Journal of Earth Science  2018, Vol. 29 Issue (6): 1419-1430 PDF     0
Crustal Structure of Yunnan Province of China from Teleseismic Receiver Functions: Implications for Regional Crust Evolution
Fang Wang1, Shuangxi Zhang1,2,3, Mengkui Li1
1. Department of Geophysics, School of Geodesy and Geomatics, Wuhan University, Wuhan 430079, China;
2. Key Laboratory of Geospace Environment and Geodesy of the Ministry of Education, Wuhan University, Wuhan 430079, China;
3. Collaborative Innovation Center of Geospace Information Science, Wuhan University, Wuhan 430079, China
ABSTRACT: Yunnan Province is located on the southeastern margin of Tibet and represents an important marker in understanding the tectonic evolution of Tibetan Plateau. In this study, we calculated teleseismic P-wave receiver functions at 49 permanent broadband seismic stations in Yunnan Province and estimated crustal thickness and the bulk crust ratios of P-wave to S-wave velocities using the H-κ method together with more detailed crustal structural profiles from the common conversion point stacking method. There is a significant transition of Moho interface and lower crustal composition along latitude 26°N in northwestern Yunnan. Decrease of crustal thickness with a concomitant increase of Poisson's ratio occurs at station CUX. An interesting phenomenon is that a step-like Moho fashion is observed at several stations, which might correspond to local thermal activities, such as partial melt/lower crust delamination. Our results show changes in crustal properties appear to be associated with varieties in upper mantle structure and compositions. We propose the controlling factor of the dynamic processes below 26°N is the result of eastern forward subduction of the Indian Plate; the northern part is controlled by the redirected material flow from the SE Tibet.
KEY WORDS: crustal structure    receiver function    Poisson's ratio    tectonic evolution

0 INTRODUCTION

Located at the boundary of the Indo-Australian and Eurasian Plate, Yunnan is an essential part of the continent- continent collision in southeastern (SE) Tibet (red polygon in Fig. 1, inset). The area is actively deforming, best expressed by the fact that about 42.4% of 20th century earthquakes with magnitude greater than 6.0 in mainland China occurred here. Besides, central Yunnan is host to significant Quaternary volcanism and active hot springs in Tengchong. Complex distributed fault systems, such as Nujiang fault, Lancangjiang fault, Red River fault, Xiaojiang fault (Fig. 1), play an important role for regional tectonic evolution (Zhang et al., 2011; Bai et al., 2010; Huang et al., 2010; Yao et al., 2010, 2008; He et al., 2009; Li et al., 2009; Xu et al., 2005). These approximately N-S faults divide Yunnan into four major geological units, Eastern Yunnan Block (Ⅰ); Sichuan-Yunnan Diamond Block (Ⅱ); Indo-China Block (Ⅲ); Yunnan-Burma-Thailand Block (Ⅳ) (Fig. 1) (Hu et al., 2012, 2005) with the stable Sichuan Basin and several orogenic belts surrounding Yunnan Province.

 Download: larger image Figure 1. Topography of Yunnan Province and surrounding regions. White arrows in the inset denote the inferred lower crust flow (Bai et al., 2010). The orange bold lines mean faults. Blue bold dash line means predicted transition belt. Green bold lines are the boundaries between major blocks. Tectonic units: Ⅰ. Eastern Yunnan Block; Ⅱ. Sichuan-Yunnan Diamond Block; Ⅲ. Indo-China Block; Ⅳ. Yunnan-Burma-Thailand Block. There are five major faults in this study area, NJF. Nujiang fault; LCJF. Lancangjiang fault; JSJ-RRF. Jinshajiang-Red River fault; ANH-XJF. Anninghe-Xiaojiang fault; LJ-XJSJF. Lijiang-Xiaojinshajiang fault. A-A', B-B', C-C', D-D', E-E' and F-F' are locations of six crustal CCP profiles. Distribution of teleseismic earthquakes color-coded with the event depth are shown in the right panel. The two concentric circles represent the 30° and 90° distances to the center of the study area (red star), respectively. The figure is generated using Generic Mapping Tool (http://gmt.soest.hawaii.edu/).

