Rayleigh Wave Group Velocity Distribution in Ningxia

INTRODUCTION

Surface wave tomography with Green's functions estimated from the cross-correlation of ambient noise has become one of the most popular tools for seismologists to image the earth structure. The essence of this technique is that under the assumption of diffuse wavefields and evenly distributed ambient noise energy, coherent information propagating through the path between two receivers that are related to Green's functions can be extracted from ambient noise crosscorrelations since incoherent components passing through other directions have been cancelled out (e.g., Weaver, 2005; Snieder, 2004; Wapenaar, 2004; Weaver and Lobkis, 2004; Campillo and Paul, 2003; Derode et al., 2003). Indeed, spectacular applications of ambient noise tomography have recently been published (Li et al., 2009; Bensen et al., 2008; Yao et al., 2008; Zheng et al., 2008; Lin et al., 2007; Yang et al., 2007; Sabra et al., 2005; Shapiro et al., 2005). Because this method can be applied to aseismic areas with dense station coverage, work on higher frequencies and is relatively unaffected by seismic source location, ambient noise tomography usually can resolve higher precision structures in the shallow subsurface than traditional surface wave tomography.

The study region is located in the tectonic junction of three intraplate blocks, the Alashan block in the west, Ordos block in the east and Qilian block in the south, and has a complicated geological background (Fig. 1). The Qilian block consists of complexly deformed Early Paleozoic arcs developed at the southern margin of the North China craton (e.g., Yin and Harrison, 2000) and has an average elevation of 3 500 to 4 000 m. The Ordos basin is a Mesozoic basin in the Sino-Korean platform that is considered to be one of the most stable tectonic provinces in China, and the rigid Alashan block consists of Precambrian basement. The study area also shows a high seismic activity, several M≥8.0 earthquakes have occurred in the past 100 years.

Figure  1.  Topographic relief and tectonic elements of the Ningxia region and adjacent areas. The box indicates the study area; dashed lines represent main faults.

DownLoad: Full-Size Img PowerPoint

In this article, in order to increase the path density, we combined seven months ambient noise data and 10 earthquakes with magnitude larger than 4.0 recorded at the Ningxia regional seismic network and the December 20, 2007, Dafeng Mine explosion data from both temporary and permanent seismic stations to measure the dispersion curves of Rayleigh waves. Group velocity distributions of Rayleigh waves between 6–22 s were determined with Occam's inversion technique. The resulting maps display obvious lateral velocity variations in the study area, low group velocities are clearly observed beneath the Yinchuan basin, high velocities beneath the Helan Mountain and southern mountain areas, which is correlated well with surface geology.

DATA AND METHOD

Seven months of continuous vertical-component ambient noise data with 20 Hz sampling rate from January to July 2006 recorded by six stations from the Ningxia regional seismic network are used in this study (Fig. 2). In a manner similar to the one-bit noise cross-correlation processing procedure, we first cut the continuous data to one-hour segments, removed instrumental response, trend, and mean value. Then, the data were filtered in the period band from 4 to 50 s and downsampled to 20 samples per second. After that the amplitude was disregarded by considering only one-bit signals (Shapiro and Campillo, 2004; Campillo and Paul, 2003). Then, traces that were processed for one hour were cross-correlated and stacked to a single time-series for each station-pair. In order to increase path density, we also include 10 earthquakes and blasting records from the 5 500-t controlled source explosion. The explosion took place in Dafeng Mine, Ningxia, on December 20, 2007, and recorded by our temporary seismic array, which was deployed by China University of Geosciences (Beijing) (CUGB) one week before the blasting, which consisted of 15 broadband stations and 9 permanent seismic stations from the Ningxia seismic area (Fig. 2).

Figure  2.  Stations used in this study, solid triangles represent stations used for both cross-correlation and explosion data analysis, inverted triangles denote stations only with explosion data used in this study and circles are events. Thick black lines are the section lines of the vertical profiles (A-A' and B-B') shown in Fig. 7.

