Zhen Lin, Wen-chao Lü, Zi-ying Xu,*, Peng-bo Qin, Hui-qiang Yao, Xiao Xiao,Xin-he Zhang, Chu-peng Yang, Xiang-yu Zhang, Jia-le Chen

a Ministry of Natural Resources Key Laboratory of Marine Mineral Resources, Guangzhou Marine Geological Survey, Guangzhou 510075, China

b China Geological Survey, Beijing 100037, China

c Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou), Guangzhou 511458, China

Keywords:

Kyushu-Palau Ridge (KPR)

Characteristics of gravity and magnetic fields

Mindanao fault

Moho depth

Crust attribute

Philippine Sea Plate

Marine scientific survey

A B S T R AC T

The southern part of the Kyushu-Palau Ridge (KPR) is located at the conjunction of the West Philippine Basin, the Parece Vela Basin, the Palau Basin, and the Caroline Basin. This area has extremely complex structures and is critical for the research on the tectonic evolution of marginal seas in the Western Pacific Ocean. However, only few studies have been completed on the southern part, and the geophysical fields and deep structures in this part are not well understood. Given this, this study finely depicts the characteristics of the gravity and magnetic anomalies and extracts information on deep structures in the southern part of the KPR based on the gravity and magnetic data obtained from the 11th expedition of the deep-sea geological survey of the Western Pacific Ocean conducted by the Guangzhou Marine Geological Survey, China Geological Survey using the R/V Haiyangdizhi 6. Furthermore, with the data collected on the water depth, sediment thickness, and multichannel seismic transects as constraints, a 3D density model and Moho depths of the study area were obtained using 3D density inversion. The results are as follows.(1) The gravity and magnetic anomalies in the study area show distinct zoning and segmentation. In detail,the gravity and magnetic anomalies to the south of 11°N of the KPR transition from high-amplitude continuous linear positive anomalies into low-amplitude intermittent linear positive anomalies. In contrast,the gravity and magnetic anomalies to the north of 11°N of the KPR are discontinuous and show alternating positive and negative anomalies. These anomalies can be divided into four sections, of which the separation points correspond well to the locations of deep faults, thus, revealing different field-source attributes and tectonic genesis of the KPR. (2) The Moho depth in the basins in the study area is 6-12 km.The Moho depth in the southern part of KPR show segmentation. Specifically, the depth is 10-12 km to the north of 11°N, 12-14 km from 9.5°N to 11°N, 14-16 km from 8.5°N to 9.5°N, and 16-25 km in the Palau Islands. (3) The KPR is a remnant intra-oceanic arc with the oceanic-crust basement.which shows noticeably discontinuous from north to south in geological structure and is intersected by NEE-trending lithospheric-scale deep faults. With large and deep faults F3 and F1 (the Mindanao fault) as boundaries overall, the southern part of the KPR can be divided into three zones. In detail, the portion to the south of 8.5°N (F3) is a tectonically active zone, the KPR portion between 8.5°N and 11°N is a tectonically active transition zone, and the portion to the north of 11°N is a tectonically inactive zone. (4) The oceanic crust in the KPR is slightly thicker than that in the basins on both sides of the ridge, and it is inferred that the KPR formed from the thickening of the oceanic crust induced by the upwelling of deep magma in the process of rifting of remnant arcs during the Middle Oligocene. In addition, it is inferred that the thick oceanic crust under the Palau Islands is related to the constant upwelling of deep magma induced by the continuous northwestward subduction of the Caroline Plate toward the Palau Trench since the Late Oligocene. This study provides a scientific basis for systematically understanding the crustal attributes, deep structures, and evolution of the KPR.

1. Introduction

The Philippine Sea Plate is located at the intersection of the Eurasian Plate, the Pacific Plate, and the Indo-Australian Plate and is surrounded by subduction systems. Owing to its complicated tectonic environment, the tectonic evolution of the Philippine Sea Plate has always been an important topic in geoscience (Hall R et al., 1995; Li XJ et al., 2017; Zhang ZY et al., 2018; Zhang Y et al., 2020). On the east side of the plate, the Izu-Bonin-Mariana subduction zone has developed due to the westward subduction of the Pacific Plate. On the south side of the plate, the subduction zone of the Yap and Palau Trenches has developed due to the compression in the north-south direction caused by the Australian Plate. On the west side of the plate, the subduction zone of the Philippine and Manila trenches has developed due to the subduction of the Philippine Sea Plate and the Eurasian Plate in two directions. On the north side of the plate, the subduction zone of the Ryukyu Trench and the Nankai Trough has developed due to the subduction of the Philippine Sea Plate toward the Eurasian Plate. The landforms developing within the Philippine Sea Plate include the West Philippine Basin, the Palau Basin, the Oki-Daito Ridge, the KPR, the Shikoku Basin, the Parece Vela Basin, the Izu-Bonin-Mariana Island Arc, the Izu-Bonin Trench, and the Mariana Trench. The study area spans the Philippine Sea Plate and the Caroline Plate, but it is primarily located in the southern part of the Philippine Sea Plate. It is bounded by the Yap and Palau Trenches in the southeast, with the Caroline Plate lies to the south. It is bounded by the KPR in the middle and borders on the Parece Vela Basin in the east and the West Philippine Basin and Palau Basin in the west. The western two basins is bounded by the Mindanao fault (Fig. 1).

Fig. 1. Map showing the regional tectonic location of the Philippine Sea Plate and the location of the study area.

The KPR lies in the center of the Philippine Sea Plate, has a north-south strike, has a length of approximately 2600 km,and is the longest aseismic ridge in the world (Lallemand S,2016). The former KPR was an island arc formed from the westward subduction of the Pacific Plate during the Eocene-Oligocene. Afterward, the island arc cracked due to the eastward subduction and roll-back of the Pacific Plate. As a result, back-arc basins formed, such as the Parece Vela Basin and the Shikoku Basin. Thein-situresidual western portion of the island arc became the current KPR, while the eastern portion became the current Izu-Mariana island arc (Karig DE,1975; Nishizawa A et al., 2007).

