Insights from the P Wave Travel Time Tomography in the Upper Mantle Beneath the Central Philippines

Abstract

In this paper, we present a high resolution 3-D tomographic model of the upper mantle obtained from a large number of teleseismic travel time data from the ISC in the central Philippines. There are 2921 teleseismic events and 32,224 useful relative travel time residuals picked to compute the velocity structure in the upper mantle, which was recorded by 87 receivers and satisfied the requirements of teleseismic tomography. Crustal correction was conducted to these data before inversion. The fast-marching method (FMM) and a subspace method were adopted in the forward step and inversion step, respectively. The present tomographic model clearly images steeply subducting high velocity anomalies along the Manila trench in the South China Sea (SCS), which reveals a gradual changing of the subduction angle and a gradual shallowing of the subduction depth from the north to the south. It is speculated that the change in its subduction depth and angle indicates the cessation of the SCS spreading from the north to the south, which also implies that the northern part of the SCS opened earlier than the southern part. Subduction of the Philippine Sea (PS) plate is exhibited between 14° N and 9° N, with its subduction direction changing from westward to eastward near 13° N. In the range of 11° N–9° N, the subduction of the Sulu Sea (SS) lies on the west side of PS plate. It is notable that obvious high velocity anomalies are imaged in the mantle transition zone (MTZ) between 14° N and 9° N, which are identified as the proto-SCS (PSCS) slabs and paleo-Pacific (PP) plate. It extends the location of the paleo-suture of PSCS-PP eastward from Borneo to the Philippines, which should be considered in studying the mechanism of the SCS and the tectonic evolution in SE Asia.

1. Introduction

The Philippines are located in the collision and convergence region of the Eurasian plate, the PS plate and the Indo-Australian plate. The interaction of the three plates has created a complex tectonic environment in the area. The Philippines are surrounded by the bidirectional subduction of the PS plate and the SCS (Figure 1). The Philippine subduction zone plays a critical role in the tectonic evolution of the SE Asia. The kinetic energy generated by the strong convergence of the Eurasian and PS plates [1] is regulated and absorbed by the bidirectional subduction system in this region [2].

The mechanism of the SCS opening dynamics and tectonic evolution are still unresolved issues. Many kinematic models have been put forward to explain the opening process of the SCS [3,4,5,6,7,8,9]. However, there is no widely accepted model of the SCS spreading. One of the major controversies in establishing the SCS opening model is which of the eastern and southwestern basins of the SCS opened first. The Philippines lies on the eastern boundary of the SCS. The westward subduction of the PS plate beneath the central Philippine limited the eastward spreading of the SCS, and the SCS plate subducted eastward beneath the Philippines [10]. Previous tomographic results revealed the slab tear of the SCS slab [11,12]. However, there were few studies on the relationship between subduction and spreading of the SCS plate. Therefore, research on deep structure of the Philippine subduction zone will possibly contribute to study the opening of the SCS.

Previous researchers have carried many studies to understand the geodynamic process of the Philippine subduction zone [13,14,15,16]. One of the most important controversial topics in this region is the PSCS. At present, there has been an inconclusive debate about where are the PSCS slabs. Some researchers argued that the PSCS once existed in the area between the southern boundary of the SCS and Borneo [17,18]. Tang et al. found 500 km southeastward high velocity anomalies below northern Borneo, which were identified as PSCS slabs [7]. Hall et al. argued that the PSCS slabs were at 800 km between East Borneo and Southern Philippines [19]. A high-velocity anomaly was discovered at 400 km–700 km depth beneath the central Philippine, which was interpreted as the PSCS slab that generated by southward subduction [12]. However, some researchers claimed that the PSCS slab subducted northward to the present SCS [17,20]. According to the tomographic results, Shi et al. proposed that PSCS has subducted northward beneath Borneo and inferred that the Paleo-Tethys and the paleo-Pacific (PP) have never been connected by PSCS [21]. Lin et al. proposed a double-side subduction model to interpret the PSCS based on plate reconstruction [22].

To answer the above questions, a high-resolution tomographic model is needed. In this work, we selected 32,224 useful relative travel time residuals recorded by 87 receivers distributed in the central Philippines. Crustal correction was applied to remove the influence on the upper mantle. These data were adopted to image the detailed upper mantle structure in the Philippine subduction zone. Then, we present a 3-D high-resolution tomographic model in the upper mantle beneath the central Philippines. Our tomographic model clearly images the steeping and tearing of SCS slabs, which provide evidence to define the opening sequence of the SCS. The PS plate and SS slab are also imaged by the present model. The PSCS-PP slabs are also revealed by our tomographic result, which helps to determine the location of paleo-suture of PSCS and PP. The present study has significant implication for the establishment of the SCS opening mechanism and the study of the tectonic evolution of SE Asia.

