Formation of mantle “lone plumes” in the global downwelling zone — Amultiscale modelling of subduction-controlled plume generation beneaththe South China Sea
Nan Zhanga,b,⁎, Zheng-Xiang Lia
a Earth Dynamics Research Group, ARC Centre of Excellence for Core to Crust Fluid Systems (CCFS), The Institute for Geoscience Research (TIGeR), Department of AppliedGeology, Curtin University, GPO Box U1987, Perth, WA 6845, Australiab The Key Laboratory of Orogenic Belts and Crustal Evolution, Institution of Earth & Space Sciences, Peking University, Beijing, China
A R T I C L E I N F O
Keywords:Mantle plumeSouth China SeaSubduction systemGlobal downwelling zoneAdaptive Mesh Refinement
A B S T R A C T
It has been established that almost all known mantle plumes since the Mesozoic formed above the two lowermantle large low shear velocity provinces (LLSVPs). The Hainan plume is one of the rare exceptions in thatinstead of rising above the LLSVPs, it is located within the broad global mantle downwelling zone, thereforeclassified as a “lone plume”. Here, we use the Hainan plume example to investigate the feasibility of such loneplumes being generated by subducting slabs in the mantle downwelling zone using 3D geodynamic modelling.Our geodynamic model has a high-resolution regional domain embedded in a relatively low resolution globaldomain, which is set up in an adaptive-mesh-refined, 3D mantle convection code ASPECT (Advanced Solver forProblems in Earth's ConvecTion). We use a recently published plate motion model to define the top mechanicalboundary condition. Our modelling results suggest that cold slabs under the present-day Eurasia, formed fromthe Mesozoic subduction and closure of the Tethys oceans, have prevented deep mantle hot materials frommoving to the South China Sea from regions north or west of the South China Sea. From the east side, theWestern Pacific subduction systems started to promote the formation of a lower-mantle thermal-chemical pile inthe vicinity of the future South China Sea region since 70 Ma ago. As the top of this lower-mantle thermal-chemical pile rises, it first moved to the west, and finally rested beneath the South China Sea. The presence of athermochemical layer (possible the D″ layer) in the model helps stabilizing the plume root.
Our modelling is the first implementation of multi-scale mesh in the regional model. It has been proved to bean effective way of modelling regional dynamics within a global plate motion and mantle dynamics background.
1. Introduction
It has been widely accepted that the vast majority of the Mesozoic-Cenozoic mantle plumes formed above the two large low shear velocityprovinces (LLSVPs) in the lower mantle (e.g., Burke and Torsvik, 2004;Burke et al., 2008). However, there are obvious exceptions. One pos-sible case is the Yellowstone plume which is at present outside thePacific LLSVP. Another is the Hainan plume in the South China Searegion that is located within the broad global mantle downwelling zone(Fig. 1). Li et al. (2014) therefore classified the Hainan plume as a “loneplume”.
The case for the Cenozoic Hainan plume being a lower mantle-ori-ginated plume has been established by both geological and seismictomographic investigations. The late Cenozoic Hainan flood basalts,
widely spreading at the Leizhou peninsula and Hainan Island north ofthe South China Sea (Flower et al., 1992; Wang and Li, 2009) (Fig. 2A),have been demonstrated to be a typical Large Igneous Provence (LIP) ofmantle plume origin (Flower et al., 1992; Zou and Fan, 2010; Wanget al., 2013). Wang et al. (2013) were able to further demonstrate thatthe flood basalts sourced from an ancient reservoir in the lower mantle.Geophysical investigations (e.g., Lebedev and Nolet, 2003; Montelliet al., 2004; Montelli et al., 2006; He et al., 2006; He and Wen, 2011;Xia et al., 2016) identified here a plume-like seismic structure ca.250 km in diameter, some pointing to a lower mantle deep root or nearthe core mantle boundary (CMB) (Fig. 1B–C).
Wang et al. (2013) speculated that this deep-rooted plume couldhave been generated by subducting slabs into the lower mantle in re-gions surrounding the South China Sea (Fig. 2B), a mechanism similar
https://doi.org/10.1016/j.tecto.2017.11.038Received 29 May 2017; Received in revised form 18 November 2017; Accepted 27 November 2017
⁎ Corresponding author at: Earth Dynamics Research Group, ARC Centre of Excellence for Core to Crust Fluid Systems (CCFS), The Institute for Geoscience Research (TIGeR),Department of Applied Geology, Curtin University, GPO Box U1987, Perth, WA 6845, Australia.
to what had been proposed for the formation of the LLSVPs (Zhonget al., 2007; C. Li et al., 2008; Z.X. Li et al., 2008; Li and Zhong, 2009).In this paper, we examine if such a mechanism is viable for the for-mation of lone plumes such as the Hainan plume using 3D geodynamicmodelling constrained by global plate motion and subduction history.We use known plate motion history and sea floor ages to constrain theinteraction between subducting slabs and lower mantle thermal in-stabilities in and around the South China Sea region, and investigate thefeasibility of Hainan plume being driven by subduction geometry.Especially, we aim to model the timing of the Hainan plume initiation,
as well as the particular subduction event(s) or geometry that promptedthe Hainan plume. We examine conditions (physical parameters) re-quired for producing such a plume.
Our model uses the plate motion model of Matthews et al. (2016)that is built into a 3D flow model. Another character of our geodynamicmodelling is that it has a high-resolution regional domain embedded ina relatively low-resolution global domain, which is set up in theadaptive-mesh-refined, 3D mantle convection code ASPECT1.3 (http://aspect.dealii.org/; Heister et al., 2017; Kronbichler et al., 2012).
0
0.5
1.0
1.5
2.0
-0.5
-1.0
-1.5
-2.0
30ºN
30ºS
60ºS
LIPS Reconstructed
dV (%)2800 km
s
40˚
20˚
0˚
100˚ 120˚ 100˚ 120˚
40˚
20˚
0˚
A)
C) 2800 kmB) 1000 km
1.5-1.5 0.3 0.7 1.1-0.7 -0.3 -0.1 0.1-1.1
dVs (%)
Hainan PlumeYellowstone
CA
Fig. 1. (A) Location of the Hainan plume over the globalslab downwelling zone, as well as the spatial relationshipbetween the majority of the post-300 Ma LIPS (re-constructed to the locations of their generations and thetwo lower mantle LLSVPs (after Burke et al., 2008). (B)–(C)Shear wave velocity anomalies beneath where the Hainanplume sits at depths 1000 km (B) and 2800 km (C) (afterMontelli et al., 2006). The red contour in (B) is the P-waveslow anomaly contour (−0.3%) from Xia et al. (2016). (Forinterpretation of the references to colour in this figure le-gend, the reader is referred to the web version of this ar-ticle.)
100˚
100˚
120˚
120˚
140˚
140˚
−20˚−20˚
0˚0˚
20˚20˚
40˚40˚
−9000 −6000 −3000 0 3000 6000 9000Elevation (m)
tsaEtsewhtuoS
Pacific Plate
Indo-Australian Plate
Eurasian Plate
PhilippineSea Plate
CarolinePlate
Sumatra
Java
hcne
rT
anai
raM
Sulawesi
Luz
on
Molucca Sea
B)A)
HainanIsland
South China Sea
Fig. 2. (A) Location of the South China Sea and the Hainan plume, with Cenozoic basalts attributed to the Hainan plume shown in red. (B) A simplified schematic model by Wang et al.(2013) featuring the formation of the Hainan plume by slab push. Map (A) is generated by GMT software (Wessel and Smith, 1998). The bathymetry comes from ETOPO1 at https://www.ngdc.noaa.gov/mgg/global/global.html. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
We first examine the geodynamic setting surrounding the SouthChina Sea region. Before the opening of the South China Sea from~30 Ma (Zahirovic et al., 2015; Cullen et al., 2010; Braitenberg et al.,
2006) (Fig. 3E–F), several subduction events occurred along Eurasia'ssoutheast margin. These subductions include the long-lived subductionof the Tethyan-Indian oceans from the south, and that of the Izanagi-Pacific oceanic plates from the east (Hutchison, 1975; Metcalfe, 1988).The Tethyan subductions occurred after Pangea formation, which may
A)150 oE
0 o
120 oE90 oE
30 oN
70 Ma 60 Ma
50 Ma 40 Ma
30 Ma 20 Ma
0 Ma10 Ma
150 oE120 oE90 oE
0 o
30 oN
150 oE120 oE90 oE 150 oE120 oE90 oE
150 oE120 oE90 oE 150 oE120 oE90 oE
0 o
30 oN
0 o
30 oN
0 o
30 oN
0 o
30 oN
0 o
30 oN
0 o
30 oN
60 MaB)
C) D)
E) F)
G) H)
Fig. 3. Cenozoic tectonic evolution of the South China Sea and adjacent regions. The plate motion model was from Matthews et al. (2016). Sea floor age data were from GPlatesSampleData/FeatureCollections/Isochrons/Muller_etal_AREPS_2016_Isochrons.gpmlz (Muller et al., 2016).
