Focused Mantle Upwelling Beneath the Southeastern Asian Basalt Province Revealed by Seismic Surface Wave Tomography

Following the termination of seafloor spreading in the South China Sea (SCS) basin, abundant intraplate volcanism widely spreads in the Indochina block, SCS basin, and Leiqiong area, forming the Southeastern Asian Basalt Province (SABP). The geodynamic origin of the SABP has long been enigmatic and debated. Here, we present a high‐resolution 3‐D upper mantle S‐wave velocity model in the region by conducting earthquake‐based surface wave tomography with seismic data collected across Southeast Asia. The resultant images depict a plume‐like structure beneath the central area of the SABP, characterized by a continuous, sub‐vertical low‐velocity column in the upper mantle. Our new findings, combined with previous geochemical and geodynamic evidence, suggest that the extensive post‐spreading intraplate volcanism within the SABP is likely induced by this focused mantle upwelling, which could be further traced down to the core‐mantle boundary as inferred by existing global velocity models.

The erupted basaltic volcanics with a thickness up to hundreds of meters were mainly distributed in the Leiqiong area, the Indochina peninsula, and also the SCS basin, forming the so-called Southeastern Asian Basalt Province (SABP) with an emplacing area of ∼0.037 Mkm 2 (see the dashed circle in Figure 1a; Gu et al., 2019;Hoang & Flower, 1998;Yan et al., 2018).
Mantle plumes have long been conceived as the origin of Large Igneous Provinces (LIPs) (e.g., French & Romanowicz, 2015;Richards et al., 1989).Previous regional (Lebedev & Nolet, 2003) and global-scale (Montelli et al., 2006) seismic tomography studies advocated that there is a lower-mantle-rooted thermal plume beneath the Hainan island (southern part of the Leiqiong area; Figure 1a).This geophysical observation later on received increasing support from a great number of geochemical studies (e.g., An et al., 2017;Gu et al., 2019;Wang et al., 2012;Wang et al., 2013;Xu et al., 2012;Yan et al., 2018;Yu et al., 2018;Zhang et al., 2020;Zou & Fan, 2010), which have commonly suggested that the basaltic volcanism within the SABP, dominated by alkali basalts and subsidiary tholeiites, share the same isotopic and geochemical characteristics with Light Rare Element Enriched (LREE) patterns and typical Oceanic Island Basalt (OIB)-type incompatible element distributions.
However, the very existence of the proposed Hainan plume remains disputed, and if it exists, its exact location and the role it played in shaping the tectonics of Southeast Asia are still ambiguous (Li et al., 2021;Lin et al., 2019;Yang et al., 2021;Zhao et al., 2021).On the one hand, although previous tomographic surveys detected the Hainan plume with notable low-velocity anomalies, its geometry and depth range acquired by these studies are not consistent with each other (e.g., Hua et al., 2022;Huang, 2014;Huang et al., 2015;Lebedev & Nolet, 2003;Lei et al., 2009;Montelli et al., 2006;Toyokuni et al., 2022;Wang et al., 2022;Xia et al., 2016), likely due to different data sets and methods used.On the other hand, previous geophysical efforts to investigate the origin of the SABP have largely been confined to the southernmost edge of the South China block (i.e., Hainan island and its neighboring; e.g., Huang, 2014;Lei et al., 2009;Liu et al., 2018;Lu et al., 2022;Xia et al., 2016), making it hard to explain how the Hainan plume model relates to volcanism within the Indochina block and the SCS basin.As a result, for the vast region in the SCS and surroundings covering the whole SABP, more robust observational evidence that could directly verify the presence of as well as characterize the nature of the expected mantle plume structure is still lacking.

Seismic Data
Broadband vertical recordings of earthquake data were retrieved from three groups of seismic stations (Figure 1b), including 28 CEA (China Earthquake Administration) stations (Zheng et al., 2010) (Yang et al., 2015) from November 2009 to December 2012.Teleseismic earthquakes with surface wave magnitude (Ms) greater than 5.5, focal depth less than 50 km, epicentral distance between 10° and 150°, period ranging from November 2009 to December 2018 were compiled, generating 3,172 seismic events in total (Figure 1c).