2. Data and Methods

2.1. Data

The range of study area is (8° N~18° N, 118° E~128° E). There are 87 receivers distributed in the study area (Figure 2a). In this work, we picked travel time data recorded by these receivers from 1960 to 2020, which was primarily derived from International Seismological Center (ISC) [23]. The selection of teleseismic tomography data should satisfy the following conditions: (1) The magnitude of events is greater than 4.5; (2) The epicentral distance is 30°–90°, which reduces the influence of deep structures such as lower mantle; (3) Only events received by more than five receivers can be used for inversion calculation. The application of relative travel time residuals to seismic imaging is intended to eliminate the effects of teleseismic events localization errors and lateral homogeneities outside the study area.

To obtain the relative travel time residuals, the mean value of each event over the whole receiver array is calculated and subtracted from the absolute travel time residuals for each receiver [24]. Finally, a total of 2921 teleseismic events and 32,224 relative travel time residuals were picked for tomographic computation (Figure 2c). Figure 2b shows the local seismic events occurred in the study area in recent 10 years.

2.2. Crust Correction

According to the application effect of predecessors [25,26], the velocity imaging results will be affected by complex crustal structure when utilizing teleseismic tomography method to image the upper mantle structure. Some researchers have conducted different methods to solve this problem [27,28,29]. In this paper, the following method were utilized to eliminate the impact of crust. More details can be found in the reference [28].

(1)

Selecting 1-D and 3-D crust velocity models. We selected ak135 [30] as 1D crust model, and crust 1.0 [31] as 3-D crust model, respectively.

(2)

Calculating travel time and travel time residuals in 1-D and 3-D crust models, respectively. In this paper, the average depth of the Moho surface of 30 km is taken as the thickness of the crust, the crust contains upper, middle and lower layers. The calculation formula is as follows:

$$\delta T_{crust}=T_{3-D}-T_{1-D}=({\displaystyle \sum _l\frac{h_l}{\cos {\theta }_l\times V_l}{)}_{3-D}-}({\displaystyle \sum _k\frac{h_k}{\cos {\theta }_k\times V_k}{)}_{1-D}}$$

(1)

where δT_crust represents the relative travel time residuals in the crust, T indicates the travel time in the crust, h indicates the thickness of each layer in the crust, θ represents the incident angle of rays at each interface and V denotes the velocity in each layer. The 3-D and 1-D subscripts indicate 3D crust model and 1-D model, respectively.

(3)

The real data used in the tomographic calculation are the measured travel time minus the theoretical travel time and then minus the travel time residuals in the crust. The expression of the formula is as follows:

$$t=(T_{obs}-T_{cal})-\delta T_{crust}$$

(2)

where t represents the relative travel time residuals, T_obs indicates the measured travel time and T_cal the theoretical travel time.

The method described above was applied to the filtered selected teleseismic data. Figure 3 presents the mean values of the picked teleseismic data and the crust-corrected relative travel time residuals at each receiver, respectively.

It can be seen that the numerical range of mean relative travel time residuals at each receiver is from −2 s to 2 s. Whether crustal correction was applied or not, most values of the mean relative travel time residuals in the study area are positive. This phenomenon is due to the existence of multiple subduction plates beneath the study area. When the rays pass through the high-velocity body, especially a subducted slab, the travel time of rays will become much shorter, so the relative travel time residuals at these regions are positive. The negative values on mean relative travel time residuals in the study area are mainly located in the east of the subduction zone of the SCS. This phenomenon is probably due to the subduction in the Philippines and asthenosphere upwelling. Thus, the travel time of the rays passing through this region is longer, which results the negative values on the mean relative travel time residuals.