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
3
have influenced the break-up of east Pangea (Keppie, 2015). At ca.180 Ma, when Pangea started to break up, the Paleo-Tethys oceansubducted beneath Eurasia first at a high latitude, but its eastern seg-ment gradually retreated southwards with the closure of both the Paleo-and Meso-Tethys oceans by ca. 140 Ma (e.g., Metcalfe, 1988; Mulleret al., 2016). Since the opening of the South China Sea at the southernmargin of the Eurasia plate from ca. 30 Ma, the convergent boundaryalong the southern margin of the Eurasia plate has gradually migratedto the present-day Sumatra and Java (e.g., Braitenberg et al., 2006;Zahirovic et al., 2015; Clift and Lin, 2001) (Fig. 3H).
At the east side of Eurasia continent, the Izanagi oceanic platesubducted beneath Eurasia until close to 60 Ma (Fig. 3A–B). The Pacificplate grew northward, and its subduction eventually replaced the Iza-nagi subduction (Fig. 3C). As the Philippine Sea plate initiated at ca.50 Ma (Fig. 3C–E) (Hall et al., 1995), the Pacific subduction systemretreated to the Mariana trench. The Philippine, East Java, and WestSulawesi continental fragments formed the east boundaries of the SouthChina Sea. At 10 Ma, the Luzon subduction segment (Fig. 3G) initiatedalong the southeastern margin of the South China Sea, and its eastwardsubduction polarity flipped the Philippine subduction system(Hashimoto, 1983). For detailed plate motion history of the region, werefer to Matthews et al. (2016), Muller et al. (2016) and Domeier andTorsvik (2014).
Current understanding about the opening of the South China Seadoes not involve mantle plume as a major driver, nor is the plumehypothesis universally accepted. A recent seismic tomography work(Xia et al., 2016) found a mushroom-like, continuous low-velocityanomaly characterized by a columnar northward-plunging tail with adiameter of 200–300 km extending down to the lower mantle, althoughin that study the area under Hainan is shown as blank. That work wasbased on data from local seismic stations along the southeast coast ofChina, which do not provide a good coverage on regions south and westof the Hainan Island.
Meriaux et al. (2015) explored the interaction between the Hainanplume and the Manila subduction zone using analogue modelling. Theymade an analogy of a modelled SE-plunging anomaly to an observed SE-plunging Hainan plume conduit as imaged by seismic tomography (Leiet al., 2009), and speculated a southeasterly origin for the lower mantleroot of the Hainan plume.
3. Model set-up and methods
Our dynamic model of the Earth's mantle has a three dimensionalspherical geometry. Within the compressible mantle, the model appliesa chemical layer with different intrinsic density above the CMB. Ourmodel focuses on the South China Sea region between 95°–135°E and−10°–30°N. This region of interest is embedded in a global mantleconvection model of coarser mesh. Details about the mesh are given inSection 3.4.
3.1. Equations and algorithms
The dynamics model is built upon the open source finite elementcode ASPECT, which is characterized with an Adaptive-Mesh-Refinement (AMR) scheme. The mantle is treated as a compressibleStokes fluid. The conservation equations of mass, momentum, energy,as well as that of compositional fields, are given by the following:
∇∙ =ρ u( ) 0 (1)
− ∇∙⎡⎣
⎛⎝
− ∇∙ ⎞⎠
⎤⎦
+ ∇ =η ε u u I p ρ g2 ( ) 13
( )(2)
⎛⎝
∂∂
+ ⇀⋅∇ ⎞⎠
− ∇⋅ ∇ =
+ ⎛⎝
⇀ − ∇⋅⇀ ⎞⎠
⎛⎝
⇀ − ∇⋅⇀ ⎞⎠
+ ⇀⋅∇
ρC Tt
u T k T ρH
η ε u u I ε u u I
αT u p
2 ( ) 13
( ) : ( ) 13
( )
( )
p
(3)
∂∂
+ ∙∇ =Ct
u C 0 (4)
where = ∇ + ∇ε u u u( ) ( )T12 is the strain rate, u is velocity, p is
pressure, T is temperature, C is a compositional field that is used torepresent the proportion of the chemical layer material above the CMB(value ranges from 0 to 1, with 0 representing normal mantle material,and 1 the chemical layer material), ρ is the density that is thermal- andchemical-dependent, η is the viscosity, k is the heat conductivity, H isthe internal heating rate, α is the thermal expansivity, g is the gravityacceleration, and I is the identity matrix. The energy equation in-corporates internal heating, viscous heating and adiabatic heating (e.g.,Leng and Zhong, 2008). The chemical density ΔρC is proportional to theconcentration C as ΔρC = γC. We use the buoyancy ratio B to prescribethe chemical density, which is the density anomaly to the maximumthermal density anomaly in the system, defined as
=∆
∆B
ρρ α T
,C
0 0 (5)
where ΔρC, ρ0, and ΔT are the chemical density anomaly, density, andthe total temperature drop across the mantle, respectively. Quantitiessubscripted with “0” represent dimensional reference values. TheRayleigh number is defined as
=∆
Raρ α Td g
η κ,0 0 0
30
0 0 (6)
where d0, η0, and κ0 are the mantle thickness of the Earth, viscosity, andthermal diffusivity, respectively.
For the discretization, a second order element is used for tempera-ture and velocity, but first order elements for the pressure to satisfy LBB(Ladyzenskaja–Babuska–Brezzi) conditions (Brezzi and Fortin, 1991).The Stokes and energy equations are solved only once each time stepusing an IMPES (Implicit Pressure, Explicit Saturation) time steppingscheme. In the energy and chemical advection equations, the BackwardDifferentiation Formula scheme of order 2 (BDF) (Iserles, 1996) timestepping is used to replace the time derivative. To stabilize the BDFdiscretization, ASPECT uses an Entropy Viscosity method (Guermondet al., 2011). The time step is automatically adjusted to fit CFL (Cour-ant–Friedrichs–Lewy) conditions. Instead of a fully compressible for-mulation, the compressibility has been simplified as vertically com-pressible (i.e. the compressibility is treated as dependent on staticpressure only) to keep the Stokes equations symmetric. For moretechnical details regarding such compressible mantle convection codes,we refer to work by Heister et al. (2017), Zhang and O'Neill (2016), andKronbichler et al. (2012).
3.2. Progressive data assimilation, the plate motion model, the boundarycondition, and the initial conditions
We drive the mantle flows with the reconstructed plate motionmodel of Matthews et al. (2016), which was derived from Muller et al.(2016) and Domeier and Torsvik (2014). The plate motions are imposedas the surface kinematic boundary conditions (e.g., Zhang et al., 2010;McNamara and Zhong, 2005; Bower et al., 2015), extracted at 1 mil-lion year intervals (e.g., Hassan et al., 2015) using GPlates (Boydenet al., 2011). Our plate motion input has a 1 × 1 degree resolution inlongitude and latitude. The ASPECT mesh for convection is totallydifferent from the surface plate motion mesh. The ASPECT mesh formantle convection is a multi-scaled one, with a refined resolution for
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
4
our interested region (see Section 3.4). ASPECT interprets the1 × 1 degree plate motion mesh to the convection mesh linearly. Anexample is provided online in ASPECT benchmark documents (https://geodynamics.org/cig/software/aspect/aspect-manual.pdf).
