In this study, we collected the continuous waveform data recorded by three temporal seismic arrays: 195 stations of the ChinArray (X1) between October 2011 and October 2012, 114 stations of the Chuanxi array (CX) between October 2006 and July 2009 (Liu et al., 2014), and 50 stations of the Xiaojiang array (XJ) between January 2009 and December 2010 (Wu et al., 2013). In addition, we also use data recorded by 42 permanent stations of the Sichuan and Yunnan regional seismic networks for 1 year in 2009. There are 401 seismic stations in total in our study region. The average interstation space is about 30 km. The dense station distribution enhances the azimuthal coverage and path density, especially in DZB, compared with previous studies (e.g., Lu et al., 2014; Wang et al., 2015). Figure 1B shows the station distribution. Most stations are equipped with Guralp CMG-3ESPS seismometer (60s–50 Hz) and Reftek 130 digitizer.

The data processing technique for ambient noise tomography is currently in a relatively advanced stage. It mainly includes single station data preprocessing, empirical Green’s function extraction, phase velocity dispersion measurement, phase velocity tomography, and surface wave azimuthal anisotropy inversion. Empirical Green’s functions (EGFs) are derived from the cross-correlation functions between each station pair using the vertical component of continuous waveform following the method of Bensen et al. (2007) and Fang et al. (2009). Phase velocity dispersions from EGFs are measured using the far-field representation of the surface-wave Green’s function and image transformation technique following Yao et al. (2006). Figure 2 shows example of Z-Z cross-correlation functions (CCFs) between stations randomly selected from XJ array. The CCFs are filtered at 5–40 s period band. Figure 3 shows an example of the dispersion measurement of the phase velocity.

In the case of weak anisotropic medium and ignoring 4 φ terms, Rayleigh wave phase velocity c ω , M , φ at an arbitrary point M for each angular frequency ω and azimuth φ can be expressed as (Smith and Dahlen, 1973): c ω , M , φ = c 0 ω 1 + α 0 ω , M + α 1 ω , M cos ⁡ 2 φ + α 2 ω , M sin ⁡ 2 φ (1) Where c 0 ω is the reference phase velocity, α 0 and α i i = 1,2 are the isotropic phase velocity perturbation and the azimuthal anisotropy coefficients, respectively. During the inversion, these parameters are determined using the continuous regionalization method of Montagner (1986) and the generalized inversion algorithm of Tarantola and Valette (1982) for each period. The magnitude of the anisotropy, Λ , and the Fast-wave Polarization Direction (FPD), Φ , are determined with Equations 2, 3, respectively. Λ = α 1 2 ω , M + α 2 2 ω , M 1 / 2 1 + α 0 ω , M (2) Φ = 1 2 tan − 1 α 2 ω , M α 1 ω , M (3)

The inversion for α i i = 0,1,2 is controlled by three parameters: the standard error of phase velocity measurements, δd, the a priori parameter error, δp (which constrains the anomaly amplitude), and the correlation length Lc (which constrains the smoothness of the resulting model). In our inversion, following Yao et al. (2010), we set δd to 2% for all measurements. For a given α ₀ , δp is set to be twice that of the standard deviation (in percent) of all observed phase velocities at each period with a minimum value of 0.15 km/s. For α ₁ and α ₂ , δp is set to be 1.5% of the average phase velocity at each period. In this study, we focus on short and intermediate period surface wave. It may result in artificial velocity anomaly if the correlation length Lc is too small. So we set the correlation length Liso=max (30 km, C ₀ *T/2) for the isotropic term, where C ₀ is the average phase velocity at a certain period, T. The correlation length for the azimuthally anisotropic parameters is set to be 2*L _iso at the corresponding period.

The knowledge of the resolution is vital in the interpretation of the tomography results. Artificial values may be introduced in inversion due to insufficient data and poor ray path coverage. Following Simons (2002) and Yao et al. (2010), we performed checkerboard tests to assess the resolution of our model. We calculated the synthetic data from input models with a 1°×1° anisotropic pattern along with the fast azimuthal axis-oriented NE and NW, with 5% isotropic anomalies at each period (Figure 4A). We added the random error with a maximum of 1% to the synthetic data. We performed the inversion with the same parameters as for the actual data.