Many studies have been previously conducted on the KPR, including its topography, landforms, characteristics of gravity and magnetism, crustal structure, sedimentary characteristics, rifting and detachment time, and tectonic evolution. The KPR is a ridge developing in a deep ocean basin and consists of a series of narrow and discontinuous seamount chains composed of hundreds of seamounts (Tang Y et al., 2011). Among them, only the Palau Islands and Okinotori Reef are exposed above the water, while the majorities are present at depths of more than 3000 m below the water. The seamount chains are asymmetric, being steep on the east side and gentle on the west side. The seamounts in the seamount chains are relatively independent, and the height difference between them and the deep basins on both sides is up to thousands of meters (Li JB et al., 2013). In terms of the characteristics of gravity anomalies, the KPR is generally characterized by low positive gravity anomalies that are horizontally narrow and vertically discontinuous, which correspond to the wavy characteristics of the topography (Wu SG et al., 2013; Dong DD et al., 2017; Zhang ZY et al.,2018). The KPR shows changing positive and negative magnetic anomalies overall, as well as a wide distribution of magnetic anomalous bodies that were formed during the basin expansion on both sides of the KPR (Zhang ZY et al., 2018).The crustal structure of the KPR changes significantly from north to south (Nishizawa A et al., 2007; Zhang J et al.,2012), which divided into three sections. The southern and northern sections of the KPR each show a three-layer crust consisting of upper, middle, and lower crust, while the middle section lacks the middle crust. The crustal thickness is over 10 km in the southern section, approximately 13 km in the northern section, and 7.2-20 km in the middle section. The upper, middle, and lower crust in the northern and southern sections of the KPR have similar crustal velocities, while the crustal velocity in the middle section greatly differs from that in the southern and northern sections (Zhang J et al., 2012).The middle section has a deep Moho, obvious mountain roots,an upper crust with thickness of 1-5 km, and a lower crust with a thickness of 5-15 km, showing the distinct crustal structure of a remnant arc (Zhang ZY et al., 2018).

The KPR possesses noticeable characteristics of oceanic crustal, with rifting of the remnant arc occurring at 25 Ma(Dong DD et al., 2017). As revealed by DSDP Site 448 in the KPR (Kroenke L, 1981), volcanic rocks with a thickness of 1.9 km developed before the Middle Oligocene, and were 2.5 km thick during the Middle Oligocene. The thickness increased by 600 m due to the accumulation of K-rich basalt flows and related breccia. The KPR is a remnant arc that was active during the Middle Oligocene and, thus, has the features of island arc tholeiites, which is proven by magnetic anomalies on the surface of the KPR.

Overall, despite many studies on the KPR, the research completed is still sporadic due to limited data obtained from surveys. In addition, the previous studies were mostly focused on the north of 15°N and rarely cover the portion from the south to 15°N of the KPR. This portion is at the intersection of the Philippine Plate, the Indo-Australian Plate, and the Pacific Plate and is an important area for studying the tectonic evolution of marginal seas in the Western Pacific Ocean. How has the northwestward subduction of the Caroline Plate, the rotation of the Philippine Sea Plate, and the westward subduction and eastward roll-back of the Pacific Plate created the current structure of the southern part of the KPR? What are the characteristics of the geophysical fields and deep structures of the KPR? Are geologic structures in the KPR continuous? The resolution of these scientific questions will help to understand the tectonic evolution of the marginal seas in the Western Pacific Ocean. Given this, this study aims to finely depict the gravity and magnetic anomaly characteristics in the southern part of the KPR and to reveal characteristics and genesis of deep and shallow structures in the KPR using data on gravity, magnetic, topography, and seismic transects obtained from geological and geophysical surveys. The purpose of this study is to provide a scientific basis for fully understanding the formation and evolution of deep structures in the KPR and marginal seas in the Western Pacific Ocean.

2. Sources and analysis of gravity and magnetic data

As shown in Fig. 1, the study area is in the scope of 129°E-136°30′ E and 6° N-13°30′ N and includes the Palau Islands, the southeastern part of the West Philippine Basin,the southwestern part of the Parece Vela Basin, and most areas of the Palau Basin. This study utilizes gravity data from Sandwell DT V29.1, magnetic data from the Earth Magnetic Anomaly Grid (2-arc-minute resolution; EMAG2), and gravity and magnetic data obtained from shipborne surveys.

The regional gravity data were obtained from the global marine gravity anomaly model V29.1 published by Sandwell DT et al. (2014, 2019) (https://www.earthbyte.org/category/resources/data-models/gravity), with a grid spacing of 1 arc minute. Lemenkova P (2020) and Bayley D (2019) used this model to conduct an interpretation of seabed landforms and to study the changes in marine communities near the Philippine Sea Basin and the KPR. The magnetic data were obtained from EMAG2 (https://geomag.colorado.edu/emag2-earth-magnetic-anomaly-grid-2-arc-minute-resolution.html). EMAG2 was compiled from satellite, marine, and aeromagnetic magnetic surveys, with a data size of more than 100 pieces(Maus S, 2009), an altitude of 4 km above mean sea level, and a resolution of 2 arc minutes. The effectiveness and accuracy of this model have been widely verified in the Philippine Sea Basin and the surrounding areas (Shin YH et al., 2012;Zubaidah T et al., 2014; Wu J et al., 2016).

The gravity and magnetic data obtained from shipborne surveys were acquired from the 11th expedition of the deepsea geological survey of the Western Pacific Ocean conducted by the Guangzhou Marine Geological Survey, China Geological Survey using the R/V Haiyangdizhi 6 from August to October 2021. For these data, the survey network was 12 km×45 km, and the precision of gravity and magnetic surveys were 0.47×10-5m/s2and 2.2 nT, respectively.