Seismic Surface Wave Tomography
We employed the two-station approach (Legendre et al., 2014;Meier et al., 2004) to measure the Rayleighwave phase velocity dispersion curves.To ensure that the variation in waveforms at two selected stations mainly reflects the structure in between, this method requires a small angle between the great circle connecting the two stations and the circle connecting a certain event to the station pair.To produce an adequate data set for dispersion measurements and avoid too large deviations from the great circle path criterion at the same time, we chose an upper limit of 5° for this angle in the selection of earthquakes and station pairs.More details related to the dispersion measurements can be found in Text S1 and Figure S1 in Supporting Information S1.
It has long been recognized that heterogenous velocity structures could lead to off-great-circle arrivals.According to Foster et al. (2014), the differences in estimated phase velocities resulting from corrections for arrival angle mostly fall in the range of 0%-1%, which indicates the discrepancies are generally negligible from the point view of interpreting the first-order tomographic features (as displayed in Figure 12 in their article).Furthermore, as documented in Foster et al. (2014) and Magrini et al. (2020), the arrival angle correction mainly produces significant improvement (∼1% of the phase velocity) at the period band of 20-50 s.For longer period Rayleigh waves (>50 s), the deviations of the measured phase velocities between the results with/without correction are systematically below 1%, which can be explained by the weaker lateral heterogeneity at relatively longer periods (sensitive to deeper structures).Since this study largely focuses on the upper mantle structures that are presumably constrained by longer period Rayleigh waves, the influence of the angle deviations, in our opinion, should be limited and is unlikely to produce substantial biases in the final tomographic images.Nevertheless, the off-great-circle angle deviations need to be corrected for better accuracy in future investigations.
For each station pair, all the phase velocity measurements from available earthquakes were averaged to produce a single path-specific dispersion curve.We repeated this picking procedure for each station pair and obtained a total of 1,819 inter-station dispersion curves (Figure 1d).All the retrieved dispersion curves were then inverted using a LSQR scheme (Deschamps et al., 2008) to derive the 2-D phase velocity maps at the period band of 20-167 s (Text S2 and Figures S2 and S3 in Supporting Information S1).The phase velocity uncertainty (Figure S4 in Supporting Information S1), which is highly correlated with the ray path coverage (Figure S5 in Supporting Information S1), is estimated to be mostly less than 30 m/s throughout the period band.Figure S6 in Supporting Information S1 indicates that a purely isotropic inversion is generally sufficient to explain our two-station measurements.Local dispersion curves at various grid knots were extracted based on the inverted 2-D phase velocity maps.Afterward, a Markov chain Monte Carlo inversion algorithm (Guo et al., 2016) was applied to invert the 1-D Vs profiles, which were finally assembled to create the composite 3-D Vs model (Figures 2 and 3).Examples of 1-D Vs inversion at three selected grid knots (denoted with green stars in Figure 2f) are demonstrated in Figure S9 in Supporting Information S1.More detailed descriptions of the Vs inversion scheme can be found in Text S4 in Supporting Information S1.

Resolution Analysis
To verify our tomographic methods as well as evaluate the resolution achieved by our data set, we conducted checkerboard and spike resolution tests for the inverted 2-D phase velocity maps (Text S3, Figures S7 and S8 in Supporting Information S1), and designed a custom restoring test for the interpreted columnar low Vs shape (Text S5 and Figure S11 in Supporting Information S1).As we can see from Figure S7 in Supporting Information S1, the input checkerboard phase velocity model could be generally well recovered throughout the period band in well-sampled regions, except in the SCS at short periods and near the edges of the study region across the period band where certain smearing effects are observed owing to relatively insufficient ray path coverage (Figure S5 in Supporting Information S1).The outputs of the spike resolution test and the synthetic restoring test (Figures S8 and S11 in Supporting Information S1) indicate that a plume-shaped low-velocity structure resembling the one in Figure 3 beneath the central SABP is indeed resolvable in this study.

