Roadmap | The review process | Home
  Transition zone
Searching for Mantle Plumes Using High-Resolution Radon Transforms

Yu Jeffrey Gu & Yuling An

Dept. of Physics, University of Alberta, Edmonton, AB, Canada, T6G2G7


Please attend our EGU talk:
Gu, Y.J. & An, Y., Complexities in the Upper Mantle Transition Zone Beneath Hotspot Locations, abstract EGU2007-A-11008, 14:45-15:00, Friday 20th April 2007, Lecture Room 23.



The origin of melting anomalies (or “hotspots”) has long been a focal point in the discussion of two competing classes of convection models, the bottom-up (“plume”) model that routinely draws comparisons with a pot of water on a hot stove (Morgan, 1971) and the top-down (“plate”) model that emphasizes the role of lithosphere, plate stresses and recycling of the enriched crust (see Foulger, 2007 for a detailed review; Ed: See also Plate Tectonic Processes page). This debate is reinvigorated by recent reports of a plethora of narrow low-velocity columns beneath hotspot locations based on high-resolution, finite-frequency tomography (Montelli et al., 2006). While the effect of the finite-frequency approach over traditional ray-based techniques has been disputed (de Hoop & van der Hilst, 2005; Ed: See also Banana Doughnut page), the need to understand the existence and vertical extent of hot mantle plumes has not. 

Technical details

The main goal of this study is to improve the resolution of the structure of mantle seismic discontinuities using the High-resolution Radon transform method (Sacchi & Ulrych, 1995). This method utilizes frequency-domain inversions to simultaneously constrain the differential times and ray parameters of SS precursors. Figure 1 illustrates the basic procedure in our analysis of sub-hotspot mantle. We analyze the transverse components of teleseismic (100-160°) recordings and group the source-receiver pairs whose geometrical reflection points fall within 10° of a given hotspot location (Figure 1a). Rather than relying on a low-resolution cap averaging approach, we compute the running average of records with distances of 20° to pre-condition each data set (Figure 1b). Then, through regularized least squares inversions, we determine the time-distance relationships of SS precursors based on the energy foci in the τ - p model. The recovered (predicted; Figure 1d) and spatially interpolated (Figure 1e) time series represent the “cleaned up” versions of the original data, while random noise and interfering phases outside of the ray parameter and time ranges of the SS precursors (Figure 1f) have been minimized. We measure the differential time τ  and the ray parameter p of SdS-SS (d for a discontinuity) exclusively in the Radon domain.

Figure 1. The process of Radon transform for the Hawaii hotspot. (a) The original time series after alignment and normalization on SS. Topographic, crustal and heterogeneity corrections have been applied. (b) The time series after partial stacking using 20° distance windows. (c) A non-quadratic Radon transform is performed to constrain the τ - p model. (d) The predicted data with Radon operator in (c). (e) Reconstructed seismograms after resampling and interpolation with same Radon operator shown in (c). (d) The residual between the data and Radon model predictions. Signals have been exaggerated for clarity. Click here for enlargement.


Using the abovementioned approach we examined 17 hotspots characterized by a score of 2 or higher in Courtillot et al. (2003). The depth of the 410-km discontinuity, as inferred from our measurements, is generally deeper at the hotspot locations (416 km) relative to the global average of 410 km (Figure 2a). This observation corroborates the time-domain observations of Gu et al. (2003) and Deuss (2007). Out of all the examined hotspots, a shallower-than-average 410-km discontinuity is only observed beneath Iceland, which we interpret as evidence of a “shallow” hotspot (Du et al., 2006). The transition zone thickness is dominated by the locally depressed 410-km discontinuity and its average beneath hotspot locations (237 km) is 5-km thinner than the global average (Gu et al., 2003; Figure 2b). The regional thickness variations inferred from Pds (higher resolution) and SdS (this study) are remarkably consistent (see Figure 2b), and underline the need for the sixth criterion (see Deuss, 2007) in differentiating deep-rooted hotspots from shallow ones.


Figure 2. (a) The depth of the 410-km discontinuity. The contour map shows the interpolated results of Gu et al. (2003). The solid circles represent the results of this paper (only sign information relative to the global average of 410 km is given). The unfilled circles show the results from Deuss (2007) (without amplitude information). (b) Transition zone thickness variations. The contour map shows the results of Gu & Dziewonski (2002; low resolution). The black symbols compare the recent thickness map of Lawrence & Shearer (2006; from high resolution receiver functions) with the regional map of this study (colored circles). The results from all three studies are well correlated. (c) Proposed plume depth from this study. TZ – transition zone or deeper, UM – upper mantle, LITHO – lithospere, Unclear – depth not well resolved.


