Insights on the African Upper Mantle From Quasi‐Love Wave Scattering

The African upper mantle is diverse, featuring several cratonic roots, active and magmatic continental rifting, abundant intra‐plate volcanism, and several oceanic hotspots offshore. Relatively small‐scale mantle convection processes, some possibly related to lithospheric thickness variations, have been proposed to account for patterns of volcanism and topography. A key constraint on such features and processes can be found via seismic anisotropy, but limited coverage of seismograph stations across the continent has resulted in sparse observations. Quasi‐Love waves, produced by scattering of Love to Rayleigh energy at lateral gradients in upper mantle seismic anisotropy, can provide information about seismic anisotropy well away from seismograph stations. We catalog 525 observations of Quasi‐Love waves across the region and back‐project to scattering points, revealing locations of lateral gradients in upper mantle seismic anisotropy. Quasi‐Love wave scattering occurs at craton edges, likely due to the contrast between lithospheric and asthenospheric anisotropy, and close to ancient orogenic belts, indicating changes in fossilized lithospheric anisotropy from past collisional events. Scattering surrounding the proposed Al‐Kufrah cratonic remnant in north‐east Africa supports its existence as cratonic lithosphere. Scattering also occurs below thin lithosphere, suggesting deviations in asthenospheric flow patterns (such as localized upwellings), notably beneath the East African Rift, oceanic hotspot tracks, and the Cameroon Volcanic Line. Quasi‐Love scattering is abundant in the Indian Ocean and beneath Madagascar, consistent with small‐scale dynamic processes in the asthenosphere that likely relate to the complex rifting history of this ocean basin and the dispersed micro‐continents within.


Introduction
The solid Earth's outer layers consist of a relatively cold, stiff and conducting lid-the lithosphere-sitting above a relatively warmer, weaker and convective layer-the asthenosphere.The thickness of the lithosphere varies laterally from <50 km to >300 km (e.g., Conrad & Lithgow-Bertelloni, 2006;Hoggard et al., 2020;Pasyanos et al., 2014;Priestley & McKenzie, 2013).The thickest lithospheric mantle tends to underlie the oldest continental crust, and deformation fabrics can be preserved in such old and thick lithosphere for billions of years, keeping a record of tectonic events in the deep past (e.g., Eakin et al., 2021;Liddell et al., 2017;Silver, 1996;Zhao et al., 2008).By contrast, the asthenosphere continuously deforms as a result of ongoing mantle convection.The interaction of the asthenospheric flow-field with the topography of the base of the lithosphere is proposed to have a strong influence on the distribution of volcanism and topography on Earth (e.g., Ball et al., 2019;Conrad et al., 2011;Davies et al., 2019;Duvernay et al., 2021;King & Anderson, 1998;Lees et al., 2020).
Mapping such deformation in the upper mantle can be done through observations of seismic anisotropy, the phenomenon whereby seismic velocities are direction-dependent, primarily due to the alignment of crystallographic axes, or lattice preferred orientation (LPO), in mantle minerals, notably olivine (Hess, 1964;Nicolas & Christensen, 1987).Upper mantle seismic anisotropy is commonly constrained by observations of shear-wave splitting, which measures azimuthal anisotropy primarily either beneath the seismograph station where it is measured or (in the case of source-side splitting) beneath the earthquake (e.g., Kaneshima & Silver, 1992;Russo & Silver, 1994;Silver & Chan, 1991).This naturally limits such measurements to locations where seismograph stations have been deployed, or where large earthquakes occur.Seismic tomography, especially of surface waves, can also place some constraints on upper mantle seismic anisotropy, but the resolution of lateral variations tends to be poor, particularly when the coverage of sources and receivers is limited, and there are significant trade-offs between anisotropic and isotropic structures (e.g., Faccenda & VanderBeek, 2023;Huang et al., 2015;Rawlinson et al., 2010;Simons et al., 2002).
Such limitations in observations are especially apparent for the African continent and its surroundings.The vast continent of Africa comprises many distinct tectonic terranes, formed and deformed by tectonic events stretching back billions of years, with much variation in lithospheric thickness (Figure 1; e.g., Begg et al., 2009;  (Global Volcanism Program, 2013).Volcanoes or seamounts associated with oceanic hotspots are labeled (Ma: Madeira; Ca: Canary; As: Ascension; SH: St. Helena; TdC: Tristan da Cunha; G: Gough; V: Vema; Ré: Réunion; Co: Comoros).Black lines mark tectonic plate boundaries from Bird (2003), with purple lines to mark the major boundaries of the East African Rift System (EARS) from Stamps et al. (2021), dashed to mark the uncertain southern continuation of the boundary.(b) Overview of existing azimuthal anisotropy measurements for the same region.Station-averaged teleseismic shear-wave splitting measurements from the updated compilation of Wüstefeld et al. (2009), with additional data from Qaysi et al. (2018), Komeazi et al. (2023), and Ebinger et al. (2024), are plotted in pink.The orientation and strength of anisotropy at 150 km depth from the anisotropic global surface wave tomography model 3D2018_08Sv (Debayle et al., 2016) is represented by the black bars.This is plotted against the lithospheric thickness map of Hoggard et al. (2020) based on the SLNAAFSA model, which in this region is a combination of the global SL2013sv (Schaeffer & Lebedev, 2013) and regional African AF2019 (Celli, Lebedev, Schaeffer, & Gaina, 2020) shear-wave velocity tomographic models.The separate cratonic "cores" of the West African and Congo cratons identified by Celli, Lebedev, Schaeffer, and Gaina (2020) are labeled (Re: Reguibat; ML: Man-Lèo; BK: Bomu-Kibale; GC: Gabon-Cameroon; Ka: Kasai; Cu: Cubango) as is the Angolan Shield (Ag) which lacks a deep root, and the locations of the Niassa (Ni) cratonic root proposed by Celli, Lebedev, Schaeffer, and Gaina (2020) and Al-Kufrah (AK) cratonic remnant proposed by Sobh et al. (2020).Fishwick, 2010).In addition, the continent hosts a variety of volcanism and uplift that is indicative of dynamic processes occurring in the asthenosphere (e.g., Ball et al., 2019;Paul et al., 2014;Reusch et al., 2010;Stephenson et al., 2021).However, the seismic anisotropy of the African upper mantle is relatively poorly known, due to the lack of seismograph station coverage (Figure 1b), and this is all the more true for the surrounding oceanic regions due to the paucity and quality of ocean-bottom seismometer deployments.
An additional constraint on seismic anisotropy can be gained from observations of Quasi-Love waves.These are surface waves generated by scattering of Love waves into Rayleigh-wave-type energy due to a coupling between spheroidal and toroidal normal modes of the Earth's vibration (e.g., Park & Yu, 1992, 1993).Scattering of this type occurs when a Love wave passes through a lateral gradient in seismic anisotropy (Park & Yu, 1992;Yu et al., 1995;Yu & Park, 1993).This has typically been observed in fundamental-mode Love waves with periods >100 s, primarily sensitive to the 100-200 km depth range.Thus observations of Quasi-Love waves have been used to infer such gradients in the upper mantle, and detect changes in anisotropy where direct measurements of anisotropy are unavailable (e.g., Chen & Park, 2013;Cheng et al., 2021;Levin et al., 2007;Rieger & Park, 2010;Servali et al., 2020;Yu et al., 1995).
While most studies of Quasi-Love waves have focused on targeted areas of interest, a recent study by Eakin (2021) performed a broad search for evidence of Quasi-Love wave scattering throughout the Australian plate, to identify where the major sources of scattering are.Lateral gradients in seismic anisotropy were inferred beneath major geological terrane boundaries and beneath the passive continental margins, suggestive of both lithospheric and asthenospheric processes.Here we follow a similar approach to identify seismic anisotropy gradients in a region including active rifting and intra-plate volcanism, with many geological terrane boundaries (some poorly constrained), and a relative lack of previous seismic anisotropy measurements.

