In the past few decades, a large number of geophysical explorations were carried out in the Xiaojiang fault zone and adjacent areas, mainly including GPS, seismic geology, fluid geochemistry, seismicity, historical earthquakes and coseismic displacement of large earthquakes, etc. The results of these studies helped us have a better understanding of the fault structure characteristics, movement attributes, seismogenic environment and dynamic mechanism of the Xiaojiang fault zone. In terms of deep structure, the existing researches are limited by factors such as the density of observation stations, and most studies focused on the structural background on the regional scale, and few are specifically on this fault zone. The implementation of Phase I of the China Earthquake Science Array(ChinArray)detection project provides a good data basis for the study of the fault structure in Yunnan. It is of great practical significance for earthquake prevention and disaster mitigation to carry out deep structural detection of the Xiaojiang fault zone and clarify the fine crustal structure of the fault and its adjacent areas.

S-wave velocity is an important parameter to determine the crustal structure, physical state difference and tectonic evolution process. Extracting the P-wave receiver function from teleseismic body-wave waveform data and inverting it is one of the important methods to obtain the crustal S-wave velocity structure at present. The traditional receiver function S-wave velocity structure inversion relies heavily on the selection of the initial model, which results in strong non-uniqueness inversion results. The two-step inversion method, which takes into account the low and high frequency receiver functions at the same time, effectively suppresses the dependence of the inversion process on the initial model, and improves the reliability of the inversion results.

Based on the three-component waveform data of 238 teleseismic events with epicentral distances ranging from 30°~90° and magnitude M≥5.8 recorded by 48 broadband seismic stations in the Xiaojiang fault zone and adjacent areas from September 2, 2011 to January 16, 2014, this paper calculates the low-frequency(α=1.0)and high-frequency(α=2.5)radial P receiver functions, respectively. Then, on this basis, the S-wave velocity structure below each station is inverted using the two-step inversion method and Bootstrap resampling technique and the deep crustal structure of the Xiaojiang fault zone and its adjacent areas is studied. The following conclusions are drawn:=

(1)The crustal S-wave velocity in the study area is obviously non-uniform in both lateral and vertical directions. The overall distribution is as follows: In the near surface, there is a low-velocity layer about 2~4km thick, which may be related to the distribution of shallow sedimentary rocks or Cenozoic soft overburden; The S-wave velocity in the middle and upper crust is alternately distributed with high and low velocity; There is an obvious low-velocity layer in the depth range of 20~35km, mainly intermittently distributed in the Sichuan-Yunnan diamond block west of the Xiaojiang Fault and the Indosinian block south of the Honghe Fault; Besides, there is also local distribution near the Shizong-Mile Fault.

(2)The low-velocity layer in the middle and north segments of the Xiaojiang fault zone are relatively developed, and it is most prominent in the middle segment, with a maximum thickness of about 28km. There is an obvious high-velocity zone in the depth range of 15~25km in the southern segment.

(3)The Poisson’s ratio in the study area is generally low(average 0.24), unevenly distributed, and has drastic lateral changes. The Poisson’s ratio in the Xiaojiang fault zone generally has a segmental feature of higher in the northern segment, the southern segment coming second, and lower in the middle segment. The corresponding relationship between the distribution of low velocity in the crust and Poisson’s ratio in the study area is not obvious, and most of the low velocity layers seem to lack the conditions for partial melting. The differences and inconsistencies in the geophysical results indicate that the deformation evolution mechanism and physical properties of the low velocity layers in the crust are relatively complex.

