1 INTRODUCTION

The conjugate continental margin of the South China Sea (SCS) is often compared with the Atlantic continental margin (Clerc et al., 2018; Song et al., 2019; Ding et al., 2020). The SCS conjugate continental margins were often classified as magmapoor rifting margins (Song et al., 2017; Liu et al., 2021) or intermediate-type rifting margins (Franke et al., 2011; Larsen et al., 2018; Ding et al., 2020). The SCS continental margin is characterized by a rapid transition from continental breakup to igneous oceanic crust and intermediate magmatism (Larsen et al., 2018; Sun et al., 2019; Zhang et al., 2021), local lower crustal high-velocity layers (e.g., Li et al., 2019; Liu et al., 2021), warm pre-rifting lithosphere (Dong et al., 2020) and a very weak lower crust (Brune et al., 2017; Song et al., 2019; Zhang et al., 2020b).

These distinctive structural features in the SCS may be related to prolonged Mesozoic Paleo-Pacific subduction and magmatic emplacement. A Mesozoic volcanic arc ran parallel to the East Asian continental margin and extended from offshore eastern China to southern Vietnam (Li and Li, 2007; Li et al., 2018). The opening of the SCS is generally considered to be associated with the rollback/retreat of the slab or reversal of the subduction direction (Li and Li, 2007; Shi and Li, 2012; Li et al., 2020), eventually leading to lithospheric break-up and incipient seafloor spreading in the early Oligocene (Briais et al., 1993; Li et al., 2015). From the Mesozoic convergent continental margin to the Cenozoic divergent continental margin, the SCS was inevitably influenced by pre-existing tectonics and exhibits strong inherited characteristics. Previous studies suggest that delamination and partial melting of the lower crust (Yan et al., 2010; Xiao et al., 2019), regional uplift and erosion (Taylor and Hayes, 1980), and widespread arc magmatism (Li et al., 2018) have occurred at the SCS margin prior to rifting. In such a complex geological setting with strongly inherited features, the rheological and mechanical properties of the crust/lithosphere may change significantly (Zhang et al., 2020a), and present significant inhomogeneity in thermal conditions, crustal stretching and rock rheology (Clift et al., 2002; Dong et al., 2014, 2020; Ding and Li, 2016; Zhang et al., 2020b).

The Nansha Block, also called Dangerous Ground, extends eastwards from the Tinjar Fault to the Reed Bank (Fig. 1) and is the southern continental margin of the V-shaped Southwest Sub-basin (SWSB). The Nansha Block has an ultra-wide rifting domain of over 400 km, much wider than the SE margin of the SCS (NW Palawan margin, ~180 km) (Hayes and Nissen, 2005; Dong et al., 2020; Nirrengarten et al., 2020). The wider tectonic domain implies the potential existence of variable crustal architecture. Previous studies have reported crustal heterogeneity and extensional differences along strike (Ding and Li, 2016; Zhang et al., 2020b), Mesozoic arc-related granites (Li et al., 2018; Xiao et al., 2019), post-spreading lower crustal high-velocity layers (Li et al., 2019), and scattered thick Mesozoic strata (Wang et al., 2016) in the Nansha Block. The current understanding of the crustal-scale structure of the Nansha Block mainly relies on (ocean bottom seismometer) OBS observations because of their good imaging for the crustal velocity structure (e.g., Qiu et al., 2011; Niu et al., 2014; Pichot et al., 2014; Wei et al., 2020). However, previous studies have focused on a single seismic profile to describe the change in crustal structure from continental to oceanic domains along seismic lines. An overall perspective is lacking and how the crustal architecture of the Nansha Block varies along strike is also not well understood, which limits comparative study of the crustal architecture of the two conjugate continental margins of the SCS.

Based on five published OBS/MCS profiles (Qiu et al., 2011; Niu et al., 2014; Pichot et al., 2014; Wei et al., 2020; Zhang et al., 2020b) in the region from the Reed Bank in the east to the Tinjar Fault in the west (Fig. 1), we carried out gravity modelling for the crustal structure of the Nansha Block. The joint magnetic anomalies also allow a comprehensive interpretation of the thick crustal blocks that are prevalent near the continent-ocean transition (COT). The integrated interpretation can compensate for the crustal structure revealed by the OBS velocity profiles.

2 DATA AND METHOD 2.1 Data source

The five published OBS/MCS profiles across the Nansha Block from east to west are AA' (OBS973-2, Niu et al., 2014), BB' (OBS-CFT, Pichot et al., 2014), CC' (OBS973-1, Qiu et al., 2011), DD' (MCS-DZ02, Zhang et al., 2020b), and EE' (OBS-DZ01, Wei et al., 2020).