The Geologic Evolution of Africa
Africa and Arabia are formed of several distinct pre-Cambrian terranes separated by orogenic belts representing tectonic events stretching over 2 billion years, from the late Archean to the Paleozoic.The major blocks are the West African, Congo, Kalahari and Tanzania Cratons and the Saharan Metacraton (Figure 1a).Aside from the Saharan Metacraton, these ancient blocks are all underlain by high seismic velocity (and, by inference, cold, strong and lithospheric) roots in the mantle, to depths of 150-300 km (Figure 1b; e.g., Begg et al., 2009;Celli, Lebedev, Schaeffer, & Gaina, 2020;Fishwick, 2010;Hoggard et al., 2020).Of the largest cratonic blocks, the internal structure of the Kalahari Craton is well-known, being formed of the Kaapvaal and Zimbabwe Cratons separated by the Limpopo Belt that represents a late-Archean continental collision (e.g., van Reenen et al., 1992), but the larger West African and Congo Cratons, and the Saharan Metacraton, are in great part buried beneath Phanerozoic sediments and their internal structure is less well-known.
The Saharan Metacraton, which underlies most of north-east Africa but is mostly buried beneath the Sahara Desert, differs from the cratons elsewhere in Africa in two ways: it underwent significant deformation in the Pan-African orogeny in the Neoproterozoic (e.g., Abdelsalam et al., 2002;Black & Liégeois, 1993), and almost entirely lacks a thick mantle root (Figure 1b).Abdelsalam et al. (2002) named it a metacraton, meaning a remobilized craton that still retains some of its previous characteristics, based on evidence that it is formed dominantly by Archean-Paleoproterozoic terranes that have behaved as a coherent block throughout the Phanerozoic.Liégeois et al. (2013) proposed that there exist three stable cratonic cores within the metacraton that resisted deformation, below the Murzuq, Al-Kufrah and Chad basins, naming these as "cratonic remnants."Modeling of lithospheric structure using a variety of geophysical, geological and petrological constraints led Sobh et al. (2020) to propose that a significant lithospheric root remains beneath the Al-Kufrah cratonic remnant, which can be seen to an extent in other LAB maps including that in Figure 1b (Hoggard et al., 2020).
The Pan-African orogeny of 800-550 Ma, that was responsible for the remobilization of the Saharan Metacraton, also brought together most of the constituent blocks of Africa and Arabia into their current arrangement, resulting in a number of orogenic belts that form the current lithosphere between the cratons (e.g., Begg et al., 2009).The continental lithosphere all the way from Arabia south through East Africa to Mozambique is the East Africa orogenic belt, representing the Pan-African collision between the eastern African cratons and India.
Since the Mesozoic, the tectonics of Africa have been dominated by rifting.First, the Central Atlantic began to open at around 185 Ma, following emplacement of the Central Atlantic Magmatic Province.The Paranà-Etendeka igneous province erupted at 135 Ma across present day Angola and South America, and the opening of the South Atlantic followed at around 100 Ma (e.g., Granot & Dyment, 2015).On the other side of the continent, the Karoo igneous province erupted in southern Africa and Antarctica at around 180 Ma, and this was followed by the opening of the Indian Ocean in a number of steps.Antarctica rifted first from southern Africa, followed by Madagascar and India together.Indian Ocean spreading then migrated east of Madagascar, which begin rifting from India at around 85 Ma, before migrating again east of the Seychelles following Deccan Traps volcanism at around 65 Ma (e.g., Collier et al., 2008).
Most recently, in East Africa, the Afar plume began to emplace the Ethiopian Flood Basalts from around 45 Ma.Subsequent to this, Arabia began to rift from Africa along the Red Sea and Gulf of Aden rifts from ca. 24 Ma (ArRajehi et al., 2010), and continental rifting has developed along the East African Rift System (EARS), which splits the African plate into the western Nubia plate and the eastern Somalia plate.The EARS extends southwest from the Afar triple junction through Ethiopia and Kenya and bifurcates around the Tanzanian Craton (Figure 1a).Further south, the system becomes more diffuse and uncertain, with a western branch along the Malawi rift and branches in the Indian Ocean including a possible broad deformation zone around northern Madagascar (e.g., Saria et al., 2014;Stamps et al., 2021).
The major rifting events have often been associated with mantle plumes that have been proposed to result in extensive magmatism and possible removal of lithospheric roots (Afonso et al., 2022;Celli, Lebedev, Schaeffer, & Gaina, 2020;Stephenson et al., 2023).Specifically, the Central Atlantic Magmatic Province produced extensive magmatism around 200 Ma in West Africa where the lithosphere is now >200 km thick, the Karoo LIP in southern Africa erupted at around 180 Ma where lithosphere is now 150-180 km thick, the Etendeka LIP around 135 Ma where lithosphere is around 80-120 km thick, and the Afar plume around 45 Ma where lithosphere is <50 km thick (Stephenson et al., 2023).Thus the activity of past and present mantle plumes may have a lasting effect on the lithosphere that influences present-day dynamics.Isolated intraplate volcanism is ongoing elsewhere in continental Africa, only where the lithospheric thickness is ≲100 km (Figure 1; Ball et al., 2019).
Several active hotspots produce volcanism in the South Atlantic Ocean west of Africa, some of which are associated with dated seamount trails testifying to their longevity, including the Tristan da Cunha hotspot believed to represent the mantle plume that was responsible for the Etendeka LIP (e.g., O'Connor & Duncan, 1990).In the western Indian Ocean, the Réunion hotspot has been the subject of much study, believed to be responsible for the Deccan Traps volcanism at ca. 65 Ma (e.g., Courtillot et al., 1988), and mapped as a complex whole-mantle plume feeding large high-temperature asthenospheric reservoirs as well as the Central Indian Ridge (e.g., Barruol et al., 2019;Mazzullo et al., 2017;Wamba et al., 2021Wamba et al., , 2023)).

