Advances in seismological methods for characterizing fault zone structure

1.   Introduction

Movement along faults in the Earth’s crust leads to extensive fragmentation of rocks on each side of the fault, giving rise to a distinct low-velocity region known as a fault zone. The fault zone comprises two primary components: the fault core and the surrounding damage zones (Ben-Zion and Sammis, 2003). The fault core, typically a narrow zone ranging from tens of centimeters to a few meters in width, consists predominantly of cataclasite and ultracataclasite. The majority of displacement along the fault occurs within the fault core. The fault damage zones typically include localized fracture zones or subsidiary faults that can extend tens to hundreds of meters away from the fault core (Chester et al., 1993; Evans and Chester, 1995). Large earthquakes often occur within structurally complex fault zones, where the geometry, continuity, and seismic velocity structure of the fault zone are important for determining seismic rupture and slip localization. Therefore, accurate and comprehensive information about the fault zone is paramount for understanding seismic processes and long-term fault evolution (Share et al., 2023; Yang HF and Zhu LP, 2010; Yang HF et al., 2011; Yu HY et al., 2018).

Detailed characterization of fault zone structure offers crucial insights into past fault behavior and potential future earthquake characteristics (Allam et al., 2014). Irregular fault geometries, such as bends, bumps, or jumps, can act as either initiation or termination points of earthquake ruptures (King and Nábělek, 1985; Elliott et al., 2015; Yao SL and Yang HF, 2022). Relatively high and low seismic velocities may provide valuable information about areas of high moment release and rupture termination, respectively (Michael and Eberhart-Phillips, 1991). Moreover, velocity contrasts along and across the fault strike exert a significant influence on the dynamic migration and stagnation of earthquake ruptures (Ben-Zion and Andrews, 1998; Bennington et al., 2013; Abdelmeguid and Elbanna, 2022; Weng HH et al., 2016; Zhao P et al., 2010). Analysis of v_P/v_S ratios in the vicinity of faults and their surrounding areas offers essential insight into rock damage, fluid conditions, and mechanistic processes within different fault segments (Allam et al., 2014; Christensen, 1996; Jiang XH et al., 2021; Lin GQ and Shearer, 2009; Share et al., 2023). The presence of low-velocity zones close to faults amplifies ground vibration amplitudes and leads to a substantial increase in ground motion near fault zones (Ben-Zion and Aki, 1990; Ben-Zion and Andrews, 1998; Chen X and Yang HF, 2020; Kurzon et al., 2014; Luo S et al., 2023; Wirth et al., 2019; Wu CQ et al., 2009). Additionally, segmented low-velocity zones can result in asymmetric rupture propagation, which amplifies ground motion and triggers seismic activity in the direction of rupture. This information is important for seismic risk assessment (Somerville et al., 1997; Gomberg et al., 2001; Weng HH et al., 2016).

An accurate assessment of the location, magnitude, extent of rupture, ground motion, and hazard distribution of future major earthquakes within a fault zone is key for urban planning and protecting human life and property. Urban planners can use this information to make informed decisions regarding infrastructure development and land-use zoning while considering potential seismic hazards. Dynamic simulations based on detailed three-dimensional (3D) structural models offer reliable tools for studying seismic rupture and fault slip, as well as predicting strong ground motions such as peak ground velocity (PGV) and peak ground acceleration (PGA). These simulations project seismic intensity distribution, providing a robust reference for seismic hazard assessment within the fault zone (Cheng J et al., 2021; Luo S et al., 2023; Yao SL and Yang HF, 2022).