2.3. Methods

In this paper, we conducted a teleseismic tomography routine to image the upper mantle structure beneath the central Philippines, which is so-called fast marching teleseismic tomography (FMTT) [32,33]. This routine is composed of the forward and inversion step. In the forward step, the calculation of the travel time from the teleseismic event to the bottom of the local model is based on the ak135 (Figure 4). FMM is performed to calculate the travel time of rays from the bottom to the receivers in the local model [34,35]. In inversion step, a kind of subspace method, is conducted to calculate the velocity structure beneath the receiver array. The relative travel time residuals are adopted to determine the 3-D velocity structure beneath the station arrays. The initial models for inversion are parameterized by a grid of nodes with tricubic B-spline interpolation. The calculation of this method is performed in spherical coordinates [32]. The Moho depth of the initial model was modified to 30 km, which aims to satisfy the condition for using the tomographic routine [36].

3. Resolution Tests and Results

In this section, the checkerboard resolution tests are first conducted to find the optimal inversion grid intervals. Then, the optimal grid intervals were applied in the following real computation and the final tomographic model were obtained. Finally, we propose a 3-D tomographic model obtained from the inversion.

3.1. Resolution Tests

To qualify the resolving ability and quality of our travel time data, we completed resolution tests by using checkerboard tests method. We applied a series of different grid intervals to find the optimal inversion parameterization. The input model of checkerboard tests is consisting of +6% or −6% positive and negative perturbations based on the initial model generated by the ak135 model. Then, the input checkerboard model is used to compute the synthetic travel time. The synthetic relative travel time residuals are used to invert the recover checkerboard model. The spatial distribution of the receivers and events used in the theoretical tests are the same as that in the real observations. The optimal lateral grid spaces are 0.65° × 0.65°, and the optimal vertical grid interval is 40 km. Checkerboard resolution tests results show good recovery effect for the velocity perturbations (Figure 5), which demonstrates the observing system composed of receivers and teleseismic events can recover the velocity anomaly patterns very well. It can be seen that when the depth is less than 300 km, the velocity perturbations beneath the station arrays can be well-resolved. However, the velocity perturbations on the edge of the study area cannot be well recovered due to the poor crossing of ray paths. When the depth is greater than 300 km, the velocity perturbations on the edge of the study region are well-resolved because of the enhancement of ray paths crossing. The last subplot of Figure 5 shows the distribution of rays in horizontal direction. It shows that the observation system used in this study is effective for studying upper mantle structures.

3.2. Results

According to the result of checkerboard resolution test, an optimal grid was utilized to parameterize the initial model. After several inversion calculations, we obtained a 3-D tomographic model with a lateral resolution of 0.65° and 0.65° and vertical resolution of 40 km beneath the central Philippines. The numerical range of relative travel time residuals is significantly reduced from (−4.5 s, +4.5 s) to (−1.4 s, +1.4 s) after tomographic inversion (Figure 6). Both before and after the inversion, the values of the relative travel time residuals show a normal distribution.

The present tomographic model is obtained by using the aforementioned approaches. Since the crust correction removes the influence of the complex crustal structure on the upper mantle structure, the velocity structure of the crust is not shown in our tomographic model. Figure 7 shows the velocity anomalies relative to background in different depths. Figure 8 and Figure 9 show the slices along the different latitude.

The present tomographic model identifies some prominent features. There are obvious continuous high velocity anomalies beneath the Manila trench and Luzon Island (white dotted line circled area in Figure 7a–i), which have also been revealed in the previous studies [12,16]. These high velocities are interpreted as the subduction of the SCS slab beneath the Philippines, which is consistent with previous views. It can be seen that the subduction of the SCS plate is parallel to the Manila trench and distributed in a NS direction (Figure 7a–h). With increasing depth, the horizontal position of the SCS subduction slab moves from west to east, reflecting that the SCS is subducted from west to east below the Philippines. The relationship between subduction plate and spreading of the SCS will be discussed in the following chapters.

The present tomographic model reveals a prominent high velocity body at depth range from 420 km to 720 km, which is circled by red dotted lines (Figure 7g–l). The high velocity anomalies are distributed in a NE direction, and its area increases with the increase of depth. The present study suggests that these high velocity anomalies are the subduction of the PSCS slab and the PP plate. it will be discussed in the next section.

Another two features identified in the present model are PS plate and SS slab. The high velocity anomalies indicating the subduction of the SS slab and the PS plate are circled by the purple (Figure 7a–f) and black (Figure 7a–i) dotted lines, respectively. According to the results, it can be seen that the area of the SS slab gradually becomes larger as the depth increases. Combining the characteristics of the horizontal and vertical velocity anomalies, the deepest part of the SS slab is above the MTZ. On the contrary, the area of the high velocity anomalies representing the PS plate is gradually smaller with the increase of depth, and its position is also moved from the west side of the Philippine Trench to the east side.

