Crustal thicknesses and Poisson's ratios beneath the Chuxiong‐Simao Basin in the Southeast Margin of the Tibetan Plateau

In the Southeast Margin of the Tibetan Plateau, low‐velocity sedimentary layers that would significantly affect the accuracy of the H‐κ stacking of receiver functions are widely distributed. In this study, we use teleseismic waveform data of 475 events from 97 temporary broadband seismometers deployed by ChinArray Phase I to obtain crustal thicknesses and Poisson's ratios within the Chuxiong‐Simao Basin and adjacent area, employing an improved method in which the receiver functions are processed through a resonance‐removal filter, and the H‐κ stacking is time‐corrected. Results show that the crustal thickness ranges from 30 to 55 km in the study area, reaching its thickest value in the northwest and thinning toward southwest, southeast and northeast. The apparent variation of crustal thickness around the Red River Fault supports the view of southeastern escape of the Tibetan Plateau. Relatively thin crustal thickness in the zone between Chuxiong City and the Red River Fault indicates possible uplift of mantle in this area. The positive correlation between crustal thickness and Poisson's ratio is likely to be related to lower crust thickening. Comparison of results obtained from different methods shows that the improved method used in our study can effectively remove the reverberation effect of sedimentary layers.


Introduction
Yunnan area, located in the southeast margin of the Tibetan Plateau, is to the east of the continental collision zone of the Indian and Eurasian plates. It is one of the most active seismic zones in continental China, with frequent occurrences of volcanoes, hot springs, strong earthquakes, and severe earthquake hazards. The two major faults in the area are the Red River and Xiaojiang Faults. The south end of the Xiaojiang Fault converges with the east segment of the Red River Fault, forming a convergence area. The Chuxiong-Simao Basin, like a wedge, is located to the north of this convergence area. Both the Xiaojiang and the Red River Fault are boundaries of major tectonic blocks within the Chinese continent ( Figure 1).
Many previous studies of this area's tectonic features and deformation mechanisms have been performed. Bai DH et al. (2010) found two major channels of high electrical conductivity at depths of 20 to 40 km in the eastern Tibetan Plateau, supporting the hypothesis that crustal flow can occur in orogenic belts and contribute to uplift of the plateau.  found the existence of low resistance anomalies in the mid-lower crust of the Qiangtang terrane and the eastern Tibetan Plateau, as well as the Sichuan-Yunnan rhombic block, which indicates southeast movement of crustal materials in the Tibetan Plateau. Fu YV et al. (2017) observed low velocity anomalies along or near the major faults in the middle crust and forming a broad zone in the lower crust of the southeast margin of the Tibetan Plateau, indicating that crustal flow exists in the lower crust in this area. Accurate knowledge of deep seismic structure will improve understanding of crustal flow and movement of mantle materials in the southeast margin of the Tibetan Plateau. Therefore, it is of great importance to achieve accurate seismic structure and Moho depth in this convergence area. During the past several decades, numerous studies have investigated various aspects of this particular area. Kan RJ et al. (1986) studied crustal structure of Yunnan from seismic refraction profiles; Wen XZ et al. (2008) studied historical patterns and behaviors of earthquake ruptures around the Red River Fault, the eastern boundary of the Sichuan-Yunnan faulted-block. Zhang X and Wang YH (2009) used a finite-difference travel time inversion to achieve crustal and upper mantle P wave velocity structure in Yunnan. Wang Q and Gao Y (2014) investigated the relationship between velocity structure and earthquake activity on the southeast margin of the Tibetan Plateau. Specifically, there are plenty of studies that use receiver function technique to invert S wave velocity structures (Wu JP et al., 2001;He CS et al., 2004;Hu JF et al., 2005;Wang CY et al., 2008) or to detect the crust and upper mantle structure around this area (Li YH et al., 2008Zhang HS et al., 2009;Liu QY et al., 2014).
It has always been one of purposes of seismological research to determine the interior structure of the Earth from seismic waves, which are affected by the joint influences of source time function, travel path, media under the station, instrument's response, and other factors. Receiver functions-Presented as time series with the impacts of earthquake source, travel path, and other factors removed-can show the relative response of Earth structure under the stations. They mainly contain the information of P-to-S converted waves and their multiples (including PmS, PPmS and PSmS, as shown in Figure 2) generated from the velocity discontinuities under the receivers (Phinney, 1964;Jordan et al., 1975;Vinnik, 1977;Langston, 1977Langston, , 1981. With the quick development of digital seismic observation techniques, receiver functions method has been widely used in recent years to obtain descriptions of crustal and upper mantle structures under seismic stations (Kosarev et al., 1999;Yeck et al., 2014;Hansen and Schmandt, 2017). Clearly, receiver functions analysis is an effective method to study the structure of Earth crust and upper mantle.
Previous studies have shown that the existence of a low-velocity sedimentary layer in the upper crust will strongly affect the determination of seismic structure achieved by conventional receiver function method (Cassidy, 1992(Cassidy, , 1995Sheehan et al., 1995;Zelt and Ellis, 1998). Velocity discontinuities caused by low-velocity layers can give rise to strong reverberation in the near-surface area, significantly masking seismic P-to-S phases associated with the Moho. Consequently, applying the conventional H-κ stacking method (Zhu LP and Kanamori, 2000) would likely lead to erroneous results of crustal thickness and V P /V S ratio. For a crust model with sedimentary layer, there are several possible Ps phases and their reverberations (PbS, PmS, PPmS and PSmS, see Figure 2). Some previous studies have shown that the crustal thickness of the Yunnan area reaches its highest value in the northwest, and decreases toward the northeast, southeast and southwest (He RZ et al., 2014;Li YH et al., 2014). However, these studies barely took into consideration the effects of low-velocity sedimentary layers.
In this paper, we conduct receiver function analysis using teleseismic data recorded in the Chuxiong-Simao Basin and adjacent area. In order to achieve better and more precise results, we apply a sediment removal filter to the selected receiver functions (Yu YQ et al., 2015). Furthermore, instead of directly using the conventional H-κ stacking method (Zhu LP and Kanamori, 2000), we employ a time-corrected H-κ stacking method (Yu YQ et al., 2015) to eliminate the effect of reverberation in sedimentary layers.