This paper assessed the accuracy of gravity data from Sandwell DT V29.1 and magnetic data from EMAG2 according to the gravity and magnetic data obtained from shipborne surveys. The gravity data was assessed using a point mass method. In detail, the boundary of the surveyed gravity data scope was extended externally by 0.5° to extract the gravity data of 1 ′×1 ′ grid cells. Then the differences between the surveyed gravity data and the gravity data from Sandwell DT V29.1 were calculated using the point mass method (Sunkel H, 1983; Wu XP, 1984; Wu X et al., 2009;Wang JQ et al., 2010; Huang JX et al., 2017). According to the calculation results, the mean difference and standard deviation between the two groups of gravity data were 0.948×10-5m/s2and 1.829×10-5m/s2, respectively. The results of comparison between both groups of gravity data of the study area are shown in Fig. 2, in which Figs. 2a, b show the planar comparative results and Fig. 2c shows the sectional comparative results. As shown in these figures, two groups of gravity data share roughly the same anomalous morphology,amplitude, and strike. Furthermore, slight differences occur in local anomaly peaks, where the surveyed data clearly highlights the subtle changes of anomalies. Therefore, the accuracy of the gravity data from Sandwell DT V29.1 fully meets the demand for small-scale research. The processing and interpretation software Oasis Montaj developed by the Canadian company Geosoft was used for upward continuation of the magnetic data obtained from the shipborne survey by 4 km, and then the surveyed magnetic data were compared with the magnetic data from EMAG2. Numerically, the values of the surveyed magnetic data after upward continuation are-71 - 78 nT, while the magnetic data EMAG2 are -95 - 108 nT.Therefore, the values of the two groups of magnetic data vary in the same order of magnitude. Moreover, the two groups of magnetic data have consistent shape and strike, which show high coincidence (Fig. 3). Therefore, magnetic data from EMAG2 also fully meet the demand for small-scale regional geological research.

Fig. 2. Diagrams showing the comparison results of gravity data. a-gravity data from Sandwell DT V29.1, where the red box denotes the scope of shipborne surveys area; b-data outside the red box are gravity data from Sandwell DT V29.1, while the data inside the red box are the gravity data from shipborne survey (the two groups of data were stitched and then gridded); c-comparison between gravity data from Sandwell DT V29.1 and surveyed gravity data along section AA′ in the study area.

Fig. 3. Diagrams showing the comparison results of magnetic data. a-magnetic data from EMAG2, where the red box denotes the scope of shipborne surveys area; b-data outside the red box are magnetic data from EMAG2, while the data in the red box are shipborne surveyed magnetic data after 4 km-upward continuation to (the two groups of data are not sutured); c-comparison between magnetic data from EMAG2 and shipborne surveyed magnetic data along section AA′ in the study area.

3. Characteristics of gravity and magnetic fields

3.1. Characteristics of gravity anomalies

The gravity anomaly values in the study area are-205×10-5m/s2-250×10-5m/s2(Fig. 4). Among them, the anomaly maximum is approximately 250×10-5m/s2,occurring in the Palau Islands, and the anomaly minimum is approximately -205×10-5m/s2, occurring in the Palau Trench. Overall, the gravity anomalies change gently in oceanic basins but change dramatically and show different strikes in oceanic ridges and trenches. As analyzed from the characteristics of the strike, amplitude, gradient, and shape of the gravity anomalies, as well as the upward continuation of gravity anomalies to different heights (Figs. 5a-d), the gravity field in the study area is characterized by zoning and can be roughly divided into four gently-changing anomalous zones and one dramatically-changing anomalous zone (Fig. 4).

Fig. 4. Free-air gravity anomalies and zoning. I-gently changing anomalous zone in the West Philippine Basin; II-gently changing anomalous zone in the Parece Vela Basin; III-gently changing anomalous zone in the Palau-Parece Vela Basin; IV-anomalous zone in the Caroline Plate;V-dramatically changing anomalous zone in the Palau.

Gently changing anomalous zone in the West Philippine Basin (I). This gravity anomalous zone is dominated by positive anomalies, with amplitudes of 10×10-5m/s2-20×10-5m/s2and anomaly values gradually rising from east to west. Local anomalies are mostly banded in shape, and the anomalies in this zone have an overall NNW strike. A series of closed contours of negative anomalies in nearly NS strike occur at the junction of the east side of the zone and the KPR.Meanwhile, an arc linear gravity high belt with the strike turning from EW to NE lies on the south side of the anomalous zone. This high-gravity belt starts in the Philippine Trench, passes through the KPR toward the east, and ends in the southwestern part of the Parece Vela Basin, with geographical coordinates of roughly 128°E -135°30 ′E and 8°N -11°30 ′N. After upward continuation of the gravity anomalies to 15 km (Fig. 5c), the high-gravity belt shows large differences in anomaly values between its north and south sides, exhibiting a noticeable linear gradient belt. This indicates that there is a deep fault, which is the fault boundary between the West Philippine Basin and the Palau Basin.

Fig. 5. Distribution of free-air gravity anomalies and faults. a-free-air gravity anomalies continued upward 5 km and faults; b-free-air gravity anomalies continued upward 10 km and faults; c-free-air gravity anomalies continued upward 15 km and faults; d-free-air gravity anomalies continued upward 20 km and faults. The white lines denote faults.

Gently changing anomalous zone in the Parece Vela Basin(II). This gravity anomalous zone is also dominated by positive anomalies, with general amplitude of 10×10-5m/s2.They have a NNE strike and have no noticeable local anomaly morphology. A series of discontinuous closed contours of anomalies occur on the west ridge of the zone, with anomaly values of -30×10-5m/s2-71×10-5m/s2. They are composed of alternating positive and negative anomalies and have a NE strike.

Gently changing anomalous zone in the Palau - Parece Vela Basin (III). This gravity anomalous zone is also dominated by positive anomalies, with a general amplitude of 10×10-5m/s2-40×10-5m/s2. This zone is divided into the western and eastern parts with the KPR as the boundary. The western part of this anomalous zone is the Palau Basin. The gravity anomalies in this part change gently, and their amplitude is relatively high and increases from east to west,with a maximum amplitude of more than 60×10-5m/s2. The eastern part is the Parece Vela Basin. The anomalies in this part change gently and have a NEE strike, with an amplitude of approximately 20×10-5m/s2. Intermittent closed contours of positive anomalies in a nearly NS strike occur along the oceanic ridge.

Anomalous zone in the Caroline Plate (IV). This gravity anomalous zone consists of positive anomalies and high-value closed contours of anomalies near trenches.

