Shallow Magmatic System of Arxan Volcano Revealed by Ambient Noise Tomography with Dense Array

Abstract

The Arxan Volcanic Field (AVF) is an active volcanic region in Northeast Asia, and its last eruption occurred approximately 2000 years ago. Its eruption mechanism remains unknown. To investigate the shallow magma system beneath the volcanic cones in the AVF, we deployed a dense seismic array consisting of 227 portable seismographs and conducted high-resolution ambient noise tomography (ANT). The results of checkerboard test (CRT) and restoring resolution test (RRT) demonstrate that our imaging results are reliable. These results reveale significant slow-velocity anomalies at depths of 5~9 km below the Tianchi caldera and GD1213 volcano in Arxan, with the highest anomaly reaching up to approximately 15%. These anomalies suggest partial melting in a shallow magma chamber, indicating ongoing volcanic activity in the AVF. The velocity of the magma chamber corresponding to a melt fraction of approximately 7.4~12.9%. Therefore, the presence of the magma chamber poses potential hazards to the Arxan region, including volcanic eruptions and their associated risks.

1. Introduction

Northeast Asia, comprising countries such as China, Japan, and Korea, is influenced by the interactions of multiple tectonic plates, making it one of the most complex geological regions in the world [1]. This complexity primarily arises from the ongoing subduction of the western (Paleo-)Pacific Plate and the consequent formation of back-arc basins, such as the one in the Sea of Japan. These geological processes have contributed to the emergence of intraplate volcanoes, most notably Changbaishan and Arxan (also referred to as the Halaha volcano). Volcanic activity was widespread during both the Mesozoic [2] and the Cenozoic [3,4], resulting in the extensive distribution of volcanic rocks across the region. Previous studies have attributed the formation of intraplate volcanoes in Northeast Asia to processes such as the upwelling of wet and hot materials, dehydration of the big mantle wedge, and movement of asthenosphere materials. However, while the velocity structure beneath the Changbaishan volcano has been extensively documented [5,6,7,8,9,10], information regarding the velocity structures beneath other volcanoes remains scarce.

The AVF is located in the eastern part of the Central Asian Orogenic Belt (CAOB) and the central–west segment of the Great Xing’an Mountain in northeastern China, adjacent to the western side of the North–South Gravity Lineament, which extends westward to Mongolia and eastward to the Songliao Basin (Figure 1a). Geologically, it lies at the southeastern edge of the Siberian Craton and has been influenced by the Mongolia–Okhotsk tectonic belt and the Pacific tectonic domain [2]. The tectonic history of the region involves the closure of the Okhotsk oceanic basin and a late Jurassic to early Cretaceous left-lateral strike-slip and back-arc extension between the Izanagi Plate and the Northeast Asian continent. This tectonic activity led to the development of large-scale north–northeast-trending left-lateral strike-slip fault systems and accompanying volcanic activities across Northeast Asia, with a succession of intracontinental rift-related volcanic depression basins forming during the Mesozoic, gradually shifting from west to east over time. Despite significant research, there remains considerable controversy surrounding our understanding of tectonic evolution, particularly in the western region of the Great Xing’an.

The Great Xing’an volcanic group, located in Northeast China, represents a significant Mesozoic volcanic rock belt [2]. Intense volcanic activity occurred during the Mesozoic era, spanning from the Jurassic to the Early Cretaceous. This period was characterized by the intrusion and eruption of intermediate acid magma [11]. The volcanic rocks predominantly comprise medium-acidic volcanic melt rocks and volcanic detrital rocks, with the latter being the dominant rock type. Geological strata in the region consist of Jurassic volcanic rock formations overlaid by quaternary volcanic deposits, exhibiting a general northeastward distribution. During the Cenozoic era, coinciding with the east–west gradient formation in China, the Great Xing’an experienced significant uplift accompanied by extensive basaltic magma activity. Volcanic activity in the Arxan Volcanic Field (AVF) during the Cenozoic era was multi-phased. While previous geological surveys primarily suggested volcanic activity during the Pliocene and Pleistocene periods, evidence from geological and geomorphological features suggests the possibility of Holocene volcanic eruption activities [12,13]. Bai et al. [14] argue that the active periods of the Yanshan and Gaoshan volcanoes should be classified as Holocene, suggesting their classification as broadly active volcanoes.

