Sharp lateral Boundaries in the D'' Region

Akiko To and Barbara Romanowicz


Global shear velocity tomographic models show the existence of two superplumes lying in the lower mantle. The large-scale low velocity structures are observed under southern Africa and the mid-Pacific (`Romanowicz , 2003). Recently, sharp lateral transitions in the velocity structure have been reported at the borders of the African superplume(Ni et al., 2002) and Northeast side of the Pacific suplerplume (Bréger and Romanowicz, 1998). Here we report that a sharp lateral boundary also exists at the southern edge of the Pacific superplume. The set of SHdiff waveforms, which graze the South Pacific, have similar features to those observed at the southeastern edge of the African superplume. The arrival time shifts rapidly, with the emergence of multiple pulses, depending on the azimuth. The coupled mode/spectral element method (CSEM in what follow) (Capdeville et al., 2003) is used to construct synthetic waveforms. We show the first order features of these waveforms from Africa and the South Pacific can be produced by a very simple structure model with strong, sharp lateral heterogeneity lying almost parallel to the ray paths.

Data and Modeling

Figure 32.1(a) shows the locations of the events and stations. Only deep earthquakes(depth > 450 km) are used. Figure 32.1(b) left panel shows the the SHdiff waveforms which graze southeastern edge of African anomaly (hereafter AFA) recorded at the Tanzanian array.They are also shown in other studies (Wen, 2001; Ni and Helmberger, 2004). Figure 32.1(b) right panel shows the SHdiff records of Fiji-Tonga events recorded at the station BDFB Brazil. They graze the Southern part of Pacific slow velocity anomaly region (hereafter PSA). These two sets of SHdiff waveform, are very similar in the following ways. First of all, the onset times of the first arrivals show large delay as the raypath enter the slow region that lies on the northern part. The first arrivals shift about 20 seconds and 12 seconds in the case of AFA and SPA respectively. Second, the waveforms, which graze the transition show an additional pulse indicated with red lines in Figure 32.1(b). This later phase is the feature we model in the following section. The travel time shift observed in PSA is due to a heterogeneity at the base of the mantle, because the differential travel times measurements of Sdiff-SKKS increase steeply for 8 seconds with respect to the back azimuth.

We used a coupled mode and spectral element method for the waveform modeling. The spectral element method is most appropriate technique at present, because it can can handle 1) the propagation of seismic waves in 3D models with strong lateral variations and in spherical geometry and 2) diffracted waves along the core mantle boundary. The drawback of the method is the large numerical cost. The CSEM is a hybrid method that couples spectral element computations with a normal mode solution, so that the spectral element method is used only in the target strongly heterogeneous regions, which is the bottom 370km of the mantle in this study. We compute the synthetic waveforms down to 12 seconds with a corner frequency at 18 seconds. The model has the 1D structure of PREM (Dziewonski and Anderson, 1981) down to a depth of 2591 km, and the 3D model below 2591km, down to the CMB. The model of the 3D part is shown in Figure 32.1(c). The boundaries are displaced 15 degrees from the great circle (shown by a yellow line) that goes though the source.

Figure 32.1(d) shows the synthetic waveforms constructed by CSEM. Each trace is normalized by its maximum amplitude. The source is located on the fast side in Figure 32.1(d) left panel and slow side in the right panel. Multiple pulses are observed in both cases. The possible explanation for each pulse is shown in the caption.

According to the tomographic models, the configurations of the raypaths and the interface of the observed cases are likely to be those of Figure 32.1(d) left panel. The synthetic waveforms capture the following features of the observations: 1) they consist of multiple pulses in the transition region 2) when moving from the fast to the slow region, the arrival of the first and last pulses become closer and finally merge for the stations located in the slow region. This features is observed more clearly in AFA case, but can only be suggested in PSA case.

The observed and synthetic waveforms are presented in different time scale and frequency ranges, to show the qualitative similarity between the waveforms. The time scale of the synthetics is much larger than that of the observed waveforms. However, this is because of the frequency limitation of the present SEM calculations, which is dictated by the computer power available to us. Calculations of the CSEM synthetics to higher frequencies would allow a better separation of these pulses for paths close to the vertical boundary, as seen in the observations. We also note that only a 3% velocity jump is sufficient to explain the observed time shift of 20 sec in AFA. A large contrast of 6% at the boundary is chosen in the models in order to observe each pulse clearly, since a smaller velocity contrast makes the amplitude of the second pulse much smaller.