Data
The study area of this paper is 21°N-28°N, 97°E-105°E, located in the southeast margin of the Tibetan Plateau. This rectangular zone contains the Chuxiong and Simao Basins, and the majority of China's Yunnan Province (Figure 1)  Tibetan Plateau. In this study, we focus on data from 97 of these, located within our study area ( Figure 1). We choose teleseismic events with M>5.0 and epicentral distances within the range of 30° to 90° from these stations. The total number of events selected is 475. The distribution of selected events is shown in Figure 3.
Seismic waveform data were filtered in the frequency range of 0.04-0.8 Hz to enhance the signals. Filtered seismograms with SNR of 4.0 or greater were then converted into receiver functions using a water level deconvolution method (in the frequency domain), following Ammon (1991). The water level and Gaussian width factor used in the method are 0.05 and 5.0, respectively. We further applied an SNR-based procedure to reject low quality re-

Method
Following Ammon (1991), the P-to-S converted phases (hereinafter referred to as Ps phases) in the receiver functions can be expressed as where δ is a Dirac delta function; A s and t s represent the amplitude and time delays of the Ps phases, respectively. For a crust model with low-velocity sedimentary layer, the primary and multiples of the converted S waves can be expressed as where r 0 is the strength of the reverberations in the sedimentary layer, n is the index of the nth reverberation of the Ps phases, ∆t is the two-way travel time of the reverberations in the sedimentary layer, and F(t) is the receiver function without the influence of the low-velocity sedimentary layer.
This equation can be expressed in the frequency domain as where i is the complex symbol.
Note that ; from the geometric series, this equation can be further expressed as Therefore, receiver function F can be obtained in the frequency domain from . ( In other words, the reverberations caused by a low-velocity sedimentary layer can be eliminated by applying a filter in the form (Yu YQ et al., 2015).
We then use the time delay and two-way travel time of the PbS phase, the first Ps phase from the bottom of the sedimentary layer ( Figure 2e), to time-correct the H-κ stacking equation. The resulting formula can be expressed as where A is the stacking amplitude, N is the number of receiver functions participated in the stacking, S m (t) represents the amplitude of the receiver function at the time t after the direct P wave, w 1 , w 2 and w 3 are weighting factors that satisfy w 1 +w 2 +w 3 =1 (Zhu LP and Kanamori, 2000), and δt m and ∆t m are, respectively, the time delay and the two-way travel time of the PbS phase.
The major difference between this time-corrected H-κ method and the conventional method is the time terms. To better understand these time terms, we now consider a simplified situation of vertical incidence. Figure 8a shows the PbS phase under this circumstance. Apparently, the time delay for the PbS phase is and the two-way travel time is where H d is the thickness of the sedimentary layer, V P and V S are By using this time-corrected H-κ stacking method, we remove the travel time associated with the sedimentary layer, so that the stations are downward projected to the bottom of the sedimentary layer. Consequently, the optimal sub-sediment crustal thickness and V P /V S ratio can be obtained (Yu YQ et al., 2015).