A number of geophysical studies have been conducted in Yunnan and adjacent areas (e.g., Bai et al., 2010; Yao et al., 2010, 2008), which have mainly focused on understanding the role that lower crustal flow may have in accommodating strain related to the continent-continent collision. However, debate still exists over the extent of the low velocity zones (LVZs) (Xu et al., 2013; Wang et al., 2010; Yao et al., 2008; Clark and Royden, 2000) and the connectivity (Li et al., 2016; Bao et al., 2015) of specific low velocity channels (Chen et al., 2014; Bai et al., 2010).

Several models have been proposed to account for the low velocity lower crustal anomalies in SE Tibet and Yunnan, which are the most commonly associated deformation of the Tibetan Plateau (Clark and Royden, 2000; Royden et al., 1997). Channel flow model (Li et al., 2016; Bao et al., 2015) argues that the continuous channelized middle-to-low crust flow is capable of flowing on geological time scale and two or more distinct flows are located in Sichuan and Yunnan following major geological unit or faults (Chen et al., 2014; Bai et al., 2010). The resulting low S-wave velocity structure is used to correspond to the eastward escaping material from central Tibetan Plateau (He et al., 2014; Hu et al., 2013, 2012; Sun et al., 2012). Regional flow model (Xu et al., 2013; Wang et al., 2010; Yao et al., 2008) posits the lower crust flow is observed within constrained regional area. The poor connectivity of LVZs suggests impossibility of extensive channel flow. The third model (Li and van der Hilst, 2010; Li et al., 2008) argues LVZs in the weakened lower crust beneath the Red River fault region are due to upper-mantle upwelling processes further southeast and connected to deep anomalies beneath the South China fold belt and South China Sea instead of the influence of the southward crust channel flows.

The existence of LVZs or crust flow in central Tibetan (Xu et al., 2015), eastern Tibetan (Chen et al., 2014; Bai et al., 2010; Wang et al., 2010; Zhang et al., 2009), and SE Tibetan (Chen et al., 2014; Bai et al., 2010) has been proved, while a further knowledge of redirected crust channel flow awaits exploration. Most of the convincing examples of lower crustal flow (He et al., 2014; Wang et al., 2010; Zhang et al., 2009) show the average continental crust VP/VS ratio ranges at 1.8 (i.e., high Poisson's ratio of ~0.28) with low S-wave velocity ranging from 2.4 to 3.2 km/s.

It's not rigorous to correspond the LVZs in Yunnan to the redirected lower crust channel flow because many rocks with different physical states have similar seismic velocities (Lees and Wu, 2000) especially in such a geologically complex area. Besides, increased fractures also cause low S velocity (Sanders et al., 1995). The widely distributed faults of Yunnan lead to many fractures. Thus seismic velocity alone is not a better indicator of variable lower-to-upper crust rock property. For this reason, we consider the Poisson's ratio, a powerful crustal parameter (Owens and Zandt, 1997), to differentiate the lateral variation of the low crust flow, considering the thick (i.e., 10–15 km), hot and viscous flow.

To evaluate the plausibility of the redirected lower crust channel flow (Hu et al., 2013, 2012; Royden et al., 2008; Clark and Royden, 2000) to Yunnan, a thorough knowledge of the lower crustal composition and crustal structure beneath Yunnan are crucial for the existing models. Although previous studies have obtained the distribution of the crustal thickness and VP/VS ratio (Li M K et al., 2016; Hu et al., 2012, 2005; Sun et al., 2012; Zhang et al., 2011; He et al., 2009; Li Y H et al., 2009; Xu et al., 2007), seismic anisotropy (Sun et al., 2012; Shi et al., 2008), LAB boundary (Hu et al., 2012), upper mantle discontinuity of the 410 and 660 km (He et al., 2014; Hu et al., 2013), Bouguer gravity anomalies (Lou and Wang, 2005; Wang et al., 1982) and heat flow distribution (Hu et al., 2000) in this area and the neighboring regions, the detailed lateral variation of crustal structure by P-wave receiver functions (RFs) in depth domain has not been investigated. Hu et al. (2012) gave three common conversion point (CCP) profiles using S-wave receiver functions (SRFs), while it cannot resolve fine crustal structures than P-wave receiver functions (Hu et al., 2012) due to its strong attenuation in mantle.