The 1-arc-minute grid global marine gravity anomaly from Sandwell et al. (2014) and the magnetic anomaly from Magnetic Anomaly Map in Southern China of Wu et al. (2019) were used for profiles AA', BB', and CC'. In 2015, the R/V Xiangyanghong 10 of the Second Institute of Oceanography, Ministry of Natural Resources, acquired two seismic lines (DZ01 and DZ02) on the Nansha Block, and also acquired gravity and magnetic data on the survey lines at a sampling interval of 1 s. The shipboard gravity and magnetic data were used in profiles DD' and EE' after re-interpolating to 1-km intervals. Bathymetry and sediment thickness were from the 15-arc-second global topographic relief data SRTM15+ (Tozer et al., 2019) and the new 5-arc-minute global total sediment thickness grid GlobSed (Straume et al., 2019), respectively.

2.2 Modelling method

The high-resolution P-wave velocity structure provides a good initial constraint on the Moho and the thicknesses of the upper and lower crusts.

A six-layer initial structure is set up (Table 1), including seawater, sediments, crust and upper mantle, and the crust is further divided into a pre-rift layer, the upper crust and the lower crust. The initial morphology of these three top boundaries and the Moho is constrained by the P-wave velocity structure (OBS973-2, OBS-CFT, OBS973-1, and OBS-DZ01). A sedimentary basement identified from the OBS profile is combined with a new global sediment thickness data (Straume et al., 2019) to identify the bottom boundary of the sediment layer, i.e., the top of the pre-rift layer. The P-wave velocity of the top upper crust is considered to be 5.6 km/s (Christensen and Mooney, 1995; Li et al., 2021; Fan et al., 2022). A layer known as the pre-rift layer is sandwiched between the top of the upper crust and the sedimentary basement and may consist of consolidated pre-Cenozoic sediments and lightly metamorphosed rocks (Vijayan et al., 2013; Pichot et al., 2014). Previous seismic reflection and refraction data in the SCS region also reveal that the P-wave velocity of pre-Cenozoic sediments can reach up to 5.0–5.5 km/s (Nissen et al., 1995; Pichot et al., 2014; Wang et al., 2016; Fan et al., 2022). The P-wave velocity of the top of the lower crust is defined to be 6.4 km/s, which roughly divides the upper felsic rocks from the lower mafic rocks.

The crustal structure of profile DD' is poorly constrained due to the absence of a published OBS velocity structure. However, multi-channel seismic reflection (Zhang et al., 2020b) and regional Moho depths (Wu et al., 2017) can be used for interpreting the sediment basement and crustal bottom boundary, respectively. A joint inversion of the crustal structure incorporating gravity, magnetic, and seismic data was performed, based on the established 6-layer initial model. With reference to previous gravity inversions of the crustal structure in the SCS (Hao et al., 2011; Vijayan et al., 2013; Gao et al., 2015) and the non-linear relationship between P-wave velocity and density (Christensen and Mooney, 1995), the initial density values of the six-layer model were set to 1.03, 2.30, 2.55, 2.70, 2.90, and 3.30 g/cm³, respectively (Table 1).

3 RESULT: CRUSTAL STRUCTURES OF THE NANSHA BLOCK

The bulk of the continental lower crust is chemically equivalent to gabbro and its components are generally considered to be mafic high-grade metamorphic rocks (Christensen and Mooney, 1995). It is ductile-deformable and generally isotropic, so we set a uniform density to the lower crust. The lateral density variation is concentrated in the upper crust and ranges from 2.60 to 2.80 g/cm³. The pre-rift layer is a brittle layer containing sedimentary, metamorphic, and volcanic rocks, and maybe densely cut by faults. The lateral density variation in the pre-rift layer varies between 2.45 and 2.65 g/cm³. The top of the upper mantle is assumed a uniform density of 3.30 g/cm³. In inversion of the magnetic anomalies, we only inverted magnetic susceptibility for a few interesting geological bodies in the profiles, rather than along the whole profile (The extent of the magnetic inversion is marked in light grey shading in Figs. 2–5). Due to the major contribution of shallow magnetic sources to the magnetic anomaly, we set the magnetic susceptibility in the upper crust only. Large-scale faults in the model are based on previous reflection seismic profiles and model interpretation (Ding and Li, 2011; Dong et al., 2014; Pichot et al., 2014; Liang et al., 2019; Wang et al., 2019; Zhang et al., 2020b; unpublished MCS-DZ01), and the gravity inversions also show significant density differences across the faults. The continental-ocean boundary (COB) marked in the model (thin dotted-dashed line in Figs. 2–5) is delineated based on marked gravity and magnetic differences between the oceanic crust and continental crusts.