A variety of seismological techniques can be employed to investigate fault zone structures, including seismic imaging methods, such as travel-time tomography and ambient noise tomography, which enable visualization and characterization of subsurface structures (Luo S et al., 2023; Mordret et al., 2019; Ozer et al., 2019; Shao XH et al., 2022; Taylor et al., 2019; Wang YD et al., 2019; Yang HF et al., 2020; Yang Y et al., 2020). Additionally, fault zone trapped wave analysis can provide valuable information about the properties and behavior of waves confined within the fault zone (Ben-Zion et al., 2003; Li YG and Leary, 1990; Li YG et al., 2000, 2016; Shao XH et al., 2022; Zhang Z et al., 2022). Other techniques, such as head wave (Bennington et al., 2013; Najdahmadi et al., 2016; Share et al., 2017, 2023; Share and Ben-Zion, 2018; Zhao P and Peng ZG, 2008) and reflected wave analysis (Li HY et al., 2007; Yang HF et al., 2011, 2014) can offer insights into wave propagation and interaction with the fault zone. Furthermore, the spatial distribution of earthquakes can serve as a powerful tool for studying fault zone structures (Cheng YF et al., 2023; Fang LH et al., 2015; Feng T et al., 2022; Kim et al., 2016; Li YQ et al., 2019; Shelly and Hardebeck, 2019; White et al., 2019). The distribution of seismic events can provide valuable information about the geometry and connectivity of fault zones, and contribute to understanding the fault zone structure and the implications for seismic processes. Overall, combining these seismological techniques enables investigation and characterization of the structures within fault zones, which provides essential insights into fault zone properties and behavior, as well as the influence of fault zone structure on the seismic rupture process.

In recent years, dense-array seismic detection technology has emerged as an increasingly important tool for investigating deep fault zone structures. Wu JP et al. monitored fault zone behavior in China as part of the National Key Research & Development Program for Major Natural Disasters from 2019 to 2021. They established key monitoring infrastructure through a large-scale dense array within the Anninghe-Xiaojiang fault zone (AXFZ) in China. The project successfully obtained high-resolution 3D velocity structures and precise seismic spatial distribution images of the AXFZ, providing key information on the spatial geometry of the fault zone. The results enhanced our understanding of the seismogenic environment and deep tectonics and facilitated numerical simulations of seismogenic processes. In this study, we focused on reviewing advancements in seismological methods used to study fault zone structures and highlight notable results obtained from the dense seismic array deployed in the AXFZ.

2.   Seismic methods for resolving fault zone structures

2.1   Seismic tomography

High-resolution seismic imaging techniques can inform on the velocity structure of the Earth’s interior and are therefore used to study fault zone structures. Techniques include local earthquake travel-time tomography, ambient noise tomography, and tectonic interface imaging.

2.1.1   Travel-time tomography from local earthquakes

Travel-time tomography, an imaging technique that utilizes P and S travel times, was first proposed in the 1970s and has since become a common method for studying the Earth’s internal tectonics and geodynamics (Ajala et al., 2019; Aki and Lee, 1976; Huang JL and Zhao DP, 2004; Ozer et al., 2019; Thurber, 1983; Wu JP et al., 2013; Zhao DP et al., 1992). Ajala et al. (2019) utilized travel-time tomography to construct the 3D velocity structure of the Coachella Valley at 0–10 km depth. By analyzing lateral velocity variations, they identified a previously unrecorded fault. To enhance the capabilities of travel-time tomography, Zhang HJ and Thurber (2003) developed a joint inversion method called double-difference tomography (TomoDD) by combining double-difference earthquake location algorithm of Waldhauser and Ellsworth (2000) and travel-time tomography. TomoDD utilizes relative travel time differences between neighboring earthquakes and absolute travel times between earthquakes and seismic stations for source location and velocity structure determination. This approach has been used extensively to obtain high-resolution crustal velocity structures at local and regional scales (Deng WZ et al., 2014; Feng T et al., 2022; Hutchinson et al., 2019; Shelly et al., 2006). Feng T et al. (2022) employed machine learning seismic detection techniques to obtain a high-precision seismic catalog and seismic phase data. Subsequently, they utilized TomoDD to obtain a high-resolution 3D velocity structure at 0–20 km depth beneath parts of the AXFZ and characterized the fault’s dip angle. It is important to note that the application of travel-time tomography relies on high spatial and temporal resolution in the earthquake catalog, so it is less suitable for regions with low seismic activity.

2.1.2   Ambient noise tomography

Ambient noise tomography is commonly used to image the S-wave velocity structure (Jin JQ et al., 2023; Li DD et al., 2019; Luo S et al., 2023; Mordret et al., 2019; Ni HY et al., 2022; Shapiro et al., 2005; Taylor et al., 2019; Wang YD et al., 2019; Yang HF et al., 2020) and offers several advantages over travel-time tomography. Unlike travel-time tomography, ambient noise tomography is not constrained by the earthquake distribution, making it applicable to regions with low seismicity. This method can effectively study the shallow crustal structure of fault zones by utilizing dense array data and extracting surface wave signals from ambient noise. Mordret et al. (2019) used a dense seismic array to image the 3D velocity structure of the San Jacinto fault zone to a depth of approximately 500 m depth (Figure 1). Their model revealed significant lateral velocity variations across fault strike, providing crucial information about the fault zone’s internal dynamics. Yang HF et al. (2020) used teleseismic P-wave arrivals to determine a 3.4 km wide low-velocity anomaly within the Chenghai fault zone and subsequently expanded the detection of the low-velocity anomaly to a depth of 1.5 km through noise tomography. This demonstrates the capability of ambient noise tomography in characterizing subsurface structures and identifying anomalies within fault zones.