Analysis and Results
Previous studies (Zhu LP and Kanamori, 2000) show that a 0.1 km/s uncertainty in V P could produce about 0.5 km uncertainty of crustal thickness. According to the studies of P wave velocity structure in the study area (Cui ZZ et al., 1987;Xiong SB et al., 1993;Wang CY and Gang, 2004), we set crustal P wave velocity at 6.30 km/s during the H-κ stacking of receiver functions. shows crustal thicknesses and V P /V S ratios in the study area obtained by the improved method.

Determination of Crustal Thickness by Improved Method
We use both the conventional method and the improved method to calculate the crustal thicknesses. With the conventional method, we directly apply the conventional H-κ stacking to the receiver functions. With the improved method, we first process the receiver functions by applying the resonance removal filter, and then apply the time-corrected H-κ stacking. In order to achieve a clear comparison of results between these two methods, we present results of four stations as examples to illustrate the ad-vantage of the improved method.
Station 53060 (see Figure 1) is located in Chuxiong Basin. There are 43 high quality receiver functions at this station after the SNRbased selection, as previously shown in Figure 4. The H-κ stacking plots of both methods for this station are shown in Figure 9. Crustal thickness beneath station 53060 is calculated as 53.0 km by the conventional method, and as 50.2 km by the improved method. Station 53181 (see Figure 1) is located in Simao Basin, the RFs of which can be found in Figure 5. The H-κ stacking plots of both methods for this station are shown in Figure 10. Crustal thickness beneath station 53060 is 36.0 km by the conventional method, and 32.7 km by the improved method. For these stations, the effect of the low-velocity sedimentary layer is effectively elim-  inated.
Station 53086 and station 53193 (see Figure 1), located out of basin area, have 116 and 91 high quality receiver functions, respectively, after the selection procedure (previously shown in Figure 6 and Figure 7). The H-κ stacking plots for these two stations are shown in Figures 11 and Figure 12. For station 53086, crustal thickness is 33.2 km by the conventional method and 33.1 km by the improved method; for station 53193, crustal thickness is 36.9 km by both methods. Since these stations have no sedimentary layer underlying, the improved method gives nearly the same results as the conventional method.
There are 31 stations located within the basins in this study. We compare the crustal thickness obtained by conventional and improved methods at all these stations ( Figure 13).      in detail, comparison of crustal thickness calculations from data at each of these stations. The results show that, for most of these stations, the improved method appears to eliminate the effect of any low-velocity sedimentary layer, resulting in crustal thickness estimates 2-4 km thinner than those obtained by the conventional method. However, at some stations, there are no apparent correlations between results by the two methods. The reason for this phenomenon might be that the result from the conventional H-κ stacking is dominated by the multiples instead of by Moho-associated signals. In other words, there is no way to obtain "the apparent crustal thickness" at these stations due to very large sedimentary effects.
To further evaluate our crustal thickness results, we compare them with those published by Wang WL et al. (2017). The Pearson correlation coefficient of the two sets of results is 0.829, which indicates a relatively good consistency. Table 3 shows the difference between these two sets of results. For most stations, our results are similar to those obtained by Wang WL et al. (2017). At some stations, however, the crustal thicknesses obtained by our study are thinner since we consider the influence of the low-velocity sedimentary layer in the upper crust.
High SNR records are required to achieve accurate travel time differences between different phases, which in turn are essential to accurate calculation of crustal thickness in H-κ stacking. In addition, the existence of dipping interfaces will compromise results, since events with the same epicentral distance but different azimuths might have different travel time differences. The teleseismic events in our study are not evenly distributed on azimuth (Figure 3), which might cause uncertainties in the results.