Previous studies on the AVF have focused on various disciplines, including geology, volcanology, geothermal studies, seismology, and hydrology [3,4,14,15,16,17,18,19,20,21,22,23,24]. The age of the Tianchi basalts is estimated to be 0.34 ± 0.203 million years, while the age of the Yanshan and Gaoshan charred woods is 2000 years, suggesting that the AVF is relatively young and still has eruptive potential [12,14]. Quaternary lavas from the AVF are characterized by alkaline olivine basalt and olivine–nepheline basanite compositions, typical of island basalt with trace and rare-earth elements and an alkaline, within-plate sodium basalt composition [17]. Studies by Cui et al. [22] on the fluid geochemistry of hot springs in the Arxan region indicated that the overflow gaseous substance primarily originate from the upper mantle or crust such as the He (3–23%) in spring gas bubbles or crater water, and overflow gaseous substance (N₂, He, Ar and CO₂) in hot spring. Studies have demonstrated that the He level in the AVF is significantly higher than in other dormant volcanoes in China, suggesting ongoing volcanic activity [24]. Additionally, the soil CO₂ emission level in the AVF is comparable to that of island volcanoes, indicating significant volcanic degassing activity.

Previous seismic studies have identified a slow velocity (P-wave and S-wave) area in the Arxan upper mantle and crust, yet discrepancies exist in the understanding of its depth and extent, attributed to differences in data coverage and methodology [5,6,19,25,26,27,28,29,30]. Tang et al. [15] employed magnetotelluric methods to identify two magma channels leading to the mantle beneath the volcano, shedding light on subsurface volcanic structures. Han et al. [31] discovered numerous low-resistance anomalies in the lithosphere interior of the eastern section of the CAOB, indicating a possible connection to the upward intrusion of asthenospheric material, suggesting dynamic geodynamic processes. Yan et al. [32] utilized mineral physics and geothermal methods to investigate the Halaha volcanic areas, revealing higher temperatures, a thinner thermal lithosphere, and lower rheological strength, likely tied to the small-scale upwelling of hot asthenospheric material due to lithospheric delamination under the surface of the Songliao Basin. In a recent study, researchers carried out seismic tomography in the AVF, offering insights into the magmatic system and seismicity [23]. However, uncertainties remain regarding the presence of upward-transported magma channels or localized subsidence in the upper mantle of the AVF, highlighting the need for further research to understand the magma system and its associated hazards.

In recent decades, ANT has rapidly progressed, with numerous scholars conducting imaging studies at various scales of velocity structures [33,34,35,36,37,38]. ANT has been particularly successful in imaging the distribution of magma chambers beneath volcanoes, as demonstrated by studies such as those by Masterlark et al. [39], Nagaoks et al. [40], and Seats and Lawrence [41]. Moreover, the continuous development of short-period dense seismic arrays has significantly enhanced the capabilities of ANT by providing a better data source for refined imaging. This method has the advantages of a high resolution and cost-effectiveness, making it increasingly popular for detecting small-scale crustal structures [42,43,44]. Therefore, this study utilized arrival time data from a high-density seismometer array. A detailed 3D S-wave (Vs) model of the upper crust in the AVF was successfully obtained, which provided a high-resolution 3D velocity model to further understand the shallow crustal structure and magma system in the AVF.

2. Data and Methods

Coupled with the ongoing use of short-period seismometers in seismology, research into the shallow crustal structure is becoming increasingly sophisticated. In September 2020, researchers from Jilin University set up a linear dense seismic array in the study region, consisting of 227 stations that comprised 197 EPS-2-M6Q and 30 Smartsolo seismographs (Figure 1b). In order to detect the magmatic system beneath the AVF, most of the craters in the AVF were covered with seismometers. The TCV and GD1213 craters were situated close to the center of the array. The station interval increased to approximately 500 m as the distance from the TCV volcano cone increases, providing enough tomographic resolution below the TCV crater. This research focused exclusively on vertical component data. According to the modified method of data processing and analysis put forward by researchers [35], the raw data fragments were first merged, cut, band-pass filtered, de-instrumented, and downsampled, and then the daily SAC [45] format files were created for every station. The daily waveform data from all stations underwent preprocessing steps, containing demeaning, detrending, normalization in the time domain, and spectral whitening in the frequency domain.