We focused our modelling for the past 200 Ma. Initializing themodel from 200 Ma, ~150 Myrs ahead of Hainan plume generation, islong enough for modelling mantle structures (e.g., Bunge et al., 1998).Nonetheless, we also examined the effect of initial two-dome mantlestructures (i.e., African and Pacific LLSVPs), by pushing the startingtime back to 300 Ma.
In addition to the kinematic boundary conditions, we assimilatedthe time-dependent thermal structure of the lithosphere based on theage of the sea floor (Muller et al., 2016), to a depth of 65 km. FollowingFlament et al. (2014), we used a half-space cooling model to derive atemperature field of the oceanic lithosphere based on the global platereconstruction. In particular, the continental lithosphere is assigned to aProterozoic age (> 540 Ma). This continental age is a simplifiedaverage value (e.g., Artemieva, 2006; Flament et al., 2014), which al-lows us to limit the development of Rayleigh-Taylor instabilities. Be-cause the background resolution is ~80 or ~160 km, the top gridcannot fully resolve our assigned lithosphere thermal structure. Thecold structure in the top element is hence overestimated outside ourinterested region. A testing model with a much younger continental age(400 Ma) showed no significant difference to our modelling results (seecase PL9 in the Results section). For the implementation of such im-posed lithospheric thermal structures to ASPECT modelling, seeGassmoller et al. (2016) and ASPECT modelling benchmark (https://github.com/gassmoeller/aspect/blob/tristan_plume_model/source/initial_conditions/plume.cc). These approaches capture the essentials ofsubduction, which injects top cold thermal boundary layer (TBL) withrealistic thickness and mass flux, although we do not artificially insertthe weak zones between the overriding plate and the subducting plate.
3.3. Important parameters
We assume an a priori mantle adiabat augmented by TBLs. The topand the bottom TBLs have temperature drops of 1250 K and 1100 K,respectively (e.g., Steinberger and Calderwood, 2006). The initialadiabatic temperature profile has a potential temperature of 1300 °C(Fig. 4A and B). We use this a priori mantle temperature profile tocompute our reference thermal expansivity profile.
Our thermal expansivity profile is based on the analytical para-meterization of α(T, z), given in eq. (2) of Tosi et al. (2013). Basically,this parameterization gives a maximum expansivity of 2.8 × 10−5/Kfor the upper mantle, and 1.1 × 10−5/K for the CMB.
Our density profile and sound bulk modulus are derived using thePREM density model (Dziewonski and Anderson, 1981). The densitylinearly increases from 3.3 kg/m3 in the upper mantle to 5.3 kg/m3 on
the CMB, about 1.6 times larger.We use Arrhenius laws to describe the variation of viscosity with
temperature and depth in the upper and lower mantle, which takes thefollowing nondimensional form:
⎜ ⎟= ⎛⎝
− ⎞⎠
η T r η r ERT
ERT
( , ) ( ) expref
0(7)
where E is the effective activation energy, Tref is a reference tempera-ture and taken as 1573 K in this study, η0(r) is the reference viscosity atTref, and R is the gas constant. In general, thermally activated creeprheology may be stress and grain size dependent. Stress dependencedepends on deformation mechanism (dislocation versus diffusion creep)which is a function of grain size; however, grain size itself is determinedby stress (e.g., Behn et al., 2009). While of great interest in the study ofmantle dynamics, such a complex rheological description is beyond thescope of the present study which is restricted to the simple Arrheniustemperature-dependence for a range of η0 and E. The active activationenergy E is chosen to be ~160 kJ/mol, which induces viscosity varia-tions of ~103. Although the activation energy is smaller than that fromlaboratory studies for olivine (~330 kJ/mol), it is used here for nu-merical stability, and also to account for other weakening effects suchas brittle deformation and non-Newtonian rheology. This is going to befurther discussed in Section 5. The depth-dependent viscosity prefactorη0(r) is specified to give rise to a viscosity jump for the lower mantle bya factor of 30 or 100 (Hassan et al., 2015; Hager, 1984), and also topossible continuous variations in viscosity with depth (i.e., pressuredependence). The representative viscosity profiles are shown in Fig. 5.
An internal heat generation rate of 9 × 10−12 W/kg is applied toyield ∼60–65% internal heating rate (Zhong, 2006; Leng and Zhong,2008), which is similar to that used in earlier studies (e.g., Bunge, 2005;Zhang et al., 2010).
When we search for the plume centre within our interest regionfrom the CMB to the top TBL, we use an excess temperature of 300 K aswell as positive vertical motion (Zhong, 2006; Leng and Zhong, 2008)to identify the plume location. This excess temperature is the upperbound of petrological estimations, which range between 200 and 300 K(Schilling, 1991). We find that a constant excess temperature is ade-quate to quantify the Hainan plume, although Bunge (2005) showedthat plume excess temperature systematically decreases from the CMBtowards the top TBL.
3.4. Mesh for regional modelling embedded in a global model
Our interest region is within a 40° × 40° dimension, between95°–135°E and −10°–30°N. In order to cover our interest region with ahigh resolution mesh but built in the appropriate global mantle flowmodel, we embed our regional dynamic model in a low resolution
Rad
ius
(km
)
Temperature (oC)
6000
4000
5000
400 24000 200016001200800 2800 36003200 4000
(A)
Temperature (oC)
1000
(B)
3500
3000
2500
2000
1500
0
Fig. 4. The initial temperature profile (A) and in a cross-section (B) for all models. The view in (B) is from the African hemisphere.
global mantle shell (Fig. 6). ASPECT's AMR scheme enables users to puttwo spatial-scale meshes together to cover the entire Earth mantle.
ASPECT's AMR scheme applies a strict rule of 2-to-1 grid spacechange (i.e., octree) for each refinement operation (Fig. 6; Kronbichleret al., 2012). ASPECT is not only capable of updating the mesh withtemperature gradient based on the AMR scheme, but can also refine agiven interest region with the 2-to-1 grid space change. In our model-ling, we fix a high-resolution mesh covering our interest region with 2levels of mesh refinement compared to the global mesh (Fig. 6). For theglobal mantle shell, we let the initial mesh resolution to be 2 levelslower than that of our interest region, and temperature gradient de-pendent. We allow the global mesh to be refined to 1 level higher onlywhen needed, which is still 1 level lower than that of our interest region(Fig. 6), in order to save computational time. This refinement can beseen at the Tonga region, North America, and north Eurasia (Fig. 6A).We also apply a 1-level refined mesh for the 10-degree-wide transitionzone surrounding our interest region (Fig. 6). This highest refinementlevel for our interest region has 68 vertical elements. Outside the in-terest region, the vertical mesh is determined by the AMR scheme(Fig. 6B), and released to the coarser mesh following the 2-to-1 gridspace change. The resolutions for such a 3-level meshes are ~41 km,~82 km, and ~165 km, respectively, from the finest to the coarsestmeshes (Fig. 6A). The higher resolution grid resolves the boundary ofchemical material better (Fig. 6B). More details regarding mesh plottingin ParaView (Ayachit, 2015) are discussed in the Supplementary ma-terial.
To check the influences of mesh refinement on plume generations,we run a test model without any refinement in our interest region.Because the plume location appears to be largely determined by theposition of chemical piles, we compared the positions of chemical piles
between cases with and without the mesh refinement. The comparisonresults are reported in case PL8.
4. Results
In this section, we first illustrate how the Hainan plume grows at theCMB and ascends to the lithosphere in our reference case PL1. We willexamine the roles of subducting slabs in the generation of the Hainanplume, and the effects of other parameters including chemical piles,convective vigor, initial plate motion, and model mesh.