DownLoad: Full-Size Img PowerPoint

After obtaining the cross-correlation functions and blasting waveform records, the group velocity dispersion measurements were made using the multiple-filter technique (Herrmann, 1973; Dziewonski et al., 1969). The waveforms were narrow bandpass filtered with the operator exp[-α(ω−ω₀)²/ω₀²], where ω₀ is the center frequency. Since there is a trade-off between the time domain resolution and frequency domain resolution with such filtering, that is, a large α provides better frequency domain resolution at the expense of reduced resolution in the time domain. In this study, the selection of the α filter parameter is considered as a function of distance, as Levshin et al. (1989) suggested.

Figure 3 shows cross-correlation functions and group velocity dispersion curves measured from two station-pairs along paths through distinct geological regions. As seen in Fig. 3, because of the nonuniform noise distribution in the real-world cross-correlation functions shows asymmetric features; however, the coherent waveforms are still obvious. With the multiple-filter technique, dispersion curves measured from station-pairs SZS-YCI showed lower group velocity values than those from JYU-TXN in Fig. 3.

Figure  3.  Top: seven months cross-correlations for the station-pairs JYU-TXN and SZS-YCI. Bottom: group velocity curves measured from the paths shown on the top: JYU-TXN (left), SZS-YCI (right).

DownLoad: Full-Size Img PowerPoint

In Fig. 4, the left panel displays the dispersion curve extracted from the vertical broadband waveforms recorded by the CUGB temporary station No. 100 at 200 samples per second. The shot point is located at Dafeng Mine, Helan Mountain, in the junction of Ningxia and Inner Mongolia, and station No. 100 is located in the Yinchuan basin, therefore, waves mainly propagate in the Yinchuan basin. The right panel gives the dispersion curve measured from the explosion waveforms recorded at station XSH with 100 samples per second. Since the station is located outside the Yinchuan basin, as seen in Fig. 4, higher group velocities are obviously shown at station XSH than station No. 100.

Figure  4.  Group velocity curves measured from the paths: the shot to temporary CUGB station No. 100 (left); the shot to station XSH (right).

DownLoad: Full-Size Img PowerPoint

It is well-known that surface waves at different periods are sensitive to the earth structure at different depth ranges. A good rule of thumb is that the depth of maximum sensitivity of Rayleigh waves is approximately one-third of their wavelength. Low group velocities at short periods (< 15 s) usually are related to sedimentary layers in the shallow crust because seismic velocities of sediments are much slower than those of bedrocks. Therefore, we noticed that the paths SZS-YCI and shot-to-station No. 100 are through the Yinchuan basin, a region with relatively thicker sediments. At longer periods, Rayleigh waves become primarily sensitive to the middle-to-lower crustal structure. The dispersion curve between stations JYU and TXN still shows faster group velocities at periods between 15 and 25 s when compared with the path SZS-YCI.

GROUP VELOCITY DISTRIBUTIONS AND SHEAR WAVE VELOCITY STRUCTURES

Rayleigh wave dispersion measurements obtained from the previous section were used to invert for group velocity distributions at periods between 6 and 22 s. The study region was divided into 0.1º×0.1º grid, and velocities between grid nodes were computed with bilinear interpolation. The Occam's inversion technique (Constable et al., 1987) was applied to invert for group velocity distributions in this study.

The coverage and the azimuthal distribution of paths generally control the resolution of surface wave tomography. In Fig. 5, we show the path density at 8 and 18 s. The Ningxia regional seismic network has a much narrower span in west-east direction than in south-north direction; accordingly, the path distribution is not uniform in the study region, and more paths are in south-north direction. Path coverage is generally good for the northern part and become worse for the southern part. Consequently, our group velocity distributions have better resolution along the latitude than longitude, and the resolution is better for northern Ningxia than southern Ningxia.

Figure  5.  Path densities at 8 and 18 s. Path density is defined as the number of rays intersecting a 0.1°× 0.1° cell.

DownLoad: Full-Size Img PowerPoint

Figure 6 presents the results of group velocity tomography at periods 8 and 18 s. At 8 s period, the lowermost group wave velocities are found beneath the Yinchuan basin, and the Helan Mountain and southern Ningxia are featured by high wave velocities. At 18 s, the distribution of low velocities beneath the Yinchuan basin widens to the Alashan block, and high velocities persist in the southern mountainous area, but the overall pattern of high and low velocities at 18 s is similar as that at 8 s.