After preprocessing, cross-correlation calculations were conducted on each station pair to derive daily noise cross-correlation (NCC) functions. Subsequently, all daily NCC functions were linearly overlaid and compute the mean across negative and positive lags. This process aimed to enhance the signal-to-noise ratio to create the last CC function. Notably, Rayleigh waves became evident in the NCC functions following the application of band-pass filters in the 0.08–2 Hz and 0.2–1 Hz ranges (Figure 2). The SNR demonstrates variability dependent on both the station interval and the frequency band of the applied band-pass filter, as depicted in Figure S1. Rayleigh waves exhibit a notably great value of the signal-to-noise ratio in the low-frequency band, particularly evident with greater distances between stations. To enhance the detection and analysis of Rayleigh waves, multiple filtering techniques were executed to the ambient noise CC function, as outlined by Herrmann [46], ultimately facilitating the calculation of the Rayleigh wave group velocity.

Figure 3a presents the Rayleigh wave group velocity dispersion data obtained from the NCC functions using the frequency–time analysis method [35,47,48,49,50], focusing on a time range of 1–6 s. Data were filtered using band-pass filter ranges of 1–20 s and 4–10 s, respectively, to isolate Rayleigh wave signals, and the group velocity was calculated for each frequency band. The resulting velocity data were integrated and concatenated to generate the final dispersion curves. Quality control measures were implemented by selecting dispersion data from NCC functions with an SNR exceeding 3. After screening out dispersion curves with significant velocity fluctuations, we obtained 1894 group velocity curves with a great SNR value from 18,749 station pairs (Figure 3a). Statistical analysis reveals notable peaks in ray path numbers, with one located in periods of approximately 2–5 s (Figure 3a). Additionally, Figure 3b illustrates a relatively even distribution of ambient noise power across the research region.

As a result of intense scattering and attenuation, extracting high-frequency Rayleigh waves (period 0.25–2 s) from station pairs separated by hundreds of kilometers presents considerable challenges [34,36,37]. However, dense station arrays enable the detection of high-frequency Rayleigh wave signals (0.25–2 s) even with small station spacing separated by a few kilometers or even hundreds of meters. This capability is demonstrated to be valuable for imaging shallow, near-surface 3D shear wave velocity structures [42,43].

After measuring Rayleigh wave group velocities, we conducted an inversion using these dispersion data to construct a three-dimensional high-resolution S-wave velocity model of the upper crust (0–9 km). Utilizing the velocity model provided by Li et al. [23] as a reference, we derived a one-dimensional average velocity initial model (Figure S2a).

According to the empirical formula of Brocher [51], the density and pressure wave velocity were coupled with the shear wave velocity inversion. In this study, the surface wave direct inversion method proposed by Fang et al. [43] was used to invert the three-dimensional shear wave velocity structure. Compared with the traditional surface-wave inversion method, this method does not need to invert the group velocity diagram but instead utilizes frequency-dependent ray tracing [52] and direct tomographic inversion based on wavelet sparse constraints. This approach accounts for the effects of the non-great-circle path propagation of surface waves in complex media, addressing the limitations of traditional surface wave imaging methods. We conducted an analysis of depth sensitivity kernels for the 1–6 s period Rayleigh wave group velocity dispersions at S-wave velocity, while the most sensitive depth for waves in the 6 s period range was 9 km (Figure S2b). Hence, the inversion incorporates S-wave velocity data down to 9 km under the surface, although the further resolution of deeper structures remains essential. The line chart depicts the root mean square and standard residual subsequent to each iteration, showing a rapid decreasing trend until it stabilizes at 0.84 (Figure S3a). Figure S3b displays histograms of the RMS time residuals from the reference and last-result models after 13 iterations. The results indicate a significant reduction in travel time residuals after inversion compared to the initial model, with a normal distribution near 0, signifying the reliability of the model after 13 iterations. The derivative weight sum (DWS) provides an average relative measure of the density of seismic rays near a given velocity node. This measure of seismic ray distribution is superior to an unweighted count of the total number of rays influenced by a model parameter, since it is sensitive to the spatial separation of a ray from the nodal location [53]. The DWS serves as an indicator of inversion result stability [54]. In this study, we established a criterion of DWS values over 10 to make sure the stability of the inversion results (Figure S4).