4.1. Formation of the Hainan plume from case PL1
Reference case PL1 starts with an initial 250 km-thick chemicallayer above the CMB with a buoyancy number B of 0.55, and a platemotion model starting from 200 Ma till the present day. The averageviscosities used for the upper mantle and lower mantle are~1 × 1019 Pa s and ~4 × 1020 Pa s (Fig. 4B), respectively. Otherparameters are given in Table 1, including a Ra value of 1.3 × 108.Internal heat generation rate is 9 × 10−12 W/kg, yielding a 61% in-ternal heating ratio, implying a significant fraction of heating from thecore (e.g., Leng and Zhong, 2008).
We first describe the general evolution of global mantle structuresaccording to this model. This case reproduces two large chemicalanomalies (Figs. 7A and 8D) similar to those of previous studies (e.g.,Zhang et al., 2010; Bower et al., 2013; Bull et al., 2014; Hassan et al.,2015), which is in agreement with the global tomography models to thefirst order (e.g., Ritsema et al., 2011; Auer et al., 2014; Fig. 7A and 8D).The eastern, northern, and western edges of the modelled Pacific che-mical pile are more discontinuous compared to the tomography models,and exhibit discrete blobs. The modelled chemical piles with high to-pography (Fig. 8D) coincide with the locations where the Hawaiianplume and the Ontong Java plateau are situated (Condie, 2001). In theAfrican hemisphere, our model shows the long east-west-oriented
6000
4000
5000
Viscosity (Pa.s)1018 102210201019 1021
Radius (km
)
Case PL1
Case PL7
6000
4000
5000
Rad
ius
(km
)
Fig. 5. Initial viscosity profiles for cases PL1–6 (black) and PL7 (red). (For interpretationof the references to colour in this figure legend, the reader is referred to the web versionof this article.)
W
E
B)A)
W
E
Fig. 6. The multi-scale mesh used for our modelling. (A) is a view from the top. The mesh as well as the composition at 70 Ma along a W-E transect in (A) are shown in (B).
Table 1Model parameters and material properties.
Parameters Symbols Value
Earth's radius R 6370 kmCore radius Rc 3470 kmMantle density ρ0 3300 kg/m3
eastern arm of the African chemical pile in the southern Indian Ocean,thickening to the west, and a more discontinuous northern armstretching towards Iceland (Fig. 7A). This pattern is consistent with thetomography models. The ‘Perm Anomaly’ (Lekic et al., 2012) beneathpresent-day western Siberia as shown in tomography models is alsopresent in our model (Fig. 7A). This comparison suggests that our modelcaptures the basics of lower mantle dynamics, although modelledchemical pile edges do not match exactly the higher-order geometricfeatures found in mantle tomography models.
Subduction-induced large-scale mantle flow and incipient plumesdevelop within the first 50 Myrs of model time. Our modelled mantlestructures show many small-scale thermal and chemical anomaliesabove the CMB, such as those in the western Pacific (Fig. 8D), thesouthern India, and the central Atlantic oceans. Our model shows moresmall-scale structures than previous studies (e.g., Hassan et al., 2015)largely because of our higher Rayleigh number. The slab and plumestructures outside our interested region (Fig. 8) are less resolved due tothe lower resolution of ~165 km for the background mantle. The slabsinside our interested region is clearly better resolved than those outside(Fig. 8A and B). The fine details of plumes outside our interest region,especially the plume heads in the upper mantle, are not well resolved(Fig. 8). Beside the resolution issue, the lacking of well-developedplume head in the upper mantle is probably also due to the strongviscous heating caused by the imposed surface plate motion.
We now focus on the South China Sea region where the Hainanplume grows. The initially homogeneous 250 km-thick chemical pile isfirst swept south- and east-wards by the Tethys and Izanagi subduc-tions. As the Eurasia continent moved south, the Tethys subductionsystem retreated correspondingly. During the first 100 Ma, south-eastward downwellings beneath the South China Sea region quicklypushed the chemical material southward by more than ~25°. Theclosure of the Paleo- and Meso-Tethys oceans is replaced by the openingof the Indian Ocean at 140 Ma. After this time period, the Izanagisubduction pushes the chemical pile more westwards, resulting in thetransition from a phase of rapid southward motion to a slower
westward motion. The local mantle structures do not change much(Fig. 8A) until the Izanagi plate has been subducted (Fig. 8B). ThePacific plate started to subduct at the Mariana trench at ca. 70 Ma,keeping a distance from the Eurasia by the Philippine plate (Supple-mentary movie). The retreat of the subducting Pacific plate beneath theMariana trench splits a small part of the hot chemical pile towardsbeneath the Philippine plate and the South China Sea (Figs. 8B, C, and9A; Supplementary movie), thus promoting a weak Hainan plume fromthat time. Between about 30 Ma and 16 Ma, the plume graduallyreaches the bottom of the lithosphere, and its root simultaneously mi-grates ~13° to the northwest to beneath the Palawan Island (Figs. 8D,9B, C, E, 11) with a temperature of ~2920 °C (excess temperature480 K) at 100 km above the bottom TBL (Fig. 9C). As the plume headascends towards the northwest (Figs. 8D and 9E) to the bottom of thelithosphere, the plume head approaches the coastal zone of South China(Fig. 9D). The modelled present-day chemical heterogeneity under theSouth China Sea is relatively weak and flat in comparison to that of thePacific region (Figs. 8D and 9B).
In this model, the lateral advection of plume source from the se-parated chemical ridge controls plume tilt in the deep lower mantle(Fig. 9E). On the other hand, surface plate motion determines the tilt ofthe plume conduit in the upper mantle (Fig. 9E). Overall, the modelledplume conduit tilts ~11° northwestward from the root to the litho-spheric bottom in the present day (Fig. 9E), which is consistent with theseismic tomographic observations (e.g., He and Wen, 2011), especiallyin the upper mantle (e.g., Montelli et al., 2006; Lebedev and Nolet,2003).
4.2. Influences of different parameters on the generation of the Hainanplume
We conduct eight additional cases PL2–PL9 (Table 2) in order toexamine the influence to modelling results by other parameters. Theseinclude the presence of a chemical layer above the CMB, the thicknessof the chemical layer, the buoyancy ratio of the chemical layer, the
−60˚−60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚60˚
A)
B)
−60˚−60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚60˚
−60˚−60˚
−30˚ −30˚
0˚
30˚ 30˚
60˚60˚
−60˚−60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚60˚
C)
D)
0.00 0.04 0.06 0.08
Composition
0 1000 1500 2000 2500 3000 3500
Temperature (oC)
0.60.40.20.1
Fig. 7. (A) Comparison of present-day chemical structures 100 km above CMB between that of tomographically “observed” (S40RTS, shown in yellow outline, plotted with 1% anomaly)and our standard case PL1 (shown in purple shade). (B) Comparison of modelled present-day chemical structures 100 km above CMB between that of case PL1 (purple outline) and PL3(blue outline), both representing compositional 0.045 variance contour. (C) The present-day chemical structure at 100 km above the CMB for PL4 and (D) the thermal structure at 300 kmabove CMB for PL2. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
7
starting age of the plate motion model, the convective vigor, the utili-sation of the multiscale mesh, and the age of the continental litho-sphere. When we determine the consistence between the model resultsand the observations, we check if the global degree-2 structure is wellreproduced to the first order, and then how close to the South China Seathe plume root location is.
Case PL2 differs from case PL1 in having the chemical layer abovethe CMB removed (i.e., a purely thermal convection case). With theremoval of the chemical layer, mantle convection in case PL2 is morevigorous (Fig. 7D) than in case PL1, and the internal heating ratio is53%. A high temperature anomaly is identified with a root southeast ofthe Caroline plate (Figs. 2A and 7D) and a quasi- plume head to thesoutheast of the Philippine islands (Fig. 11). The conduit above the660 km phase boundary is not clearly visible, although identifiable inthe depth range of 200–400 km. This plume was initiated at ~50 Maunder the Pacific plate, and it tilts gradually towards the northwest dueto the fast moving Philippine Sea plate. The plume head disconnectedfrom the root at ~30 Ma. However, such a plume structure is not re-flected in any seismic study (e.g., Lebedev and Nolet, 2003; C. Li et al.,2008; Z.X. Li et al., 2008; Xia et al., 2016). This case demonstrates that,under the latest plate reconstruction, a chemical layer on the CMB helpsthe formation of a plume root closer to the South China Sea. Thenorthwest motion of the Philippine Sea plate drives a strong mantlewind, delivering the temperature anomaly in the upper mantle from thePacific side to areas closer to the South China Sea.