3.1 Profile AA'

Profile AA' is located on the eastern edge of the Reed Bank, where the crustal thickness thins from ~20 km in the continental domain to ~6 km in the oceanic domain. The profile shows a slowly rising Moho and a wide crustal necking zone (130–280 km, Fig. 2). The density of the pre-rift layer decreases laterally from the Liyue Basin to the sides, with the lowest density of 2.45 g/cm³ at the seaward end of the necking zone. There are also lateral variations in the thickness and density of the upper crust, which is characterized by a boudin-like structure that is more pronounced in the seaward direction. The lower crust thins sharply in the necking zone, from a thickness of 12 km to a minimum of 3 km. At the seaward end of the necking zone (~130 km), the crustal thickness is only 6–7 km. The upper and low crustal density in this region is 2.63 and 2.80 g/cm³, respectively. It is uncertain whether the crust is continental or oceanic in this region. This region is located at the junction of the Eastern Subbasin and the Southwestern Subbasin, and is near the Zhongnan Fault, where the magnetic anomalies are not significant and the crustal density may have been reduced by fracturing.

3.2 Profile BB'

Profile BB' crosses the northeastern part of the Zhenghe Massif and the crustal architecture is distinctly different from that of profile AA'. Here, the crustal necking zone is narrow, only approximately 40 km in width, and the crustal thickness thins dramatically from 17 to 5.5 km over a short distance from the continental to the ocean (Fig. 3). There is a small lateral variation in the density of the pre-rift layer and a large density variation in the upper crust (Fig. 3). From 60 to 180 km in the model, a block of the upper crust has not been significantly thinned. The thickest upper crust block is immediately above the thinnest lower crust. The ends of the block correspond to the thickening of the lower crust and dense faults, indicating that it is a relatively rigid block. The block shows strong negative magnetic anomalies up to -200 nT. Inversion and modeling results show that the thick crust block has a high magnetic susceptibility of 0.062–0.078 (SI) and a density of 2.80 g/cm³. Similar to that in the profile AA', the lower crust thinning remains the most significant, reaching a minimum thickness of only 2.5 km.

3.3 Profile CC'

Profile CC' is located to the southwest of the Zhenghe Massif, where the crustal necking zone is also relatively narrow. Within a distance of 80 km, the crust thins from 19 to 5.5 km (Fig. 4). The thinning of the lower crust is still the most pronounced within the necking zone, to only 2.5 km. However, outside the necking zone, the thinning of the crust is relatively uniform. Setting the sediment layer to a uniform density of 2.30 g/cm³ in the model does not fit the observed values well. Therefore, we set a Mesozoic sedimentary layer at a density of 2.50 g/cm³ (yellow layer in Fig. 4), consistent with the observations of Wang et al. (2016). The thick upper crust at a model distance of 70 km also exhibits a relatively high density and magnetic susceptibility of 2.80 g/cm³ and 0–0.033 (SI), respectively, similar to that in profile BB'.

3.4 Profile DD'

Profile DD' crosses the entire Nansha Block from the southern edge of the Southwest Subbasin of the SCS to the Nansha Trough. Here thinning of the crust is relatively uniform and a necking zone is approximately 300 km wide. The crust thins from a maximum of 18.5 to 5.5 km (Fig. 5). There is a large lateral variation in crustal thickness and density, with boudin-like structures becoming more pronounced towards the COB. The crustal density is high in the middle of the Nansha Block and low on both sides, and decreases monotonically towards the sea. There are more high-frequency components in the gravity and magnetic anomalies, suggesting shallow magmatism. The crust is thinnest at the Nansha Trough, with sediment thicknesses exceeding 5 km.

3.5 Profile EE'

Profile EE' is near the Tinja Fault (West Baram Line), the western boundary of the Nansha Block. Here the spreading center of the Southwest Subbasin does not reach and the profile is entirely in the thinned continental crust. The upper crust is undulating in thickness. The thinning of the lower crust is still striking. An upwelling Moho and a 1.7 km thick lower crust are found in the Nanweixi Basin (NWXB, Fig. 6). The Nansha Trough has the thickest sediments and the thinnest crust in the entire profile.