3. Results

3.1. Checkerboard Resolution Test and Restoring Resolution Test

In tomographic inversion, the DWS is utilized for the preliminary estimation of the three-dimensional velocity model. This parameter is associated with the number of rays intersecting surrounding a grid point [54]. We expect the DWS to show high values, as a higher DWS value at the speed framework mesh point indicates a better resolution. The results suggest that the DWS distribution (Figure S4) is excellent beneath regions with a high concentration of stations, particularly below the vicinity of the GD1213 and TCV calderas, indicating that this part of the model has a higher resolution. However, in the marginal areas of the AVF, relatively low DWS values are observed as a result of the wide spacing between seismic stations. Moreover, the value of the DWS for the model generally diminish below a depth of 9 km due to data limitations, although a sufficient resolution is maintained at the TCV caldera.

We further accomplished a couple types of synthetic tests for the velocity model to evaluate the resolution. One of them was the CRT. In the CRT, negative and positive perturbations of 20% relative to the 1D initial velocity model were alternately allocated to the 3D grid nodes in vertical and lateral directions. We investigated the model resolution with grid intervals of 0.055° × 0.035°, 0.075° × 0.055°, and 0.115° × 0.08° to define the most suitable grid spacing for inversion (Figures S5–S7). A 5% perturbation was applied to the grid nodes in the depth direction, with increased layer thickness to avoid velocity-discontinuity interfaces. The checkerboard resolution results in the depth direction are shown in Figure 4.

At the same time, we introduced a parameter V to represent the recovery degree of CRT [55]. The value of V ranges from 0 to 1, with larger values indicating better recovery, and vice versa. The value of V can be obtained from

V = (V_in − V_ref + V_out − V_ref)²/2 * [(V_in − V_ref)² + (V_out − V_ref)²]

(1)

where V_in and V_out represent the input and output velocity models of the CRT, respectively, and V_ref denotes the 1D initial velocity model. Figure 5 and Figures S8–S12 effectively illustrate the CRT results by plotting the recovery degree of the CRT values derived from Equation (1). The outcomes illustrate that the model recovers well at depths above 9 km below the TCV and GD1213 calderas from the result.

However, in the peripheral areas of the AVF, the model cannot fully recover due to sparse seismic ray coverage. Generally, all models with varying grid intervals are effective in the vicinity of the volcanic craters of interest, demonstrating a higher resolution compared to when used in marginal areas. Therefore, considering both the resolution results and the checkerboard resolution test results, a grid interval of 0.055° × 0.035° in the horizontal directions was chosen for the inversion.

We also conducted an RRT to estimate the reliability of the result model. To establish the true model, we added velocity anomalies of −10% and 10% into the 1D initial velocity model. In this case, the high- and low-velocity perturbations in the true model are similar to those in the S-wave velocity model (Figure 6 and Figure S13). The input velocity anomalies at a depth of ~9 km were accurately restored in principal components of the S-wave velocity from the RRT results, which were covered by a packed seismic arrangement. However, speed variations at the edges of the packed seismic arrangement could not be adequately resolved. The test results indicate that our model beneath the TCV volcanic region is high-resolution, and the key characteristics, as talked about below, are robust. The resolution around the GD1213 and TCV calderas is sufficient for all cycles, as it is almost at the focal point of the seismic arrangement.