Existing Constraints on Seismic Anisotropy
The most reliable, high-resolution measurements of upper mantle seismic anisotropy can be made by shear-wave splitting analysis; where seismograph station coverage is dense, this can lead to excellent lateral resolution.However, for the most part, Africa is only relatively sparsely covered by seismograph instruments, especially in the north (Figure 1a).As such, the shear-wave splitting observations are restricted mainly to southern Africa, the East African Rift, Madagascar, Cameroon and Morocco (Figure 1b), leaving most of the West African Craton, Sahara Metacraton and Congo Craton with almost no measurements.
Notably, strong, rift-parallel seismic anisotropy is inferred around the Main Ethiopian Rift, interpreted as representing aligned melt intrusions in the lithosphere (Kendall et al., 2005).Laterally variable anisotropy is constrained further south, where the East African Rift bifurcates around the Tanzania Craton (e.g., Walker et al., 2004).Ebinger et al. (2024) recently re-analyzed shear-wave splitting throughout the East African Rift and found that relatively shallow sources of rift-parallel anisotropy, from oriented melt pockets in the lithosphere and channelized base-of-lithosphere flow, dominate beneath the rift, being stronger where the lithosphere is thinner.To the south, the results of both shear-wave splitting (Silver et al., 2001) and surface wave dispersion analysis (Adam & Lebedev, 2012) in the thick lithosphere of southern Africa revealed a contrast across the Limpopo Belt, where there is strong, orogen-parallel anisotropy, suggesting that present-day anisotropy was controlled by Archean tectonic processes.
A dominance of null results from shear-wave splitting analysis have been used to infer vertical mantle flow in both Cameroon (De Plaen et al., 2014), associated with the Cameroon Volcanic Line, and Madagascar (Paul & Eakin, 2017).Other stations in Cameroon indicate anisotropic fast directions parallel to pre-Cambrian suture zones (De Plaen et al., 2014), whereas Madagascar appears to have a very laterally variable anisotropic structure that may indicate both lithospheric and asthenospheric complexities (Ramirez et al., 2018;Reiss et al., 2016).
Upper mantle seismic anisotropy is particularly poorly constrained in the oceans.However, a number of oceanbottom seismometers, in combination with nearby land-based instruments, have been used to investigate the western Indian Ocean.In this region, shear-wave splitting combined with anisotropic surface wave tomography has been used to infer flow of Réunion plume material toward the Central Indian Ridge (Barruol et al., 2019;Mazzullo et al., 2017).Elsewhere within the oceans, results from source-side splitting that utilizes earthquakes from the global mid-ocean ridge system strongly suggest a narrow zone of anisotropy directly beneath the plate boundary that is distinct from the pattern beneath the ocean basins (Eakin et al., 2018;Nowacki et al., 2012).
As a whole, the African Plate is covered by global anisotropic seismic tomography models, such as D2018_08Sv (Figure 1b; Debayle et al., 2016), which suggest some very large-scale variations in anisotropy, for instance differences between the West African Craton, Congo Craton and the intervening Cameroon Volcanic Line (Figure 1b).However, such results are very low resolution and not sensitive to sharp contrasts in anisotropy.
So far only one previous investigation of Quasi-Love wave scattering has been conducted in Africa.A recent study using a temporary deployment near Lake Malawi, on the East African Rift found scattering occurred in the vicinity of the lake, which was interpreted as edge-driven convection (Cheng et al., 2022), due to the presence of the Niassa cratonic root as mapped by Celli, Lebedev, Schaeffer, and Gaina (2020).

Quasi-Love Wave Analysis
The presence of seismic anisotropy in the Earth results in coupling between toroidal and spheroidal normal modes of vibration, which has been shown to result in waveform anomalies known as Quasi-Love and Quasi-Rayleigh waves (Park & Yu, 1992, 1993;Yu & Park, 1993).In essence a Quasi-Love wave represents scattering (or conversion) from Love to Rayleigh wave energy.These Quasi-Love waves travel with elliptical Rayleigh-type particle motion and arrive on the seismogram after the Love wave but before the Rayleigh wave.They retain a similar waveform shape to the Love waves from which they were generated.Such Quasi-Love waves are produced, at periods of 50-250 s, when lateral gradients in seismic anisotropy are present, and are sensitive primarily to the 100-200 km depth range (e.g., Yu & Park, 1993;Yu et al., 1995).Thus, when a Love wave encounters a gradient in seismic anisotropy, a scattered Quasi-Love wave can be produced, traveling at the Rayleigh wave velocity for the remainder of the raypath; therefore, the location of the lateral anisotropic gradient can be back-projected from the arrival time of the Quasi-Love wave relative to the Love wave (e.g., Rieger & Park, 2010).
In this study, we identify Quasi-Love waves arriving at seismograph stations in and around Africa.We downloaded publicly available data from permanent deployments for all earthquakes of magnitude M W > 6.5, depth <50 km, and epicentral distance >70°, that were available at each station (Figure 2).The magnitude and depth criteria select for high-amplitude fundamental mode surface waves, and the epicentral distance restriction is necessary to ensure sufficient separation of the Love and Rayleigh waves to distinguish the Quasi-Love wave arrival.To remove any interference from higher mode surface waves, we use a low-pass filter of 0.01 Hz (100 s).Depth sensitivity kernels suggest that Quasi-Love waves at these frequencies are sensitive to a broad upper mantle depth range, with peak sensitivity at 100-200 km; see Figure S1 of Levin et al. (2007) or Figure 9 of Chen and Park (2013).
To identify a Quasi-Love wave, we visually inspect the transverse component of a seismogram for the fundamental Love wave (G1) arrival, and the vertical and radial components for the Quasi-Love (QL) and fundamental Rayleigh (R1) waves (Figure 3a).Previous studies have sometimes used only the vertical component to search for Quasi-Love waves (e.g., Chen & Park, 2013;Levin et al., 2007), despite the expected particle motion being retrograde elliptical in the vertical-radial plane.Including information from the radial component allows us to improve detection and confirm elliptical particle motion (e.g., Eakin, 2021;Servali et al., 2020).In Rayleigh (and Quasi-Love) waves, the phase of the radial (R) component of particle motion is π/2 radians ahead of the vertical (Z) component.Thus, applying a Hilbert transform, which effectively shifts the phase of a signal by π/2, to the radial component, is useful for detection and polarization analysis of Rayleigh waves (e.g., Baker & Stevens, 2004;Chael, 1997): the Hilbert transform of the radial component, H(R), should be in phase with the vertical component Z (Figure 3a).An acceptable Quasi-Love wave must have in-phase H(R) and Z, such that R-Z particle motion is elliptical (Figure 3e) and H(R)-Z particle motion is linear (Figure 3f).
To measure the Quasi-Love arrival time, and confirm its similarity to the Love wave, we window 100 s of the transverse component either side of the maximum Love wave amplitude (t G ), and correlate this with the stacked Z and H(R) components (since stacking will amplify the Quasi-Love arrival relative to noise).The time lag corresponding to the maximum correlation co-efficient, that falls within the first 50% of the time window between the Love and Rayleigh maximum amplitudes, is taken as the Quasi-Love wave delay time, δt (Figure 3b).
All seismograms were visually inspected so that results only passed quality control when the Quasi-Love waveforms were distinct, elliptically polarized, and similar to the Love waveform.The conditions of elliptical polarization and similarity to the Love waveform reduce the possibility that we are observing other waves, such as multiply reflecting S waves.Results with non-transversely polarized Love waves were rejected.With the remaining results, the Quasi-Love delay times δt can then be used to back-project the scattering point along the event-station path, based on the equation, where δx is the distance traveled between the scatterer and the receiver, and V R and V L are the Rayleigh and Love wave velocities, respectively; re-arranging gives Rather than using precise estimates for V L and V R , which vary along ray paths as well as from path to path, we can substitute an approximation for the epicentral distance Δ which uses only the delay time between the Love and Rayleigh waves (t R t L ) and between the Love and Quasi-Love waves (δt), which are easy to measure from the seismograms based on maximum amplitudes.
Given the long wavelength (∼400 km) of the Love waves used, and the constant velocity approximation used above, the locations are by nature somewhat approximate.The analysis can only resolve features on the scale of at least ∼1/4 of the wavelength (e.g., Chen & Park, 2013), giving a lateral resolution of around 100 km.Additionally, the average phase velocities along the raypath before and after scattering may differ, which would result in an inaccurate estimation of δx, for example, a 5% difference in average velocities would result in a 5% error in δx.Thus, scattering that occurs further from the receiver is less well-resolved; 95% of scatterers are within 6,000 km of their receiver (Figure S1 in Supporting Information S1), corresponding to a possible error of up to 300 km.
Errors in back-projection can also be introduced by deviations from the assumed great-circle path, due to heterogeneity in the velocity structure of the Earth, which again will be a larger effect with greater δx.We take care not to include results where the polarization of the surface waves differs by more than 10°from that expected from the source-receiver great-circle path.Nevertheless, small deviations in backazimuth from the great-circle path can imply mislocation of scatterers perpendicular to the path with a magnitude of ∼δx sin ϕ, where ϕ is the angular deviation.A maximum deviation of 10°would lead to a possible error ∼17% of the magnitude of δx.However, most deviations are <5°, corresponding to <9% of δx error.
In addition to the back-projected locations of scatterers, we also record the amplitude of Quasi-Love waves relative to the Love wave.The amplitude depends on both the strength of anisotropic gradient and the geometry of the anisotropy with respect to the wave propagation path: maximum scattering occurs when the propagation is at 45°to the anisotropic symmetry axis, and zero scattering when parallel or perpendicular (Park, 1997;Rieger & Park, 2010).Thus, while higher amplitudes imply greater confidence in the existence of a strong contrast in anisotropy, lower amplitudes do not necessarily imply weaker contrasts or less confidence, and indeed a lack of scattering cannot be taken as evidence for a lack of anisotropic gradients.