4 DISCUSSION 4.1 Variations and nature of the thick upper crust along the southern margin

Thick and relatively less stretched upper crust is prevalent on the landward side near the ocean basin in profiles AA' to DD'. These masses may have influenced the thinning regime of the crust, exhibiting a distinct depth-dependent crustal extensional feature. The planar locations of the areas of the thick upper crust in the models are shown in Fig. 7 and are based in part on the gravity anomaly and topography data.

Although these thick upper crust blocks are prevalent and affect the thinning of the lower crust to various degrees, they appear to have different density and magnetic susceptibility properties. The thick upper crust of profiles AA' and DD' appears to have the same origin, both with low densities and weak to moderate magnetic susceptibilities. The local high magnetic susceptibility up to 0.04 (SI) of the thick upper crust in profile DD' may be influenced by magmatic intrusions along large-scale faults (model 80 km). The magnetic susceptibility of these thick upper crust blocks is consistent with that of granites from the SCS region (Lang et al., 2011; Tang et al., 2021). The density and P-wave velocities also match those of granitoids measured in the laboratory (Christensen and Mooney, 1995). Mesozoic in-situ granites (including granodiorite and monzonitic granite), intruded during the Mesozoic paleo-Pacific subduction (Li and Li, 2007), were dredged from the Xiaozhenzhu Rise in the northern Reed Bank (Xiao et al., 2019).

The properties of the thick upper crust block of profile BB' are distinct from those of profiles AA' and DD'. Although samples of Mesozoic tonalitic and monzogranitic rocks were dredged (Yan et al., 2010) and interpreted (Pichot et al., 2014), they show strong magnetism (magnetic susceptibility of 0.062 (SI)) and high density (2.80 g/cm³). Although tonalitic rock may can have high density and magnetic susceptibility, it is difficult to form a large-scale tonalitic body, and monzogranitic rock usually has a low density and magnetic susceptibility (Christensen and Mooney, 1995; Lang et al., 2011). Geochemical analyses by Yan et al. (2010) suggest that these Mesozoic granite samples were derived from partial melting of the Precambrian basement. We suggest that this thick upper crust block in profile BB' may be from an ancient rigid unit similar to the Xisha Massif, which underwent extensive intermediate-acid magmatic intrusion during the Mesozoic (Zhu et al., 2017). The thick upper crust block in profile CC' adjacent to the Zhenghe Massif exhibits similar density (2.80 g/cm³) and magnetic susceptibility (up to 0.033 (SI)) to that in profile CC', and it is postulated that they are of the same origin.

The prevalence of granitic and/or granite-intruded thick upper crust at the seaward tip of the Nansha Block support extensive Mesozoic magmatism and later continental break-up along an island arc (Li et al., 2018; Xiao et al., 2019). However, differences in the nature of these thick upper crusts blocks attest to the strong heterogeneity of the Nansha Block.

4.2 Ductile shear of the upper crust: boudinage

Models show that the thick upper continental crust near the COT seems to depict a boudin-like structure. Crustal boudinage is frequently observed at some rift margins, such as the central South Atlantic margins (Clerc et al., 2018), the North Atlantic margins (Reston, 1988, 2007), the Northern Pyrenees Zone (Clerc and Lagabrielle, 2014), and the northern margin of the SCS (Deng et al., 2020).

Boudinaged crust is generally considered to be in a semi-ductile state, with a ductile shear zone in the inter-boudin area taking up most of the extensional offset (Clerc et al., 2018; Zhang et al., 2020a). The interior of the block is only slightly deformed, but the inter-boudin area is strongly necked and the crustal block moves to the sides as if it was extracted, creating a continuous lenticular shape. The ductile flowing lower crust fills under the inter-boudin area via isostatic compensation (Figs. 8–9). The ductile shear zone is dominated by banded mylonites, which have low viscosity and friction and are not conducive to block stability. The greenschist mylonites recovered at IODP Site U1504 on the distal margin of the northern SCS also indicate ductile deformation behavior within the continental basement (Sun et al., 2020; Zhang et al., 2020a).

Crustal boudinage is closely related to the thermal regime of the crust during rifting. The level and scale of boudinage varies between different rifting margins (Table 2). For example, a crust-scale ductile shear and boudinage are observed at the western edge of the Barents Sea, where low-angle ductile shear zones cut through the crust to reach the Moho (Gernigon et al., 2014; Clerc et al., 2018). An upper crustal boudinage covered by very thick salt sediments with a shear zone extending into the lower crust (Clerc et al., 2018) is observed at the Gabon margin in the central South Atlantic. In the eastern North Pyrenees Zone, analogous to the most distal part of the passive continental margin, the crust layer together with the overlying pre-rift sediments transformed into a boudin-like structure, related to rifting in a localized high thermal anomaly (Clerc and Lagabrielle, 2014). At the western Iberian margin, the boudinaged lower crust is interpreted to be related to the cold lithosphere of the region (Reston, 2007).