Overall Features of Crustal Thicknesses and Poisson's Ratios
Our calculations of crustal thicknesses and V P /V S ratios under each station in the study area are shown in Table 1. Figure 14 presents a better, clearer, picture of these results. Figure 15 employs spatial smoothing to show the overall distribution of crustal thicknesses and Poisson's ratios.
Overall, from Figures 14a and 15a, we find that, the crustal thickness decreases from about 55 km in the northwest toward northeast, southwest and southeast. The thinnest crust appears in the south of the study area, with a thickness of about 30 km. The crustal thicknesses are a little higher in the northeast, at about 37 km. These general variation features are consistent with previous studies using different methods, including joint inversion (Li YH et al., 2014) and receiver functions (He RZ et al., 2014;Wang WL et al., 2017).
Moreover, our results show that the Red River Fault, a large strikeslip fault situated at a NW-SE orientation, draws an apparent boundary for crustal thickness in the study area ( Figure 15a). Crustal thicknesses are thinner to the west of the Red River Fault and thicker to the east, indicating that the fault cuts the crust (Xu MJ et al., 2006). However, it is interesting that the crustal thickness between Chuxiong City and the Red River Fault is a little thinner than the surrounding area nearby, indicating that there seems to be an uplift of upper mantle in this area, which could result from movements of deep mantle materials. This phenomenon is consistent with previous studies (Zhang XM et al., 2011;Deng JM et al., 2014).
On the west side of the study area, the variation of crustal thickness is relatively gentle, with most of the stations colored red (Figure 14a), indicating that the crustal thicknesses are less than 40 km. On the east side of study area, crustal thickness varies greatly across the Red River Fault, from thicker north to thinner south, which suggests that the Sichuan-Yunnan rhombic block, the southwest boundary of which is the Red River Fault, possibly intakes a large amount of the southeast-direction escape of the Tibetan Plateau. However, in the middle part of the Red River Fault, there is no obvious variation in crustal thickness. In the east side of the study area, it seems that the Xiaojiang Fault is a boundary to abrupt variations in crustal thickness. Crustal thickness appears to be thicker within the Sichuan-Yunnan rhombic block and thinner outside, suggesting that the Xiaojiang Fault blocks the material flow of the Tibetan Plateau in the southeast direction.
Poisson's ratio is an elastic constant that measures the compressibility of material perpendicular to applied stress, or the ratio of latitudinal to longitudinal strain. For regular rocks, the Poisson's ratio ranges from 0.20 to 0.35, and is very sensitive to the composition of rocks. According to the relationship between Poisson's ratio and V P /V S ratio , we calculate the Poisson's ratio in the study area from V P /V S results (Table 1 and Figure 14b). Poisson's ratios are shown beneath each station in Figure 10b. Figure 15b shows the overall distribution of Poisson's ratios. We find that the Poisson's ratios beneath the stations in this area range from 0.18 to 0.30, which is generally consistent with other studies (Li YH et al., 2009;Wang WL et al., 2017).
Crustal thickness difference of different methods (km) −4 −2 0 2 4 Figure 13. Crustal thickness difference between conventional and improved methods. The blue inverted triangles indicate that for these stations, crustal thicknesses obtained by the improved method are less than those by the conventional method; the red triangles indicate that crustal thicknesses obtained by the improved method are higher than those by the conventional method. The shadow part indicates the Chuxiong-Simao Basin, same as in Figure 1.
Overall, Poisson's ratios distribute unevenly in the study area, with a general characteristic decrease from north to south, although with local complications positively correlated to the variation of crustal thickness. The correlation between crustal thickness and Poisson's ratio can provide a valuable constraint on the tectonic process of continental crust. Ji SC et al. (2009) concluded that if crustal thickening is caused mainly by the lower crust, the Poisson's ratio would be positively correlated to crustal thickness; whereas if crustal thickening is caused mainly by the upper crust, the Poisson's ratio would be negatively correlated to crustal thickness. Our results indicate that the variation of Poisson's ratio in the study area is likely related to lower crustal thickening, which is consistent with other studies (Zhang ZJ et al., 2005).

Discussions and Conclusions
Using seismic data from a dense temporary seismic array, ChinArray Phase I, we obtained crustal thicknesses and Poisson's ratios beneath 97 seismic stations within the Chuxiong and Simao Basins and adjacent area. In particular, our data were from filtered receiver functions of 475 earthquake events, analyzed by a timecorrected H-κ stacking method. The method's resonance removal filter and time-corrected H-κ stacking can eliminate reverberations and travel time associated with the sedimentary layer.
We find that the crustal thickness varies strongly in the study area. The northwest part has a relatively thicker Moho, with crustal thickness over 50 km. Crustal thickness becomes lower from     northwest toward southwest, southeast and northeast, the geometry and range of which seems to be related to the Sichuan-Yunnan rhombic block surrounded by the Red River Fault and the Xiaojiang Fault, indicating that the crust is thicker within this rhombic block than in the adjacent area. Also, we find that the Red River Fault separates the study area in terms of crustal thickness, suggesting that the Sichuan-Yunnan rhombic block takes in the escape material of the Tibetan Plateau or leads the direction of escape. In addition, relatively thin crustal thickness in the center of study area indicates possible uplift of upper mantle. Poisson's ratios, inferred from V P /V S ratios obtained by H-κ stacking, distribute unevenly in the study area. Overall, Poisson's ratios decrease from northwest to southeast direction, which is correlated with the variation of crustal thickness in this area. This correlation is likely to be related to lower crust thickening.
Comparison of H-κ stacking results from the conventional and the time-corrected (i.e. improved) method shows that the improved method can effectively remove the effect of low-velocity sedimentary layers. After applying the improved method, crustal thickness results beneath most of the stations are 2-4 km thinner than those obtained from conventional methods. Some stations show no correlation of the results by these two methods, the reason for which might be that the reverberation in these sedimentary layers is especially strong. Results at other stations show no change, perhaps indicating that no low-velocity sedimentary layer exists beneath those stations, or that the sedimentary layer has solidified.