In this paper, we use teleseismic receiver function method (Zandt and Ammon, 1995; Langston, 1979) based on detailed CCP profiles, which can not only denote the crust geometry but also can be used to indicate the lower crust thermal activity represented by sharp intra-crust impedance, such as the steplike/double Moho (Huang et al., 2014) indirectly, to study the detailed crust structure of Yunnan Province. Assuming the uniform and extensive soft material redirected from SE Tibet, we will see the strong variation of crust composition. Especially we focus on the details of the major channel of lower crustal flow in southeastern Tibet (Li et al., 2016; Bao et al., 2015; Bai et al., 2010; Xu et al., 2007). The detailed lateral variations of crustal parameters might give us more knowledge about the regional tectonic evolution.

1 DATA AND METHODOLOGY

Teleseismic receiver functions are a useful tool for estimating crustal thicknesses and average crustal VP/VS ratios beneath seismic stations (He et al., 2014; Zhu and Kanamori, 2000; Zandt and Ammon, 1995). It can sample the interior structure of crust and upper mantle directly beneath a seismic station and thus provide evidence of variations of crustal structure in the study area. The difference between the VP/VS values of bulk crust and that of lower crust is only about 0.02 (He et al., 2014). Thus bulk crustal VP/VS can be approximated as the lower crust VP/VS ratio (He et al., 2014; Thompson et al., 2010; Niu and James, 2002).

The mean VP/VS ratio and the Poisson's ratio, which are related with the composition of the lower crust (Christensen, 1996), can be derived from the relationship (σ=0.5[1− ((VP/VS)2−1)-1]) and might be better diagnostic of crustal composition than either P or S wave velocity alone (Wang et al., 2010; Christensen and Fountain, 1975), constraints of thermal anomalies and possible partial melt in the study region for common situation. If the crust anisotropy is strong, we cannot use VP/VS to relate to media composition (Hammond, 2014).

1.1 Pre-Processing of Raw Three-Component Seismic Data

The waveform data in our study were provided by Data Management Center of Yunnan Seismic Network, China Earthquake Administration. There are 49 permanent broadband seismic stations deployed in Yunnan Provincial Seismic Network (Fig. 1 and Table 1). Since we used continuous waveform data and had to cut the event out. The presence of P-wave arrival is near (±5 s) the predicted time. We chose relative long time window (40 s) before P-wave phase. All vertical-component waveforms were visually inspected to see if they have a clear primary phase and small pre-event noise to ensure relative good signal-to-noise ratio (Huang et al., 2014). Three- component seismic data were filtered with a band-pass filter from 0.05–2 Hz. Eventually we selected 828 earthquake events of magnitude larger than 5 during the period from 2011 to 2014 at epicentral distances from 30° to 90° to secure near-vertical incidence of the P waves (Fig. 1).

Table 1 Station locations and crustal thickness and VP/VS estimate

Most seismicity occurred around the circum-Pacific seismic belt (Fig. 1). And events from Europe and the Indian Ocean provide good back-azimuthal coverage. Our data used a total of 160 s time series (40 s before P, and 120 s after) to calculate teleseismic receiver function at each station, using a time domain iteration deconvolution method (Li et al., 2016; Bao et al., 2015; Hammond, 2014; Huang et al., 2014; Ligorría and Ammon, 1999). A 1 Hz Gaussian low-pass filter was applied to the deconvolution results. Occasionally, low signal-to-noise ratio waveforms result in a poorer quality RF but the abundance of available data can somewhat compensate the lack of quality.

1.2 Selection of Receiver Functions and the H-κ Method

Ideally, for receiver function study, we expect the resulting RFs for each station to show a high level of coherency. While for a dipping Moho interface, amplitude and arrival time of the crustal phases vary periodically with the back-azimuth. And stacking results from up-dip zone will better fall in with actual crustal structure (Wang P et al., 2010). Besides, anisotropy (Savage, 1998) and sediments (Yeck et al., 2013) could also affect the resulting RFs. In the H-κ stacking technique a flat Moho is assumed (Zhu and Kanamori, 2000), thus ignoring these effects will lead to various results.