Discussion and Conclusion

When the wavepath in the D" is quasi-parallel to a sharp vertical boundary, the Sdiff waveforms are accompanied by secondary phases (red lines in Figure 32.1(b) and Figure 32.1(d)). The synthetic tests from the models of Figure 32.1(c) give only a qualitative constraint on the model, which is the existence of a sharp lateral boundary in the D'' region. However, because SEM includes the 3D effects from strong heterogeneous structures, the order of magnitude of the effects to the waveforms is well captured with the simple model.

We show that sharp lateral boundaries, which rise almost like a vertical wall, exist not only in the border of the African plume but also under the south Pacific. This indicates that the low velocity region in the lower mantle under Pacific and Africa, observed as the strong degree-2 pattern in shear velocity tomographic models, have the similar nature also in the finer scales. Unlike the African superplume where the shape and the location of much of the boundaries are revealed, large uncertainty remains in the shape of the Pacific superplume.


We thank the National Science Foundation for support of this research.


Bréger L. and B. Romanowicz, Three-dimensional structure at the base of the mantle beneath the central Pacific. Science , 282 , pp. 718-720,1998.

Capdeville, Y., A. To and B. Romanowicz, Coupling spectral elements and modes in a spherical earth: an extension to the "sandwich" case, Geophys. J. Int., 154 , 44-57, 2003

Dziewonski A.M. and D.L. Anderson, Preliminary reference Earth model. Phys. Earth Planet. Inter. , 25 , pp. 297-356,1981.

Mégnin C. and B. Romanowicz, The three-dimensional shear velocity structure of the mantle from the inversion of body, surface and higher-mode waveforms.Geophys. J. Int., 143 , pp. 709-728,2000.

Ni S. and D. Helmberger, Seismic Modeling Constraints on the South African Superplume draft 2004

Ni S., E. Tan, M. Gurnis and D.V. Helmberger, Sharp sides to the African super plume. Science 296, pp. 1850-1852,2002.

Romanowicz B, Global Mantle Tomography: Progress Status in the Past 10 Years Annu. Rev. Earth Planet. Sci., 31 303-328, 2003

Wen L., Seismic evidence for a rapidly varying compositional anomaly at the base of the Earth's mantle beneath the Indian Ocean.Earth Planet. Sci. Lett., 194 , pp.,2002


Figure 32.1: (a) Earthquakes (stars), stations (triangles), and projections of the raypaths. Background model is the shear velocity model SAW24b16 (Mégnin and Romanowicz,2000) at the CMB. The thick yellow lines show the diffracting portion of the paths on the CMB. (b) Observed velocity waveforms. A bandpass filter with corner frequency at 0.01 to 0.125 (Hz) is applied. Left; Waveforms from the event 19970904 in Fiji-Tonga region recorded at South Africa. Right; Waveforms form 12 events in Fiji-Tonga region recorded at the station BDFB in Brazil. (c) The shear velocity model used in the CSEM synthetic waveform calculation. Each quadrant has either -3% or +3% constant anomaly with respect to PREM. (d) Synthetic waveforms calculated using CSEM for the model shown in (c). The numbers on the left side of each trace indicates the station number. Left;the source is located on the fast side of the interface Right; the source is located on the slow side of the interface The blue lines follow the trough of the first arrivals. When the source is located in the slow anomaly region (Right panel), large postcursors (green lines) are observed at the receivers located in the slow regions. They correspond to paths turning within the velocity gradient. They are observed at stations 221-203. When the source is located in the fast region (Left panel), a reflected wave is observed at the stations in the fast regions and refracted wave, which first enters the fast region and then enters the slow region by refraction, is observed at stations in slow regions. Both waves are shown by red lines. They are observed at stations 81-105.
\epsfig{file=toh04_1.eps, width=14cm}\end{center}\end{figure*}

Berkeley Seismological Laboratory
215 McCone Hall, UC Berkeley, Berkeley, CA 94720-4760
Questions or comments? Send e-mail:
© 2004, The Regents of the University of California