Results
We report 525 observed and back-projected Quasi-Love scatterers located throughout continental Africa, Arabia, India and the Atlantic and Indian oceans (Figure 4; Table S1).Due at least in part to the distribution of suitable earthquake-station pairs (Figure 2), a large number of these are located off the east coast of Africa, in Madagascar or the Indian Ocean.However, we find evidence for Quasi-Love scattering from many different regions.It should √ km, where 100 km is the minimum resolution relative to the wavelength and the (0.05 × δx) factor scales with distance δx from the receiver considering the potential error associated with the velocity and raypath assumptions.
be noted that each measurement is likely associated with a considerable location uncertainty on the order of 100s of km (Figure 4), meaning that some of the observed scatter may be due to mislocations (thus effectively "blurring") of the same features; yet, overall, there is clearly a wide range of scattering features recorded.
To understand the overall spatial distribution of scatterers, and where they are preferentially located, it can be instructive to compare with random distributions (e.g., Eakin, 2021).To account for the heterogeneous distribution of available ray paths (Figure 2), we generated pseudo-random distributions of possible scatterer locations by randomly selecting a path from all available ray paths, and selecting a point at random along the receiver-side half of that ray path; this was repeated until the same number of random scatterers were generated within the study area as there were observed scatterers.We repeated this process 100 times to get 100 sets of random scatterers with which to compare our results.The cumulative distribution function of the distances of observed scatterers from certain features can then be calculated and compared with the same cumulative distribution functions for the random scatterers.A two-sample Komolgorov-Smirnoff (K-S) test (Komolgorov, 1933) can be used to compare two cumulative distribution functions and calculate the probability that they come from the same population; this effectively tests the statistical significance of an association between Quasi-Love scatterers and particular features (e.g., Davies et al., 2015;Eakin, 2021;Hoggard et al., 2020).For example, we tested a possible association of Quasi-Love scatterers with geological terrane boundaries as mapped by Hasterok et al. (2022), but we were unable to identify a statistically significant relationship based on K-S testing (Text S1 in Supporting Information S1).

Lithospheric Thickness
Quasi-Love scatterers are found in locations that span the full range of lithospheric thickness (Figures 5a and 5b).In this case we compare lithospheric thickness from the Hoggard et al. ( 2020) SLNAAFSA model which uses as a proxy for the lithosphere-asthenosphere boundary the 1,175°C isotherm, calculated from shear-wave velocities following the anelastic calibration of Richards et al. (2020).Although the choice of isotherm for the LAB depth is debated, with higher temperatures suggested by others (e.g., Fullea et al., 2021), the chosen isotherm will not affect the relative thicknesses and outlines of thinner/thicker regions.The SLNAAFSA model is derived from the global tomographic shear-wave velocity model SL2013sv (Schaeffer & Lebedev, 2013), updated with regional models for Africa (Celli, Lebedev, Schaeffer, & Gaina, 2020) and the South Atlantic (Celli, Lebedev, Schaeffer, Ravenna, & Gaina, 2020).
With this independent model, more Quasi-Love scatterers are found to occur in the 130-180 km thickness range than expected from randomly generated points (Figure 5b).This depth range corresponds to a local minimum in the overall distribution of lithospheric thicknesses in the region, representing the transition from the dominant "thin" lithosphere to the thick, cratonic lithosphere.A correlation of scatterers with the edges of the thick cratons is apparent, as can be seen, for example, along the northern boundary of the West African craton and the northern boundary of the Congo craton (Figure 5a).The 150 km contour in LAB depth, according to the Hoggard et al. ( 2020) SLNAAFSA model, tends to follow the edges of thick lithosphere, and 29% of scatterers occur within 200 km of this contour, compared with 22.5 ± 1.9% expected from randomly generated distributions (dashed red line; Figure 5c).A two-sample K-S test for the randomly generated cumulative distribution function and the observed dataset gives a p-value of 0.015, well within the threshold of 0.05 (i.e., 95% percent confidence) for rejecting the null hypothesis that the distribution is random.Thus an association of Quasi-Love scattering with the edges of thick lithosphere can be inferred.

The Ocean-Continent Boundary
Of the 525 scatterers with the study region, 140, or 27%, are located within 200 km of the ocean-continent boundary (Figure 6a).From 100 randomly generated distributions, the average proportion within 200 km of the ocean-continent boundary is expected to be 22.8 ± 1.8% (Figure 6b).Thus the ocean-continent boundary is over-represented in our Quasi-Love wave dataset, as also observed by Eakin (2021) in Australia.However, a twosample K-S test for the overall distribution with respect to the random distributions results in a probability of 0.11, which does not fall below the usual threshold of 0.05 that would allow us to reject the null hypothesis that the distribution is random with respect to the ocean-continent boundary.As such, a relationship between Quasi-Love scattering and the ocean-continent transition may still be considered a possibility but should not be treated as significant.