It is relatively easy to pick a small amount of radial receiver functions (RFs), however, we need to automate the selection process due to the large amounts of data used in this study. Typically, RFs selected by hand might cause the distortion and loss of representative features (Li et al., 2016; Tkalčić et al., 2011). We use the cross correlation matrix (CC-M) method (Li et al., 2016; Tkalčić et al., 2011, 2006). Here we give a brief description of this statistical method (for more details, please refer to Tkalčić et al., 2011). Assuming that we have M RFs after deconvolution and each of them has N sample points, the cross-correlation pairs are any selected time-series (RFs), and a cross-correlation matrix can be constructed in the form of Eq. (1), where rij are the cross-correlation coefficients between two RFs. The matrix is symmetric with rij equal to rji.

 $C = \left[ {\begin{array}{*{20}{l}} {{r_{11}}}&{{r_{12}}}& \cdots &{{r_{1j}}}& \cdots &{{r_{1M}}} \\ {{r_{21}}}&{{r_{22}}}&{}&{}&{}&{} \\ \vdots &{}& \ddots &{}&{}&{} \\ {{r_{i1}}}&{}&{}&{{r_{ij}}}&{}&{{r_{iM}}} \\ \vdots &{}&{}&{}& \ddots &{} \\ {{r_{M1}}}&{{r_{M2}}}& \cdots &{{r_{Mj}}}& \cdots &{{r_{MM}}} \end{array}} \right]$ (1)

There are two steps to use the algorithm. First we need to compute the pre-defined matrix given above. Then there are two pre-determined control parameters χ and τ that should be satisfied. They are the thresholds for the every cross-correlation coefficient in the matrix and a given percentage of the selected waveforms to all the RFs, respectively. Empirically, the value χ and τ are set to 0.9 and 25 (Li et al., 2016; Tkalčić et al., 2011, 2006), using the trial-and-error approach, in the cases when the number of RFs is larger than 50 (a common case). One can adjust these values due to the practical situation. An example for stations CAY and CUX is shown below in Fig. 2.

 Download: larger image Figure 2. Radial RFs computed for stations CAY and CUX without selecting waveforms are shown in gray dash lines. The blue lines are the RFs using the cross-correlation matrix approach with χ=0.95 and τ=0.2. The bold red line is the average of our selected RFs. The Gaussian low-pass filter is 1.0 as is mentioned above.

After this selection process we obtained 8 721 RFs (Table 1). To suppress the influence of random noise, RFs with similar back-azimuths and ray parameters were stacked at a station (Huang et al., 2014). The number of RFs is limited at the stations deployed in northwestern Yunnan compared with other stations. Comparison of the RFs obtained in northwestern stations with adjacent stations shows a more complex RFs, particularly the multiple phases (PpPs and PpSs+PsPs phases). Thus at these stations RFs were selected manually to maximize the number of RFs included in the analysis.

The classical H-κ stacking method (Li M K et al., 2016; He et al., 2014; Huang et al., 2014; Sun et al., 2012; Wang C Y et al., 2010; Li Y H et al., 2009; Zhu and Kanamori, 2000) was used to estimate the crustal thickness H and VP/VS ratio κ beneath each station. This method calculates the time delays of Moho P-to-S conversions and the following crustal multiples to determine H and κ simultaneously. The H-κ domain stack function, s (H, κ), is defined as

 $s(H, \kappa) = {\omega _1} \times RRF({t_1}) + {\omega _2} \times RRF({t_2}) - {\omega _3} \times RRF({t_3})$ (2)

where RRF(t) is the radial receiver function, t1, t2 and t3 are the predicted Ps, PpPs and PpSs+PsPs arrival times. ωi (i=1, 2, 3) are weighting factors that obey ω1+ω2+ω3=1. The RFs of each station were then stacked to get maximum stacking amplitude at the optimum values of H and κ. The uncertainties of crustal thickness and VP/VS ratio were given by (He et al., 2014; Wang C Y et al., 2010; Zhu and Kanamori, 2000)

 $\sigma _H^2 = \frac{{2{\sigma _s}}}{{{\partial ^2}s/\partial {H^2}}}$ (3)
 $\sigma _\kappa ^2 = \frac{{2{\sigma _s}}}{{{\partial ^2}s/\partial {\kappa ^2}}}$ (4)

where σs is the estimated variance of s(H, κ) from stacking. Here we used weighting factors of 0.6, 0.3 and 0.1 (Li et al., 2016; He et al., 2014; Wang C Y et al., 2010) for ω1, ω2 and ω3, respectively. To perform the H-κ method, one should specify the average crustal P velocity because a 5% uncertainties in VP can cause ~2 km thickness change (Wang C Y et al., 2010; Zhu and Kanamori, 2000). A crustal P velocity of 6.2 km/s was used referring the refraction exploration in Yunnan (Zhang and Wang, 2009).