At 21:48:34 on May 21, 2021, a M_S6.4 earthquake occurred in Yangbi County(25.67°N, 99.87°E), Dali Prefecture, Yunnan Province, with a focal depth of 8km. The epicenter is located in Cangshanxi Town, Yangbi County. The earthquake occurred on a secondary fault zone on the west side of the Weixi-Qiaohou-Weishan Fault. The focal mechanism is right-lateral strike-slip with a small amount of normal faulting, and the earthquake sequence is the foreshock-mainshock-aftershock type. The Yangbi M_S6.4 earthquake is the largest earthquake in Yunnan Province after the Jinggu M_S6.6 earthquake on October 7, 2014, and the largest earthquake in the northwestern Yunnan after the Yongsheng M_S6.0 earthquake on October 27, 2001.
The preparation and occurrence of earthquakes are related to changes of regional stress field. The accurate measurement of changes in seismic wave velocity to monitor the changes in crustal stress over time is an effective way to make physical earthquake predictions. Studies have shown that before a strong earthquake, the physical properties of the medium in the crust will change, and this change can be reflected by changes in seismic wave velocity. Therefore, according to the temporal and spatial change characteristics of the seismic velocity in the crust, the physical properties of the crustal medium can be known. Ambient noise has a higher time sampling rate than natural seismic sources, is much more economical than artificial seismic sources, and has the advantages of repeatability, economy and environmental protection, so it is very suitable for tracking changes in the physical properties of the internal structure of the earth's crust.
In the previous work, we explored the application of seismic data and ambient noise cross-correlation technology to earthquake tracking, analysis and prediction. Based on the continuous waveform data of the Yunnan seismic network, the Rayleigh wave relative travel time offset time series of the station pairs are generated regularly and applied to the daily earthquake situation analysis and judgment. The occurrence of the Yangbi M_S6.4 earthquake is a good test for applying the results of wave velocity measurement from ambient noise to the earthquake analysis and prediction practice. Therefore, based on the continuous broadband vertical component waveform data recorded by the station pairs near the epicenter of the Yangbi M_S6.4 earthquake, and using the ambient noise wave velocity measurement method, the relative travel time offsets of the Rayleigh wave between the stations are obtained. A retrospective study of the regional wave velocity changes before the Yangbi M_S6.4 earthquake was performed to gain deeper insights into the mechanism of earthquake preparation. This study also provides a good reference sample for earthquake cases to measure the regional underground medium wave velocity changes and capture the characteristics of wave velocity anomalies before strong earthquakes in Yunnan area by using ambient noise technology.
Based on the continuous broadband vertical component waveform data recorded by 7 stations near the epicenter of the Yangbi M_S6.4 earthquake(4 stations of Yunnan network and 3 of Xiaguan network)from January 1, 2019 to May 21, 2021, the empirical Green's functions between 17 station pairs were extracted from the background noise cross-correlations. In the frequency range of 0.1~0.5Hz, the relative travel time offsets of the direct Rayleigh waves in the very day's empirical Green's function and the reference empirical Green's function of the station pair are measured directly using the cross-correlation time delay calculation method. On this basis, based on the Rayleigh wave relative travel time offset time series of 17 station pairs, with ±1.5 times the standard deviation as the anomaly threshold, the abnormal changes in the regional wave velocity around the epicenter of the Yangbi earthquake five months before the Yangbi earthquake were analyzed. In addition, based on the YUL--TUS, HDQ--BAS, EYA--CHT and YUL--CHT station pairs, the methods of “single station pair” and “multi-station pair combined average” are used to analyze the time-varying characteristics of the relative travel time offset near the epicenter. Conclusions are drawn as follows:
(1)From the perspective of the overall change of the abnormal station pairs, in the five months before the Yangbi M_S6.4 earthquake, the distribution of abnormal station pairs in relative travel time offsets of the Rayleigh wave experienced a change process from scattered distribution in the outer periphery to concentration to the epicenter, and during the whole change process, the abnormal station pairs mainly show negative anomalies(the relative travel time deviation exceeds -1.5 times the standard deviation line).
(2)The analysis of the relative travel time offset of Rayleigh waves based on the combined stations of YUL-TUS, HDQ-BAS, EYA-CHT and YUL-CHT near the epicenter shows that about 5 months(158 days)before the Yangbi M_S6.4 earthquake, the wave velocity of the underground medium near the epicenter showed an obvious accelerating trend, the amount of relative travel time offset change was -0.35%.
(3)The abnormal station pairs before the earthquake were concentrated near the epicenter, and dominated by negative anomalies. The wave velocity of the underground medium near the epicenter before the earthquake showed a significant acceleration trend, which was an obvious phenomenon observed before the Yangbi earthquake. Whether the above-mentioned change characteristics have to exist before strong earthquake requires further examination with more earthquake cases.
(4)Based on seismic waveform data, this paper uses ambient noise cross-correlation and cross-correlation time delay calculation methods to obtain the relative travel time offset of Rayleigh waves of station pairs and analyze the changes in wave velocity near the epicenter. The physical meaning is clear. The wave velocity measurement by ambient noise does not depend on a specific seismic source, and the data results have a high time sampling rate and can be updated every day. The method of measuring wave velocity from ambient noise is expected to become a new method and new approach for earthquake prediction.