Lithospheric Scattering: Southern and West Africa
Given the Quasi-Love scatterers are expected to be sensitive to anisotropic gradients in the 100-200 km depth range, any scattering where the lithosphere is >180 km thick is presumably due to a change in fabrics within the lithosphere, rather than the asthenosphere.The major areas where thick lithosphere is present are southern and West Africa.
Scatterers within thicker lithosphere are seen commonly in southern Africa, where many are associated with lithospheric terrane boundaries (Figure 7).For instance, an ∼E-W trending band of scatterers at ∼23°S coincides with the Limpopo Belt, a late-Archean orogenic belt formed by the collision between the Zimbabwe and Kaapvaal cratons.Shear-wave splitting observations from this area also indicate a change in the pattern of seismic anisotropy across the belt (Silver et al., 2001), as do surface wave velocity inversions (Adam & Lebedev, 2012;Debayle & Ricard, 2013, Figure 1b).Both suggest the fast direction is generally orientated NE-SW within the Kaapvaal craton and closer to N-S over the Zimbabwe craton, with E-W orientations in the Limpopo Belt.Previous shear-wave splitting results also indicate sharp transitions in fabrics around the edges of the Tanzania craton, where the East African Rift bifurcates to surround the craton; correspondingly, Quasi-Love scattering can be found within the vicinity of the craton, with several results of relatively large amplitude (Figure 7).As a rule, through southern Africa, scatterers tend to occur either within the pre-Cambrian orogenic belts or close to the edges of cratons, with notable exceptions within the Congo craton, where internal boundaries are obscured by the thick Congo basin.In West Africa, scatterers are clearly associated with the boundaries of the thick lithosphere of the West African craton, notably its northern and western boundaries (Section 4.1; Figure 5a), while a further six scatterers are observed within the craton interior, beneath the Taoudeni Basin.

North-East Africa and Arabia
Many Quasi-Love scatterers occur close to the East African Rift (Figure 8).Scatterers are observed throughout the northern part, the Main Ethiopian Rift, where the lithosphere is <70 km thick.Scatterers that may be associated with the rift also occur in thicker lithosphere further south, along both Eastern and Western Rifts, as well as along the Malawi Rift Zone and Davie Ridge (the latter coincident with the ocean-continent boundary) (Figure 8).To the north of the Afar triple junction, scatterers are less commonly seen along the active Red Sea rift, but eight scatterers fall along the low-seismic-velocity zone that underlies western Arabia, beneath the Neogene-Recent lava fields known as Harrats (Figure 8).Other scatterers in Arabia commonly fall close to the boundary of the thick lithosphere underlying eastern Arabia, with some in its interior.
Relatively few Quasi-Love scatterers are observed in northern Africa, due at least in part to the paucity of stations.Apart from those associated with the edges of the West African and Congo Cratons, many scatterers appear to be associated with the zones of active volcanism, such as the Cameroon Volcanic Line and the Hoggar, Harouj and Darfur domes (Figure 8).A further set of scatterers appears to form a ring around the edge of the proposed Al-Kufrah cratonic remnant (Figure 8), a region of locally thick lithosphere proposed by Sobh et al. (2020) to be a remnant of the ancient Saharan Metacraton root.

The South Atlantic and Indian Oceans
The lithosphere in the oceans is almost always <120 km thick, and thus observed Quasi-Love waves are most likely to be due to changes in anisotropy within the asthenospheric mantle, that is, representative of the present day active flow field.In the South Atlantic, away from the ocean-continent boundary and the mid-ocean ridge, Quasi-Love scatterers appear to be mostly associated with known hotspots and/or their tracks (Figure 9).For example, a set of scatters line up along the St. Helena hotspot track between the island of St. Helena and Cameroon, while others lie close to the Tristan da Cunha-Gough hotspot track further south (Figure 9).Further scatterers may be associated with the Cape Verde and Vema hotspots (Figure 9).However, the uncertainty in the nature and location of hotspot tracks, and the small number of scatterers in this region, precludes any conclusive statistical analysis of associations.
Quasi-Love scatterers in the western Indian Ocean are widespread, although, given location uncertainty (Figure 4), many nearby scatterer locations may be recording the same gradient.Nonetheless, they concentrate mostly around certain features: these include Madagascar, the Seychelles microcontinent and the intervening Mascarene Basin, as well as the mid-ocean ridges (Figure 10a).There may also be an association with lowseismic velocity bodies imaged by tomography such as that of Wamba et al. (2021) (Figure 10b).Scattering in Madagascar coincides with very variable results from shear-wave splitting results (Figure 10b; e.g., Paul & Eakin, 2017;Ramirez et al., 2018;Reiss et al., 2016), as well as complex tectonic boundaries, where Stamps et al. (2021) suggest a diffuse deforming zone extends across much of northern Madagascar.(Ball et al., 2019), with notable magmatic centers marked as Ho: Hoggar; Ai: Air; Ha: Harouj; Ti: Tibesti; Da: Darfur; CVL: Cameroon Volcanic Line.Branches of the East African Rift System (Stamps et al., 2021) are marked in red as MER: Main Ethiopian Rift; WR: Western Rift; ER: Eastern Rift; MR: Malawi Rift; DR: Davie Ridge.Other plate boundaries in blue (Bird, 2003).Congo Craton (CC) and West African Craton (WAC) also marked.

Discussion
Similar to that previously seen for other continents (Eakin, 2021), observed Quasi-Love wave scattering in and around Africa is widespread, indicating abundant, and in large part previously unmapped, lateral gradients in seismic anisotropy.It is not possible to determine precisely the cause of any particular scattering observation, given the large spatial errors and range of processes that can give rise to seismic anisotropy.However, the overall collective distribution of scatterers allows us to make a series of inferences.For example, the generation of Quasi-Love waves is clearly influenced by both asthenospheric and lithospheric processes, as, while the Quasi-Love waves are sensitive primarily to the 100-200 km depth range (e.g., Yu & Park, 1993;Yu et al., 1995), scattering occurs in regions of both thin (<100 km) and thick (>200 km) lithosphere (e.g., Figures 5 and 11), and, notably, at the transition between the two (Section 4.1, Figure 5).This overall distribution remains very similar even when considering only the highest amplitude Quasi-Love waves with respect to Love wave amplitudes: Figure S3 in Supporting Information S1 shows only those scatterers with Quasi Love wave amplitude more than 0.2 times the Love wave amplitudes, and the same major features are highlighted.Below, we discuss what these distributions may signify, for the continental, ocean-continent boundary, and oceanic regions in turn.