2 RESULTS 2.1 Crustal Thickness and Poisson's Ratios in Yunnan

RFs of 6 representative stations and their H-κ stacking results are displayed in Fig. 3. The best estimate H and κ are denoted by the error ellipse computed using Eqs. (3) and (4). RFs at each station along with the predicted arrival times of Ps and other crustal multiples are quite clear. In general, RFs of all stations have clear Moho Ps phase and the coherence of receiver function is fine, suggesting that the CC-M algorithm introduced above selects coherent RFs well. The standard deviations of crustal thickness and VP/VS ratios are estimated to be less than 3.0 km and 0.1 for all stations (Table 1).

 Download: larger image Figure 3. RFs and H-κ stacking results of 6 example stations. The red ellipses indicate the best estimates of H and κ and their uncertainties. The black dash lines are the predicted arrival times of Moho Ps and its crustal multiples using the estimated H and κ values.

In general, the variation of Moho depth with latitude is weaker than that along longitude, except for station CUX, where anomalously thin crust (~33 km) occurs (Fig. 4). Moho depth increases from the southeast to northwest. The crustal thickness is relatively thinner (~33 km) in the southwestern Yunnan-Burma-Thailand Block. While, to the north, the western Sichuan-Yunnan Diamond Block has a thicker Moho depth at about ~50 km. At the northernmost station ZOD, the crust is about 52 km. To the east of Sichuan-Yunnan Diamond Block, the crustal thickness decreases to ~45 km. With the exception of station CUX, which will be discussed later, the Moho topography is flat (~40 km) in central Yunnan.

 Download: larger image Figure 4. Results of the Moho depth and Poisson's ratio beneath each station of Yunnan Province. Blue dashed line represents the transition belt. Profiles A-A' to F-F' are the same with those in Fig. 1.

Crustal Poisson's ratios range from 0.19 to 0.33, showing systematic spatial variations. Figure 4 shows stations (e.g., stations LUS, TUS, DAY, KMI, HLT, MAL and XUW) along latitude 26°N are of low Poisson's values ranging from 0.21 to 0.24 interspersed in the clearance of high anomalies and normal values except for stations YUL, YUM and LUQ where Poisson's ratios are 0.280, 0.257 and 0.266, respectively. There exist three obvious high Poisson's ratio regions with values ranging from 0.28 to 0.32. Western part of the Yunnan and Sichuan-Yunnan Diamond Block are characterized by higher Poisson's ratios of 0.28–0.3. There are high Poisson's ratios in Tengchong volcano area, Chuxiong basin and Tonghai area too. Poisson's ratios of 0.26–0.27 are observed in the eastern Yunnan Block, whereas in the southwestern part (Yunnan-Burma- Thailand Block), lower (< 0.22) ratios are observed (Fig. 4). For most areas, the value of the Poisson's ratio belongs to normal level (0.27).

2.2 Detailed Crustal Structures by CCP Stacking Method

The H-κ stacking method stacks RFs of all azimuths at a single station and assumes a simple isotropic horizontal layer overlying a homogeneous half space. Due to this, results are limited by the station spacing (~50 km). To get more detailed spatial images we employed the CCP stacking method of Zhu (2000) to image the crustal structure in greater detail along several profiles. The choice of profiles (Fig. 4) was motivated by predictions from the channel flow model (Li et al., 2016; Bao et al., 2015).

Amplitudes of all RFs of stations near each profile were projected to their Ps conversion points (Huang et al., 2014; Zhu, 2000) based on their time delays with respect to the first arrival and using 1D IASP91 background velocity model (Kennett and Engdahl, 1991) and H-κ stacking results under individual stations. The crustal volume was divided into 1-km-wide and 0.5-km-high bins and amplitudes of all RFs were stacked in each bin. We used the first Fresnel zone size as the width of the ray. It determines the lateral resolution of the CCP image which varies with depth, ranging from ~5 km in the upper crust to ~10 km at the Moho depth (Huang et al., 2014) using the dominant period of 1 s of RFs.