Figure  6.  Estimated Rayleigh wave group velocity maps at 8 (left) and 18 s (right).

DownLoad: Full-Size Img PowerPoint

From the group velocities between 6 and 22 s, shear wave velocity structure inversions were carried out at each node of the 0.1º×0.1º grid by using a program developed at Saint Louis University. We adopted isotropic layered models that are composed of 50 layers with 1-km thickness of each layer overlying a halfspace. The S-wave velocity in each layer is referred to deep seismic sounding profiles and tomography results (Zhao et al., 2007; Li et al., 2002; Jin et al., 1999; The Committee Compiled Earth Science Profiles, SSB, 1992), and P-wave velocity and density are computed from S-wave velocity based on empirical formulas. Since the maximum period we measured for group velocity is 22 s in this study, the resolving power below about 25–35 km is very limited, and the resulting structure below that depth is of little significance.

Figure 7 shows S-wave velocity structure along two vertical profiles from the surface to 35 km depth. In the profile A-A', it is clear that the low-elevation Yinchuan basin is clearly marked by low S-wave velocity to about 5–8 km, and high velocities are shown beneath the Helan Mountain and Alashan block with relatively high topography. In the B-B' profile crossing the Ningxia from south to north, at shallow depth, the velocities are relatively higher in the southern segment of the profile than those in the northern segment, and the lowermost velocities are observed in the southernmost part of the profile that is corresponding to the Yinchuan basin. In both profiles, no middle-crustal low-velocity zones are revealed in our study.

Figure  7.  Vertical cross sections of shear wave velocity along the lines A-A' and B-B' in Fig. 1. Topography is plotted above each profile (black area).

DownLoad: Full-Size Img PowerPoint

DISCUSSION AND CONCLUSIONS

Due to the limitation of the geometry of Ningxia regional seismic network, there are a few tomographic studies and several deep seismic sounding studies in this region. Tomographic studies using earthquake data from Jin et al. (1999) and deep seismic sounding profiles (Li et al., 2002; The Committee Compiled Earth Science Profiles, SSB, 1992) all showed that lateral velocity variations in the Ningxia area are evident, and the Yinchuan basin is featured with low velocities at a depth down to at least 7.5 km, and high velocities are revealed beneath the mountainous region. The Yinchuan basin is a trough-like Cenozoic graben-basin located in the west edge of Ordos massif, it has very thick sediment layers, and the relatively high velocity beneath the Helan Mountain and southern mountainous region is probably due to their shallow bedrock depths. The lateral velocity variation at shallow depth is correlated well with surface geology and relief, which is consistent with our tomographic results. At deep depth, results from Jin et al. (1999), Li et al. (2002), and Zhang et al. (2009) found low-velocity layers in the middle-to-lower crust beneath the Yinchuan basin that are interpreted as partial melting in this area; however, they are not observed in our study. Since in this study the longest period we measured for the group velocity is 22 s, which corresponds to the resolving depth around 25–5 km, in order to detect the middle-to-lower crustal structures, we have to rely on group velocity measurements from longer periods. Meanwhile, denser datasets with better azimuthal coverage by including adjacent regional networks are necessary to address the middle-to-lower low-velocity layer issue in the near future.

In summary, by combining continuous ambient noise data and 5 500-t controlled source explosion data, we measured the dispersion curves of fundamental mode Rayleigh waves. Group velocity distributions between 6–22 s were determined with the Occam's inversion technique by dividing the study area into grids with 0.1°×0.1°. In the study region, the lateral velocity variation is significant between different tectonic blocks, and the velocity structures are correlated well with surface geology and relief. The Yinchuan basin is clearly revealed with low velocities, and the Helan Mountain and southern mountain areas are featured with high velocities. The Helan Mountain marks not only the distinct tectonic blocks between the Yinchuan basin and Alashan block and also lateral velocity variation between the two tectonic units.