4. Discussion

4.1. Subduction of South China Sea

The phenomenon of slab tearing was first discovered in the study of the interaction between seafloor spreading ridges and trenches [37,38]. Dickson and Snyder proposed the concept of slab window, if the mid-ocean ridge continues to expand during the subduction to the trench, and a continuous expansion gap is formed between the two sides of the plate. This gap is called the slab window, and the phenomenon is widely distributed in the circum-Pacific subduction zone [37]. Later, it was found that the hotter, younger plates are susceptible to stretching and tearing during subduction to form plate tear zones [39,40,41,42,43,44]. The present tomographic model clearly reveals a steeping subduction of SCS slab. The SCS slab shows a near-vertical morphology along the east-west spreading profile north of 16° N (Figure 8a–d). Especially, it shows a near-vertical subduction angle on a slice along 17.5° N (Figure 8a). However, in the south of 16° N, the subduction angle of SCS slab gradually flattens out as its position moves from north to south (Figure 8e–i and Figure 9a–d). The subduction angle is only 45° in the southernmost slice, which is along 11.5° N (Figure 9d). The above phenomenon reflects the tearing of the subduction slab in the SCS, which have been also revealed by others [11,44]. A decrease of the subduction angle along an island-arc was also observed in Central America [45,46], which is important as the subduction angle has an impact on the stress field and island-arc and the earthquake distribution. According to the model of this study, the slab window is located near 16° N. It can be seen that a seismic blank zone is formed near the slab window (Figure 2b), which is probably due to the existence of the slab window that prevents stress concentration in this region [44]. A series of volcanoes distributed along the island-arc are formed on both sides of the plate window. High velocity anomalies indicating the SCS slab are also found between 13° N and 11° N in our tomographic model, while the previous results indicated the SCS slab only existed north of 13° N [12].

Previous researchers have suggested that the steepening of the SCS slab was mainly associated with the collision between the Palawan microcontent block and Philippines [12]. The above factors do influence the subduction angle of SCS slab, while we argue that the more important cause of this phenomenon is the difference in the timing of the eastward subduction of the SCS. The opening of the SCS was blocked by the PS plate, which resulted in the subduction of the SCS beneath the Philippines [10]. Different scholars hold different views on the opening sequence of the SCS. Some scholars believed that the eastern sub-basin opened earlier than the southwestern sub-basin [3,8,47], while others held the opposite opinion [48,49]. On the slice along 17.5° N (Figure 8a), the subduction depth is even down to 800 km. The depth of the SCS slab is only 450 km along a 12° N slice (Figure 9d). It demonstrates that the subduction depth of SCS slab changes from deep to shallow from north to south. Our tomographic model indicates that the subduction of the SCS was developing from north to south. It is derived that the northern part of the SCS subducted earlier than the southern part. Furthermore, it can also be inferred that the northern part of the SCS opened earlier than the southern part. This provides evidence of deep tectonics to determine the opening sequence of the SCS. We have drawn a cartoon diagrams to illustrate the slab tearing of the SCS plate (Figure 10). The location of the slab tear is below Luzon Island, which is close to the mid-ocean ridge in the eastern sub-basin of the SCS. As the slab tearing forms a slab window, it provides a channel for the upwelling of the high temperature asthenosphere material [44]. It may also be one of the reasons for the development of volcanic activity in the region.

4.2. Subduction of Proto-South China Sea

Previous researchers have conducted a large number of studies on the PSCS and formed different understandings [11,12,17,19,21]. Some scholars believe that the PSCS once existed in the area between the southern boundary of the SCS and Borneo [17,18]. Tang et al. interpreted a 500 km southeastward high velocity anomaly below Borneo as the slab remnant of PSCS [7]. Zahirovic et al. suggested slabs found beneath north Borneo and the SCS at a depth less than 1000 km were the remnant of the PSCS [50]. Hall et al. claimed that the PSCS slabs were at 800 km between East Borneo and South Philippines [19]. Palawan was classified as a PP subduction accretion zone, which was different from previous results [8]. Zhou et al. determined the location of the paleo-suture of PSCS and the PP subduction accretion zone based on the study of the paleogeographic evolution of the Late Mesozoic rocks in SE Asia [51]. Therefore, the PSCS is basically widely recognized subducted below Borneo. The more controversial point is whether the PSCS extended to the east.