3.2. Three-Dimensional S-Wave Velocity Model

This study acquired a precise spatial shear wave velocity model beneath the AVF at 1–9 km below the surface. The model was clipped dependent on the DWS distribution (DWS > 10) and the recovery degree of the CRT (V > 0.6), as shown in Figure S14. The outcomes of this study indicate significant horizontal variations in the shallow crustal S-wave velocity compositions below the AVF. Low velocities and high velocities are evident in each profile. Figure 7 illustrates, parallel to the horizon, profiles of velocity perturbations at 1~9 km below the surface, which more clearly show lateral velocity than absolute velocity variations. Low S-wave velocities along the northeast direction beneath the AVF align with the distribution of volcanic cones. The 3D S-wave velocity model and velocity perturbations reveal unique features closely associated with volcanic activity. The model displays two different magnitudes of velocity perturbations at shallow and deeper depths. At depths of 1–3 km, there are significant low-velocity perturbations concentrated near the volcanic cones of WSLZ, TCV, YM, GD1308, SHGB, SFS, XDG, GD1153, GD1325, GD1213, GD1201, and DC, which are related to volcanic eruptions. Significant low-speed perturbations can also be seen around Dujuan Lake. At greater depths of 5–9 km beneath the GD1213, more significant low-velocity perturbations become apparent, especially at 7 km below the surface, where the velocity anomaly attains up to nearly 15%. These large low-velocity perturbations are correlated with the residual low-speed bodies left after volcanic eruptions and with magma pipes and chambers that can transport and store partially melted material.

Figure S15 illustrates different vertical profiles of the results traversing the AVF calderas. Additionally, Figure 7 presents three vertical cross-sections of velocity perturbations through the AVF, which highlight more pronounced velocity variations. Furthermore, Figure 8 displays the distribution of seismic earthquakes [23], showing that most seismic events are situated at the transition zones between different perturbation bodies. Predominantly, large negative perturbations under the TCV caldera and GD1213 caldera are observed at depths of approximately 5 km, with the width of velocity perturbations beneath the GD1213 caldera at 7 km being the broadest, at approximately 6–10 km. At depths of 5–9 km, the negative perturbation remains larger than the adjacent component at the equal level. Similarly, notable low-velocity perturbations are present below GD1213 caldera. Additionally, low-velocity anomalies distinct from the surrounding material are identified at 0–3 km depths beneath the volcanic cone of the SHGB and WSLZ, although these anomalies are less pronounced than those under the TCV caldera, DC caldera, and GD1213 caldera. Weaker low-velocity perturbations are also observed at depths of 5 km beneath the GD1286 caldera and WSLZ.

4. Discussion

Following the latest volcanic eruption in the AVF, the magma system still exhibits partial melting and high temperatures in the crust around the volcanic cones. In the late 20th century, researchers inferred the upper limit of the Quaternary magma chamber’s depth using pressure values derived from the crystallization of the first crystalline phase olivine in the most recently erupted lava flow at WDLC [56]. In response to the distinct features of the velocity perturbation anomaly structure in the AVF, various explanations have been proposed for different velocity anomalies at various depths beneath the AVF volcanic cone (Figure 8).

In this study, the S-wave model is highly consistent with surface features. There are many intersecting faults distributed in the study area (Figure 1b), predominantly trending in the northeast and north–northeast directions, with the majority situated near the volcano cone (although no exposed faults have been observed on the surface [24]). The region features several volcanogenic lakes, including TCV Lake, WSLZ Lake, and DC Lake. TCV Lake, located at the summit of the volcano, formed from the accumulation of water within the caldera. Researchers detected He from the mantle in bubbling gasses and proposed that deep fluids transport heat upward through magma pipes [24]. WSLZ is a volcano created by phreatomagmatic activity that re-erupted after caldera breaks in the Mesozoic. DC Lake may have origins similar to those of the Mauna Ulu lava lake [57] and Aloi lava pond [57,58] in the Kilauea volcano, Hawaii, and was provisionally attributed to a collapsed lava lake.

In contrast to the Yellowstone volcanic group [59], the low-velocity zone observed at shallower depths (approximately 3 km) may result from fluid accumulation within fractures in the area, rather than from the migration of magmatic fluids (such as gasses and hydrothermal fluids) and melts from the magma chamber to shallower depths. In the Yellowstone volcanic area, shallow (0–3km), low-velocity bodies are linked with the main magma chamber located at a depth of 5–14 km under the Yellowstone volcano, aiding the transport of lava flow to shallow depths [59,60]. Otherwise, no significant hydrothermal variations were observed around the volcanic cones (XDG, GD1286, GD1202, SHGB, and GD1201) in the AVF. Consequently, we inferred that the slow velocity perturbation at shallower depths (1 km) under GD1201, as well as the slow velocity regions at ~3 km under the surface below calderas GD1213, GD1153, and SHGB, can be caused by the presence of fluids (such as surface water circulating through cracks or atmospheric precipitation) instead of lava flow [44,60,61].