Drastically changing gravity anomalous zone in the Palau(V). This gravity anomalous zone consists of positive and negative anomalies that have a staggered distribution and change dramatically, with anomaly values of -205×10-5m/s2-250×10-5m/s2. The anomalies in this zone are mostly blocky or banded in shape. They show unstable strike dominated by NE, NEE, and nearly NS directions. The maximum and minimum gravity anomalies in the study area are located in the Palau Islands and the Palau Trench,respectively. A complete high-amplitude linear positive anomalous zone in nearly NS strike occurs in the Palau Islands, and a linear negative anomaly belt associated with that in Palau Islands exists in the Palau Trench. It is inferred that such association phenomenon was caused by the gravitational effects on the boundary.

3.2. Characteristics of magnetic anomalies

The study area is located in a low-latitude region. Owing to the effects of oblique magnetization, the locations of magnetic bodies reflected by the shape of magnetic anomalies vertically deviate from their real locations to a certain extent.However, since this study is a small-scale regional geological study, the deviation degree does not affect the comprehensive geological and geophysical understanding of the study area.

The magnetic anomalies have values of -210 nT-175 nT and different strikes, and different zones show different characteristics of magnetic anomalies (Fig. 6). As identified according to signs, shapes, amplitudes, gradient changes, and strikes of anomalies, the magnetic field in the study area features zoning and can be roughly divided into two banded magnetic anomalous zones, one magnetic quiet anomalous zone, and one alternating positive and negative magnetic anomalous zone.

Fig. 6. Magnetic anomalies and zoning. I-banded magnetic anomalous zone in the West Philippine Basin; II-banded magnetic anomalous zone in the Parece Vela Basin; III- magnetic quiet anomalous zone in north Palau; IV-positive and negative magnetic anomalous zone in south Palau.

Banded magnetic anomalous zone in the West Philippine Basin (I). With an amplitude of 50-170 nT, the magnetic anomalies in this zone have a NW strike and a banded shape.Meanwhile, positive and negative anomalies are alternately distributed, showing the typical characteristics of oceanic crust magnetic lines and reflecting different expanding directions and speeds of oceanic basins. The oceanic basins expanded during 50-30 Ma (Deschamps A and Lallemand S,2002).

Banded magnetic anomalous zone in the Parece Vela Basin (II). As indicated by surveyed magnetic data, the magnetic anomalies in this magnetic anomalous zone have a nearly NS strike and are distributed in a banded pattern, with a maximum amplitude of less than 100 nT. Owing to the limited scope of the study area, this zone does not show noticeable characteristics of oceanic-crust magnetic lines. The crustal age is 24 Ma (Scott R et al., 1981), corresponding to magnetic lineation No. 8 (Mrozowski C and Hayes E, 1979).Positive and negative magnetic anomalies are alternately distributed along the oceanic ridge in the western part of this area, with surveyed magnetic anomaly values of -150-110 nT.

Magnetic quiet anomalous zone in north Palau (III). This magnetic anomalous zone roughly corresponds to the gently changing gravity anomalous zone in the Palau-Parece Vela Basin (III), with weak magnetic field. The magnetic anomalies in this zone are -50 - 50 nT, showing a small anomaly gradient that gently changes. They have a NE strike.As a relatively magnetic quiet zone overall, this zone indicates weak-magnetism sources of the magnetic fields.

Positive and negative magnetic anomalous zone in south Palau (IV). This magnetic anomalous zone roughly corresponds to the dramatically changing gravity anomalous zone in the Palau (V). The magnetic anomalies in this zone are distributed in a disordered and saltatory manner, with positive and negative anomalies alternately distributed.Meanwhile, they have no stable strike. All these indicate that this zone features frequent magmatic activities and strong magnetism, which is closely related to tectonic activity such as the collision, subduction, compression, and rotation of the Philippine Plate, the Indo-Australian Plate, and the Caroline Plate.

4. Inversion of the Moho depth

Welford JK et al. (2007, 2010), Geng M et al. (2019) and Li HL et al. (2020) used 3D density inversion to study the characteristics of regional deep structures. With the data on water depth and sediment thickness as constraints, this study built the 3D density model of the study area using 3D density inversion. Based on this, the Moho depth in the study area was obtained. The water depth data were sourced from the 15 arc-second grid dataset of the global bathymetric product released by the General Bathymetric Chart of the Oceans(GEBCO_2020 Grid; https://www.gebco.net/data_and_products/gridded_bathymetry_data/gebco_2020/). Meanwhile,the sediment-thickness data originated from the global five arc-minute total sediment thickness grid distributed by the National Geophysical Data Center (NGDC) (Whittaker JM et al., 2013; https://www.ngdc.noaa.gov/mgg/sedthick/#version2).

4.1. Inversion method

The gravity inversion is a linear problem. LetAdenote the sensitivity matrix,mrepresent the underground density distribution, andddenote the data. Then, the gravity inversion problem is expressed as follows:

where, φsrepresents the stabilizing function;mirepresents an element inm;landurepresent the lower and upper limits of an element inm, respectively, and α represents the regularization parameter, the values of which are used to balance the effects of the misfit function and the stabilizing function on the regularized equation. A high α value indicates that the stabilizing function produces greater effects on the regularized equation than the misfit function, while a lowα value indicates that the misfit function produces greater effects on the regularized equation than the stabilizing function. Oldenburg DW (1984) suggested that 1-norm or 2-norm of the model parameters be used for a model that gently changes. According to the practical condition, this study selected 2-norm of the model parameters as the misfit function, which is expressed as follows:

Since the gravity anomalies attenuate with an increase in distance, the inversion results tend to concentrate near the surface (Li Y, 1996, 1998). To eliminate such effect, it is necessary to introduce the depth weighting functionW. This study selected the sensitivity-matrix-based weighting function(Portniaguine O et al., 1999, 2002; Li Y et al., 2000), which is expressed as follows:

where,wjis an element of the depth weighting functionW;Aijrepresents an element in the sensitivity matrixA;irepresents the number of the observation point;jrepresents the number of the underground grid, and β is a constant used to control the strength of the weighting function, which isβ=2 for gravity inversion in general. Substituting the depth weighting function into the objective function will y:

This paper solved the equation using the conjugate gradient algorithm, which requires small memory and is suitable for solving large-scale operations (Pilkington M,1997; Nocedal J and Wright SJ, 2006). The key step in the algorithm is to calculate the gradient of the objective function,which is as follows in this study:

The process for the conjugate gradient algorithm is as follows:

(i) Set the maximum iterationsite_num,j=0, and the initial modelm0. Then calculatemw(0)=W-1m0and

(ii) For the first iteration, set the search direction toL0=-I0;

(iii) Calculate step lengthk:

(iv) Calculatemw(j+1)=mw(j)+kLj,mj+1=Wmw(j+1), and then

(v) End whenris small enough, or whenj＞ite_num;otherwise, setj=j+1, and go to (ⅵ);

In the process of the iterative calculation, a penalty function was used to constrain the resultmobtained from each iteration, to prevent values of model parameters from exceeding the scope (Portniaguine O et al., 1999). Given thatmirepresents a parameter inmand the maximum and minimum of density areandl, respectively, the expression of the penalty function is:

Regularization parameter α is critical for an inversive calculation (Zhdanov MS, 2002; Zhang L et al., 2012). Set α0=0if the initial model is 0; otherwise calculate α0using the following formula:

αgradually decreases in the subsequent iterations, and its expression is:

4.2. Parameter setting

The data coordinates were projected to the spatial coordinate system based on the Mercator projection, and then data on gravity, water depth, and sedimentary layers were gridded, with a grid spacing of 8.4 km×8.4 km and grid number of 100×100. The underground portion was divided into cuboids with a size of 8.4 km×8.4 km×0.5 km. In this way, the study area was divided into a total of 0.8×106small cuboids. Referring to Li HL et al. (2020), the initial model consisted of three layers, namely the seawater, sediment, and crust-mantle layers, and the background density was set at 2.9 g/cm3. The seawater layer was set based on the water depth data, and the density difference of this layer was set at-1.87 g/cm3. That is, the seawater density was 1.03 g/cm3.Afterward, the density difference of the sedimentary layer was set at -1.3 g/cm3. That is, the density of the sedimentary layer was 1.6 g/cm3. Moreover, the density difference of the crustmantle layer was uniformly set at -0.2 g/cm3. Based on these parameters, a 3D density model was obtained using the above-mentioned inversion method (Fig. 7).

The Moho is the boundary between Earth’s crust and mantle, which was first discovered as a discontinuity in the seismic velocity. Owing to the inherent relation between the seismic wave velocity and density, the Moho is also a density discontinuity. Welford JK et al. (2007, 2010) and Li HL et al.(2020) specified an appropriate isopycnic surface in the density anomaly model to represent the Moho. In this study,the Moho was defined as an isopycnic surface with a density difference of 0.2 g/cm3(density=3.10 g/cm3), of which the extracted depth was the Moho depth (Fig. 8).

Fig. 8. The Moho depth (black lines denote faults, and the pink line represents the position of section BB′ in Fig. 9).

4.3. Characteristics of the Moho

The Moho in the study area generally rises from the KPR to both oceanic basins (Fig. 8). The Moho in the West Philippine Basin is at a depth of 6-8 km. It is generally shallow in the west and deep in the east and is up to 7.5-8.5 km near the KPR in the east. The Moho in the Palau Basin is also shallow in the west and deep in the east. It is 7-10 km in the western part of the basin and is relatively stable. The Moho in the Parece Vela Basin is at a depth of 7-12 km and is deep in the west, thick in the east and south, and thin in the north. It is up to a maximum of approximately 12 km to the south of fault F2.

The Moho in the KPR rises from south to north overall.The Moho in the portion from the Palau Islands to fault F3 northward (8.5°N) along the KPR (section BB´ in Fig. 9;corresponding to the section from Palau to the position at a distance of 200 km on section BB´) is at a depth of 16-25 km.It is up to a maximum depth of 25 km under the Palau Islands.The Moho from fault F3 to fault F2 northward along the KPR(8.5°N-9.5°N; corresponding to the section at a distance of 200-310 km on section BB´) is at a depth of 14-16 km. The Moho in the portion from fault F2 to fault F1 (9.5°N-11°N;corresponding to the section at a distance of 310-440 km on section BB´) is at a depth of 12-14 km. The Moho to the north of fault F1 is at a depth of 10-12 km. Overall, with F3 and F1 as boundaries, the Moho rapidly rises to the south of F3, steadily rises between F3 and F1, and remains relatively stable overall to the north of F1.

Fig. 9. The Moho depth in the KPR from south to north.

5. Discussion

5.1. Gravity and magnetic anomaly response of faults

To obtain data on deep structures in the study area, the gravity and magnetic data were processed through upward continuation, vertical derivative, and Euler deconvolution method using two pieces of software, namely FUGRO-LCT and Oasis Montaj. FUGRO-LCT is a gravity, magnetic, and seismic comprehensive processing and interpretation system developed by the American company FUGRO, and Oasis Montaj is the processing and interpretation software developed by the Canadian company Geosoft. The results identified nine large and deep faults F1-F9 in the study area according to the analysis of the characteristics of gravity and magnetic fields, the characteristics of the Moho, and the processing results of gravity and magnetic data. Among them,four large and deep faults are described and discussed below,including the F1, F2 and F3 in a NE-NEE strike and F8 in a nearly NS strike.

Fault F1—the Mindanao fault. This fault starts from the Philippine Trench near the Mindanao Islands in the west and extends to the Parece Vela Basin in the east. The western part of F1 is in a nearly EW strike, the middle part is in a NE strike, and the eastern part enters the Parece Vela Basin and is in a NEE strike. F1 extends for approximately 910 km in the study area and is the boundary between the West Philippine Basin and the Palau Basin. The relationship between F1 and the gravity and magnetic anomaly response is as follows:

(i) Both sides of F1 show different characteristics of gravity fields in deep. As shown in the second vertical derivative map of gravity anomalies upward continued to 10 km (Fig. 10a), F1 shows noticeably linear gravity highs,which correspond to the boundaries between gravity anomalous areas. The north side of F1 is characterized by low-amplitude and gentle gravity anomalies, while the south side of F1 features high-amplitude gravity anomalies. After the gravity anomalies are upward continued to 20 km (Fig.5d), the differences in gravity anomalies between the north and south sides of F1 are clearer, indicating that F1 is a large and deep fault that at least cuts through the Moho. The differences are more noticeable at the position of the KPR. In detail, the north side of F1 in this position shows a series of discontinuous closed contours of anomalies, with positive and negative anomalies alternately distributed, while the south side presents an intermittent linear gravity high zone.