Case PL3 differs from case PL1 in having a reduced initial thicknessof the chemical layer to 200 km. We conduct this case because previousglobal models (Hassan et al., 2015; Burke et al., 2008) argued that thetotal volume of LLSVPs only allows for a global chemical layer of200 km. With such a reduced initial thickness for the chemical layer,the resultant present-day Pacific chemical pile only shows a finger-shaped blob beneath the Molucca Sea, south of Philippine, but notunder the South China Sea (Figs. 7B, 10A, and 11). This result is similarto the distribution of chemical pile in Fig. 8A of Hassan et al. (2015),but our Pacific chemical pile does not expand to under western Aus-tralia as in Hassan et al. (2015).
Case PL4 features an increased buoyancy ratio of 0.6. It is wellknown that a larger buoyancy ratio leads to less deformation to thechemical layer (e.g., Davaille, 1999; Tackley, 1998; Kellogg et al., 1999;Zhang et al., 2010). The resultant present-day chemical piles are notfully separated into two LLSVP domes at CMB (Fig. 7C). Instead, themodelled African and Pacific piles connect under the India Ocean andthe South America, although the Pacific pile reaches to beneath theSouth China Sea. A similar chemical structure was achieved in a pre-vious study (Zhang et al., 2010). We consider the outcome of this caseinconsistent with the global seismic tomographic results.
Cases PL5 and PL6 examine the influence of plate motion history.Case PL5 applies plate motion from 150 Ma, 50 Myrs less than that forcase PL1. The plume generation does not show much difference fromthat of case PL1. In contrary, case PL6 applies plate motion from
B) 50Ma
0.1
Composition
1.00.5 0.750.25
∆T(oC)
5500.0-400 300-200
C) 30Ma
A) 70Ma
D) 0Ma
Fig. 8. 3D Hainan-hemispherical views of thermochemicalstructures for case PL1 at (A) 70 Ma, (B) 50 Ma, (C) 30 Ma,and (D) the present day. The thermal structures are plottedas iso-volumes (ParaView) of residual temperature withcontour levels of −100 °C (blue) and 350 °C (red), and thechemical structures are plotted as iso-volumes of compo-sitional field (purple). The core mantle boundary is plottedas a dark red spherical surface. The present-day coastlinesare plotted for reference. (For interpretation of the refer-ences to colour in this figure legend, the reader is referredto the web version of this article.)
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
8
300 Ma, 100 Myrs earlier than that for case PL1. This case leads to theformation of a two-dome (LLSVP) initial condition (e.g., Zhang et al.,2010; Bower et al., 2013) prior to the initiation of the Hainan plume.The two-dome mantle structure is formed after ~110 Myrs of modelcalculation at ca. 190 Ma. Case PL6 leads to a slightly different present-day chemical structure. The modelled Pacific chemical pile has a smallportion split and pushed to beneath the southern tip of the PhilippineSea plate (Figs. 10B and 11), not westwards enough for the plume headto reach the South China Sea. The modelled Pacific chemical pile in thiscase differs from that in case PL1 in that it is more easterly positioned at~40 Ma, possibly caused by longer duration for the Izanagi subduction.Such a more easterly position for the Pacific chemical pile reduces the
volume of chemical pile that is pushed to the South China Sea.In Cases PL7, the Rayleigh number is reduced by a factor of 10 to
1.3 × 107, otherwise this case is identical to case PL1 (Fig. 5). We re-duce Rayleigh number by increasing mantle viscosity uniformly. Basi-cally, the viscosities of the upper and lower mantle are increased to1 × 1020 and 4 × 1021 Pa·s, respectively. A more viscous mantle meansless vigorous convection, which leads to a thicker cold thermalboundary layer on top. When the subducted thicker cold slab reachesthe chemical layer, it reduces the number of small structures generated.Hence, the resultant mantle structures show fewer small piles. This casedoes not produce any chemical root and small thermal plume under theSouth China Sea, or nearby areas. It neither produces a clear Perm
80˚ 90˚ 100˚ 110˚ 120˚ 130˚ 140˚ 150˚ 160˚
−20˚ −20˚
−10˚ −10˚
0˚ 0˚
10˚ 10˚
20˚ 20˚
30˚ 30˚
40˚ 40˚
80˚ 90˚ 100˚ 110˚ 120˚ 130˚ 140˚ 150˚ 160˚
Eurasia
Sunda Shelf
South China
Sea
Indian Ocean
Philippine
Plate
Indochina
80˚
80˚
90˚
90˚
100˚
100˚
110˚
110˚
120˚
120˚
130˚
130˚
140˚
140˚
150˚
150˚
160˚
160˚
−20˚ −20˚
−10˚ −10˚
0˚ 0˚
10˚ 10˚
20˚ 20˚
30˚ 30˚
40˚ 40˚
0.05 0.08 0.1 0.2 0.3Composition
80˚
80˚
90˚
90˚
100˚
100˚
110˚
110˚
120˚
120˚
130˚
130˚
140˚
140˚
150˚
150˚
160˚
160˚
−20˚ −20˚
−10˚ −10˚
0˚ 0˚
10˚ 10˚
20˚ 20˚
30˚ 30˚
40˚ 40˚
80˚
80˚
90˚
90˚
100˚
100˚
110˚
110˚
120˚
120˚
130˚
130˚
140˚
140˚
150˚
150˚
160˚
160˚
−20˚ −20˚
−10˚ −10˚
0˚ 0˚
10˚ 10˚
20˚ 20˚
30˚ 30˚
40˚ 40˚
A
A’
A A’10o20o
(C) (D)
(E)
Eurasian
Plate
Pacific
Plate
Indian Ocean
Plate
Australian
Plate
Philippine
Sea Plate
Caroline
Plate
Eurasian
Plate
Indian Ocean
Plate
Australian
Plate
Philippine
Sea Plate
Caroline
Plate
Eurasian
PlatePacific
Plate
Indian Ocean
Plate
Australian
Plate
Philippine
Sea Plate
Caroline
Plate
Pacific
Plate
(A) (B)aM0aM03
Pacific
Plate
01000
1500
2000
2500
3000
3500
Temperature (oC)
0.03 0.04 0.06
Fig. 9. Modelled results from our standard case PL1: chemical field 100 km above the CMB at 30 Ma (A) and at the present-day (B) from our standard case PL1; temperatures at 100 kmabove CMB (C), 200 km depth (D), and under the South China Sea region along the AA′ transect (E; position of the transact is shown in D). The dashed-line box delineates our interestregion.
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
9
Anomaly.Case PL8 is a case without mesh refinement under the South China
Sea, so to test the effects of two-level mesh refinement on the globalmantle structures, especially the chemical piles. Otherwise the case isthe same as case PL1. The resultant chemical piles in this case are al-most identical to these in case PL1. The contours of CMB chemicalmaterial from cases PL8 and PL1 have a 95% overlap to each other(Fig. 12). As expected, the main difference occurs at the local SouthChina Sea region. Case PL8 still exhibits a small piece of chemicalheterogeneity beneath the Philippine islands, the east boundary of theSouth China Sea, but all the chemical and thermal anomalies are situate~4° east of the South China Sea (Fig. 12). This case indicates that themesh refinement in the interest region helps resolving the fine struc-tures more accurately.
Case PL9 is a simple test on the effect of assigned continental age(now 400 Ma) on the mantle structures under the South China Sea(Table 2). The calculated results do not show any difference from casePL1.