The surrounding AVF area is characterized by high-frequency seismic activity. According to statistics, from 1991 to 1999, a total of 1946 earthquakes occurred, with 70% registering below a magnitude of 2.0. This low-magnitude, high-frequency seismicity is similar to volcanic seismicity [14]. The upward migration of magma is accompanied by a decrease in pressure and a change in volume, leading to a phase transition between liquid and gas [44], which may also coincide with seismic activity [62]. The depth of volcanic earthquakes often reflects the proximity of magma to the surface [62,63]. Figure 8 shows the focal depth [23], indicating conditions conducive to the existence of a magma chamber in the shallow crust. Therefore, the low-velocity body at a 5–9 km depth below some craters (TCV, GD1213, and GD1153) (Figure 8) can be classed as a magma chamber with partial melting and magmatic fluids at high temperatures. These features also correspond to the anomalous great value of the wave velocity ratio in this area shown by researchers [23]. Bodies resembling slow-velocity bodies have been observed beneath Mt Asama volcano in Japan, interpreted as a magma chamber [40], as well as beneath Yellowstone volcano [41,59,60,64], Toba in Indonesia [65], Lastarria volcano in Chile [61], and Laoheishan volcano in Wudalianchi. XDG is a Holocene volcano, and multiple volcanic earthquakes occur around the low-velocity body beneath the XDG volcanic cone. We consider these low-velocity bodies as pipes that can transport magma fluid and partially melted material at high temperatures. Similarly, they may be described as pipes beneath WSLZ volcanic cones that can transport magmatic fluid and partially melted material at high temperatures. In this study, no magma chamber or pipe was found beneath GD1201. Additionally, the weaker low-velocity zones beneath the SHGB volcanic cones can be interpreted as cooling magma pipes. Regarding whether there is a magma chamber beneath GD1286 and how it is spatially distributed (Figure 8), further geophysical surveys are needed to boost the resolution and verify our results due to the relatively low resolution in this area.

Based on our model, we assesed that there is a magma chamber with a bulk of no less than 200 cubic kilometers (~8 × 7 × 4 km) beneath the GD1213 and TCV near the crater. Although many volcanoes have magma chambers directly below the caldera [39], a horizontally offset magma chamber from the volcano’s summit is a common feature [40]. The volcanic area contains a total of 20 craters within approximately 40 × 40 square kilometers. It is possible that there are several magma chambers under the AVF. Based on our present tomographic results and previous research, the magma chamber beneath GD1213 and TCV in Arxan may have pipes connecting it to the lower crust. This may be the primary cause of volcanic eruptions near craters (TCV, GD1213, DC, XDG, GD153, and GD1325) in the AVF. If this hypothesis is confirmed, it would suggest that although the migration channel of the magma reservoir is blocked by high-speed material at 3–5 km (Figure 8), this obstruction could significantly impact the physical mechanisms of magma migration. Consequently, the risk of volcanic eruption remains due to the magma reservoir’s connection to the lower crust.

Due to the limitations of tomographic imaging using short-period Rayleigh wavelengths (<6 s), it is not possible to obtain magma distributions at greater depths using these imaging results. However, previous geophysical and geochemical studies can still provide valuable information regarding the structure of the lower and middle crust, as well as the upper mantle.