Fig. 10. Maps showing the distribution of faults in the study area. a-second vertical derivative of gravity anomalies upward continued to 10 km and the faults; b-second vertical derivative of gravity anomalies upward continued to 15 km and the faults; c-inversion results of gravity anomalies using the Euler deconvolution method and the faults (black crosses represent Euler deconvolution solutions of depths more than 5 km);d-first vertical derivative of EMAG2 magnetic anomalies upward continued to 5 km and the faults (black crosses represent Euler deconvolution solutions of depths more than 5 km, and white and red lines represent faults).

(ii) Both sides of F1 show noticeably different characteristics of magnetic fields. As shown in the first vertical derivative map of magnetic anomalies upward continued to 5 km (Fig. 10d), magnetic anomalies in both NS and EW strikes differ with F1 as the boundary. To the north of F1, the West Philippine Basin shows obviously strong banded magnetic anomalies in a NW strike, while the Parece Vela Basin shows weak banded magnetic anomalies in a nearly NS strike. A weak magnetic anomaly area with a NE strike exists to the south of F1 and it is a magnetic quiet zone overall. The magnetic field to the south of the magnetic quiet zone becomes strong and shows disordered cloddy magnetic anomalies with unstable strikes. Furthermore, the north and south sides of F1 in the position of the KPR show noticeably different magnetic anomalies. In detail, the north side shows high-amplitude strong magnetic anomalies (Figs. 6, 11b),while the south side shows low-amplitude weak magnetic anomalies (Fig. 6).

(iii) Both sides of F1 show noticeably different depths of the Moho. According to Fig. 8, to the north of F1, the Moho in the West Philippine Basin is shallow (depth: ＜ 10 km) and that on the east side of the Palau Basin is deep (depth: ＞ 10 km).The Moho in the position of the KPR is shallow in the north and deep in the south.

(vi) As indicated by results of Euler deconvolution method, vertical vein rocks are highly developed in the deep parts of F1. Fig. 10c shows the results of Euler deconvolution of gravity anomalies with a structural index of 1 (the calculated results with a depth of less 5 km are excluded),indicating strong magmatic activities in F1. The oceanic basins on the north and south sides of F1 show different ages.The West Philippine Basin on the north side was formed during the Eocene, with an age of 49-33 Ma (Ozima M et al.,1983; Karig DE, 1975). In contrast, the oceanic crust of the Palau Basin on the south side of F1 is noticeably older than the West Philippine Basin, with the ages of granites and metamorphic rocks being 125-118 Ma (Hall R et a1, 1995;Malyarenko AN and Lelikoy EP, 1995). Meanwhile, midocean ridge basalts have developed in the Palau Basin.

The second vertical derivative results of gravity anomalies upward continued to 10 km and 15 km still show clear information on deep structures of F1 (Figs. 10a, b). Therefore,the Mindanao fault is a lithospheric-scale deep fault, of which the western section is the boundary between the West Philippine Basin and the Palau Basin, the middle section cuts off the KPR from its shallow to deep parts, and the eastern section is the inner boundary of the Palau Basin.

Fault F2. This fault is located near 9.5°N and on the southwestern end of the Parece Vela Basin. It extends for approximately 230 km in the study area, with the position corresponding to the eastern section of the boundary between the gently changing anomalous zone in the Parece Vela Basin(II) and the gently changing gravitational-anomalous zone in the Palau-Parece Vela Basin (III). As shown in the second vertical derivative map of gravity anomalies upward continued to 10 km and 15 km, F2 shows linear gravity highs in the NEE strike. Moreover, the gravity and magnetic anomalies on both sides of F2 are noticeably different.Specifically, the north side of F2 shows gentle gravity highs and gentle magnetic anomalies in the NE-NEE strike. In contrast, the south side of F1 exhibits disordered and dramatically changing cloddy gravity and magnetic anomalies with unstable strikes. The results of inversion using the Euler deconvolution method indicate that the vertical vein rocks are highly developed alone the deep part of F1 (Fig. 10c).

Fault F3. This NEE-strike fault is located near 8.5°N and on the northernmost end of the Palau Trench. It extends eastward to the southernmost end of the Parece Vela Basin and extends for approximately 260 km in the study area. As shown in the second vertical derivative maps of gravity anomalies upward continued to 10 km and 15 km, F3 exhibits noticeable NEE-strike linear gravity highs (Figs. 10a, b).Furthermore, the gravity and magnetic anomalies on both sides of F3 are noticeably different. Specifically, the north side of F3 shows disordered and cloddy anomalies with unstable strikes, while the south side of F3 generally shows banded gravity anomalies and NW-strike magnetic anomalies due to the effects of the Palau and Yap subduction zones.Moreover, the results of inversion using the Euler deconvolution method indicate that the vertical vein rocks are highly developed in the deep part of F3 (Fig. 10c). Since both the Palau and Yap trenches extend northward until F3, F3 is considered an important boundary fault.

The second vertical derivative results of gravity anomalies upward continued to 10 km and 15 km still show clear characteristics of F3 (Figs. 10a, b). Therefore, F2 and F3 are lithospheric-scale deep faults and cut off the Moho of the KPR. Moreover, F2 is a fault inside the Parece Vela Basin and F3 is an important boundary fault.