The present tomographic model reveals prominent high velocity anomalies in the MTZ (Figure 8h–i and Figure 9a–i). Yumul et al. have ever suggested that there preserved an ancient subducted slab beneath the central Philippines [52]. The ophiolite belt in the Sunda continent is considered as evidence of the extinction of the PSCS [53]. Fan et al. interpreted the high velocity anomalies revealed in the central Philippines as the PSCS slabs [12]. The above evidence suggest that the high velocity anomalies imaged in MTZ is the PSCS slab. The morphology of the remnant of the PSCS slabs identified by our model differs somewhat from that of predecessors in the study area [12]. The high velocity anomalies representing the PSCS slabs found in this study contain two parts, one part is between 410 km–800 km in depth beneath the Philippine Trench in the range of 14° N–8° N (Figure 8h–i and Figure 9a–i). This part high velocity anomalies are basically consistent with the previous studies [12]. The other part is between the MTZ beneath Palawan Island (Figure 9a–e), which have been partially destroyed by lava activity. Sibuet et al. have ever defined the location of the paleo-suture of PSCS in south of Palawan and north Borneo [8]. We speculate that these two parts were connected at the beginning. According to the present tomographic model and other evidence, we sense that the PSCS subduction extended eastward to the Philippines. Combining the present results with previous study [8,21,51], we propose that the high velocity anomalies identified in MTZ is the eastward extension of the PSCS slabs and the PP plate. Zhao et al. suggested that on the time scale, the superposition of the subduction of PP plate and the subduction of the PSCS might play an important role for the opening of the SCS [54], which also supports the rationality of our interpretation. This view helps to determine the location of paleo-suture of PSCS-PP from Borneo to the Philippines. Lin et al. also mentioned the possibility that the PSCS extended to the Philippines [22].

4.3. Subduction of Philippine Sea and Sulu Sea

Our model reveals two high-velocity anomalies subducted along the Philippine Trench (black and purple dotted line circles in Figure 7a–i). One of them subducted eastward in the range of 11° N to 8° N (Figure 9e–i). Based on its horizontal distribution and previous studies [12,13], we interpret this high velocity anomaly as subducted slab of SS. The other high velocity anomalies are in the range of 14° N–8° N, which are interpreted as subduction of the PS plate. It is not obviously imaged in the range of 13° N–11° N, which might be result by magma upwelling and the melting of the subduction slab. The PS plate subducted westward south of 11° N, while north of 11° N it turns to subducted eastward. Its subduction depth is much greater in the southern part than in the northern part, with the maximum subduction depth up to 600 km. This indicates that the subduction of the PS plate along the Philippine trench occurs from the south to the north. The subducting slab of both the SS slab and the PS plate are above the PSCS slab, which demonstrates that the subduction of the PSCS predates the SS and the PS plate.

5. Conclusions

In this study, by applying FMTT program based on teleseismic travel time data derived from ISC receivers, we present a high-resolution 3-D tomographic model, which imaged detailed upper mantle structure beneath the central Philippines. Considering the results of teleseismic tomography, the main scientific problems in the study area are studied in detail.

The present tomographic model identifies high velocity anomalies beneath the Manila trench, which is interpreted as the subduction of SCS slab. The steeping and tearing of SCS slab are clearly imaged by our tomographic model. The subduction angle of the SCS slab changes from steep to gentle from north to south, and the depth of subduction decreases from the deepest 800 km to about 450 km. This indicates that the subduction at the eastern boundary of the SCS developed from the north to the south, and it can also be inferred that the SCS opened earlier in the north than in the south. Our tomographic model images high velocity anomalies representing the SCS slab between 13° N and 11° N, which is different from previous study. The most prominent high velocity anomalies revealed by the present tomographic model are in the MTZ between 14° N–8° N, which is interpreted as the PSCS slab and PP slab. It extends the location of the paleo-suture of PSCS-PP eastward from Borneo to the Philippines, which should be considered in the study of mechanism of the SCS and tectonic evolution in SE Asia. The PS plate and SS slab are also identified in the present tomographic model. The PS plate subducted at 14° N–8° N, which subducted westward to the north of 13° N and eastward to the south of 11° N. The subduction depth in the south is greater than that in the north. The deepest point is about 650 km below the MTZ. The disappearance of the PS subduction plate between 13° N and 11° N might be due to mantle melting and the SS slab subducted at 11° N–8° N.