The checkerboard tests indicate that for the azimuthal anisotropy parameters, the 1°×1° patterns are well recovered throughout our study area (Figures 4B–D). The azimuthal anisotropy is not well recovered in the margin due to relatively poor azimuthal path coverage. Generally, checkerboard tests provide qualitative information about spatial resolution. However, we must remember that the checkerboard test differs from the inversion of actual data due to its inherent limitations, as pointed out by previous studies (e.g., Simons, 2002). For example, the limits of ray approximation, the horizontal smoothing imposed in the inversion, and the trade-off between lateral heterogeneities and anisotropy (Pandey et al., 2015).

Tarantola and Valette (1982) pointed out that calculating the posterior covariance operator of the model parameters after inversion can estimate the posterior error and resolution of the inversion parameters. If the posterior error is significantly smaller than the prior error, the inversion results are reliable, and the resolution is high. Conversely, when the posterior error is close to the prior error, it indicates that the data cannot resolve the parameter. Figure 5 shows the posterior error distribution of the anisotropic intensity after inversion. Its amplitude is much smaller than the prior error of the anisotropic intensity and the anisotropic intensity in the study area. Taking 10s as an example, in the central, the posterior errors of the anisotropic intensity Ac are all less than 0.3%. Therefore, the inversion results are reliable for regions with anisotropy strength greater than 2%.

The SMB has frequent seismic activities, dominantly about magnitude 6.0, showing the characteristics of stronger and less large earthquakes. The present crustal deformation and intense earthquake activities in this area are closely related to the collision of the Indian plate and the Eurasian plate and the lateral extrusion of the TP. Therefore, understanding the velocity structure and anisotropy characteristics will help us further understand the dynamic relationship between the present crustal deformation process and the material extrusion in the southeastern TP.

The velocity anomaly in the SMB is heterogeneous. At 14–32 s periods, it is mainly sensitive to the S-wave velocity structure in the depth range of 20–45 km (Figure 7). The area near the Wuliangshan fault zone (WLSF) continues to exhibit low-velocity anomalies, and the other areas show relatively high-velocity anomalies relative to short periods. Regarding azimuthal anisotropy (Figures 8A, B), the FPDs in the SMB show noticeable regional variance. In the central, the FPDs are generally similar to the fault strikes and the orientation of the principal compressive stress (Sheng et al., 2022). Among them, the FPD near the LCJF in the west is NE-SW direction, the FPD near the WLSF zone in the middle is near the N-S direction, and the FPS near the RRF zone in the east is NW-SE direction. These phenomena are consistent with the phase velocity azimuthal anisotropy of Wang et al. (2015), the group velocity azimuthal anisotropy of Lu et al. (2014), and Pms anisotropy (Chen et al., 2013; Sun et al., 2015; Cai et al., 2016) (Figure 9B).

Previous studies have shown that the RRF experienced a left-lateral strike-slip movement at 32–17 Ma and transformed into a right-lateral strike-slip movement at about 5 Ma (Gilley et al., 2003). In contrast, the XJF zone started to experience an initial left-lateral strike-slip movement from the Middle Miocene to the Early Pliocene (about 17–5 Ma) (Roger et al., 1995). At this time, the Chuandian diamond block (including the CXB and DZB) began to rotate and extrude clockwise along the XJF. During this process, the Chuandian diamond block was bound to have a robust southward extrusion on the SMB. Since the Oligocene (37–21 Ma), the Lincang granite in the southern SMB has experienced a gradual uplift from south to north (Shi et al., 2006). Furthermore, with the obstruction of the Lincang granite belt, the crustal structure of the central SMB was deformed strongly, forming a northwardly protruding honeycomb-shaped structure.

Therefore, when subjected to compressive stress in the southwest direction of the Chuandian diamond block, the crust of the eastern SMB underwent shear deformation, forming crustal anisotropy parallel to the strike of the RRF. Meanwhile, the crust of the western SMB was blocked by the Lincang granite belt and strongly deformed, forming crustal anisotropy parallel to the strikes of the Lincang granite belt and the LCJF. Likewise, the central is in the transition zone of differential deformation on both sides, forming near N-S trending crustal anisotropy, which is almost consistent with the extensive development of folds and thrust structures in the Mesozoic-Cenozoic strata in the block (Guan et al., 2006).