The origin of Cenozoic volcanism in Northeast China is closely linked to the subduction of the western Pacific plate rather than to deep mantle plum-type hotspot volcanos [66]. Body wave tomography has revealed significant slow-velocity bodies extending 200–400 km under the surface beneath Northeast China. This suggests that Cenozoic volcanic activity is closely associated with deep processes such as the subduction and dehydration of the Pacific Plate beneath the continental mantle transition zone in Northeast Asia, as well as convection circulation within the BMW, leading to the partial melting and upwelling of asthenospheric material [67]. Geochemical studies indicate that the magma system of the AVF originates from the mantle. Basaltic melts can ascend from the upper mantle to the Moho and subsequently rise to the lower crust, where they provide partially melted siliceous fuel to the magma chamber in the upper and middle crust [17,22,24]. Zhang et al. [18,19,20] found a low-velocity body extending to the vicinity of the mantle transition zone beneath Arxan, which is connected to the low-velocity bodies in the southern Songliao Basin and related to volcanic origins. Furthermore, the imaging results reveal prominent low-velocity bodies in the crust [19]. The southern and northern thicknesses, as well as the degree of partial melting of the asthenosphere in the eastern segment of the CAOB, exhibit heterogeneity [31]. Meanwhile, tomographic imaging results indicate east-thick, west-thin characteristics in the asthenosphere of the CAOB. Receiver function results indicate significant lithosphere thinning in the eastern part of the north–south gravity gradient zone [68]. Arxan has an ascending mantle flow that partially melts, exhibiting a lower degree of partial melting compared to the Changbaishan volcano [69]. This study’s results on the lithosphere’s temperature structure suggest that the temperature beneath the volcanic area is high, possibly due to the small-scale upwelling of thermal material brought about by lithospheric layering beneath the Songliao Basin [32]. The high-speed subsiding flow beneath the Songliao Basin triggers secondary convection in the western part, causing local upwelling of the asthenosphere and resulting in a slow velocity in the mantle and lower crust below the Arxan volcano [28].

Studies have carried out a magnetotelluric survey around Arxan, identifying varying degrees of low-resistance bodies in the middle-lower crust near XDG and GD1213 [15]. They have suggested that these anomalies are likely remnants of crustal magma pockets and cooling magma channels associated with the most recent eruption, revealing two magma channels leading to the mantle below the two volcanic belts. Li et al. [23] proposed the existence of crustal pipes for transporting magma fluids and partially molten magma systems at high temperatures beneath the AVF, indicating a crust-scale magma system. These findings suggest that basaltic magma can ascend from the upper mantle to the Moho and subsequently rise to the lower crust, providing fuel for the partial melting of the continental crust in the middle to shallow crust magma chamber.

Utilizing the method proposed by Chu et al. [70], we determined the correlation between the melt fraction and shear velocity based on the velocity in the magma reservoir. By substituting rhyolite with basalt, we estimated the melt fraction for the AVF. The velocity of the magma chamber lies between 2.894 and 3.136 km/s, corresponding to a melt fraction of approximately 7.4~12.9% (Figure 9). Given that the low-velocity anomaly is attributed to partial melting and the limitations of seismological methods, we calculated the melt fraction to be between 7.4% and 12.9%. Although the melt fraction has not yet reached the threshold for volcanic eruption [71], it is imperative to closely monitor the eruption risk.

5. Conclusions

Tomography based on a dense array offers novel insights into the distribution of the magma system in the AVF. Volcano eruptions have significantly advanced our comprehension of the magma system compared to previous studies.

This study identified low-velocity bodies at a depth of 5–9 km below the AVF, coinciding with a series of volcanic earthquakes near these anomalies, indicating the potential existence of a magma chamber in the crust with a volume of at least 200 km³. Extensive volcanic activity during the Quaternary period erupted along the deep rift system in the AVF, possibly related to the stagnant subduction of the Pacific Plate beneath northeastern China, leading to the continuous accumulation and upward thickening of the lithosphere. Moreover, the presence of magma channels and chambers in the middle-lower crust and upper mantle suggests ongoing magma replenishment, potentially contributing to the eruption potential near the TCV crater and GD1213. However, due to limited detailed information, a reliable resolution is primarily observed in the central AVF area and some surrounding volcanoes. Therefore, further seismic observations spanning larger spatial and longer temporal scales are necessary to comprehensively delineate the entire magma system in the lower and middle crust and upper mantle, thereby facilitating the analysis of its relationship with shallow magma chambers. The melt fraction of the magma chamber was found to be between 7.4% and 12.9%. Although the threshold for a volcanic eruption has not yet been reached, it is still essential to closely monitor the potential eruption risk.