Figure  1.  Horizontal slices of the (a, c) 3D velocity structure and (b, d) velocity uncertainties beneath the San Jacinto fault zone using ambient noise tomography. Slices taken at depths of (a, b) 70 m and (c, d) 200 m. The main Clark fault trace is indicated by thick black curves. From Mordret et al. (2019).

DownLoad: Full-Size Img PowerPoint

Traditional ambient noise tomography relies on base-order surface wave dispersion data. However, higher-order surface waves exhibit increased sensitivity to S-wave velocity variations, particularly in the presence of soft layers and for deeper subsurface investigations. Joint inversion of both base-order and high-order surface waves offers advantages in reducing the non-uniqueness of the inversion process (Luo YH et al., 2008).

2.1.3   Ambient noise interferometry

Ambient noise interferometry is used to reconstruct the Green’s function between pairs of stations by cross-correlating the received signals. This technique can enable the use of ambient noise as a valuable signal, with one station acting as a source and another as a receiver. This approach enables accurate estimation of the Green’s function between stations, encompassing both body and surface waves. Body waves have higher frequencies compared to surface waves, and extracting body wave signals in ambient noise interferometry improves the resolution for imaging subsurface structures (Nakata et al., 2015). She YY et al. (2022) employed ambient noise interferometry to extract reflected wave information, thus identifying shallower fault structures in the Chenghai fault zone. Ambient noise interferometry also provided insight into fault structures that complemented the low-velocity anomalies obtained through noise tomography (Yang HF et al., 2020). Results obtained through noise interferometry were found to be consistent with those obtained through receiver function imaging (Jiang XH et al., 2021), validating the effectiveness and reliability of the method.

2.1.4   Receiver function imaging

Although ambient noise tomography shows good lateral resolution, it has limited sensitivity to sharp vertical velocity boundaries. Therefore, ambient noise tomography is not suitable for accurately determining the depth of the sensitive cores within the low-velocity zone (LVZ). Receiver function imaging, however, assumes an equivalent source and a horizontal laminar earth model. Therefore, the method effectively removes the contributions of the seismic source, propagation path, and instrument response by deconvolving the vertical component from the radial component. The resulting receiver function provides valuable phases, including the Moho converted Ps wave and its multiple reflected waves. These waves are particularly sensitive to velocity discontinuities and the v_P/v_S ratio. The H-κ stacking method is widely used for estimating crustal thickness and the v_P/v_S ratio (Zhu LP and Kanamori, 2000; Wang P et al., 2010; Wang WL et al., 2017; Fadel et al., 2018).

The H-κ method typically assumes that the v_P/v_S ratio remains constant throughout the fault zone. However, this assumption can have implications for accurately determining the geometry and velocity variations within the LVZ. Therefore, Jiang XH et al. (2021) proposed a novel approach using primary and secondary Ps-converted waves (${{\mathrm{P}}}_{{\mathrm{b}}}{\mathrm{s}}$ ${{\mathrm{P}}}_{{\mathrm{b}}}{\mathrm{pPs}}$) generated by the low-velocity structures within the fault zone (Figure 2). This approach allows for the estimation of depth and v_P/v_S ratio variations within the LVZ over time. Jiang XH et al. (2021) showed that their LVZ model could be successfully recovered even in the presence of noise. By incorporating primary and secondary Ps-converted waves and considering their interactions with the low-velocity structures, this method offers a more comprehensive understanding of the LVZ properties, including depth and v_P/v_S ratio variations.