Figure 5 displays the detailed CCP profiles, which sample the possible branches of lower crustal flow as proposed by Li et al. (2016) and Bao et al. (2015). Though some gaps exist due to station spacing, we still can see a clear Moho interface along almost all profiles. In profile A-A', a bending Moho is shown at ~30 km depth beneath station CUX, while the Moho interfaces under the southeastern stations are relative flat except that station YIM has a step-like velocity boundary or two interfaces close to the Moho depth, which is perhaps related with lower crust delamination (Huang et al., 2014). For TOH station, we observed another weak velocity discontinuity above Moho, which can be seen in two mostly perpendicular CCP profiles (Fig. 5, A-A' and B-B'). Note the relative high Poisson's ratio occurs beneath this station with the existence of the interface. Similar results can also be acquired beneath SIM (Fig. 5, C-C'). Profiles B-B' and C-C' cross the Red River fault and Lancangjiang fault respectively. Both of them show strong, coherent and consistent Moho interfaces. However, for the rest of CCP profiles, the identification of Moho is clear and it shows much spatial variability. For example, step-like (indicated by blue arrows) interfaces are present in profile F-F'. These significant features will be discussed in the next section.

 Download: larger image Figure 5. Crustal structure imaged by CCP stacking of teleseismic RFs along six profiles shown in Figs. 1 and 4. EYB. Eastern Yunnan Block; SYDB. Sichuan- Yunnan Diamond Block; ICB. Indo-China Block; YBTB. Yunnan-Burma-Thailand Block. Red bold interfaces denote the observed Moho interface and blue dashed lines the predicted velocity discontinuities. The blue arrows point to the possible step-like locations. The ellipse using blue dashed line outlines the suspicious layer and a higher Poisson's anomaly occurs under that station. There is no vertical exaggeration except for surface topography on the top of profiles A-A' to F-F' in meters.
3 DISCUSSION 3.1 Comparison of Crustal Structures with Previous Studies

Figure 6 shows the H and κ values obtained in this study and previous seismic results. Results of each station are grouped by the geological units (Fig. 1) they belong to. In the Yunnan-Burma-Thailand Block, the crust thicknesses show great coherency. While VP/VS value under station TNC deviates with others, showing an intermediate result. The crust parameters beneath stations GOS and YUL in Indo-China Block coincide with the result of Li et al. (2016) but differ greatly from others. The results at station JIG in this study agree well with most results (He et al., 2014; Sun et al., 2012) except that the VP/VS value of Li et al. (2016) is higher than ours.

 Download: larger image Figure 6. Comparison with previous seismic results. Asterisk represents the result of the joint inversion of RFs and surface wave data. DSS is short for deep seismic sounding, a project conducted in the 1986s (Li et al., 2009).

Things become complex in Sichuan-Yunnan Diamond Block. The crust thicknesses under stations DOC and QIJ are similar while the VP/VS results of Li et al. (2016) are obvious lower than ours. There are two stations where the crust parameters differ greatly. One is station DAY and the other is station CUX. Our results at station DAY are relatively consistent with those of He et al. (2014) and Sun et al. (2012). And tomography result (Lei et al., 2009, Fig. 14b) shows a thick crust beneath station DAY, supporting our results.

Crustal parameters at station CUX are disputable. Results by Li et al. (2016) and Sun et al. (2012) show a relative thick crust (~50 km) with a VP/VS ratio of 1.7 (e.g., a low Poisson's ratio 0.23). While other results (Hu et al., 2012, 2005; Zhang et al., 2011) argue the Moho depth beneath station CUX is ~33 km, which coincides with ours. In DSS results (Kan and Lin, 1986), he points out there exists an upper mantle rise centered at Chuxiong Basin. At the same time, our CCP profile (Fig. 5, A-A') captures almost the same detail features in profile 5 of Kan and Lin (1986). A weak positive impedance at ~42 km under station CUX and two close interfaces in lower crust under station YIM conform to the credibility of our results. Wang P et al. (2010) also observed big discrepancies between reflection seismic results near Chuxiong basin. Perhaps the complex tectonic structure (Wang P et al., 2010; Kan and Lin, 1986) in Chuxiong rift basin and the limited active seismic energy fail to get a consistent and strong Moho interface under Chuxiong shooting point.