Continental Africa and Arabia
Where the lithosphere is thick, scattering is often most easily explained by contrasts between adjacent blocks of differing tectonic origin.Lateral gradients in seismic anisotropy are already mapped across the Limpopo Belt in southern Africa by shear-wave splitting (Silver et al., 2001) and surface-wave dispersion analysis (Adam & Lebedev, 2012), and indeed we see Quasi-Love waves produced at this ancient orogenic belt (Figure 7).There is evidence that similar gradients, deep deformation fabrics preserved from ancient tectonic events, may be widespread in southern Africa, with scattering in the Damara-Irumide Belt and at other terrane boundaries (Figure 7).Scattering within craton interiors, in the Congo (Figure 7) and West African (e.g., Figure 4) Cratons, may be due to interior structure that is not mapped at the surface due to the overlying thick intracratonic basins.Cratons being commonly formed of smaller anisotropically distinct pieces is suggested by Chen et al. (2021) based on receiver function evidence from the Yilgarn and Superior cratons.
A strong association is revealed between scatterers and the edges of thick lithosphere (Section 4.1, Figure 5).The association may reflect a contrast between anisotropy within the lithospheric and asthenospheric mantle, or lateral changes in asthenospheric flow direction, caused, for example, by edge-driven convection generated by the step in the LAB (King & Anderson, 1998).Edge-driven convection at the northern margin of the Congo craton has been proposed as a mechanism to generate the observed volcanism at the Cameroon Volcanic Line (Reusch et al., 2010), while others have proposed lithospheric instabilities (Milelli et al., 2012).
Our results add new constraints to the area occupied by the Saharan Metacraton, where poor seismograph coverage has meant that mantle structures have been poorly constrained.In this region shear-wave splitting measurements are almost entirely absent, and little variation in the anisotropic pattern is evident from global anisotropic surface-wave tomography (Figure 1b).Our new results corroborate previous inferences made about the region, notably the possible presence of the Al-Kufrah cratonic root, as modeled by Sobh et al. (2020).A ring of Quasi-Love wave scatterers appears to encircle the entire boundary of the proposed lithospheric root (Figure 8), consistent with a similar process to that causing craton-edge scattering in the rest of Africa, and perhaps indicating a local underestimation in the lithospheric thickness in this region of poor seismological coverage (the lithosphere is mapped to be <150 km thick throughout most of the region according to Hoggard et al., 2020).The Al-Kufrah remnant may therefore be a strong, long-lived root that survived the proposed partial destruction of Saharan lithospheric mantle roots in the Pan-African orogeny (Sobh et al., 2020).Across the rest of the Metacraton, scatterers are generally associated with areas of Cenozoic volcanism, which are proposed to be the surface expression of small-scale mantle upwellings (e.g., Ball et al., 2019;McKenzie, 2020); this is the first evidence from seismic anisotropy that could support this.2011)) in magenta.Red lines mark the active boundaries of the East African Rift System (EARS), with yellow dashed lines marking uncertain boundaries, and the yellow region marking an area of diffuse deformation (Stamps et al., 2021).Yellow stars mark hotspot locations.MB: Mascarene Basin; RR: Rodrigues Ridge.(b) Plotted against the 171 km depth slice from the SEMINDO V S tomography model (Wamba et al., 2021).The purple line marks the hotspot track of the Réunion plume after Maher et al. (2015).Shear-wave splitting results from the Wüstefeld et al. (2009) compilation are shown in green; those in the Indian Ocean and Madagascar come primarily from the studies of Hammond et al. (2005), Barruol and Fontaine (2013), Reiss et al. (2016), Eakin et al. (2018), Ramirez et al. (2018), andScholz et al. (2018).
Geochemistry, Geophysics, Geosystems 10.1029/2023GC011385 Quasi-Love scattering throughout the northern part of the East African Rift, where the lithosphere is thin, is most likely due to focused upwelling, and/or along-axis channelized flow, in the asthenosphere (e.g., Civiero et al., 2022;Ebinger et al., 2024).Evidence from shear-wave splitting indicates the strongest azimuthal anisotropy is beneath the thinnest lithosphere with longest history of volcanism, meaning hot asthenosphere has been concentrated in these areas (Ebinger et al., 2024).This would be consistent with contrasts in upper mantle anisotropy between the rift zone and the surrounding areas, and between the thinner, northern parts and the thicker lithosphere further south associated with the Tanzanian craton, as is suggested by the scattering (Figures 7 and 8).Scattering where the lithosphere is thicker may also be influenced by pre-existing anisotropic gradients, such as those across craton boundaries, which the rift has subsequently exploited.Scatterers along the Malawi Rift Zone, as found here (e.g., Figure 7), were also found by Cheng et al. (2022), who argued that the lateral anisotropic gradient signified edge-driven convection at the eastern edge of the putative Niassa craton, whose lithospheric root was identified by Celli, Lebedev, Schaeffer, and Gaina (2020).
To the north of the rift, scattering appears to occur along the low-velocity asthenospheric channel in Western Arabia (Figure 8).Extensive volcanism and topographic highs along this axis have been previously explained by either upwelling or channelized transport of hot material from the Afar plume (Camp & Roobol, 1992;Hansen et al., 2006;Wilson et al., 2014), both of which would represent deviations in mantle flow that could produce gradients in seismic anisotropy.

The Ocean-Continent Boundary
A possible association between scatterers and the ocean-continent boundary (Section 4.2, Figure 6) is consistent with the findings of Eakin ( 2021) who find a strong association with the Australian ocean-continent boundary, and Servali et al. (2020) who find scattering at the North American Atlantic margin.In places, such as the west coast of West Africa or the east coast of southern Africa, the ocean-continent boundary is coincident with the edges of thick lithosphere (Figure 5), so a lateral gradient in seismic anisotropy is expected for reasons not related to the ocean-continent boundary.At the north-west coast of Western Africa, the lithospheric root steps back from the coast, and here there appears to be a double-line of scatterers, one at the lithospheric root edge and one below the ocean-continent boundary (compare Figures 5 and 6), although those at the ocean-continent boundary may also be associated with the Canary Islands hotspot track (Figure 9; Maher et al., 2015).Unlike in Australia (Eakin, 2021), there are few scatterers associated with the ocean-continent boundary where the lithosphere is thin on both sides, except in the microcontinents of the western Indian Ocean.Given this, and the relatively low statistical significance of the association in our dataset, we are not able to further constrain whether the oceancontinent boundary is typically associated with deviations in asthenospheric flow regardless of lithospheric thickness.