The SMB has undergone 50°–70° clockwise rotational deformation under the combined action of the Chuandian diamond block since the Pliocene (Sato et al., 2007). The GPS velocity field shows that, relative to the South China block, the NNE pushing caused by the collision of the Indian plate and the Eurasian plate (Wang and Shen, 2020) and the gravitational potential energy generated by the plateau uplift resulted in the eastward extrusion of the TP (Li et al., 2021). After encountering the obstruction from the stable South China block, the TP material moved southeast and southward, causing the Chuandian diamond block to rotate clockwise around the eastern Himalayan tectonic knot. Therefore, we infer that the compressive stress in the southwest direction of the Chuandian diamond block has a controlling effect on the crustal deformation of the SMB, and the west of the block was blocked by the Lincang granite belt and experienced intense tectonic deformation (Figure 10).

In the middle RRF and its vicinity, the anisotropic FPD at 4–24 s periods (corresponding to the crustal depth range, Figure 7) is near the N-S direction (Figure 5), showing a large angle with the fault strike. The azimuthal anisotropy of the surface wave group velocity, Pms anisotropy, and S-wave azimuthal anisotropy (Figure 9B) also show almost the same characteristics (Lu et al., 2014; Cai et al., 2016; Liu et al., 2019). The RRF zone, as a critical plate boundary cutting the Moho surface (Wang et al., 2014), did not cause the FPD of the anisotropy in the crust to be parallel to the strike of the fault. The tectonic stress field obtained from the focal mechanism (Xu, 2001; Wu et al., 2004; Xu et al., 2016) and the GPS velocity field (Jin et al., 2019) show that the maximum principal stress in the vicinity of the middle RRF zone is in the near N-S direction, which is consistent with the FPD of azimuthal anisotropy (Figure 9A). The anisotropy of the upper crust is generally considered to be related to the shape preferred orientation of microstructures (Rabbel and Mooney, 1996; Crampin and Peacock, 2008). Secondly, fault zones may also affect it, but the influence is limited (Gao et al., 2011). The GPS velocity field (Figure 1A) and seismic activity suggest that the middle RRF zone is completely locked from the surface to 8 km depth (Zhao et al., 2015; Wang et al., 2022). We speculate that the anisotropic material in the upper crust of the middle RRF zone may not have a dominant arrangement along the fault strike but is related to the action of the regional tectonic stress field (Figure 10).

At 4–32 s periods, the Rayleigh surface wave phase velocities in the eastern and western sides of the SZMLF have noticeable velocity differences (Figures 8A, B, D), which correspond well with the S-wave and P-wave velocity variations (Wu et al., 2013; Yang et al., 2020). Unlike the velocities on the west side of the fault zone with the periods changing, the east side of the fault zone continues to show relatively high phase velocity anomalies. The apparent velocity differences may reflect the inhomogeneity of the crustal structures and lithologies on both sides. The receiver function showed the crustal thickness on both sides of the SZMLF differs up to 10 km (Wang et al., 2017) (Figure 8D). The geochemical study of the volcanic rocks in the northern SZMLF zone shows that it is mainly alkaline basalt, characterized by low TiO₂ and high Al₂O₃, which is different from the Emeishan continental overflow basalt with high TiO₂ and low Al₂O₃ characteristics (Dong et al., 2002). The Bouguer gravity anomaly and the equilibrium gravity anomaly along both sides of the SZMLF zone are high in the north and low in the south, which shows that the SZMLF is the block boundary separating the YZB and CYB with different crustal structures. In addition, the thermochemical states of the upper mantle in South China based on multi-observation probability inversion (Yang et al., 2021) show that in the lithospheric mantle, the average density in this region also takes the SZMLF zone as the western boundary to separate the YZB and CYB. Coupled with the complex tectonic stress at the junction of the blocks leading to prominent regional characteristics of the FPDs, we infer that the SZMLF zone may be the western boundary between the YZB and CYB in the crust (Figure 10).