Comparison With Existing Tomographic Models
Pronounced lateral variations in Vs across the depth range are observed in the tomographic results (Figure 2).
The Vs model at a shallow 30 km depth (Figure 2a) correlates well with known tectonic units and is consistent with previous regional Vs tomographic findings.For example, significant low velocities are imaged beneath the southeastern margin of the Tibetan Plateau where a relatively thick crust is present, whereas high velocities are observed across the SCS and Celebes Sea that could be explained by their relatively thin crust (e.g., Chen et al., 2021;Tang & Zheng, 2013;Zhao et al., 2019).In particular, the Celebes Sea is imaged with a more prominent high velocity in our model compared to the Vs model of Huang and Xu (2011) (Figure S12 in Supporting Information S1), which indicates that our Vs model achieves higher imaging resolution, especially in resolving relatively small-scale anomalies thanks to the dense regional ray-path coverage (Figure S5 in Supporting Information S1).The northern part of Borneo and the Taiwan region are characterized by notable low velocities, which are documented in previous tomographic images as well (e.g., Chen et al., 2021;Huang & Xu, 2011).(Figure S12 in Supporting Information S1).A recent surface wave tomography study by Chen et al. ( 2021) also detected a low-velocity anomaly in this region at a depth range of ∼100-150 km.As exhibited in Figure 3 and Figure S10 in Supporting Information S1, the two vertical profiles transecting the imaged low Vs body in W-E and S-N directions show this is a spatially concentrated low-velocity formation that vertically extends in the upper mantle.Strikingly, this low Vs anomaly is also observable in an early multimode waveform inversion study (Lebedev & Nolet, 2003).The main difference between our results and this classical work (Figure S13 in Supporting Information S1) is that the primary low-velocity zone is currently imaged beneath the central region of the SABP rather than below the Hainan Island, which we think is attributable to more available data used in this study compared to that from nearly two decades ago.Moreover, this low-velocity feature can also be found in the well-accepted MIT08 global model of mantle Vp (Li et al., 2008) as illustrated in Figure 1 of Wu and Suppe (2018) despite its comparatively low resolution in our study region.Compared to recent mantle Vp models (Hua et al., 2022;Toyokuni et al., 2022), our Vs model achieves generally higher imaging resolution in the shallow depths (<300 km) with significantly larger amplitudes of velocity anomalies as shown in Figure S14 in Supporting Information S1.

Multidisciplinary Evidence for the Focused Mantle Upwelling
The Hainan plume model is originally and repeatedly invoked by previous geophysical means to account for the Late Cenozoic intraplate volcanism erupted in the Leiqiong area after the cessation of seafloor spreading in the SCS basin (e.g., Huang, 2014;Lebedev & Nolet, 2003;Lei et al., 2009;Lu et al., 2022;Montelli et al., 2006;Xia et al., 2016), but it is difficult to interpret the coeval basaltic volcanism that occurred in the Indochina block and the SCS basin given the large geographic distance among these volcanic sites (more than 500 km as shown in Figure 1a).In contrast, the vertically extended low Vs structure resolved in this study, which is beneath the central region of the SABP, better explains the spatial distribution of the widespread post-spreading basaltic volcanism emplaced in a circular manner on the surface (Figure 4).It also correlates with the high mantle potential temperature (∼1,440-1,550°C) estimated for this flood basalt province (e.g., An et al., 2017;Hoang & Flower, 1998;Wang et al., 2012) and agrees with the commonly reported OIB-type isotope geochemistry of the erupted basalts (e.g., An et al., 2017;Gu et al., 2019;Wang et al., 2012;Wang et al., 2013;Xu et al., 2012;Yan et al., 2018;Yu et al., 2018;Zou & Fan, 2010).Meanwhile, our presented tomographic images are well aligned with the recent geodynamic modeling outputs (Figure S15 in Supporting Information S1) that revealed a dome-shaped broad upwelling zone beneath the central SABP in the upper mantle (Lin et al., 2019).These independent validations are particularly important since our inversion is driven only by seismic surface wave data.Thus, these different lines of evidence collectively imply that this plumelike low-velocity anomaly is likely to be present below the central part of the SABP.
It should be noted that the depth extent of this low-velocity anomaly is not constrained in our current study due to the surface waves' imaging ability (from the uppermost mantle down to ∼300 km).To the upper limit, understanding the connection between the imaged low velocities and the widely distributed volcanism at the surface certainly requires more detailed crustal images in the future, which calls for adequate geophysical instrumentation in the SCS.To the lower limit, previous global mantle Vs tomographic models (Figure S16 in Supporting Information S1; Houser et al., 2008;Montelli et al., 2006;Simmons et al., 2010) commonly show that prominent low velocities are discernible from the upper mantle down to the core-mantle boundary (CMB) below the SABP, indicating that the low-velocity anomaly imaged in this study could be traced down to the CMB.