5. Discussions and conclusions
Our geodynamic modelling utilises the latest plate motion model ofMatthews et al. (2016) to constrain the mantle evolution under theSouth China Sea. This plate reconstruction has the best resolution at theSoutheast Asia region compared to previous reconstructions (e.g.,Lithgow-Bertelloni and Richards, 1998; Scotese, 2001). Though, alter-native reconstruction (e.g., Domeier and Torsvik, 2014) for the time
period of Mesozoic-Cenozoic, particularly concerning our interestedregion, is similar to Matthews et al. (2016). Hence, the resultant mantlestructures at Hainan Island region should show no significant differencebetween reconstruction models. We are able to produce the observedAfrican and Pacific thermochemical anomalies, and the modelled globalmantle structures are similar to those in previous studies (Zhang et al.,2010; Bower et al., 2013; Hassan et al., 2015) to the first order. In theinterested region surrounding the South China Sea, our model alsoproduced a concentration of chemical materials beneath the OntongJava plateau (Figs. 7A and 8D), as in Hassan et al. (2015). In our caseswith a chemically anomalous lower mantle, the chemical layer thickensin response to large subduction-induced flows and gets swept into anetwork of circular flat domes and sharp ridges. Although our modelscannot resolve the fine structures of background plumes due to the lowresolution outside the interest region, we still can observe “plumes”, ortemperature anomalies, anchored to the peaks of domes and ridges inthe deep mantle (Davaille et al., 2005; Tan et al., 2011). The temporalevolution of the chemical structures is guided by slab push forces onone hand, and the structures are gradually eroded by entrainment onthe other. Our modelled Hainan plume conduit shows necking in theupper mantle and spanning under the lithosphere (e.g., Fig. 9E), whichis generally consistent with the behaviour of modelled plumes in pre-vious studies (e.g., Hassan et al., 2015).
The plume root and conduit under the South China Sea have beenshaped by the subduction of the Tethyan-Indian oceanic slabs in thewest and north, and the subduction of the Izanagi-Pacific slabs from theeast. The eastward retreat of the subduction system from the Izanagi to
Table 2Model parameters and key outputs.
Case Chemical layer thickness (km) Rayleigh number Buoyancy ratio Plate motion history (Ma) Continent age (Ma) Plume diameter (km)a
a The plume diameter is measured at 800 km depth.b This plume conduit at 800 km depth crosses the transition resolution.c This plume is within a large LLSVP.
0.1
Composition
1.00.5 0.750.25
B) PL6A) PL3
∆T(oC)
5500.0-400 300-200
Fig. 10. 3D Hainan-hemispherical views of present-daythermochemical structures for cases PL3 (A) and PL6 (B).The thermal and chemical structures are plotted the sameway as in Fig. 7.
Fig. 11. Model plume conduits from cases PL1, PL2, PL3,and PL6 at present day.
−60˚−60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚60˚
A)
−60˚−60˚
−30˚ −30˚
0˚ 0˚
30˚ 30˚
60˚60˚
0.00 0.04 0.06 0.08
Composition
B)
0.60.40.20.1
Fig. 12. A comparison of calculated chemical piles betweencases without (PL8; A) and with the mesh refinement (PL1; B)in our interest region. The chemical fields are taken at depth100 km above the CMB.
N. Zhang, Z.-X. Li Tectonophysics 723 (2018) 1–13
11
the Pacific plates (Fig. 3A–C) plays a key role in driving a small che-mical pile to underneath the South China Sea (Figs. 8C and 9A). Thenorthwest motion of the Philippine plate also helps the plume head tomove to the South China Sea region. The modelling results show si-milarities to the mechanism adapted by Wang et al. (2013) for thegeneration of the Hainan lone plume (Li et al., 2014). However, ourmodelled plume shows that chemical heterogeneity under the SouthChina Sea is a feature separated out of the proto-Pacific chemical pileduring the Pacific subduction retreat, whereas Wang et al. (2013)suggested that a stagnant chemical heterogeneity produced the Hainanplume.
Our model is characterized by a high-resolution mesh for the SouthChina Sea region embedded in a low-resolution global mantle mesh,which allows us to resolve the local fine structures yet still capture theglobal evolution. This multi-scale modelling approach for a regionalmodel was proposed by Tan et al. (2006) who used two solvers forgrids, and interpolated the fields between the two grids. ASPECT pro-vides a direct multi-scale solver for mantle convection modelling. Itapplies a strict rule of 2-to-1 grid space change. Our modelling is thefirst to implement such an approach to a real geological modelling.
Our Hainan plume modelling favours a thicker chemical layer and ahigher Rayleigh number. When the thickness of the chemical layer isreduced to 200 km (case PL3), the chemical root retreats southeastwardto beneath the Molucca Sea, instead of a plume root exactly beneath theSouth China Sea. Case PL3 uses the same thickness of chemical layer asHassan et al. (2015), hence produces a global structures similar to theirs(Hassan et al., 2015). Although a thickness of 200 km is comparable tothe total volume estimation of the two LLSVPs (Burke et al., 2008;Hassan et al., 2015), such estimations are by no mean accurate due tothe resolution limit of seismic tomography (Wang and Wen, 2004), ourpoor understanding of both the composition and boundary definition ofLLSVPs, and the neglect of LLSVP-like dense chemical materials in re-gions outside the two LLSVPs in such estimations. Our model has thelargest Rayleigh number compared to previous studies (Bower et al.,2013; Hassan et al., 2015) (CitcomS has a factor of 11 scaled with theregular Rayleigh number). When the Rayleigh number is reduced byone order of magnitude in case PL6, the resulting thicker boundarylayer leads to a sharper pile boundary, and depressed small-scalethermal- and chemical-heterogeneities, especially under the SouthChina Sea area.
Among the three parameters we discussed above (i.e., plate mo-tions, chemical layer thickness, and Ra number), the plate motion,especially the Izanagi plate replacement by the Pacific subduction, isthe key factor for driving the Hainan plume production. The influenceof this plate motion change is manifested in most of our cases (e.g., LP1,2, 4, 5, and 6). For example, this plate motion change pushes thechemical root towards beneath the South China Sea in PL1, but beneaththe Molucca Sea, the Philippine Sea plate, and the Caroline plate in ourcases PL2, 4, 5, and 6 (Fig. 11). These root locations are all close to theSouth China Sea, showing a systematic trend (Fig. 11). Our modelsprefer a large Ra to produce the Hainan plume within the downwellingzone because only thin subducting slabs (from a thin top thermalboundary) allow the chemical root of plume to survive between them.Once the Ra is reduced, the plume-scale thermochemical anomaly isdestroyed by the thick slabs (e.g., case PL7). Higher activation, hencemore thin slabs, may also help the survival of the chemical root for theHainan plume. However, we were unable to increase the activationenergy further because this would lead to extremely slow computationusing the current ASPECT code. A testing case similar to case PL3, butwith temperature-dependent viscosity variation to slightly higher than104, has a convergence problem. We noted that Heister et al. (2017)recently updated the Stokes solver for the compressible mantle flow,which should improve the computational capability of ASPECT forlarge viscosity variation in future. The chemical layer thickness doesnot seem to be critical for maintaining a plume in the downwellingzone, though the layer volume affects the specific plume location. Cases
PL2 and PL3 both show plume structures close to the South China Seaalthough their locations are more to the southeast.
Supplementary data to this article can be found online at https://doi.org/10.1016/j.tecto.2017.11.038.
Acknowledgements
This work was supported by Australian Research Council AustralianLaureate Fellowship grant (FL150100133) and Curtin University re-search fellowship support through the Australia-China Joint ResearchCentre for Tectonics and Earth Resources (ACTER) to ZXL, and byChina's Thousand Talents Plan (2015) and NSFC41674098 to NZ. Wethank careful reviews and comments by Ramon Carbonell, CedricThieulot, and an anonymous reviewer. Computational Infrastructure forGeodynamics is thanked for distributing the software ASPECT that isused in this study. We thank Juliane Dannberg, Rene Gassmoeller, andSiqi Zhang for ASPECT technical helps. Yebo Liu is thanked for helpingwith data formatting. GPlates software created by the EarthByte groupwas used. Computational work was supported by resources provided bythe High-performance Computing Platform of Peking University and bythe Pawsey Supercomputing Centre with funding from the AustralianGovernment and the Government of Western Australia. This is CCFScontribution 1033, and a contribution to IGCP 648. The public datarepository of ASPECT can be accessed at https://github.com/geodynamics/aspect/.
