the lowermost mantle can be explained well by models of slab sinking constrained by subduction history, assuming that slabs sink vertically from the trench (Ricard et al., 1993), even when using only the most recent 200 Myr of subduction history. Although the thermochemical nature of LLSVPs is sometimes debated (e.g., Davies et al., 2015), several lines of evidence now suggest that the LLSVPs are both warmer and compositionally distinct from the surrounding lower mantle. The anti‐correlation of shear‐ and bulk sound speed (Su and Dziewonski, 1997; Masters et al., 2000), sharp boundaries imaged by detailed waveform modeling (e.g., He and Wen, 2009; Wang and Wen, 2007), tidal constraints (Lau et al., 2017), and inferences of density from full‐spectrum tomography (e.g., Moulik and Ekström, 2016) are all consistent with a thermochemical rather than purely thermal origin of the LLSVPs. Additional evidence for a thermochemical origin is provided by the distribution of present day hotspots near the interiors and margins of the LLSVPs (Thorne et al., 2004; Burke et al., 2008; Austermann et al., 2014), expected on the basis of laboratory analogue experiments and numerical simulations (Davaille et al., 2002; Jellinek and Manga, 2002), as well as the observation that primitive helium isotope ratios in ocean island basalts are associated with mantle plumes rooted in the LLSVPs (e.g., Williams et al., 2019). The reconstructed eruption locations of large igneous provinces also fall near the present‐day boundaries of the LLSVPs (Burke and Torsvik, 2004; Torsvik et al., 2006), which supports the idea that the LLSVPs have been relatively stable over at least the past 200 Myr. Geodynamic models that impose time‐dependent, paleogeographically constrained plate motions can produce chemical piles whose large‐scale features are consistent with the morphology of the LLSVPs (e.g., McNamara and Zhong, 2005; Bower et al., 2013; Rudolph and Zhong, 2014).
Figure 1.1 Correlation between structure at 2800 km depth and other mantle depths for each of four tomographic models. Correlations for spherical harmonic degrees 1–2 are shown in blue and degrees 1–4 are shown in yellow. Where the curves are thicker, the correlation is significant at the p = 0.05 level.
All recent global tomographic models are generally in very good agreement at long wavelengths in the lowermost mantle (e.g., Becker and Boschi, 2002; Cottaar and Lekic, 2016). The long‐wavelength structure, comprising spherical harmonic degrees 1–4, is consistent in the upper and lowermost mantle across four recent global VS tomographic models (Figure 1.2). However, shear velocity variations across the mantle transition zone and shallow lower mantle appear to shift to a pattern that is weakly anticorrelated with structure in both the asthenosphere and lowermost mantle. This change in large‐scale structure, evident in the global maps of VS variations (Figure 1.2), is also captured by the mantle radial correlation function (RCF), shown in Figure 1.3 (e.g., Jordan et al., 1993; Puster and Jordan, 1994), which measures the similarity of the pattern of structures between every pair of depths in the mantle. Radial correlation analyses of global tomographic models reveal a zone of decorrelation across the transition zone and shallow lower mantle. Some (but not all) recent global shear‐wave (e.g., French and Romanowicz, 2014; Durand et al., 2017) and P wave (Fukao and Obayashi, 2013; Obayashi et al., 2013) tomographic models show a dramatic change in the radial correlation structure near 1,000 km depth. While the precise depth of the change in the long‐wavelength correlation is affected by the data constraints and model parameterization choices of the tomographic models (discussed later), it appears to coincide with other shorter‐wavelength features such as scatterers and deflected plumes, suggesting that they may share a common dynamical origin.
Figure 1.2 Shown here are plots of Voigt VS variations at spherical harmonic degrees 1–4 from four recent tomographic models at selected depths across the transition zone and shallow lower mantle and within the lowermost mantle and the lithosphere. In each subpanel, the amplitude of the color scale is indicated in percent.
Figure 1.3 Radial correlation functions computed for four recent global tomographic models. The color scale indicates the value of the correlation between structures at different mantle depths. In each panel, the lower‐left triangle is the RCF for only spherical harmonic degrees 1–2, while the upper right triangle shows the RCF for degrees 1–4.
Changes in the continuity and shape of upwelling and downwelling features have been identified in both global and regional tomographic models. The whole‐mantle waveform tomographic model SEMUCB‐WM1 (French and Romanowicz, 2014) reveals striking lateral deflections of upwellings beneath several active hotspots. In particular, the imaged plume conduit beneath St. Helena appears to be deflected 1,000–650 km, in contrast to its more vertical shape in the lower mantle below 1,000 km and in the upper mantle and transition zone (French and Romanowicz, 2015). While St. Helena provides perhaps the most striking example, there are also possible deflections near 1,000 km beneath the Canaries and Macdonald (French and Romanowicz, 2015). Regional tomography of the North Atlantic shows some evidence of deflection of the Iceland plume above 1,000 km depth (Rickers et al., 2013), though this is not seen in all regional and global models (French and Romanowicz, 2015; Yuan and Romanowicz, 2017). Regional tomography indicates that the Yellowstone plume is laterally shifted near 1,000 km (Nelson and Grand, 2018). Slabs imaged in the global P wave model GAP‐P4 (Obayashi et al., 2013; Fukao and Obayashi, 2013) appear to stagnate within or below the mantle transition zone. The Northern Mariana slab is imaged as a nearly horizontal feature above 1,000 km, the Java slab is stagnant or less steeply inclined above 1,000 km, and the Tonga slab is imaged as a horizontal fast anomaly above 1,000 km (Fukao and Obayashi, 2013). This behavior is not universal; in Central America, there is no evidence for slab ponding or stagnation, and many slabs in the Western Pacific and South America are imaged above 650 km depth.
Seismic scatterers have been identified in the mid‐mantle, and their depth‐distribution peaks near 1,000 km. Using P‐to‐S receiver function stacks, Jenkins et al. (2016) located scatterers beneath Europe, Iceland, and Greenland. The scatterers are distributed at depths of ∼800–1400 km, with a peak near 1,000 km. In global surveys of seismic reflectors, scatterers are observed in all regions with data coverage, and with no clear association to tectonic province or subduction history (Waszek et al., 2018; Frost et al., 2018). The global distribution of reflectors imaged by Waszek et al. (2018) is quite broadly distributed across the depth range of 850–1300 km with a peak close to 875 km depth. Changes in the distribution of small‐scale heterogeneity are also supported by analyses of the spectra of tomographic models. A decrease in the redness in mantle tomographic models occurs below 650 km, coincident with the observations of short‐wavelength scattering features, stagnant slabs, and deflected upwellings.
The