Figure  2.  (a) Stacked synthetic receiver functions with random noise for a hypothetical low-velocity-zone (LVZ) model with a 1.5 km sunken shape in the center. The red dashed lines denote the possible range of PbpPs phase arrivals. The blue and green solid lines show the picked arrival time of ${{\mathrm{P}}}_{{\mathrm{b}}}{\mathrm{s}}$ and ${{\mathrm{P}}}_{{\mathrm{b}}}{\mathrm{pPs}}$ phases, respectively. (b–c) Synthetic receiver function travel time inversion results. Red lines represent the (b) inverted v_P/v_S ratio and (c) depth extent of the LVZ. Black lines denote the hypothetical model. From Jiang XH et al. (2021).

DownLoad: Full-Size Img PowerPoint

Receiver function migration imaging can be used to investigate crustal and upper mantle velocity discontinuities. This method can determine the depth and morphology of these discontinuities by converting phase information contained in the receiver function from the time domain (or frequency domain) to the depth domain, thereby enhancing the spatial resolution of receiver function imaging (Chen JH et al., 2005). A common method for imaging these seismic discontinuities is Common Conversion Point (CCP) stacking (Chen L et al., 2006; Hong DQ et al., 2021; Li W et al., 2022; Liu Z et al., 2017; Tian XB et al., 2021; Zhu LP, 2000). CCP stacking sums receiver functions from different seismic stations that share a common conversion point corresponding to a specific depth within the Earth. By stacking the receiver functions at each common conversion point, the resulting image provides valuable insights into the structure and characteristics of crustal and upper mantle discontinuities. Liu Z et al. (2017) used CCP stacking to image the crustal structure across several faults in the eastern margin of the Qinghai-Xizang Plateau (Figure 3). The resulting images clearly showed the velocity discontinuities around the fault system, shedding light on the fault geometry and dynamic fault processes.

Figure  3.  Pms Common Conversion Point (CCP) stacking of the crustal structure across the faults (a) and illustrated deformation cartoon (b). Dashed lines in (a) show amplitude bands. From Liu Z et al. (2017).

DownLoad: Full-Size Img PowerPoint

2.1.5   Spectral ratio method

The spectral ratio method has emerged as a promising approach for determining the depth of the LVZ (Chen ZQ et al, 2023; Molnar et al, 2018; She YY et al., 2022; Song JH and Yang HF, 2022). This method extracts information about the impedance frequency response by employing the horizontal-to-vertical spectral ratio (HVSR), which compares the spectra of horizontal and vertical components of a waveform. This enables estimation of the shallow S-wave velocity structure and facilitates frequency-depth conversion to image the continuous seismic wave impedance depth (Arai and Tokimatsu, 2004). She YY et al. (2022) applied the HVSR method to investigate the geometry of the LVZ in the Chenghai fault zone. Their findings revealed an inverted trapezoidal LVZ with a depth of approximately 1.0 km and a width of approximately 3.0 km.

Song JH and Yang HF (2022) employed the standard spectral ratio (SSR) method to investigate the seismic site response characteristics within the Chenghai fault zone by analyzing amplitudes and spectra of local and teleseismic earthquakes recorded by a dense seismic array. Their analysis revealed the presence of a new, heightened site response within the pre-existing LVZ of the fault zone. This site response exhibited a width of 500 m and indicated a higher level of ground motion amplification (Figure 4). Song JH and Yang HF (2022) attributed this localized intensification to a concentrated low-velocity zone within the fault zone.

Figure  4.  Averaged spectral ratio results across the Chenghai fault zone obtained from (a) P waves and (c) S waves of teleseismic earthquakes. The low-velocity zone identified by Yang HF et al. (2020) is indicated by thick red lines. The white dashed lines mark the boundaries between four parts that have different apparent features. The third part marks a 500 m wide zone around the fault. Corresponding uncertainty in frequency is shown in (b) and (d). From Song JH and Yang HF (2022).