Faults F8-1, F8-2, and F8-3. These faults serve as the boundary faults of the West Philippine Basin, the Palau Basin,and the Parece Vela Basin. They are in nearly NS strike and are dislocated by NEE-strike faults (Fig. 10). The gravity anomalies show noticeable segmentation along these faults.Specifically, they transition from continuous and highamplitude linear anomalies to saltatory, discontinuous, and relatively-low-amplitude anomalies from south to north.Based on this, they can be divided into four sections, which correspond to four faults. (1) The section to the south of F3 corresponds to the fault on the west side of the Palau Islands,showing high-amplitude and continuous linear gravity highs.(2) The section between F2 and F3 corresponds to fault F8-1,exhibiting continuous and medium-high-amplitude linear gravity highs. (3) The section between F1 and F2 corresponds to fault F8-2, showing intermittent and medium-low amplitude gravity highs. (4) The section to the north of the Mindanao fault (F1) corresponds to fault F8-3. The gravity anomalies in this section are in the NE strike, with positive and negative anomalies alternately distributed and local gravity highs appearing as a series of discontinuous closed contours of anomalies (Fig. 11a). The characteristics of these gravity anomalies coincide with the discontinuity of terrain revealed by the multibeam bathymetric data (Fig. 11c). Since the KPR experienced rifting at approximately 25 Ma (Dong DD et al., 2017), it is referred that the seamount chain in the study area was formed prior to at least 25 Ma. Therefore, the current seamount chain was formed from the upwelling of deep magma along early faults. The Bouguer anomaly along F8-1, F8-2, and F8-3 also shows discontinuous closed contours of anomalies (Fig. 11d). Similar to the gravity anomalies, the magnetic anomalies along these faults are also noticeably different. Specifically, the section to the north of the Mindanao fault (F1) shows alternately distributed positive and negative NEE-strike anomalies, the section from F1 to northern F3 exhibits gently changing NE-strike anomalies,and the section to the south of F3 shows dramatically changing magnetic anomalies. The segmentation of gravity anomalies and the differences in the magnetic field characteristics reveal the discontinuity of these NS-strike faults and reflect the discontinuity of geologic structures in the southern part of the KPR. The second vertical derivative results of gravity anomalies upward continued to 10 km and 15 km still show clear characteristics of the deep structures of F8-1, F8-2, and F8-3 (Figs. 10a, b). Therefore, these faults are considered lithospheric-scale deep faults.

Fig. 11. Distribution of faults in the northern KPR in the study area. a-distribution of shipborne surveyed gravity anomalies and faults; b-distribution of shipborne surveyed magnetic anomalies and faults; c-distribution of terrain and faults; d-distribution of shipborne surveyed bouguer anomaly and faults. Pink lines represent faults.

In addition, the gravity anomalies upward continued to 10 km show weak signals representing fault characteristics indicated by gravity anomalies in the positions of NEE-strike faults F4, F5, F6, and F7 and NW-strike F9 (Fig. 5b). Given that the crust on the north side of the KPR in the study area is 6-8 km thick, F4, F5, F6, F7, and F9 are crustal-scale deep faults.

5.2. Deep structure and genesis of the KPR

According to the Moho depth and crustal thickness (Figs. 8,12) obtained from gravity inversion, the crust in the KPR generally thins in a stepped manner from south to north and locally thins near 11°N (F1), where the crustal thickness is less than 6 km (Figs. 12, 13) and the direction of crustal extension is transformed from nearly NS into NNW (Fig. 12).Furthermore, the crustal structure in the KPR is discontinuous and shows noticeable segmentation. With F3 and F1 as boundaries, the southern part of the KPR can be divided into three sections, namely the tectonically active zone, the tectonically active transition zone, and the tectonically inactive zone (Fig. 8).

Fig. 12. Thickness of the crusts in the study area (the red line represents the position of section CC′).

The crust in the KPR is relatively thick below the Palau Island from south to north, with a thickness of approximately 24 km. This is related to the northwestward continuous subduction of the Caroline Plate toward the Palau and Yap trenches. The continuous NW-direction subduction of the Caroline Plate toward the Philippine Plate since the Late Oligocene caused the deep magma below the Palau Islands to continuously upwell and the crust to continuously thicken.The subduction-induced magmatic activities gradually weakened from south to north, causing the crust in the KPR to thin in a stepped manner from south to north. The constant magmatic activities in the late stage also induced highamplitude gravity anomalies and disordered high-amplitude magnetic anomalies around the Palau Island. From the Palau Islands to the north, the gravity and magnetic anomalies gradually weakened as magmatic activities weakened.Overall, the southern part of the KPR shows significantly different tectonic activity in the NS direction. In the KPR portion to the south of 8.5°N (fault F3), the crust is approximately 14-24 km thick and the Moho depth reaches a maximum of 24 km below the Palau Islands. Therefore, this KPR portion is a tectonically active zone. The crust is approximately 8-14 km thick in the KPR portion between 8.5°N and 11°N (faults F3-F1), which is, therefore, a tectonically active transition zone. Moreover, the crust is approximately 6-8 km thick in the portion to the north of 11°N (fault F1), which is, therefore, a tectonically inactive zone.

Seamounts are highly developed in the KPR. Their sedimentary characteristics are noticeably different before and after the KPR rifting, with distinct angular unconformity between the sedimentary strata formed before and after the KPR rifting. Furthermore, the sediments developing after the KPR rifting underwent little disturbance. In contrast, the sedimentary strata that formed before the KPR rifting show noticeable magma intrusion along the bedding planes, with magma vertically developing upward along volcanic rock masses (Fig. 14b). The sedimentary basement in the KPR is present as a layer of disordered high-amplitude reflectors,under which no seismic sequences have been identified.Furthermore, oblique reflectors have developed inside the crust, which is approximately 5 km thick. The basin on both sides of the KPR shows noticeable typical characteristics of seafloor spreading. It indicated that the crust in the KPR has obvious properties of oceanic crust (Fig. 14a).The crustal characteristics of the KPR are similar to the structural characteristics of the oceanic crust in the sub-basin in the eastern part of the South China Sea. Specifically, the oceanic crust of the sub-basin is approximately 5-8 km thick (Xu ZY et al., 2020) and shows noticeable oblique reflectors (Ding WW et al., 2018). Moreover, the oceanic crust below the Zhenbei-Huangyan seamount chain is approximately 10-13 km thick, which is thicker than the crust in the sub-basin on both sides of the seamount chain, showing noticeable mountain-root effects (Zhao et al., 2018). Therefore, the crust in the KPR has obvious properties of oceanic crust (Fig. 14a).During 50-40 Ma, the Pacific Plate underwent subduction toward the Philippine Plate in the west, leading to the formation of the ancient Izu-Mariana island arc. Afterward,with the eastward roll-back of the Pacific Ocean, the ancient Izu-Mariana island arc underwent back-arc spreading and rifting at approximately 25-22 Ma, and the remnant arc on the west side is the current KPR (Hall R et al.,1995; Nishizawa A et al., 2007; Fang Y et al., 2011). Based on this reasoning combined with the crustal characteristics on the multichannel seismic transects, the KPR is considered to have the tectonic background of a volcanic arc and is a remnant intra-oceanic arc with the oceanic-crust basement (Fang Y et al., 2011; Li JB et al., 2013). The water depth in the northern part of the KPR varies in the range of 3000-4500 m (Fig. 11c),indicating that the KPR is a deep oceanic ridge.