For a relative flat and simple crust case (Frederiksen and Delaney, 2015), a good azimuthal coverage, abundant data amount may lead to a desirable H-κ result. However it will become worse or error-prone if one doesn't consider some complexities, such as strong seismic anisotropy (Hammond, 2014; Savage, 1998), steep dipping Moho (Wang P et al., 2010; Lombardi et al., 2008) and sedimentary layer (Li et al., 2016; Yeck et al., 2013) under seismic station. CCP stacking technique can better indicate more precisely impedance change (Zhu, 2000) because it's based on the amplitude information of teleseismic RFs from all azimuths along their ray-paths. In this study, enough RFs and a good azimuthal coverage provide strong support to get a more stable result.

3.2 The Behavior of the Lower Crust in the Northwestern Yunnan

Poisson's ratio is sensitive to the composition, it increases with high mafic content and decreases with the silica content (Wang C Y et al., 2010; Owens and Zandt, 1997; Christensen, 1996). Values of Poisson's ratio between 0.25 and 0.30 have a variety of compositional implications and evidence suggests that the unmelted crustal rocks with Poisson's ratio greater than 0.30 are rare (Wang C Y et al., 2010; Owens and Zandt, 1997; Christensen, 1996). For common lower crustal composition, it's characterized by low Poisson's values (0.25), intermediate values (0.25–0.28), and high values (> 0.28) (He et al., 2014; Wang C Y et al., 2010; Zandt and Ammon, 1995). Also, high Poisson's ratios have often been linked with the existence of soft materials which has been used to suggest areas of lower crustal deformation and the connected channeling of crustal materials in Yunnan (Li et al., 2016; Sun et al., 2012).

The spatial distribution of the Poisson's ratio in our results shows systematic variations. High Poisson's anomalies are largely constrained above 26°N. The very high Poisson's ratios (> 0.3) at stations TNC and CUX suggest partial melting under the volcano areas (Li et al., 2009; Xu et al., 2007) and the areas with anomalously high heat flow. A band of low-normal continental Poisson's ratios along latitude 26°N indicates the strong transition of lower crustal composition. Delamination leads to a lower Poisson's ratio or features typical of felsic lower crust (He et al., 2014; Zandt and Ammon, 1995). And systematic felsic composition of continental lower crust is therefore considered to reflect the preferential removal of mafic lower crust by delamination (He et al., 2014; Christensen, 1996).

The three CCP profiles (i.e., D-D', E-E' and F-F' in Fig. 5) show strong lateral variations of Moho interface. Particularly, a rolling Moho discontinuities from station DAY to YUM with a ~20 km depth difference suggests a smooth transition between the Sichuan-Yunnan Diamond Block and Chuxiong Basin. Step-fashion Moho is related with the seismic phase transform and the thermal activity of delamination in the low crust and upper mantle (Huang et al., 2014).

The two or more redirected channels of lower crust flow previously suggested (Li et al., 2016; Bao et al., 2015; Bai et al., 2010) are not consistent with these results. These presences of normal Poisson's ratios below 26°N suggest that lower crustal deformation is limited above this latitude. Moreover, a thick crust is required for lower crustal flow (Clark and Royden, 2000; Hopper and Buck, 1996), whereas the crust below 26°N is less than 45 km.

3.3 Upper Mantle Rise near Latitude 26°N of the Study Area

The corresponding relationship between the high Poisson's ratio anomalies and the thermal fields is confirmed in this study. In high-temperature regions where partial melting occurs, the S-wave velocity decreases more significantly than the P-wave velocity (Lü et al., 2014), leading to a relative high Poisson's ratio. Thus Poisson's ratio distribution provides better constraints of thermal anomalies and possible partial melt in the study region. According to the heat flow results (Hu et al., 2000), the average thermal radiation value reaches up to 100–150 mW/m2 in Tengchong and Chuxiong region, 80 mW/m2 in Tonghai region. All the thermal anomaly areas in Yunnan agree with the high Poisson's ratio anomalies in our study.

