The thickness of the mantle transition zone beneath Ross Island, and parts of the Transantarctic Mountains and East Antarctic Craton has been mapped using data from the 2000–2003 Transantarctic Mountain Seismic Experiment to determine if, as indicated by some tomographic images, an upper mantle thermal anomaly centered beneath Ross Island is a deep-seated feature extending into the mantle transition zone. Some 2700 receiver functions have been stacked using a 3D velocity model, revealing Ps conversions from the mantle transition zone discontinuities at depths of 410 and 660 km. Results yield an average nearly uniform transition zone thickness (266 ± 10 km) that is slightly larger than the global average, implying that the upper mantle thermal anomaly does not likely extend into the transition zone. This finding favors explanations for the upper mantle thermal anomaly invoking a plume head or small-scale convection.
 While all of the seismic studies show a pronounced variation in seismic wave speeds across the East–West Antarctic boundary extending laterally over a wide region to depths of about ∼250 km in the mantle, some of them also indicate that the thermal anomaly could extend much deeper into the mantle [Sieminski et al., 2003; Watson et al., 2006]. Understanding the depth extent of the thermal anomaly is important for placing constraints on how it may have formed. For example, a shallow (i.e., ∼<250 km depth) thermal anomaly would favor models invoking a plume head or small-scale convection, whereas a deeper anomaly would point to larger scale geodynamic processes in the lower mantle, like the broad, lower mantle upwelling associated with the East African rift system [e.g., Park and Nyblade, 2006; Benoit et al., 2006; Simmons et al., 2007].
 In this study, we investigate the depth extent of the mantle thermal anomaly beneath Ross Island, and parts of the TAM and the East Antarctic Craton by mapping topography on the 410 and 660 km discontinuities using receiver function stacks from teleseismic earthquakes recorded by the Transantarctic Mountain Seismic Experiment. The mantle transition zone is bounded by discontinuities at depths of 410 and 660 km that are generally interpreted as mineral phase transformations in olivine [Bina and Helffrich, 1994]. Because the Clapeyron slopes of these phase transitions are temperature dependent and opposite in sign, the thickness of the region between the phase transformations (i.e., the transition zone thickness or TZT), as well as the topography on each discontinuity, can provide information about the thermal structure of the upper mantle (e.g., a thinner-than-average TZT indicates a warm thermal anomaly, a thicker-than-average TZT would be a cooler thermal anomaly). Consequently, by mapping topography on the 410 and 660 km discontinuities and determining the thickness between them, we can address directly whether or not, as suggested by Sieminski et al.  and Watson et al. , there could be a deep-seated thermal anomaly under the study area that extends into the mantle transition zone.
 The data used in this study were recorded between December 2000 and December 2003 by the Transantarctic Mountain Seismic Experiment [TAMSEIS], which consisted of 41 portable broadband seismometers aligned in three arrays (Figure 1). A 16-station array, running roughly northeast to southwest with an average station spacing of 80 km, crossed the TAM and extended into the interior of the EAC (North–South Array, Figure 1). A second 16-station array with a station spacing of 20 km crossed the TAM in a northwest–southeast orientation (East–West Array, Figure 1). Nine coastal stations along with two permanent stations at Scott Base and Wright Valley formed the coastal array (Figure 1). Data from earthquakes with Mb 5.5 and located at a distance between 30–90° from the stations were used for this study. The frequency domain deconvolution method [Ammon, 1991] with Gaussian filter widths of 0.6 and 1.5 was used to generate the receiver functions. The Ps conversion points for P waves at a depth of 660 km are shown to demonstrate the spatial coverage of the dataset (Figure 1).