References
Artemieva, I.M., 2006. Global 1 × 1 thermal model TC1 for the continental lithosphere:implications for lithosphere secular evolution. Tectonophysics 416 (1), 245–277.http://dx.doi.org/10.1016/j.tecto.2005.11.022.
Auer, L., Boschi, L., Becker, T., Nissen-Meyer, T., Giardini, 2014. Savani: a variable re-solution whole-mantle model of anisotropic shear velocity variations based on mul-tiple data sets. J. Geophys. Res. 119, 3006–3034.
Ayachit, Utkarsh, 2015. The ParaView Guide: A Parallel Visualization Application.Kitware (ISBN 978-1930934306).
Behn, M.D., Hirth, G., Elsenbeck II, J.R., 2009. Implications of grain size evolution on theseismic structure of the oceanic upper mantle. Earth Planet. Sci. Lett. 282, 178–189.
Bower, D., Gurnis, M., Seton, M., 2013. Lower mantle structure from paleogeographicallyconstrained dynamic Earth models. Geochem. Geophys. Geosyst. http://dx.doi.org/10.1029/2012GC004267.
Bower, D., Gurnis, M., Flament, N., 2015. Assimilating lithosphere and slab history in 4-DEarth models. Phys. Earth Planet. Inter. 238, 8–22.
Boyden, James A., Müller, R. Dietmar, Gurnis, Michael, Torsvik, Trond H., Clark, JamesA., Turner, Mark, Ivey-Law, Hamish, Watson, Robin J., Cannon, John S., 2011. Next-generation plate-tectonic reconstructions using GPlates. In: Geoinformatics:Cyberinfrastructure for the Solid Earth Sciences. Cambridge University Press,Cambridge, pp. 95–113 (ISBN 9780521897150).
Braitenberg, C., Wienecke, S., Wang, Y., 2006. Basement structures from satellite derivedgravity field: South China Sea ridge. J. Geophys. Res. 111, B05407.
Brezzi, F., Fortin, M., 1991. Mixed and Hybrid Finite Element Methods. Springer-Verlag,New York, NY.
Bull, A., Domeiera, M., Torsvik, T.H., 2014. The effect of plate motion history on thelongevity of deep mantle heterogeneities. Earth Planet. Sci. Lett. 401, 172–182.
Bunge, H.-P., 2005. Low plume excess temperature and high core heat flux inferred fromnon-adiabatic geotherms in internally heated mantle circulation models. Phys. EarthPlanet. Inter. 153 (1), 3–10.
Bunge, H.-P., Richards, M.A., Lithgow-Bertelloni, C., Baumgardner, J.R., Grand, S.,Romanowicz, B., 1998. Time scales and heterogeneous structure in geodynamic Earthmodels. Science 280, 91–95. http://dx.doi.org/10.1126/science.280.5360.91.
Burke, K., Torsvik, T.H., 2004. Derivation of large igneous provinces of the past200 million years from long-term heterogeneities in the deep mantle. Earth Planet.Sci. Lett. 227, 531–538.
Burke, K., Steinberger, B., Torsvik, T.H., Smethurst, M.A., 2008. Plume generation zonesat the margins of large low shear velocity provinces on the core-mantle boundary.Earth Planet. Sci. Lett. 265, 49–60. http://dx.doi.org/10.1016/j.epsl.2007.09.042.
Clift, P.D., Lin, J., 2001. Preferential mantle lithospheric extension under the South Chinamargin. Mar. Pet. Geol. 18, 929–945. http://dx.doi.org/10.1016/S0264-8172(01)00037-X.
Condie, K.C., 2001. Mantle Plumes and Their Record in Earth History. CambridgeUniversity Press, pp. 19 (ISBN 0-521-01472-7).
Cullen, Andrew, Reemst, Paul, Henstra, Gijs, Gozzard, Simon, 2010. Rifting of the southChina Sea: new perspectives. Pet. Geosci. 16, 273–382.
Davaille, A., 1999. Simultaneous generation of hotspots and superswells by convection ina heterogeneous planetary mantle. Nature 402, 756–760. http://dx.doi.org/10.1038/45461.
Davaille, A., Stutzmann, E., Silveira, G., Besse, J., Courtillot, V., 2005. Convective
patterns under the Indo-Atlantic “box”. Earth Planet. Sci. Lett. 239 (3), 232–252.Domeier, M., Torsvik, T.H., 2014. Plate tectonics in the late Paleozoic. Geosci. Front. 5,
303–350.Dziewonski, Adam M., Anderson, Don L., 1981. Preliminary reference Earth model. Phys.
Heine, Dietmar Müller, R., 2014. Topographic asymmetry of the South Atlantic fromglobal models of mantle flow and lithospheric stretching. Earth Planet. Sci. Lett. 387,107–119.
Flower, M.F.J., Zhang, M., Chen, C.Y., Tu, K., Xie, G., 1992. Magmatism in the SouthChina Basin: 2. Post-spreading quaternary basalts from Hainan Island, South China.Chem. Geol. 97, 65–87.
Gassmoller, R., Juliane, Dannberg, Bredow, E., Steinberger, Bernhard, Torsvik, Trond H.,2016. Major influence of plume-ridge interaction, lithosphere thickness variations,and global mantle flow on hotspot volcanism-the example of Tristan. Geochem.Geophys. Geosyst. 17, 1454–1479.
Guermond, J., Pasquetti, R., Popov, B., 2011. Entropy viscosity method for nonlinearconservation laws. J. Comput. Phys. 230, 4248–4267.
Hager, B., 1984. Subducted slabs and the geoid: constraints on mantle rheology and flow.J. Geophys. Res. 89, 6003–6015.
Hall, R., Ali, J.R., Anderson, C.D., Baker, S.J., 1995. Origin and motion history of thePhilippine Sea Plate. Tectonophysics 251 (1–4), 229–250.
Hashimoto, M. (Ed.), 1983. Accretion Tectonics in the Circum-Pacific Regions, (ISBN 90-277-1561-0).
Hassan, R., Flament, N., Gurnis, M., Bower, D., Muller, D., 2015. Provenance of plumes inglobal convection models. Geochem. Geophys. Geosyst. 16, 1465–1489. http://dx.doi.org/10.1002/2015GC005751.
He, Y., Wen, L., 2011. Seismic velocity structures and detailed features of the D″ dis-continuity near the core-mantle boundary beneath eastern Eurasia. Phys. EarthPlanet. Inter. 189, 176–184.
He, Y., Wen, Lianxing, Zheng, Tianyu, 2006. Geographic boundary and shear wave ve-locity structure of the “Pacific anomaly” near the core–mantle boundary beneathwestern Pacific. Earth Planet. Sci. Lett. 244, 302–314.
Heister, T., Dannberg, J., Gassmoller, R., Bangerth, W., 2017. High accuracy mantleconvection simulation through modern numerical methods – II: realistic models andproblems. Geophys. J. Int. 210, 833–851.
Iserles, A., 1996. A First Course in the Numerical Analysis of Differential Equations.Cambridge University Press.
Kellogg, L.H., Hager, B.H., van der Hilst, R., 1999. Compositional stratification in thedeep mantle. Science 283, 1881–1884. http://dx.doi.org/10.1126/science.283.5409.1881.
Keppie, D.F., 2015. How the closure of paleo-Tethys and Tethys oceans controlled theearly breakup of Pangaea. Geology 43, 335–338.
Kronbichler, M., Heister, T., Bangerth, W., 2012. High accuracy mantle convection si-mulation through modern numerical methods. Geophys. J. Int. 191, 12–29.
Lebedev, S., Nolet, G., 2003. Upper mantle beneath Southeast Asia from S velocity to-mography. J. Geophys. Res. 108, 2048–2061.
Lei, J., Zhao, D., Steinberger, B., Wu, B., Shen, F., Li, Z., 2009. New seismic constraints onthe shallow structure of the Hainan plume. Phys. Earth Planet. Inter. 173 (1–2),33–50.