The Atlantic and Indian Oceans
An association between scatterers and possible asthenospheric upwellings is apparent in the South Atlantic, perhaps most notably along the St. Helena Rise (Figure 9).Here a clear hotspot track is coincident with a linear low-velocity anomaly (Celli, Lebedev, Schaeffer, Ravenna, & Gaina, 2020;Schaeffer & Lebedev, 2013) and a linear positive residual topography anomaly (Davies et al., 2019;Hoggard et al., 2017).The low-velocity anomaly here was suggested by Celli, Lebedev, Schaeffer, Ravenna, and Gaina (2020) to represent a line of locally thinned lithosphere providing a conduit for hot, buoyant asthenospheric material.The St. Helena Rise consists of an 80-Myr record of volcanic seamounts tracing the path of the African plate over the St. Helena mantle plume (Maher et al., 2015).The present-day Quasi-Love wave scattering, combined with the low-velocity anomaly, suggests ongoing, if in large part avolcanic, upwelling or lateral deviations of asthenospheric flow beneath such features.In the case of the St. Helena Rise, the associated elongated low-velocity feature extends onshore beneath the Cameroon Volcanic Line, where magmatism is ongoing, indicative of mantle upwelling and coincident with numerous Quasi-Love wave scatterers.
Scattering in the rest of the South Atlantic is located for the most part close to either the mid-ocean ridge (where passive upwelling is expected), active hotspots (Cape Verde, Canary, Vema), or mapped igneous provinces that may be related to past hotspot locations, for example, around the Walvis Ridge, associated with the Tristan da Cunha and Gough hotspots (e.g., O'Connor et al., 2018), or volcanism further south (Figure 9).The association with both present and past hotspot locations suggests that, after a plate has moved across a mantle plume, the resulting thinning of the lithosphere may lead to continued localization of asthenospheric upwelling or flow, that would result in similarly localized gradients in seismic anisotropy that can be detected by Quasi-Love waves.
Less straightforward than the South Atlantic, the western Indian Ocean (Figure 10) has a complex rifting and seafloor spreading history and contains several continental fragments, notably Madagascar and the Seychelles (e.g., Plummer & Belle, 1995).Madagascar is further associated with possible internal tectonic boundaries, where Stamps et al. ( 2021) have identified a diffuse deformation zone in the north of the island, and possible plate boundary separating the northern and southern parts of the island.Quasi-Love scattering is associated with these continental fragments, which, given the lithosphere is ≲100 km thick, suggests that asthenospheric flow is disturbed by the complexity in lithospheric structure above.Indeed, existing measurements of anisotropy beneath Madagascar from shear-wave splitting have been very variable (Figure 10b), with fast directions changing from roughly E-W in the north to roughly N-S in the south, with a variety of fast directions in the center of the island, and some null results (Paul & Eakin, 2017;Ramirez et al., 2018;Reiss et al., 2016).Active upwelling of the asthenosphere, perhaps in response to lithospheric delamination, has been suggested, especially in the center of the island (Ramirez et al., 2018).Here, there is a band of consistent Quasi-Love wave scattering that may attest to the suggested vertical motions (Figure 10).Scatterers are also frequent within Mascarene Basin, between Madagascar and the Seychelles-Mauritius microcontinents, which contains a fossil mid-ocean ridge that has been inactive since the rifting of India from the Seychelles at around 63 Ma (e.g., Collier et al., 2008).Beneath this basin a broad low-velocity anomaly has been mapped in the asthenosphere, called the Mascarene Basin Asthenospheric Reservoir (MBAR; Figure 11c) suggested to be a reservoir of hot mantle sourced from the Réunion plume (Barruol et al., 2019).The Quasi-Love scatterers may thus indicate complex flow patterns associated with this reservoir.Moreover, to the east of Réunion, scatterers plot close to the Rodrigues Ridge, suggested to be the site of an asthenospheric flow channel connecting the Réunion plume to the mid-ocean ridge (Barruol et al., 2019) (Figures 10 and 11b).Elsewhere in the western Indian Ocean, scatterers are associated frequently with the present-day active mid-ocean ridge systems, consistent with localized passive upwelling.A few sets of scatterers in the northern part of the basin coincide with alternating positive and negative shear-wave velocity anomalies at asthenospheric depth (Figure 10b; Wamba et al., 2021), perhaps hinting at further small-scale convection patterns beneath this basin.

Conclusions
New observations of Quasi-Love wave scattering indicate widespread lateral gradients in upper mantle seismic anisotropy beneath Africa, Arabia, the eastern South Atlantic and the western Indian Ocean.These appear to be related to a variety of processes, both lithospheric and asthenospheric.In regions of thick lithosphere they most likely reflect anisotropic gradients associated with contrasts in frozen-in anisotropy between different tectonic blocks or associated with ancient collisional events, such as those found throughout the orogenic belts of southern Africa.Thus the spatial distribution of scatterers could potentially be used to further identify previously unmapped structures within the West African and Congo Craton interiors that are currently hidden beneath sedimentary cover, such as the Taoudeni Basin in West Africa.
Most clearly we find Quasi-Love wave scattering is preferentially associated with the edges of thick lithosphere.This is most easily explained by the contrast between lithospheric and asthenospheric anisotropy, but could also be attributed to small-scale asthenospheric convection tied to the edge of the thick lithosphere.Interestingly, scatterers surrounding the proposed Al-Kufrah cratonic fragment in North East Africa, where existing seismological models are low in resolution due to poor station coverage, may lend weight to arguments supporting a deep, localized lithospheric root to this feature.
Where the lithosphere is thin, Quasi-Love wave scattering most likely reveals gradients in asthenospheric anisotropy below.These are often associated with locations of proposed upwellings or small-scale convective dynamics induced by variations in LAB topography above.This includes below the East African Rift, underlying intra-plate magmatic centers in north Africa, and beneath oceanic hotspots.A correspondence with hotspot tracks in the South Atlantic is a notable example, suggesting localized upwelling, or channelized flow, along corridors of thinned lithosphere left behind as the plate drifted over the mantle plume.Lastly, pervasive scattering in the western Indian Ocean is consistent with small-scale convection and various upper mantle velocity anomalies beneath this ocean basin.These both likely result from the complex and piece-wise rifting history that formed this ocean basin, as well as ongoing mantle plume activity beneath it today.

Data Availability Statement
A table of all Quasi-Love wave observations with their back-projected scatterer locations can be found in Table S1.All seismological data used in this study are publicly available and were sourced via the International Federation of Digital Seismograph Networks (FDSN) webservices, from repositories hosted by IRIS (https://ds.

Figure 1 .
Figure 1.(a) Topographic map of Africa and surroundings with the major (meta)cratons highlighted (WAC: West African Craton; CC: Congo Craton; TC: Tanzania Craton; ZC: Zimbabwe Craton; KC: Kaapvaal Craton; SMC: Saharan Metacraton) as well as the intra-cratonic Congo (Cb) and Taoudeni (Tb) basins.Yellow inverted triangles are seismic stations used in this study and red triangles are locations of Holocene volcanoes (Global Volcanism Program, 2013).Volcanoes or seamounts associated with oceanic hotspots are labeled (Ma: Madeira; Ca: Canary; As: Ascension; SH: St. Helena; TdC: Tristan da Cunha; G: Gough; V: Vema; Ré: Réunion; Co: Comoros).Black lines mark tectonic plate boundaries from Bird (2003), with purple lines to mark the major boundaries of the East African Rift System (EARS) from Stamps et al. (2021), dashed to mark the uncertain southern continuation of the boundary.(b) Overview of existing azimuthal anisotropy measurements for the same region.Station-averaged teleseismic shear-wave splitting measurements from the updated compilation of Wüstefeld et al. (2009), with additional data from Qaysi et al. (2018), Komeazi et al. (2023), and Ebinger et al. (2024), are plotted in pink.The orientation and strength of anisotropy at 150 km depth from the anisotropic global surface wave tomography model 3D2018_08Sv(Debayle et al., 2016) is represented by the black bars.This is plotted against the lithospheric thickness map ofHoggard et al. (2020) based on the SLNAAFSA model, which in this region is a combination of the global SL2013sv(Schaeffer & Lebedev, 2013) and regional African AF2019(Celli, Lebedev, Schaeffer, & Gaina, 2020)  shear-wave velocity tomographic models.The separate cratonic "cores" of the West African and Congo cratons identified byCelli, Lebedev, Schaeffer, and Gaina (2020)  are labeled (Re: Reguibat; ML: Man-Lèo; BK: Bomu-Kibale; GC: Gabon-Cameroon; Ka: Kasai; Cu: Cubango) as is the Angolan Shield (Ag) which lacks a deep root, and the locations of the Niassa (Ni) cratonic root proposed byCelli, Lebedev, Schaeffer, and Gaina (2020)  and Al-Kufrah (AK) cratonic remnant proposed bySobh et al. (2020).

Figure 2 .
Figure 2. All 1284 earthquakes (stars) and 133 seismograph stations (triangles) used in this study.Gray lines indicate earthquake-station paths that did not result in a well-constrained Quasi-Love wave; red lines are those that did.