In 2018, a short-period seismic network was set up in Eryuan area of Yunnan Province to carry out continuous field observation of the sub-instability process of the earthquake. The relevant data of the Yangbi M_S6.4 earthquake sequence are mainly from the waveforms recorded by this network, combined with some other stations from Yunnan regional seismic network. The Yangbi earthquake sequence shows that the events in this area began to occur intensively on May 18. A total of 2 000 earthquakes with M>0.1 were recorded from May 18 to 23, including 770 foreshocks.

Seismicity analysis shows that two clusters of foreshocks occurred successively in the adjacent area of the main earthquake in the northwest segment of the rupture strip within 3 days, then in the subsequent impending period(within 1 hour before the main shock)about 60 events spread symmetrically from the center of the fracture zone to the ends. The spatial distribution of foreshocks in different periods shows the spatial migration of local fractures and accelerated expansion prior to the main shock. The spreading speed is about 5km/d from foreshock clustering process to 96km/d in impending earthquake period. The epicenter of the main shock is located at the edge of the cluster foreshocks and the northwest end of the final rupture zone. Subsequent aftershocks extend southeastward to the whole fracture zone in about half an hour, and the final fracture zone is more than 20 kilometers long, showing unilateral propagation of the rupture. Since 2018, b-value in the Yangbi area has been stable(0.9~1.1)for the past three years. After March this year, the b-value abnormally decreased to 0.6 before the main shock, reflecting that there was a significant process of continuous increase of local stress before the Yangbi earthquake.

The identification of short-term precursors and somehow definite information is one of the focus problems in earthquake prediction research. On the basis of the experimental results, Ma Jin proposed the theory of seismic meta-instability stage based on the characteristics of the load stress after the peak value from rock experiments and the corresponding change of related physical field, and considered that the degree of fault activity synergy was a sign to determine the stress state of the fault. When the fault activity changes from the expansion and increase of the stress releasing points in the early stage of meta-instability to the connection between the released segments at the late stage of meta-instability, that is, the quasi dynamic instability stage, the stress release on the fault will accelerate, and the acceleration mechanism is the strong interactions between the fault segments. In the context that the macroscopic stress state cannot be known directly, the original intention of the “meta-instability” test area is to try to capture the characteristic signal of the meta-instability stage described by the experimental phenomenon through the deformation and seismicity of the actual faults during the earthquake preparation process. It is clear that in this stage, the fault will continue to expand in the pre-slip zone theoretically, and it will enter into the quasi dynamic fracture expansion before the impending earthquake. This theory is obviously embodied in the foreshocks of this earthquake, forming the phenomenon of rapid migration of small earthquakes as mentioned above. From the current understanding of the meta-instability, it can be seen that the seismogenic fault is in the state of overall stress release at this stage, rather than the continuous increase of stress. Therefore, the decrease of b value before the earthquake shows that local faults have been activated and entered the final stage of nucleation process. The quasi dynamic spreading phenomenon before this kind of moderate-strong mainshock displayed by small earthquake activity can be identified as the precursor of a kind of earthquakes.