Our results indicate upper crustal boudinage in the Nansha Block. Unlike in the Gabon margin, however, the upper crust layer of the Nansha Block is not immediately overlain by thick syn-rift sediments but by a brittle pre-rift layer that is incised extensively by faults. The Nansha Block may possess a "classical" brittle-sheared-ductile rifting regime, with the crystalline upper crust bearing the role of a brittle-ductile transition zone. We suggest that such a rifting regime may be related to the thermo-mechanical state during rifting. The SCS margin developed over a Mesozoic continental arc may retain high residual heat (Clift et al., 2002). A warm and very weak lower crust of the SCS (Clift and Lin, 2001; Brune et al., 2017; Song et al., 2019; Dong et al., 2020) supports more ductile flow. Additionally, due to the lack of sediment input to the Nansha Block east of the Tinjar Fault during rifting, the lack of overlying insulation by syn-rift sediments may have made the pre-rift crust layer more brittle. An interesting phenomenon is the strong positional correlation between the roots of large-scale detachment faults and the sheared inter-boudin region in the models (Figs. 2–6), indicating a linkage between faults and shear zones. The upper crust of the Nansha Block underwent the paleo-Pacific subduction from the Triassic to Cretaceous (Li and Li, 2007), forming many thrust faults. These inherited faults may have provided weak zones for tectonic shearing during later extension. Such a thermo-mechanical setting may explain the crustal boudinage of the intermediate layer in the Nansha Block. Therefore, we suggest a brittle-sheared-ductile thinning model for the Nansha Block, constituting a fully brittle pre-rift layer, a sheared upper crust, and a ductile lower crust (Fig. 8).

Based on the thermal regime and the inherited pre-rift structure, we reconstructed an evolving process that shaped the current boudin-like crustal architecture.

(1) Mesozoic palaeo-Pacific subduction (Li and Li, 2007) created numerous thrust faults within the basement of the Nansha Block, which became strain localization zones when regional stresses shifted to extension (Fig. 9, phase 1).

(2) The crustal thickening and magmatic accretion caused by the palaeo-Pacific subduction may have resulted in a weak crustal rheology and a delayed rifting phase in the Nansha Block (Brune et al., 2017; Deng et al., 2020). Linear diffusion creeps in discrete pre-existing tectonic weak zones is probably the dominant mode of deformation within the basement of the Nansha Block (Fig. 9, phase 2), similar to the Banana split model (Bürgmann and Dresen, 2008).

(3) With the further extension, the faults in the brittle layer will eventually connected to the shear zone, forming the ramp-flat-ramp geometry common to detachment faults (Fossen, 2010). As the detachment fault developed, the shear zone continued to evolve, and the lower crust compensated for doming, contributing to the present-day crustal architecture of the upper crustal boudinage (Fig. 9, phase 3).

5 CONCLUSION

Based on gravity, magnetic and refraction/reflection seismic profiles, five 2D structural models of the crust in the Nansha Block are constructed. The modelling results reveal that the upper crust of the Nansha Block is highly variable in density and thickness. In addition, although the thick upper crust seems to be prevalent along the margin, it shows different properties, some from granitic masses formed by Mesozoic arc magmatism and others from Precambrian rigid blocks by magmatic addition.

We propose that the heterogeneous Nansha Block underwent a brittle-sheared-ductile thinning due to a warm but not so hot rift regime. The pre-rift layer experienced brittle deformation, the crystalline upper crust was boudinaged by shear, and the lower crust experienced ductile flow.

The pre-existing structures of the rifted margin played a crucial role in the formation of the hyper-extended crust. The inherited crustal thermal regime, mechanical state and material composition prior to the SCS rifting determined the initial crustal rheology and ultimately controlled the architecture of the crust.

6 DATA AVAILABILITY STATEMENT

The sediment thickness and topographic relief data were obtained from https://ngdc.noaa.gov/. The gravity data was obtained from https://topex.ucsd.edu/WWW_html/mar_grav.html. The shipboard gravity and magnetic data were available from the corresponding author upon reasonable request.

7 ACKNOWLEDGMENT

We are grateful to the three anonymous reviewers for their very constructive comments that helped to improve the paper.