 Following the method of Owens et al. , a geographical binning technique was used to stack the receiver functions that included a 3D velocity model of the upper mantle. The 3D velocity model was taken from Watson et al. , who used P and S relative travel time residuals to invert for upper mantle velocity variations. Using the 3D velocity model, the travel time and conversion point of every Ps conversion were calculated from 40 to 800 km depth for each station-event pair. The travel time for a Ps conversion at a specified depth was used to extract the amplitude from each receiver function. Then, for that depth, all of the amplitude measurements falling within a circular region (bin) were stacked. Permutations in the stacking procedure that we tried included changing the diameter of the bin, altering the minimum number of measurements in each bin, varying the minimum number of stations represented in each bin, and performing the stacking using the 1-D IASP91 model from Kennett and Engdahl .
 To obtain receiver function stacks with the best possible signal-to-noise ratios but still be able to resolve variations in discontinuity topography over horizontal distances of about 100 km, we found that the optimal stacking parameters included a minimum of ten receiver functions per bin from at least five stations, a bin radius of 75 km, and a bin increment of 25 km. Results obtained using a 3D vs. 1D velocity model were similar. Neither the choice of 0.6 or 1.5 Gaussian filter widths, nor the addition of Ps conversion from PP waves, affected the depths of the discontinuities by more than ±1 km. In addition, only Ps conversions from P waves recorded on stations deployed on rock were used. Receiver function stacks using data from ice stations do not yield clear images of the 410 and 660 km discontinuities, as discussed in the next section. Owens et al.  found that for their method of stacking receiver functions with a fixed velocity model, the precision of the depth estimate of the discontinuities is on the order of ±3 km. However, because of uncertainties in crustal thickness [Lawrence et al., 2006b] and upper mantle velocities [Watson et al., 2006] for our study area, we estimate the uncertainties in our discontinuity depths to be ±10 km.
Figure 2 shows profiles of receiver function stacks along the EW and Coastal arrays, and Figure 3 shows a profile of receiver function stacks for the NS array. The profile along the EW array (Figure 2a) reveals a distinct Ps arrival at 410–420 km depth everywhere and a Ps arrival at 680–690 km depth along the eastern half of the profile. The difference in depths yields a TZT of 270–280 km. A smaller amplitude arrival at about 540 km is also seen, possibly corresponding to the 520 km discontinuity. A stack along the 162°E line of longitude (roughly parallel to the coastal array) shows a Ps arrival at 410–415 km depth (Figure 2b) and Ps arrivals at 670–680 km depth. Similar to the EW array, the difference between the depth of the 410 km and 660 km discontinuities yields a TZT of 260–270 km.
 The results along the NS array are more difficult to interpret (Figure 3). For the 250 km long part of the profile crossing the TAM, clear Ps arrivals from the 410 km discontinuity can be seen at depths of 410–420 km (Figure 3a). And a Ps arrival from the 660 km discontinuity can be seen in the middle of the profile at depths of 680–690 km. This is consistent with the profile along the EW array (Figure 2a). However, along the remaining 1000 km of the array extending into the interior of the EAC, two Ps arrivals at depths of 400 km and 460 km are seen (Figure 3b). The arrivals have similar amplitude to each other, and neither arrival matches the depth of the Ps arrival from the 410 km discontinuity seen in the parts of the study area covered by Figures 2 and 3a.
 The origin of the Ps arrivals at 400 km and 460 km depth is uncertain. However, these Ps arrivals are not seen when only data from stations on rock are used in the stacking. Thus, they are most probably caused by ice layer reverberations masking or interfering with the Ps conversion from the 410 km discontinuity. The ice sheet can be as thick as 2.8 km under the stations and perhaps Ps conversions from the top and base of the ice layer may be affecting the Ps conversion from the 410 km discontinuity [Anandakrishnan and Winberry, 2004]. The Ps conversion from the 660 km discontinuity also cannot be seen in the ice-covered portion of the NS profile (Figure 3b), possibly obscured by ice reverberations as well.