Lekic, V., Cottaar, S., Dziewonski, A., Romanowicz, B., 2012. Cluster analysis of globallower mantle tomography: a new class of structure and implications for chemicalheterogeneity. Earth Planet. Sci. Lett. 357/358, 68–77.
Leng, W., Zhong, S., 2008. Viscous heating, adiabatic heating and energetic consistency incompressible mantle convection. Geophys. J. Int. 173, 693–702.
Li, C., van der Hilst, R.D., Engdahl, E.R., Burdick, S., 2008. A new global model for P wavespeed variations in Earth's mantle. Geochem. Geophys. Geosyst. 9, Q05018.
Li, Z.X., Bogdanova, S.V., Collins, A.S., Davidson, A., De Waele, B., Ernst, R.E., Fitzsimons,I.C.W., Fuck, R.A., Gladkochub, D.P., Jacobs, J., Karlstrom, K.E., Lu, S., Natapov,L.M., Pease, V., Pisarevsky, S.A., Thrane, K., Vernikovsky, V., 2008. Assembly, con-figuration, and break-up history of Rodinia: a synthesis. Precambrian Res. 160,179–210.
Li, Z.X., Zhong, S., Wang, X.C., 2014. Formation of mantle plumes and superplumes:driven by subduction? GAC-MAC Abstract Volime 37. In: GAC-MAC Annual Meeting,21–23 May, Fredericton, Canada, pp. 162–163.
Lithgow-Bertelloni, C., Richards, M.A., 1998. Dynamics of Cenozoic and Mesozoic platemotions. Rev. Geophys. 36, 27–78. http://dx.doi.org/10.1029/97RG02282.
Matthews, K.J., Kayla, T., Maloney, Sabin Zahirovic, Williams, Simon E., Maria, Seton,Dietmar Müller, R., 2016. Global plate boundary evolution and kinematics since the
late Paleozoic. Glob. Planet. Chang. 146, 226–250.McNamara, A.K., Zhong, S.J., 2005. Thermochemical structures beneath Africa and the
Pacific Ocean. Nature 437, 1136–1139. http://dx.doi.org/10.1038/nature04066.Meriaux, C., Duarte, J., Schellart, W., Meriaux, A., 2015. A two-way interaction between
the Hainan plume and the Manila subduction zone. Geophys. Res. Lett. http://dx.doi.org/10.1002/2015GL064313.
Metcalfe, I., 1988. Origin and assembly of south-east Asian continental terranes. Spec.Publ. Geol. Soc. London 37 (1), 101–118. http://dx.doi.org/10.1144/GSL.SP.1988.037.01.08.
Montelli, R., Nolet, G., Dahlen, F.A., Masters, G., Engdahl, R., Hung, S., 2004. Finitefrequency tomography reveals a variety of plumes in the mantle. Science 303,338–343.
Montelli, R., Nolet, G., Dahlen, F.A., Masters, G., 2006. A catalogue of deep mantleplumes: new results from finite frequency tomography. Geochem. Geophys. Geosyst.7, Q11007. http://dx.doi.org/10.1029/2006GC001248.
Muller, R.D., Seton, Maria, Zahirovic, Sabin, Williams, Simon E., Matthews, Kara J.,Wright, Nicky M., Shephard, Grace E., Maloney, Kayla T., Barnett-Moore, Nicholas,Hosseinpour, Maral, Bower, Dan J., Cannon, John, 2016. Ocean basin evolution andglobal-scale plate reorganization events since Pangea breakup. Annu. Rev. EarthPlanet. Sci. 44, 107–138.
Ritsema, J., Deuss, A., van Heijst, H.J., Woodhouse, J.H., 2011. S40RTS: a degree-40shearvelocity model for the mantle from new Rayleigh wave dispersion, teleseismictraveltime and normal-mode splitting function measurements. Geophys. J. Int. 119.http://dx.doi.org/10.1111/j.1365-246X.2010.04884.x.
Schilling, J.G., 1991. Fluxes and excess temperatures of mantle plumes inferred from theirinteraction with migrating mid-ocean ridges. Nature 352, 397–403.
Scotese, C.R., 2001. Atlas of Earth history. In: PALEOMAP Progress Rep. 90-0497. Dep. ofGeol., Univ. of Tex. at Arlington, Arlington.
Steinberger, B., Calderwood, A., 2006. Models of large-scale viscous flow in the Earth'smantle with constraints from mineral physics and surface observations. Geophys. J.Int. 167, 1461–1481.
Tackley, P.J., 1998. Three-dimensional simulations of mantle convection with a ther-mochemical CMB boundary layer: D″? In: Gurnis, M. (Ed.), The Core-MantleBoundary Region. Geodyn. Ser., vol. 28. AGU, Washington, D.C., pp. 231–253.
Tan, E., Choi, E., Thoutireddy, P., Gurnis, M., Aivazis, M., 2006. GeoFramework: couplingmultiple models of mantle convection within a computational framework. Geochem.Geophys. Geosyst. 7, Q06001. http://dx.doi.org/10.1029/2005GC001155.
Tan, E., Leng, W., Zhong, S., Gurnis, M., 2011. On the location of plumes and lateralmovement of thermochemical structures with high bulk modulus in the 3-D com-pressible mantle. Geochem. Geophys. Geosyst. 12, Q07005. http://dx.doi.org/10.1029/2011GC003665.
Tosi, N., Yuen, D.A., de Koker, N., Wentzcovitch, R.M., 2013. Mantle dynamics withpressure- and temperature-dependent thermal expansivity and conductivity. Phys.Earth Planet. Inter. 217 (0), 48–58. http://dx.doi.org/10.1016/j.pepi.2013.02.004.
Wang, P., Li, Q., 2009. The South China Sea: Paleoceanography and Sedimentology.Springer Science & Business Media (ISBN 978-1-4020-9745-4).
Wang, Y., Wen, L., 2004. Mapping the geometry and geographic distribution of a very lowvelocity province at the base of the Earth's mantle. J. Geophys. Res. 109, B10305.http://dx.doi.org/10.1029/2003JB002674.
Wang, X., Li, Z., Li, X., Li, Jie, Xu, Y., Li, X., 2013. Identification of an ancient mantlereservoir and young recycled materials in the source region of a young mantle plume:implications for potential linkages between plume and plate tectonics. Earth Planet.Sci. Lett. 377–378, 248–259.
Wessel, P., Smith, W., 1998. New, improved version of Generic Mapping Tools released,EOS, 79, 579.
Xia, S., Zhao, D., Sun, J., Huang, H., 2016. Teleseismic imaging of the mantle beneathsouthernmost China: new insights into the Hainan plume. Gondwana Res. 36, 46–56.
Zahirovic, S., Nicolas, Flament, Dietmar Muller, R., Seton, Maria, Gurnis, Michael, 2015.Large fluctuations of shallow seas in low-lying Southeast Asia driven by mantle flow.Geochem. Geophys. Geosyst. 17. http://dx.doi.org/10.1002/2016GC006434.
Zhang, S., O'Neill, C., 2016. The early geodynamic evolution of Mars-type planets. Icarus265, 187–208.
Zhang, N., Leng, S. Zhong W., Li, Zheng-Xiang, 2010. A model for the evolution of theEarth's mantle structure since the Early Paleozoic. J. Geophys. Res. 115. http://dx.doi.org/10.1029/2009JB006896.
Zhong, S.J., 2006. Constraints on thermochemical convection of the mantle from plumeheat flux, plume excess temperature and upper mantle temperature. J. Geophys. Res.111, B04409. http://dx.doi.org/10.1029/2005JB003972.
Zhong, S., Zhang, N., Li, Z.X., Roberts, J., 2007. Supercontinent cycles, true polar wander,and very long wave-length mantle convection. Earth Planet. Sci. Lett. 261, 551–564.
Zou, H., Fan, Q., 2010. U–Th isotopes in Hainan basalts: implications for sub-astheno-spheric origin of EM2 mantle end member and the dynamics of melting beneathHainan Island. Lithos 116, 145–152.