The SCB located at the northwest of the YZB. In the northeastern study area, the crustal anisotropy of the southern SCB at 4–20 s periods is dominated by NE-SW and NEE-SWW trends (Figure 8C), which is consistent with the NE-SW trend of the surface eastern Sichuan fold belt (Li et al., 2014; Xiong et al., 2016). The shear wave splitting of local earthquakes (Tai et al., 2015), Pms anisotropy (Sun et al., 2012), and S-wave azimuthal anisotropy (Zhu et al., 2021) also show the same crustal anisotropy characteristics. Since the Cenozoic, the eastward compressive stress of the TP may have been transmitted to the eastern Sichuan fold belt through the hard SCB, leading to an uplift of the east Sichuan fold belt (1–2 mm/a) in this area (Shi et al., 2016; Yuan et al., 2018). We believe that the azimuthal anisotropy of the upper crust is related to the high-angle stratigraphic directional arrangement caused by the eastern Sichuan fold belt deformation, which can be considered as the response evidence for the eastward extrusion of the TP. At 22–34 s periods (Figures 9B, C), the FPD of the azimuthal anisotropy is mainly in the NW-SE direction, and the anisotropy amplitude is greater than 2%, basically consistent with the direction and magnitude of the SKS anisotropy (Liu et al., 2020). We deduce that the lithosphere has undergone significant deformation.

At the period of 4–30 s, the area near PZH exhibits high phase velocity anomaly (Figure 8C), which is in good agreement with the characteristics of high Poisson’s ratio, high wave velocity, high resistivity, high density, positive magnetic anomaly, and low terrestrial heat flow (Chen et al., 2015; Wang et al., 2017; Wu et al., 2013; Cheng et al., 2017; Shen et al., 2015; Xu et al., 2015; Teng et al., 2019), all distributed in the inner core of the Emeishan large igneous rock province. Moreover, the FPDs in the vicinity of PZH have the characteristics of the NS direction at 4–30 s periods. Another notable feature is a large area of low-velocity anomalies near the Xiaojinhe fault (XJHF) zone in the north, which is consistent with the S-wave velocity obtained by Liu et al. (2014). The phase velocities are sensitive to the S-wave velocities of the mid-to-lower crust at 20–34 s periods (Figure 7). Within this period, the FPD near the northern XJHF zone gradually changed from NS to NE-SW direction, roughly parallel to the fault zone.

Geochemical and regional geological studies (Xu and Zhong, 2001) indicated that during the period of 263-251Ma in the Late Permian, a famous tectonic activity occurred in the western YZB, the magmatic activity of the Emeishan great igneous rock province (Xu and Chung, 2001), whose dynamic mechanism may be related to mantle plume activity. Both sources of intracrustal magmatic intrusions and eruptive basalts come from the deep mantle. According to the dome structure and basalt thickness, the active center of the mantle plume is located near PZH (Shen et al., 2007; Xu and Zhong, 2001). We speculate that the high-velocity anomaly in the crust may be formed by the cooling and crystallization of a large number of high-density basic and ultramafic rock intrusions under the PZH area during the uplifting process of the early mantle plume (Wu et al., 2013; Yang et al., 2014; Zheng et al., 2016).

The XJHF, as the western boundary of the Yangtze block, obliquely cuts the Chuandian diamond block into two sub-blocks in CXB and DZB (Xiang et al., 2002; Xu et al., 2003). The topographic differences between the two sides of the XJHF fault zone are significant. To the north of the fault zone, many peaks with altitudes of 4500–6500 m, and the average altitude exceeds 3500 m. While the average altitude to the south of the fault zone rapidly drops to about 2000 m.

Geological investigations have shown that differential movements on both sides of the fault zone are apparent. The horizontal slip rate values of the sub-block in CXB toward SE are 2 mm/a higher than those of the DZB. Since the Late Quaternary, the average differential upward and downward movement rates have reached 1.0–1.3 mm/a (He et al., 1993; Xu et al., 2003). Furthermore, there are also considerable differences in vertical movements observed by level (Liang et al., 2013). What causes the significant differential motion between the two sub-blocks within the Chandian diamond block? Seismic tomography results show (Liu et al., 2014; Wang et al., 2020) that there is a pronounced lower crustal flow in the CXB, and the flow direction is consistent with the extrusion direction of the block to the south. We suggest that the southward extrusion of the TP material is blocked by the hard intracrustal masses with PZH as the core, resulting in a rapid uplift of the northern topography.