DownLoad: Full-Size Img PowerPoint

2.2   Fault zone waves

2.2.1   Trapped waves

In general, seismic imaging techniques are limited in their ability to directly detect the structure of the fault core within a fault zone. However, analyzing fault zone trapped waves can provide valuable insights into the dimensions and characteristics of the fracture core (Zhang Z et al., 2022). When seismic waves encounter the interface between the low-velocity medium within the fault zone and the higher-velocity surrounding rocks, reflection and transmission occurs. If the angle of incidence exceeds the critical angle, the wave undergoes total reflection within the fault zone. This leads to the formation of a wave train with the properties of a Love surface wave, known as the fault zone trapped wave (Ben-Zion, 1998). This wave train arrives after the primary body waves, exhibiting low frequency, large amplitude, and dispersion characteristics (Ben-Zion, 1998). By analyzing the travel time, frequency, amplitude, and dispersion characteristics of the fault zone trapped wave, it is possible to invert and simulate waveforms to obtain the width and geometry of the fracture zone, Q values within the fracture zone, as well as velocity variations within and outside the fracture zone. This approach has been widely applied in fault studies worldwide and provides valuable information about the internal structure of fault zones (Ben-Zion et al., 2003; Li YG and Leary, 1990; Li SL et al., 2009; Li YG et al., 2000, 2016; Peng ZG et al., 2003; Qiu HR et al., 2017; Share et al., 2017; Yang HF et al., 2011; Zhang Z et al., 2022). For example, Zhang Z et al. (2022) found a deep (~4 km) and narrow (50–150 m) inner damage zone around the Haiyuan fault by using the waveform modeling of trapped waves (Figure 5).

Figure  5.  (a) Schematic diagram of fault damage zone inferred from P-wave arrival time analysis and trapped wave modeling. By analyzing teleseismic and local events, they identify an approximately 1 km wide fault damage zone and 200 m wide inner zone as shown in (c). (b) Locations of the stations and the scales of the damage zone and inner core. (c) The flower-type geometry of the fault damage zone from trapped wave modeling. Stations used in the waveform modeling of trapped waves at 3–6 Hz are shown in red, whereas stations in both red and blue are included in the modeling of trapped waves at 1–3 Hz (b and c). From Zhang Z et al. (2022).

DownLoad: Full-Size Img PowerPoint

Indeed, there is a trade-off between the depth resolution and other parameters when analyzing fault zone trapped waves. Li YG et al. (2000) investigated the Landers fault zone and determined that at the surface, the fault zone exhibited a width of 250 m with a 45% reduction in wave velocity compared to the surrounding rocks. However, at a depth of 10 km, the fault zone narrowed to 125 m width with a 35% reduction in wave velocity. Conversely, Peng ZG et al. (2003) found that the Landers fault zone had a width of about 200 m with wave velocity reduction of 30%–40% relative to the surrounding rocks at depth of approximately 2–4 km. The disparities between these results highlight the trade-off between depth-dependent fault zone width and wave velocity variation. Additional constraints in fault zone studies can be incorporated through analysis of higher-frequency body waves such as direct waves, head waves, and reflected waves (Bennington et al., 2013; Li HY et al., 2007; Qiu HR et al., 2017; Share et al., 2017; Zhang M et al., 2022). To effectively observe fault zone trapped waves, a dense array of seismic stations deployed perpendicular to the fault strike is typically required. It is important to consider the receiving distance, as excessively long distances may lead to wave attenuation, making it challenging to accurately identify and analyze the trapped waves.

2.2.2   Resonance waves

In addition to Love-type trapped waves, another type of trapped wave involving normal modes or Rayleigh-type signals has been observed in the radial and vertical components of seismic recordings (Gulley et al., 2017; Malin et al., 2006). These waves exhibit amplitudes between those of the direct P and S waves. Such waves have been detected at various stations along the San Jacinto fault zone (Qin L et al., 2018; Qiu HR et al., 2017) and the Parkfield segment of the San Andreas fault (Ellsworth and Malin, 2011). After a large earthquake, the energy released during the rupture excites free vibrations within the Earth. These vibrations occur at specific eigenfrequencies and can be described by a set of eigenfunctions (Gilbert, 1971). The observed eigenfrequencies and eigenfunctions are highly sensitive to the internal structure of the Earth and have been extensively used for global-scale mapping of Earth’s deep structure. Similar to oscillatory waves found in any finite object (Geimer et al., 2020), seismic energy trapped in a fault zone waveguide can also generate normal (or resonant) modes in a finite (width and depth) waveguide. These resonant modes provide valuable constraints on the internal structure of the fault zone waveguide. Qiu HR et al. (2020) made notable advancements in this field by analyzing long-duration resonant wave signals following the arrival of trapped waves. They employed a dense array of seismic stations across the San Jacinto fault zone in southern California with a spacing of 10–30 m. The results showed a velocity contrast along the fault, which had not been achieved by previous trapped wave studies due to resolution. The resonant waves exhibit a lower frequency component (<3 Hz) than trapped waves, which typically peak around 5 Hz. This lower frequency component offers distinct spatial sensitivities for structural imaging of the fault zone.