Fig. 13. Section spanning the crustal structure of the KPR (the survey line is shown in see Fig. 12).

Fig. 14. Characteristics of seismic transects of the KPR (see L1 and L2 in Fig. 1 for section locations; modified from Dong DD et al., 2017).

According to the crustal thickness section (Fig. 13), the crustal thickness in the KPR is 6-7 km and noticeable characteristics of mountain roots are present below the KPR.Meanwhile, the crustal thickness in the section coincides with that calculated based on multichannel seismic transects, thus proving the reliability of the crustal thickness inversion results obtained in the study. The crustal thickness is approximately 4-5 km in the Parece Vela Basin and is approximately 2-4 km in the West Philippine Basin. Both are slightly less than the crustal thickness in the KPR. Deep seismic reflection sections show that a deep normal fault that inclines eastward developed in the KPR before the KPR rifting (Fig. 14b). It is inferred that this fault formed via the action of tensile stress produced during the rifting of the remnant arc. Frequent volcanic activities occurred in the KPR before and after the Middle Oligocene, leading to the development of volcanic rock masses with a thickness of approximately 5 km (Kroenke L, 1981). Therefore, the crust in the KPR is slightly thicker than that of basins on both sides possibly because of the oceanic crust thickening, which was induced by KPR rifting and corresponding upwelling of deep magma before and after the Middle Oligocene.

6. Conclusions

This study finely depicts the gravity and magnetic anomaly characteristics in the KPR and identifies the spatial distribution of large and deep faults in the ridge based on the interpretation of gravity and magnetic anomaly characteristics. Furthermore, this study reveals the deep structural characteristics and crustal attributes in the KPR and explores their genesis according to the inverted crustal thickness and Moho depth combined with multichannel seismic transects. The following understanding is achieved.

(i) The KPR is a deep oceanic ridge. The geologic structures in the KPR are noticeably discontinuous in the NS direction, with NS-strike faults dislocated by NEE-strike deep faults.

(ii) The gravity and magnetic anomalies in the southern part of the KPR show distinct zoning and segmentation,revealing different field-source attributes and tectonic genesis.The KPR is divided into four sections from south to north,with each section showing noticeably different gravity and magnetic anomalies. The section to the south of 8.5°N shows NNE-strike, high-amplitude, and linear positive gravity anomalies and high-amplitude and variable positive and negative magnetic anomalies. The section between 8.5°N-9.5°N (F2-F3) shows medium-amplitude and nearly NS-strike linear positive gravity anomalies and variable magnetic anomalies. The section between 9.5°N-11°N (F1-F2) shows discontinuous gravity highs and gentle magnetic anomalies. The section to the north of 11°N (F1) shows both gravity and magnetic anomalies with alternating positive and negative anomalies in a NE or NEE strike. These sections correspond well to deep faults in terms of locations. The West Philippine Basin on the west side and the Parece Vela Basin on the east side of the KPR show noticeable characteristics of NS-strike magnetic lines and NW-strike magnetic lines,respectively, reflecting typical characteristics of seafloor spreading.

(iii) The Moho depth and crustal thickness in the southern part of the KPR are both discontinuous and are divided into three sections overall with deep faults F3 and F1 as boundaries. The Moho in the KPR rapidly rises from south to north, with the depth rapidly decreasing from 25 km in the remnant arc in the Palau Islands to 12-10 km. The Moho depth contours and crustal thickness isolines are twisted in the position of fault F1 and are in a NNW strike in the north and a nearly-NS strike in the south.

(vi) The southern part of the KPR is a remnant intra-oceanic arc with oceanic crust basement. The tectonic activities in this part are noticeably different in the NS direction. Based on this, the southern part is divided into three zones. In detail, the portion to the south of 8.5°N (F3) is a tectonically active zone, the KPR portion between 8.5°N and 11°N is a tectonically active transition zone, and the portion to the north of 11°N is a tectonically inactive zone. The formation of these zones is inferred to correlate with the gradual weakening of magmatic activities from south to north resulting from the continuous NW-direction subduction of the Caroline Plate toward the Philippine Plate.

(v) The oceanic crust of the KPR is slightly thicker than that in the basins on both sides of the KPR, which is related to the thickening of oceanic crust resulting from the upwelling of deep magma in the process of remnant arc rifting during the Middle Oligocene. The thick oceanic crust under the Palau Islands is related to continuous thickening caused by the constant upwelling of deep magma, which results from the continuous NW- strike subduction of the Caroline Plate toward the Palau Trench since the Late Oligocene.

CRediT authorship contribution statement

Zhen Lin and Wen-chao Lü conceived the presented idea.Zhen Lin prepared the manuscript. Peng-bo Qin processed the gravity inversion. Zi-ying Xu helped complete the idea. Huiqiang Yao, Xiao Xiao, Zhang Xin-he, Chu-peng Yang, Xiangyu Zhang, Jia-le Chen reviewed and edited the manuscript.All authors discussed the results and contributed to the final manuscript.

Declaration of competing interest

The authors declare no conflicts of interest.

Acknowledgment

This work was supported by the project of China Geological Survey (DD20191002), the Key Special Project for Introduced Talents Team of Southern Marine Science and Engineering Guangdong Laboratory (Guangzhou)(GML2019ZD0208), the National Natural Science Foundation of China (41606080, 41576068). The authors acknowledge the source of THE GEBCO Grid. The authors would like to thank Prof. Dong-dong Dong for his help to provide the seismic profile. The authors also want to thank the reviewers and editors for their contributions to the improvement of this paper.