Deep Origin of the Post-Spreading Intraplate Volcanism
Previous petrological, geochemical, and geophysical investigations in favor of mantle plume activities have proposed that the subduction of the Indo-Australian and Pacific Plates to the deep mantle stimulated and contributed to the thermochemical upwelling to form the mantle plume (e.g., Gu et al., 2019;Wang et al., 2013;Wei et al., 2021).The SABP, which consists of the imaged plume-like conduit in the center, is surrounded by broad mantle downwelling zones (i.e., deep slab subduction zones) in all directions (Figure 1a).It has been speculated that the sinking of subducted slabs to the lower mantle could not only push the dense chemical layer upward but also induce thermal instability in the lower mantle and thus lead to the formation of thermal-chemical domes (i.e., plumes or superplumes) (e.g., Steinberger & Torsvik, 2012;Zhong et al., 2007).The unique tectonic setting of this focused mantle upwelling implies that it may be genetically linked to the sinking of the subducted slabs surrounding it down to the CMB (Figure S16 in Supporting Information S1).As proposed by Wang et al. (2013), the avalanches of the subducted Indo-Australian and Pacific slabs to the CMB may have pushed up a thermal-chemical pile to form the upwelling plume, which is ultimately responsible for the abundant intraplate volcanism observed at the surface.
A recent tomographic study (Shi et al., 2023) conducted in the southeastern margin of Eurasia with a great amount of S-wave data reveals a prominent, broad low-velocity feature in and below the mantle transition zone The large-scale sub-vertical conduit located right below the central SABP is interpreted to be the major mantle upwelling trunk that is likely to feed the surface intraplate volcanism.The magenta patches that depict the surface coverage of the Late Cenozoic basaltic volcanism within the SABP and the green dashed circle that marks the surficial extent of the SABP are the same as shown in Figure 1a.
(MTZ) beneath the SCS, which is interpreted to be an upwelling mantle plume structure.Combining evidence from receiver function observations of a relatively thinner MTZ beneath this region (Wei & Chen, 2016;Yu et al., 2017), we speculate that the upper mantle plume is likely to be sourced directly from the lower mantle.However, many unknowns like the physical properties (e.g., width, temperature, rheology) of the upwelling plume, interaction styles between the upwelling plume and the stagnant slabs in the MTZ (Hua et al., 2022), as well as its fundamental driving mechanism certainly requires further investigations.More accurate geophysical images constrained by multidisciplinary observations and integration with petrological, geochemical, and 3-D numerical modeling studies in the future are envisioned to provide new insights into these in-depth questions.

Conclusions
Based on multiple sources of earthquake data recorded across Southeast Asia, we constructed a high-resolution 3-D upper mantle Vs model in the SCS and surrounding regions via seismic surface wave tomography.The tomographic results reveal a columnar low-velocity zone beneath the central region of the SABP, which is indicative of focused mantle upwelling that may be responsible for the widespread and voluminous basaltic volcanism observed at the surface.Our seismic images offer valuable clues on the origin of the extensive post-spreading intraplate volcanism emplaced in the SCS and surroundings, which is one of the most outstanding problems of Southeast Asian geology.The tomographic illumination also lays the groundwork for developing future evolutionary models that could help to examine the relationship between deep slab subduction and mantle plume generation, as well as unravel other significant geodynamic processes occurring in this spatially and temporally complex region.