2.2.3   Head waves

The presence of different rock properties at fault interfaces leads to the generation of fault zone head waves when body waves propagate within a fracture zone (Ben-Zion and Aki, 1990). These head waves travel with the velocity of the faster block and are refracted from the fault to the slower velocity block (Bennington et al., 2013). Therefore, the head waves arrive earlier than direct body waves and exhibit opposite polarity. Stations located near slower interfaces record these head waves as the first arrivals (Ben-Zion and Malin, 1991). Since head waves propagate through the fault zone interface, they provide valuable information about the contrasting properties on each side of the fault (Ben-Zion and Malin, 1991; Bennington et al., 2013; Najdahmadi et al., 2016; Qiu HR et al., 2017; Share et al., 2017, 2023; Zhao P and Peng ZG, 2008). By analyzing fault zone head waves, Share and Ben-Zion (2018) found that rocks in the northeastern San Jacinto fault zone exhibited velocities more than 7.5% lower than those in the southwest. This implies that significant seismic events occurring in the northern fault zone could propagate towards the southeast.

2.2.4   Reflected waves

The fault zone is commonly characterized by the presence of low-velocity anomalies, and the analysis of reflected waves offers an effective approach to determining the boundaries of the fault zone (Li HY et al., 2007; Yang HF et al., 2011, 2014; Yang HF, 2015). By tracing the paths of reflected waves that propagate through the fault zone and reflect multiple times at the fault zone boundaries, information is provided about the depth and width of the LVZ. Yang HF et al. (2011) used direct and reflected wave arrivals to investigate the structure of the Calico fault. Their study revealed a 1.3 km wide LVZ with a 40% reduction in v_P and a 50% reduction in v_S. The LVZ was observed to dip at an angle of 70° to the northeast, reaching a depth of 3 km. Notably, this depth was shallower than the >5 km depth determined by trapped waves. By using reflected waves, the trade-off between fault zone width and velocity can largely be avoided, allowing for enhanced depth resolution of the fault zone structure. However, the accuracy of this method is highly dependent on earthquake location. Even a small (<0.5 km) focal depth error can lead to several kilometers of error in estimating the reflection point location (Li HY et al., 2007). To improve the accuracy of seismic location and, consequently, the precision of structural studies in the fault zone, the use of dense seismic array observations can be employed.

2.3   Seismic activity

Spatial earthquake distribution plays a crucial role in constraining the geometry of fault zones and provides insights into the rupture processes occurring at the focal depth (Fang LH et al., 2015; Feng T et al., 2022; Kim et al., 2016; Li YQ et al., 2019; Ross et al., 2017, 2019; Rubin et al., 1999; Shelly and Hardebeck, 2019; White et al., 2019). This information is essential for assessing seismic hazards and understanding earthquake behavior. For example, Kim et al. (2016) used seismic activity to investigate fault dip and geometry along the Parkfield fault in California. The vertical distribution of seismic events revealed a change in fault dip from NE to SW, suggesting a twisted fault geometry (Figure 6). However, the accuracy of hypocentral parameters depends on the precision of the velocity model employed. Therefore, it is challenging to resolve earthquake locations in regions with complex velocity structures, particularly if only a one-dimensional velocity model is used.

Figure  6.  Determination of fault-plane dips based on the spatial distribution of earthquakes. The vertical distribution of seismicity suggests that the fault dips NE in the northwestern region, nearly vertical in the central region, and SW in the southeastern region. The schematic model is illustrated in the lower panel for clarity. From Kim et al. (2016).

DownLoad: Full-Size Img PowerPoint

2.4   Other innovative methods in fault zone analysis

The application of non-traditional seismic signals has opened new avenues for investigating active fault behavior and subsurface structures (Figure 7). Brenguier et al. (2019) used train traffic as a noise source for monitoring active faults via seismic interferometry. Sheng Y (2023) introduced seismic stereometry as an alternative method, illustrating the potential of seismic signals generated by cars. Similarly, Zhang H et al. (2023) explored lateral variations along the southern San Andreas fault zone by analyzing traffic signals through a dense seismic array. These studies collectively offer new perspectives for subsurface characterization and fault monitoring, showcasing the adaptability of seismic methods to unconventional noise sources. The findings emphasize the transformative potential of traffic-generated seismic signals while highlighting the need for further fault zone and subsurface study.

