Figure 1. (a) Tectonic setting of the SCB. The different colours in the background indicate the different periods of rock outcrops distributed throughout the SCB. The thick solid lines represent the convergence boundary and deep faults between blocks: GHF: Gaoyao–Huilai fault; ZDF: Zhenghe–Dapu fault; ALF: Anhua–Luocheng fault; CNF: Changle–Nanao fault; JSF: Jiangshao-Shaoxing fault. The thick dashed lines are the inferred boundaries between blocks. The thin black lines are major faults in mainland China, while the blue circles denote Mw ≥ 3.0 earthquakes that occurred from 1980 to 2021 downloaded from the Incorporated Research Institutions for Seismology (IRIS) website (http://ds.iris.edu/ieb/). The gold triangles represent the node geophones deployed along the Zhuhai–Lianzhou profile. (b) Tectonic setting of Southeast Asia. The equilateral blue rectangle denotes the main research area in panel (a). NCB: North China block; SCB: South China block; ECS: East China Sea; SCS: South China Sea; TP: Tibetan Plateau; IB: Indochina block. (c) Teleseismic events recorded by the node geophones. The dark red triangle indicates the location of the dense array in Guangdong Province, China. The circles of different sizes are the teleseismic events recorded by the dense seismic array with epicentral distances within 30~90° and Mw ≥ 5.0. The Mw 7.3 Banda Sea earthquake (black arrow) occurred on 24 June 2019 at 02:53:39 (Coordinated Universal Time, UTC) at an epicentral distance of 212.0 km (33.57°).
1.2 The dense short-period seismic array
The detection of fine subsurface structures in modern seismology always requires a good sampling frequency in the spatial domain, which may be beyond the ability of most permanent seismic stations with a large spacing of 20~100 km (Shen et al., 2019). Meanwhile, these permanent stations can be supplemented by deploying portable broadband stations with a small spacing of 1~20 km (Wang et al., 2018; Ye et al., 2017; Yuan et al., 1997; Zhu, 2000). However, broadband stations are expensive and thus are sparsely adopted, while the resolution depending on the spacing between two stations is limited. In recent years, with the development of dense short-period seismic arrays with intervals ranging approximately from 100 m to 2 km, fine crustal structures have been discovered in increasingly observations.
In the early studies, Yuan et al. (1997) proposed the moveout correction methods based on the INDEPTH short-period seismic array, and Zhu (2000) developed the common conversion point (CCP) migration method based on the LARSE short-period seismic array. More recently, based on 473 seismometers, Shen et al. (2020) suggested that a pure shear shortening model beneath the Qilian Mountains can explain the lateral growth of the northern front of the Tibetan Plateau. Moreover, based on a dense array of 340 seismometers, Tian et al. (2021) found that the lateral growth of the northeastern front of the Tibetan Plateau can be represented with a model incorporating upper crustal overthrusting and lower crustal underthrusting beneath the Liupan Shan area. However, none of these features have been reported by previous portable or permanent broadband seismic studies.
In the following, one type of seismometer named the node geophone from the petroleum industry, which was introduced to the passive source seismic studies, has gradually become the most advantageous option for local exploration surveys. The deployments of large-N experiments within small areas using the node geophones (~100 m) became highly convenient for the 3D detection of shallow crustal structures (Dougherty et al., 2019; Lin et al., 2013). Some laboratory and outdoor experiments even show that node geophones can also be adopted to effectively record teleseismic signals; however, the shortcomings, namely, that they are biased toward relatively high frequencies and are susceptible to local noise and ambient temperatures, limit their use in teleseismic studies (Ringler et al., 2018). Despite this, these geophones have been introduced for traditional teleseismic receiver function analysis (Ward & Lin, 2017), and the results confirm that the recording of nodes can provide good constrains of shallow basin and fault structures (G. Liu et al., 2018; Ward et al., 2018). Simultaneously, node geophones are also used for monitoring traffic, weather, thunder, and blasts in urban seismology (Johnson et al., 2019; Lythgoe et al., 2021).
In this research, we carried out a geophysical survey with a dense linear array of short-period node geophones. This array was oriented north-south connecting the cities of Zhuhai and Lianzhou in Guangdong Province, China (Fig 1 a). From the recording of teleseismic events, we obtained an unprecedentedly detailed stacking profile using the teleseismic receiver function method. With these high-resolution results, we can provide certain evidence for the existence of remnants from the amalgamation of the East and West Cathaysia blocks.
2 Data and Methods
Using 527 short-period, 3-channel node geophones (Fairfield Zland GEN2 5Hz), we obtained 5.75 TB of seismic waveform data over 37 days from 25 May to 2 July 2019 and a total of 41 Mw ≥ 5.5 teleseismic events were recorded (Fig 1 c). Although short-period node geophones are convenient, the quality of the recorded waveforms is greatly affected by the deployment conditions. To ensure that the seismometers are coupled with the earth, node geophones each need to be buried within a 30 cm deep hole and at least 10 cm from the surface, in this way, persistent noise signals at approximately 20~40 Hz can be suppressed (Farrell et al., 2018).
In the early stage, the Python-based seismic tool Obspy was used to process the raw data (Beyreuther et al., 2010). To better constrain the performance of the geophones, we specifically analyzed the filtering parameters by using a series of two-pass, four-pole butterworth filters. Finally, a relatively narrow and low frequency band of 0.02–1 Hz was adopted to avoid the sensitivity of the instrument to high frequencies and prevent the contamination of the teleseismic signals with ambient noise. Upon rotating the three-component waveforms into radial-transverse-vertical (RTZ) coordinates, the direct primary (P-wave) arrival could be clearly distinguished from the continuous waveforms. After applying time-domain iterative deconvolution (Ligorría & Ammon, 1999) and manual checking, a total of 991 teleseismic receiver functions were selected. During the deconvolution process, low-pass Gaussian filters with different Gaussian factors of 2.0, 5.0 and 10.0 were adopted, and the corner frequencies were 1.0, 2.4 and 4.8 Hz, respectively. To avoid the influence of decreasing slowness with increasing conversion depth on the converted waves, the moveout correction method proposed by Yuan et al. (1997) was used to correct each curve to the 65° theoretical curve calculated by the iasp91 velocity model (Kennett & Engdahl, 1991). In the following, according to the piercing points of the Moho-converted phase, hereafter referred to as the Pms phase, the receiver functions with a Gaussian factor of 5.0 were aligned along the profile (Fig 2 a, b). Stacking of receiver functions with Gaussian factors of 2.0 and 10.0 are shown in the Supplementary Materials (Fig S2). In addition, we also binned the receiver functions in each 0.08° grid interval to strength the arrival of the Pms phases. The theoretical Pms arrival times were calculated by the TauP Toolkit (Crotwell et al., 1999) at a depth of 35 km under the iasp91 velocity model. Moreover, to intuitively enhance the signals from multiple waves, we also carried out PmpPms and PmpSms moveout corrections, and the stacking results are shown in Figure 3. During stacking, a smoothing factor was employed to smooth the receiver functions between adjacent grids, while the factor was determined by dividing the amplitude by the square of the distance between two grids. The amplitude and time errors during picking are indicated by the mean value and standard deviation with 100 resamplings according to the 1-dimensional