The determination of a large scale strong lateral variations of Moho interface and lower crustal composition transition belt in northwestern Yunnan in this study have important implications for regional tectonic evolution. Tomographic image under Tengchong volcano (Lei et al., 2009) shows a striking upwelling mantle flow with a width of ~150 km, extending horizontally at ~200 km depth in the east to west (EW) direction. Hu et al. (2012) found fast S-wave direction was oriented in the EW direction across latitude 25°N–26°N in central Yunnan. And the fast S-wave direction is usually interpreted as the lattice preferred orientation of olivine crystals in lithosphere induced by mantle flow (Hu et al., 2012; Silver and Chan, 1991). In the high resolution tomography result of Huang et al. (2015) (i.e., Profile WE, in Fig. 8), LVZs under Tengchong also extend horizontally in the EW direction and have a thinned Moho under Chuxiong area, which agrees with our results. The S velocity of uppermost mantle (about 30–50 km under the Moho discontinuity) inversed by S-wave (Lü et al., 2014, i.e., Fig. 8 (B) therein) also reveals a band of LVZs in EW direction near 26°N. SKS analysis (Sol et al., 2007) reveals the fast direction of the shear wave splitting turns from almost NS direction to EW direction from north to south around 26°N. Besides previous GPS observation shows the velocity field is redirected from southward to southwestward near 26°N, even in an east-west direction in western Yunnan (Fig. 7) (Bao et al., 2015; Shen et al., 2005). All these evidences are consistent with the sudden transition of Moho interface and crustal composition along the transition belt.

 Download: larger image Figure 7. A geodynamic model for part of the study region. The volcanic icon and red star indicate location of the upper mantle rise beneath stations TNC and CUX, separately. Green arrows denote GPS velocity field relative to the South China Block (Bao et al., 2015; Shen et al., 2005). The blue dashed line denotes the transition belt of Moho discontinuity and lower crust composition. The Moho interface is obtained through CCP stacking of P receiver function method in our study. Blue arrows indicate possible mantle flow from SE Tibet and Burma (Hu et al., 2012; Lei et al., 2009).

To explain the Poisson's anomalies along the transition belt and origin of the hot material for the bending Moho in our CCP profiles, a geodynamic model is proposed in our research (Fig. 7). Integrating our CCP images and other results with previous tomographic (Huang et al., 2015; Lei et al., 2009) and SRFs (Hu et al., 2012) results, we also propose the mantle flows from Myanmar and SE Tibet interact into an eastward flow. And regional up-welling hot upper mantle materials increase the area temperature and cause the rise of Moho interface. We suggest the mechanism of the upper-mantle rises at stations TNC and CUX are influenced by the mantle flow rather than the lower crust channel flow.

4 CONCLUSION

In conclusion, we calculated teleseismic P receiver functions of 49 permanent broadband seismic stations in Yunnan Province and estimated crustal thickness and Poisson's ratio under individual station using the H-κ method. We also obtained detailed crustal structural images using the CCP stacking method. Our CCP results are significantly improved compared to previous similar studies (Li et al., 2016; He et al., 2014; Hu et al., 2013, 2012) and provide new constraints on the tectonic evolution of northwestern Yunnan.

The results show a northwest-southeast decrease in the study area from 50–54 to 30–35 km in less than 50 km of horizontal distance. The thicker crust in the north has a high Poisson's ratio, supporting the idea of weak lower crust where deformation from the Indian collision in SE Tibet is focused. However, apart from some single high Poisson's ratios that are likely linked to volcanic areas, we see values typical of normal continental crust. This suggests that channels of lower crustal deformation that have previously been suggested to exist south of 26°N are not supported by these results. Rather, we propose the controlling factor of the dynamic processes south of 26°N is the result of mantle flow related to the eastern forward subduction of the Indian Plate.

ACKNOWLEDGMENTS

We thank Hok Sum Fok and James Hammond for their serious reading of the original work and providing many suggestive advices. We thank two anonymous reviewers and editors for their detailed reviews that improve this manuscript. We also thank Robert B. Herrmann and Lupei Zhu for their CPS (Computer Programs in Seismology) package and H-κ code separately. Waveform data for this study were provided by Data Management Center of Yunnan Seismic Network. This study was supported by the 973 Project of China (No. 2013CB733303) and the National Natural Science Foundation of China (No. 41474093). The final publication is available at Springer via https://dx.doi.org/10.1007/s12583-017-0822-9.

REFERENCES CITED