Perturbations of S410S-SS ray parameters present the most revealing observations relevant to the search for hot thermal anomalies (Figure 3a). We identify both positive and negative jumps in travel time curves (and thus in ray parameters) for rays bottoming beneath hotspots (Figure 3b). This can be explained using two idealized conceptual models (Figure 4). If a low velocity column exists in the upper mantle beneath a hotspot (Figure 4a), then shorter SS waves (with larger ray angles near the reflection points) would slow down precipitously with increasing ray lengths within the column. This phenomenon is relatively short-lived, as travel times stabilize near the center of the anomaly (which is expected to be slower seismically, if it corresponds to a plume) and increase gradually at longer distances. This simple model can explain sufficiently the travel time observations associated with the Hawaii, Tahiti, Macdonald, Juan de Fuca, Cape Verde, Yellowstone, and Iceland hotspots. Should a low-velocity perturbation extend into the transition zone, its predicted effect on S410S would offset or overcome that on SS, thereby producing a “flat” or positive differential ray parameter at shorter distances. Low-velocity anomalies at the Azores, Samoa, Louisville, Canary, New England and Bowie hotspots appear to agree with this model. If due to thermal variations, the epicentral distance at which the abrupt change in differential ray parameter occurs will reflect the temperature gradient and width of the anomaly. Hotspots that exhibit smooth, PREM-like distance-time relationships, for example Pitcairn and Marqueses islands (not shown), are more likely to originate from lithospheric processes. Our data coverage is insufficient in the distance range 100-130° to determine accurately the transition zone characteristics beneath Afar and Reunion.

Figure 3. (a) An example showing a characteristic change in the differential ray parameter, equivalent to a significant jump in the travel time curve. (b) Detailed distance-time relationships for S410S-SS for the majority of hotspot locations. The dotted line represents the travel time curves based on PREM (Dziewonski & Anderson, 1981). The dark solid lines denote the travel time observation from our spatially reprocessed data.

Figure 4. Two conceptual models illustrating positive and negative changes in ray parameters of S410S-SS across the distance range of 100-160°. If the anomalies are due to plumes, then the amount of “bending” and the distance where it occurs reflect the temperature and lateral dimension of the hot thermal anomalies.

Travel time perturbations of S670S are known to correlate with the global distribution of subduction zones (Shearer & Masters, 1992), but their effectiveness in constraining anomalies at or below the upper mantle transition zone is compromised by potential compositional variations and the majorite-garnet phase transition (Hirose, 2002; Deuss, 2007). The depths inferred for the 660-km discontinuity from the Radon inversions vary substantially and are, in general, inconsistent with those expected from the olivine γ-spinel to silicate perovskite and magnesiowüstite transformation (Anderson, 1967) at high mantle temperatures. Furthermore, the differential ray parameter of S660S-SS is surprisingly uneventful throughout the examined distance range. The lack of a characteristic change in p may suggest the absence of significant anomalies immediately below the 660-km phase boundary, though complexities associated with the garnet phase transformation and waveform interference from ScSdScS and sdsS cannot be ignored (Schmerr & Garnero, 2006).

Lastly, the presence of mid-mantle reflectors have been suggested as potential evidence of compositional stratification beneath some hotspot locations (Shen et al., 2003). In an attempt to link lower/mid-mantle reflectors with hotspots, we conducted a systematic search in time and Radon domain for such reflectors (Figure 5). We identify clear τ - p maxima corresponding to 900-1000 km reflectors beneath Hawaii, Louisville, Tahiti, Juan de Fuca, Canary and Macdonald islands, a list that includes both hotspots that apparently have upper-mantle origins and those apparently rooted in the transition-zone or deeper, as discussed above. We also observed additional mid-mantle reflectors beneath non-hotspot locations in the northern Pacific, which leads us to believe that the presence of mid-mantle reflectors is more common than was previously thought and their spatial distribution does not correlate with the locations of hotspots. 


Figure 5. Examples of lower-mantle reflectors deeper than 850 km. Similar lower/mid-mantle reflectors have been observed beneath a significant fraction of hotspots as well as under non-hotspot locations. 


Our study highlights the importance of the 410-km discontinuity as a mantle plume tracer. The phase boundary is generally depressed beneath hotspots, and a discontinuous travel-time curve provides powerful constraints on the possible depths of hot thermal plumes. Among 17 potential plume locations that we studied, we confidently identify the Azores, Samoa, Louisville, Canary, New England and Bowie hotspots as being underlain by anomalies that are relatively deep, i.e., within the transition zone and potentially deeper. Pitcairn and Marqueses hotspots show no indication of underlying anomalies, as evidenced by the lack of change in the S410S-S660S ray parameters, only mild depression of the 410-km discontinuity, and relatively normal transition zone thicknesses. Our study also suggests a shallowly rooted Iceland hotspot (see also Du et al., 2006), from both an anomalously shallow 410-km discontinuity and a positive change in the S410S-SS distance-time relationship.


We thank Mauricio Sacchi for technical help. We are also grateful to Gillian Foulger for the invitation to contribute and for many valuable suggestions and comments. This work is supported by Alberta Ingenuity, Canadian Foundation for Innovation (CFI) and NSERC.


last updated 2nd April, 2007