•
A high-resolution 3-D upper mantle S-wave velocity model surrounding the South China Sea is constructed • A continuous, low-velocity column is imaged beneath the central region of the Southeastern Asian Basalt Province • The post-spreading intraplate volcanism within the Southeastern Asian Basalt Province is likely induced by the focused mantle upwelling Supporting Information: Supporting Information may be found in the online version of this article.

Figure 1 .
Figure 1.(a) Tectonic map of the study region.The red patches represent the surface coverage of the Late Cenozoic basaltic volcanism in the Southeastern Asian Basalt Province (SABP) with their ages indicated accordingly (after Yan et al. (2018) and Gu et al. (2019)).The gray dashed circle approximately outlines the affecting domain of the SABP.The bottom-left inset shows our study region in a broader view with plate motion directions and subduction zones marked by the black arrows and purple saw-toothed lines, respectively.(b) Distribution of seismic stations and the interstation ray paths for successfully retrieving Rayleigh-wave phase velocity dispersion curves.The IRIS, CEA, and VN stations are denoted by different colors as explained in the bottom-left legend.Station ZHQ and PTNR plotted with inverted light blue triangles are used to demonstrate the dispersion picking procedure as shown in Figure S1 in Supporting Information S1.(c) Azimuthal distribution of the collected 3,172 teleseismic earthquakes (blue dots) with Ms > 5.5, focal depth <50 km, epicentral distance between 10° and 150°, and period ranging from November 2009 to December 2018.The red dot represents the earthquake that is used to demonstrate the dispersion picking procedure as shown in Figure S1 in Supporting Information S1.(d) Measured Rayleigh-wave phase velocity dispersion curves (black thin lines) for all station pairs as shown in b.The red thick line is the theoretical dispersion curve calculated from the PREM model(Dziewonski & Anderson, 1981).
As the depth increases, a key feature in the Vs model is a low Vs zone that appears beneath the central region of the SABP (Figures2b-2f), which is noticeable in the Vs model ofHuang and Xu (2011) at 70-150 km depth

Figure 2 .
Figure 2. Horizontal slices of the Vs anomaly (relative to the regional mean) plotted at various depths.The abbreviations in a: SETP = Southeastern margin of the Tibetan Plateau; TW = Taiwan region; SCS = South China Sea; CS = Celebes Sea; NB = Northern part of Borneo.The purple dashed circle that delineates the influence zone of the SABP is the same as shown in Figure 1a.The two black lines in f stand for the surface locations of the vertical profiles presented in Figure 3.The green stars in f mark the three selected grid knots for illustrating the 1-D Vs inversion scheme shown in Figure S6 in Supporting Information S1.

Figure 3 .
Figure3.W-E and S-N vertical cross sections of the Vs anomaly (relative to the regional mean) from the uppermost mantle down to 300 km depth.The surface locations of the two profiles (A-A' and B-B') are given in Figure2f.Please note that the vertical profiles have a V/H scaling ratio of ∼5, which means that the vertical dimension is exaggerated by a factor of ∼5.

Figure 4 .
Figure 4.A 3-D view of the imaged low-velocity shapes with an iso-surface of −5.0%Vs anomaly beneath the study region.The large-scale sub-vertical conduit located right below the central SABP is interpreted to be the major mantle upwelling trunk that is likely to feed the surface intraplate volcanism.The magenta patches that depict the surface coverage of the Late Cenozoic basaltic volcanism within the SABP and the green dashed circle that marks the surficial extent of the SABP are the same as shown in Figure1a.

from November 2009 to December 2018, 53 IRIS
(Incorporated Research Institution for Seismology) stations from November 2009 to December 2018, 3 VN (Vietnam Network) stations