Local v_P/v_S ratios play an important role in advancing understanding of fault zones and their underlying dynamics. Lin GQ and Shearer (2007) determined relative v_P/v_S ratios of different events within the same earthquake cluster, shedding light on consistent ratios among events within the cluster. Lin GQ et al. (2015) then extended this approach to near-source v_P/v_S ratios and constrained the size and depth of the shallow magma reservoir beneath Kīlauea caldera. This application of seismic parameters showcases the potential for providing insights into subsurface volcanic processes, and the same method could also be applied to better understand fault zone properties.

Figure  7.  Summary of seismic methods used to resolve fault zone structure.

DownLoad: Full-Size Img PowerPoint

3.   Study on the AXFZ based on a dense seismic array

Dense seismic arrays are important for studying fault zones across all seismic methods, because a dense array enhances the ability to image high-resolution fault zone structures and capture seismic wave signals. For example, dense seismic arrays have been used to reveal structural intricacies of various segments within the San Andreas fault zone (Brenguier et al., 2019; Lewis et al., 2007; Li YG et al., 2004; Li YG, 2021; Zhang H et al., 2023) and the geometry of fracture zones within the San Jacinto fault zone (Mordret et al., 2019; Qiu HR et al., 2017, 2020; Share et al., 2017, 2023; Wang YD et al., 2019; Yang HF et al., 2014).

The AXFZ (Figure 8), situated on the southeastern margin of the Qinghai-Xizang Plateau, plays a pivotal role in eastward extrusion of the Qinghai-Xizang crust. This fault zone has experienced several notable earthquakes, including 11 earthquakes of M6–7 and 7 earthquakes of M7–8 (Wen XZ et al., 2008, 2011), so the AXFZ has high seismic hazard potential in the medium and long term. Increased observation of the AXFZ is required to further elucidate the processes involved in the formation and occurrence of large earthquakes and improve hazard assessment.

Figure  8.  Locations of the AXFZ and the four profiles (red lines) shown in Figures 9–12. Gray lines represent active faults. Yellow circles show locations of historical earthquakes with magnitudes >6.5. AXFZ: Anninghe-Xiaojiang fault zone; AFZ: Anninghe fault zone.

DownLoad: Full-Size Img PowerPoint

Luo S et al. (2023) used ambient noise tomography to obtain a comprehensive 3D model of the upper crustal v_S model beneath the Anninghe fault zone. They revealed the presence of a narrow LVZ along the fault at a depth of 0–3 km. Interestingly, this LVZ shifts to greater depths of 4.5–8 km in the eastern part of the fault zone. Based on these observations, they proposed a seismotectonic model that interprets the narrow LVZ as a fracture zone constrained by the presence of water (Figure 9). According to their model, the western part of the fracture zone lays beneath the eastern part, resulting in the formation of a seismogenic zone characterized by increased microseismic activity at greater depths within the eastern section of the Anninghe fault zone. Moreover, the study conducted large earthquake simulations using the 3D velocity structure obtained with the dense seismic array. The results of these simulations indicated that incorporating the detailed velocity structure into the models resulted in an increase in simulated PGA during rupture by 25%–50% compared to using a smoother velocity model. This underscores the significance of considering detailed velocity models when assessing the hazard associated with large earthquakes.

Figure  9.  3D seismotectonic model beneath the northern Anninghe fault zone. The location of the vertical v_S profile is shown as Line 1 in Figure 8. The arrow in the vertical profile indicates the direction of the interpreted thrust fault. In the horizontal slice, the arrows indicate the current sinistral strike-slip movement of the Anninghe fault. From Luo S et al. (2023).

DownLoad: Full-Size Img PowerPoint

We employed TomoDD to investigate the P-wave velocity structure around the Anninghe fault zone. Figure 10 shows a profile across the fault. The results revealed velocity variation across the fault. Notably, earthquakes predominantly occur within the high-velocity zone. The findings suggest that the fault may exhibit an eastward dip, which aligns with conclusions drawn by Luo S et al. (2023) from ambient noise tomography.

Figure  10.  A vertical v_P profile along the latitude of 29.2°N (Line 2 in Figure 8) across the Anninghe fault zone. F1: Anninghe fault. The pink line shows the interpreted dip angle of the fault. Black dots represent earthquakes.

