomaly on the Pacific side of the subducting slab is a resolvable feature (Fig. 4a), as detailed by Obayashi et al. (2002). Such an association is not visible on the conductivity profile, however. The high conductivity anomaly beneath the south Philippine Sea (Fig. 3d) is a relatively shallow feature limited in depth to about 550km, below which low conductivity anomalies dominate. This vertical pair of high and low conductivity anomalies appears to be a resolvable feature in the light of the checkerboard resolution test (Fig. 4d). The corresponding seismic crosssection (Fig. 3c) is complicated by the presence of the three subducted slabs of the Pacific plate from the Mariana trench, the Philippine Sea plate from the Philippine trench, and the Indo- Australia plate from the Java trench, which are imaged as fast anomalies. The intervening zones of low velocity anomaly and lower-than-surrounding velocity at relatively shallow depths may be correlated to the relatively shallow high conductivity anomaly beneath the Philippine Sea (Fig. 3d). The underlying low conductivity anomaly clearly corresponds to the high velocity anomalies associated with the subducted slabs of the Pacific, Philippine Sea and Indo-Australia plates that are now stagnant beneath the 660-km discontinuity (Fukao et al., 2001). Conversion to temperature anomaly field The above comparison of images from the seismic and geoelectromagnetic tomography points to a correlation between the high (low) velocity anomalies and low (high) conductivity anomalies, although the correlation remains qualitative owing to a large difference in resolution. Such a correlation indicates thermal origins for both the velocity and conductivity anomalies. Assuming that this is the case, we convert the velocity and conductivity anomalies to temperature anomalies to see the consistency (or inconsistency) between the two separately estimated temperature fields. For the velocity-to-temperature conversion we used a depth-dependent conversion factor c proposed by Karato (1993) on the ground of mineral physics: dt 3D = c[dv 3D /V 1D ] (1) where dt 3D is the lateral temperature perturbation from the radially symmetric temperature distribution T 1D, and dv 3D is the lateral velocity perturbation from the radially symmetric velocity distribution V 1D. For the conductivity-to-temperature conversion, we used the following formula: dt 3D /T 1D = ln[s 3D /S 1D ]/ln[s 1D /S O ] (2) where S 1D is the radially symmetric conductivity model used as a starting model for inversion for the three-dimensional conductivity model S 3D. S O is the pre-exponential term of the Arrhenius' formula for the representative mantle mineral in the relevant depth range: S = S O exp( H/kT) (3) where H is activation enthalpy and k is Boltzman s constant. Equation (2) may be derived from (3), assuming that dt 3D /T 1D << 1. For T(1D) at depth d, we used a model of Ito and Katsura (1989): T(1D) = 0.5d+1500, where T and d are measured in K and km, respectively. For S O, we used a laboratory value of S O for ringwoodite (Xu et al., 1989), a major mineral phase in the mantle transition zone. Fig. 5 shows the cross-sections for dt 3D obtained from the seismic tomography and geoelectromagnetic tomography, respectively. The Hawaiian hot anomaly is in excess of 200-300K (Figs. 5a and 5c) or 300-400K (Figs. 5b and 5d). It has been inferred, as in the cartoon by Nataf (2000), that the temperature of a mantle plume is about 300K higher than that of the normal upper mantle and about 500K higher than that of the normal lower mantle. The northwestern Pacific cold anomaly is 300-400 K (Fig. 5a) or 100-200 K (Fig. 5b) lower than the average. It has been inferred (e.g., Turcotte and Schubert, 1982) that subducting slabs are even more than 300-400K colder than the surrounding mantle. The Philippine Sea anomaly is a combination of a hot anomaly of 100-200K (Fig. 5c) or 300-400K (Fig. 5d) and an underlying cold anomaly with 300-400K (Fig. 5c) or 200-300K (Fig. 5d) below the normal. Discussion and conclusions Fig. 5, thus, seems to suggest that seismic tomography tends to emphasize cold anomalies and that geoelectromagnetic tomography tends to emphasize hot anomalies. This difference may be explained as follows. The seismic tomography by Fukao et al. (2002) relies mostly on first arrival times of direct P waves which propagate preferentially through faster (colder) zones as off-great circle ray path effects. Since geoelectromagnetic tomography senses conductivity anomalies as total amplitude differences of conductivity which varies by more than an order of magnitude (Table 1), it tends to sense preferentially more conducting (hotter) regions than less conducting (colder) regions. Taking into account such a sensitivity difference, as well as limitations in resolution and accuracy of inversion and uncertainties from conversion formulae (1) and 4