 Another interesting feature that is seen in parts of the profiles (except in Figure 3b) is a double peak or else a very broad peak for the Ps arrival from the 660 km discontinuity. Where there are two Ps arrivals, the second one is at ∼710–720 km depth. The second arrival could be a Ps conversion for the “720” discontinuity, which is caused by a phase transformation in garnet at the base of the mantle transition zone [Simmons and Gurrola, 2000]. Where there is just one broad arrival, this could possibly result from Ps arrivals overlapping from the 660 and 720 km discontinuities.
4. Discussion and Conclusions
 We summarize our results in Figure 4 with a map of TZT and use this map to discuss the implications of our findings for the thermal structure of the upper mantle. Since the 3D velocity model used in stacking the receiver functions was obtained from travel time tomography, the stacked Ps arrivals are adjusted for lateral heterogeneity in upper mantle structure, but the depths of the discontinuities could remain biased from the true depths. Consequently, changes in TZT provide a more robust indication of a thermal perturbation within the transition zone than do the absolute depths of the 410 km and 660 km discontinuities.
 In spite of the topography on the 410 km and 660 km discontinuities, the thickness of the TZT remains fairly consistent across the study area. Although there is a variability of ±10 km in the thickness from one stacking bin to another, this variability is within the uncertainty of the individual TZT estimates. The TZT, where imaged, is slightly greater than the global estimates of TZT, suggesting that there is no warm thermal anomaly thinning the mantle transition zone. The average TZT in our study is 266 ± 10 km, whereas global estimates range from 242 to 260 km [Chevrot et al., 1999; Flanagan and Shearer, 1998; Lawrence and Shearer, 2006].
 As discussed in the introduction, some seismic studies point to the presence of a deep-seated thermal anomaly in the upper mantle beneath the study area. Watson et al. , using P and S body wave tomography, found a velocity anomaly corresponding to a temperature anomaly of 200–300 K under Ross Island and extending laterally 50–100 km beneath the TAM from the coast. The Watson et al. model showed that the thermal anomaly is best developed in the upper 200 km, but that it could extend as deep as the mantle transition zone. The continent-scale tomography model of Sieminski et al.  also suggests that there could exist a thermal anomaly extending to depths 400 km beneath the study area. Our result showing a somewhat thicker-than-average TZT, however, demonstrates that the upper mantle thermal anomaly under the WARS and TAM does not extend as deep as the transition zone, at least not in the regions imaged surrounding Ross Island.
 Prior interpretations for the upper-mantle thermal anomaly have invoked the presence of a plume head or small scale convection related to rifting in the WARS [e.g., Behrendt et al., 1991; Kyle et al., 1992; Ritzwoller et al., 2001; Morelli and Danesi, 2004], interpretations that are consistent with a shallow thermal anomaly in the upper mantle. Our finding that the thermal anomaly does not likely extend as deep as the mantle transition zone supports these interpretations.
 This conclusion contrasts with interpretations advanced recently for other large continental rift systems, for example in eastern Africa, that invoke deep-seated processes originating from below the mantle transition zone manifest seismically as a broad and deep anomaly. The WARS has been likened to the East African rift system because both rift systems have developed over similarly large continental regions. Seismic images of mantle structure under eastern Africa show that the thermal anomaly there extends from uppermost mantle depths well into the transition zone and perhaps into the lower mantle [e.g., Park and Nyblade, 2006; Benoit et al., 2006; Simmons et al., 2007]. The difference between the depth extent of the mantle thermal anomaly beneath our study area in Antarctica and eastern Africa suggests that very different geodynamic processes are likely responsible for creating the two rift systems.
 We would like to thank Mitchell Barklage, Bruce Beaudoin, Jerry Bowling, Juliette Florentin, Audrey Huerta, Jesse Lawrence, Bruce Long, Bob Osburn, Tim Parker, John Pollack, Sara Pozgay, Moira Pyle, Brian Shiro, Rigobert Tibi, and many other individuals for assistance in preparing, deploying, and retrieving TAMSEIS stations and data. We would also like to thank Stuart Henrys and two anonymous reviewers for their constructive comments. This work was funded by the National Science Foundation under grants OPP9909603 and OPP9909648.