DownLoad: Full-Size Img PowerPoint

Shao XH et al. (2022) obtained three high-resolution P-wave velocity profiles within the Anninghe fault zone (between Mianning and Xichang) using travel-time tomography of methane gas sources (Figure 11). They particularly investigated shallow depths of 0–2 km and found this region was characterized by low velocities, primarily attributed to the presence of Quaternary sediments. Integrating data from methane gas sources and travel-time tomography offers a unique approach to investigating shallow layers surrounding a fault zone. Understanding the fault’s shallow structure provides important input data for modeling seismic events, thus contributing to improved hazard assessments in the region.

Figure  11.  (a) Location of the active shots (red stars) and stations (black dots) near the profile (rotated 2.22° counterclockwise from Line 3 in Figure 8). Black lines show the two branches of the Anninghe fault. F1 and F2 indicate the western and eastern branches of the fault, respectively. (b) The vertical v_P profile. (c) The v_P perturbation model. The black stars in (b) and (c) show the locations of the shots. The identification numbers of the shots are above the box in (b) and (c). From Shao XH et al. (2022).

DownLoad: Full-Size Img PowerPoint

Chen ZQ et al. (2023) investigated the sedimentary structure of the Anninghe fault zone using continuous waveform data obtained from two linear dense arrays, which were positioned approximately 80 m apart across the fault zone. Chen ZQ et al. (2023) used the HVSR method to estimate the thickness of the sediment layer adjacent to the fault zone of 300–600 m. The sediment layer exhibited an eastward dip, indicating a correlation with the east-west extrusion and compression that occurred during the fault’s formation. These findings have implications for the seismic vulnerability of densely populated areas, because the resonance period of multistory buildings, particularly those with 6–9 stories, aligns with the peak frequency range of the sediment layer. Therefore, such buildings are more susceptible to earthquake damage due to resonance effects.

Feng T et al. (2022) combined the machine learning-based earthquake location workflow (LOC-FLOW; Zhang M et al., 2022) and TomoDD to investigate the north-central section of the Xiaojiang fault. The LOC-FLOW method accurately detected seismic events from raw continuous waveform data, allowing the construction of a high-accuracy seismic catalog. Subsequently, the TomoDD method was applied to derive a high-resolution 3D velocity structure of the region. The seismicity pattern indicated a west-dipping Xiaojiang main fault (Figure 12). Tomographic analysis of the velocity structure demonstrated significant lateral heterogeneity, with distinct velocity variations across the fault. Notably, most earthquakes occurred at the boundary between high-velocity and low-velocity regions or within the high-velocity region situated above the low-velocity zone. Two seismic gaps were identified along the main fault zone, representing regions with relatively low earthquake occurrence, which can be significant indicators for assessing future seismic hazards in the area.

Figure  12.  v_P structure beneath the Xiaojiang fault zone along Line 4 in Figure 8. Black circles denote earthquakes. The white dashed line shows the main Xiaojiang fault, which is delineated from the seismicity distribution. The thick blue bars denote the position of the surface-mapped Xiaojiang fault. From Feng T et al. (2022).

DownLoad: Full-Size Img PowerPoint

4.   Conclusions

This review paper discusses various seismological methods used to study fault zone structure, including seismic tomography, fault zone wave analysis, and seismic activity studies. These techniques provide valuable structural information for understanding fault zone evolution and seismic rupture processes. We also discussed method limitations that result in inadequate detection capabilities for many active faults with high seismic activity. For instance, imaging resolution may be insufficient for distinguishing small-scale velocity anomalies, and important fault zone seismic waves, such as trapped and resonant waves, may go undetected. Additionally, there is room for improvement in terms of seismic location accuracy. To address these limitations, it becomes necessary to deploy dense or ultra-dense arrays in important fault zones to enhance seismic imaging resolution. By combining dense arrays with analyses of fault zone seismic waves and high-precision earthquake distribution, fault zone structures can be better constrained.

The development of complementary joint inversion can lead to a qualitative improvement in the study of fault zone structure. Moreover, different fault zone seismic waves impose different constraints on the fault zone structure. In the future, the utilization of full waveform inversion techniques, incorporating multiple seismic waves, may provide a more accurate characterization of fault zone structures. Overall, it is an ongoing challenge for continuously advancing and integrating various seismological techniques to improve our understanding of fault zone structures and their associated seismic behavior.