Figure 3 .
Figure 3. Quasi-Love wave example.(a) The transverse (T), vertical (Z) and radial (R) components of the seismogram, filtered with a Butterworth bandpass filter between 0.002 and 0.01 Hz, are shown for station SUR in South Africa relative to the source time of a 35 km-deep, M W 6.8 earthquake in Japan on 12 May 2015.The green and blue shaded areas on the T and Z components respectively show the approximate expected arrival windows for fundamental Love and Rayleigh waves.Thick dashed green and blue lines mark the times of the maximum amplitudes of the Love (G1) and Rayleigh (R1) waves respectively.Below, the Hilbert transform of the radial component (H(R), dashed red line) is superposed on the vertical component (solid blue line), showing good alignment for both the Rayleigh wave and an earlier Quasi-Love (QL) phase.Below again, the stack of Z + H(R) is shown in magenta, with the transverse component in green superposed and time-shifted to its maximum correlation with Z + H(R) in the first 50% of the time between G1 and R1, showing a waveform similarity between the G1 and QL phases.(b) The absolute correlation co-efficient between T and Z + H(R) is shown as a function of time.(c) Shows horizontal particle motion for the G1 phase, indicating linear transverse motion, while (d, e) show vertical-radial particle motion for the R1 and QL phases respectively, showing elliptical vertical-radial motion for both R1 and QL.(f) Taking the Hilbert transform of the radial component results in linear particle motion in the H(R)-Z plane for the QL phase.(g) Global map with an azimuthal equidistant projection centered on the receiver, showing the great-circle raypath between the source and receiver, with a blue circle to indicate the back-projected position of the QL phase, in Madagascar.

Figure 4 .
Figure 4. Projected Quasi-Love wave scatterers plotted as crosses against the global tectonic map of Hasterok et al. (2022).The different geological provinces are colored by their most recent tectonic event.The size of the crosses is proportional to the amplitude of the Quasi-Love wave relative to the Love wave.White circles represent indicative errors in the scatterer locations, estimated by ̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ ̅ 100 2 + (0.05 × δx) 2

Figure 5 .
Figure 5. (a) Locations of Quasi-Love wave scatterers, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave, plotted against the depth to lithosphere-asthenosphere boundary (LAB), approximated by the 1175°C isotherm, calculated by Hoggard et al. (2020) from the SLNAAFSA model (see text for details).The red dashed line marks the 150 km depth contour, and scatterers highlighted in red are those within 200 km of this contour.(b) Colored bars represent a histogram of lithospheric thickness from the same model across the region in (a).Gray lines represent the histograms of LAB depth sampled from the model by randomly generated scattering points.The red line represents the histogram of LAB depth sampled by our observed Quasi-Love scatterers.Thus, where the red line is above the gray lines, at 130-180 km, there are more scatterers than expected.(c) The cumulative distribution function in red of observed scatterers in (a) as a function of distance from the 150 km depth LAB contour, against the randomly generated sets of scatterers in gray, where the black lines indicate the mean and standard deviation of these sets.

Figure 6 .
Figure 6.(a) Quasi-Love wave scatterers, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave, with the ocean-continent boundary shown in green (Hasterok et al., 2022).Scatterers highlighted in red are within 200 km of the ocean-continent boundary.(b) The cumulative distribution function in blue of observed scatterers as a function of distance from the ocean-continent boundary, against the randomly generated sets of scatterers in gray, where the black lines indicate the mean and standard deviation of these sets.

Figure 8 .
Figure 8. Projected Quasi-Love wave scatterers in north-east Africa, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave, plotted against the AF2019 V S tomography model (Celli, Lebedev, Schaeffer, & Gaina, 2020).Areas with lithospheric thickness >150 km are obscured.The approximate outline of the putative Al-Kufrah cratonic fragment (Sobh et al., 2020) is shown by the orange dashed line.Black polygons represent Neogene-Recent volcanism (Ball et al., 2019), with notable magmatic centers marked as Ho: Hoggar; Ai: Air; Ha: Harouj; Ti: Tibesti; Da: Darfur; CVL: Cameroon Volcanic Line.Branches of the East African Rift System (Stamps et al., 2021) are marked in red as MER: Main Ethiopian Rift; WR: Western Rift; ER: Eastern Rift; MR: Malawi Rift; DR: Davie Ridge.Other plate boundaries in blue(Bird, 2003).Congo Craton (CC) and West African Craton (WAC) also marked.

Figure 9 .
Figure 9. Quasi-Love scatterers in the South Atlantic Ocean, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave, plotted against igneous provinces (orange polygons; Stephenson et al., 2023), hotspot locations (yellow stars, Az: Azores; GM: Great Meteor; Ma: Madeira; Ca: Canary; CV: Cape Verde; As: Ascension; SH: St. Helena; TdC: Tristan da Cunha; G: Gough; V: Vema; D: Discovery; S: Shona; B: Bouvet) and major hotspot tracks (purple lines, after Maher et al., 2015).Ocean-continent boundaries are in green and mid-ocean ridges in blue.

Figure 10 .
Figure 10.Quasi-Love scatterers in the Indian Ocean, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave: (a) plotted against seafloor age, with the Seychelles (Se) and Mauritius (Ms) microcontinents plotted in gray, and with active ridges marked in blue and extinct ridges (from Matthews et al. (2011)) in magenta.Red lines mark the active boundaries of the East African Rift System (EARS), with yellow dashed lines marking uncertain boundaries, and the yellow region marking an area of diffuse deformation(Stamps et al., 2021).Yellow stars mark hotspot locations.MB: Mascarene Basin; RR: Rodrigues Ridge.(b) Plotted against the 171 km depth slice from the SEMINDO V S tomography model(Wamba et al., 2021).The purple line marks the hotspot track of the Réunion plume afterMaher et al. (2015).Shear-wave splitting results from theWüstefeld et al. (2009) compilation are shown in green; those in the Indian Ocean and Madagascar come primarily from the studies ofHammond et al. (2005),Barruol and Fontaine (2013),Reiss et al. (2016),Eakin et al. (2018),Ramirez et al. (2018), andScholz et al. (2018).

Figure 11 .
Figure 11.(a) Quasi-Love scatterers, with symbol size proportional to the Quasi-Love amplitude relative to the Love wave, plotted against shear-wave velocity at 150 km depth, taken from the AF2019 tomographic model (Celli, Lebedev, Schaeffer, & Gaina, 2020) blended with the SL2013sv model (Schaeffer & Lebedev, 2013) outside of the AF2019 bounds.(b) Cross section through the blended model from A to A′ as marked in (a), with Quasi-Love scatterers plotted at an indicative depth of 150 km.Topography is shown on top, with filled dark gray for continental areas, with red triangles indicating locations of active volcanoes, and purple and magenta lines to show active and inactive spreading ridges respectively.Possible interpretations are labeled.The lithosphere-asthenosphere boundary (LAB) from Hoggard et al. (2020) is shown as a gray dashed line.Abbreviations are Ma: Madagascar; Re: Reguibat Shield; ML: Man-Léo Shield; CVL: Cameroon Volcanic Line; GC: Gabon-Cameroon Shield; Ka: Kasai Shield; ZC: Zimbabwe Craton.(c) Same as (b) with the cross-section from B to B′. AK: Al-Kufrah cratonic fragment; MBAR: Mascarene Basin Asthenospheric Reservoir.