(2), we consider that the temperature fields estimated from the seismic and geoelectromagnetic tomography are in reasonable agreement. The temperature anomalies estimated are on the order of ±200-400K in the mantle transition region. References Anderson, O. L., E. Schreiber, R. C. Liebermann, and N. Soga, Some elastic constant data on minerals relevant to geophysics, Rev. Geophys., 6, 491-524, 1968. Broyden, C. G., J. E. Dennis, Jr., and J. J. More, On the local and superlinear convergence of quasi-newton methods, J. Inst. Math. Appl., 12, 223-245, 1973. Bullen, B. Introduction to the Theory of Seismology, 381pp., Cambridge Univ. Press. New York, 1963. Forte, A. M., and H. K. Claire Perry, Geodynamic evidence for a chemically depleted continental tectosphere, Science, 290, 1940-1944, 2000. Forte, A. M., and R. L. Woodward, Seismic-geodynamic constraints on three-dimensional structure, vertical flow, and heat transfer in the mantle, J. Geophys. Res., 102, 17981-17994, 1997. Fukao, Y., S. Widiyantoro, and M. Obayashi, Stagnant slabs in the upper and lower mantle transition region, Rev. Geophys., 39, 291-323, 2001. Fukao, Y., A. Toh, and M. Obayashi, Whole mantle P-wave tomography using P and PP-P data, J. Geophys. Res., in press, 2002. Hager, B. H., R. W. Clayton, M. A. Richards, R. P. Comer, and A. M. Dziewonski, Lower mantle heterogeneity, dynamic topography and the geoid, Nature, 313, 541-545, 1985. Inoue, H., Y. Fukao, K. Tanabe, and Y. Ogata, Whole mantle P-wave travel time tomography, Phys. Earth Planet. Inter., 59, 294-328, 1990. Ito, E., and T. Katsura, A temperature profile of the mantle transition zone, Geophys. Res. Lett., 16, 425-428, 1989. Karato, S., The role of hydrogen in the electrical conductivity of the upper mantle, Nature, 347, 272-273, 1990. Koyama, T., A study on the electrical conductivity of the mantle by voltage measurements of submarine cables, Ph.D Thesis, University of Tokyo, 2001. Li, X. D., and B. Romanowicz, Global mantle shear-velocity model using nonlinear asymptotic coupling theory, J. Geophys. Res., 101, 22245-22272, 1996. Nataf, H., Seismic imaging of mantle plumes, Ann. Rev. Earth Planet. Sci., 28, 391-417, 2000. Obayashi, M., H. Sugioka, and Y. Fukao, Seismic slow anomalies in the deep upper mantle oceanward of subducting slabs, Program Abstract, Japan Earth and Planetary Science Joint Meeting, 2002. Schultz, A., and G. Pritchard, Three-dimensional inversion for largescale structure in a spherical domain, SEG, Geophysical Developments Series, 7, 451-476. Su, W. J., R. L. Woodward, and A. M. Dziewonski, Degree 12 model of shear velocity heterogeneity in the mantle, J. Geophys. Res., 99, 6945-6980, 1994. Tarits, P., and M. Alexandrescu, 3D analysis of very long period geomagnetic data: The conductivity of the lower mantle, The 14 th workshop on electromagnetic induction in the Earth, 90-91, 1998. Turcotte, D. L., and G. Schubert, Geodynamics, Applications of Continum Physics to Geological Problems, 450pp., Jone Wiley & Sons, New York, 1982. Utada, H., T. Koyama, H. Shimizu, and A. D. Chave, A semi-global reference model for electrical conductivity in the mid-mantle beneath the north Pacific region, Geophys. Res. Lett., submitted, 2002. Xu, Y., B. T. Poe, T. J. Shankland, and D. C. Rubie, Electrical conductivity of olivine, Wadsleyite, and ringwoodite under upper-mantle conditions, Science, 280, 1415-1418, 1998. Yuen, D. A., O. Cadek, A. Chopelas and C. Matyska, Geophysical influences of thermal-chemical structures in the lower mantle, Geophys. Ress. Lett., 20, 899-902, 5
Figure 1. (a) P-wave velocity anomaly distribution at depths 480-550km (Fukao et al., 2002). High and low velocity anomalies are colored blue and red, respectively in linear scale. Tomographic cross-sections will be taken along the two profiles shown here. (b) Electric conductivity anomaly distribution at depths 450-550km (Koyama et al., 2002). High and low conductivity anomalies are colored red and blue, respectively, in logarithmic scale. The solid lines with circles indicate submarine geopotential cables and the isolated circles represent magnetometer stations. Figure 2. (a) Result of checkerboard resolution test for the P-wave velocity anomaly distribution at depths 480-550km (Fukao et al., 2002). (b) Result of checkerboard resolution test for the electric conductivity anomaly distribution at depths 450-550km (Koyama et al., 2002). See Fig. 1 for other explanations. Figure 3. (a) Cross-section of velocity anomalies at depths 300-1000km along the Japan-Hawaii profile. (b) Cross-section of conductivity anomalies at depths 350-850km along the Japan-Hawaii profile. (c) Cross-section of velocity anomalies along the Japan-Hawaii profile. (d) Cross-section of conductivity anomalies along the Philippines-Hawaii profile. See Fig. 1 for the locations of the two profiles and for other explanations. 6
Figure 4. Results of checkerboard resolution tests along the two profiles shown in Fig. 1. See Fig. 3 for other explanations. Figure 5. (a) Cross-section of temperature anomalies at depths 300-1000km along the Japan-Hawaii profile from the seismic tomography. (b) Cross-section of temperature anomalies at depths 350-850km along the Japan-Hawaii profile from the geoelectromagnetic tomography. (c) Cross-section of temperature anomalies along the Japan-Hawaii profile from the seismic tomography. (d) Cross-section of conductivity anomalies along the Philippines-Hawaii profile from the geoelectromagnetic tomography. See Fig. 1 for the locations of the two profiles and for other explanations. Table 1. 7