Stepwise Lithospheric Delamination Leads to Pulsed Cenozoic Uplifts of Central Tien Shan

Intracontinental orogens play important roles in continental evolution of the Earth. On the one hand, the continental orogens mark the termination of the amalgamation between different blocks, through which a large continent or even supercontinent may be formed finally. On the other hand, the intracontinental orogens are usually characterized by weaker rheology than the adjacent blocks and thus are likely to be reactivated first under a new tectonic event. In both cases, the intracontinental orogens experience strong deformation and tectonics. Geological hazards such as large earthquakes occur frequently, which cause heavy damage in general. Therefore, understanding the mechanism of the intracontinental orogens provides insight into the continental cycles and reduces the geological hazards.

The Tien Shan (Figure 1) is a typical continental orogen located in the southern boundary of the Central Asian Orogenic Belt that was formed during the Paleozoic to the Mesozoic (e.g., [1–3]). Its basement includes rigid Precambrian blocks surrounded by various Paleozoic accretion–collision belts [4]. The Tien Shan is divided into the western, central, and eastern Tien Shan through the right lateral Talas-Ferghana Fault and the longitude of ~80°E (the east of the Lake Issyk-Kul) ([4, 5]; Figure 1).

The Central Tien Shan, being situated between Tarim Basin and Pamir (to the south) and Kazakh Shield (to the north), is a natural laboratory to study the rejuvenation mechanism of the Tien Shan orogenic belt. It is further divided into North, Middle, and South Tien Shan (abbreviated as NTS, MTS, and STS, respectively). The Cenozoic orogenesis of the Tien Shan is generally considered distant effect of Indo-Asian collision [4, 6–8] although some studies suggested a potential mantle plume under the Tien Shan (e.g., [9, 10]). Results of apatite fission-track thermochronology, structural modeling, and magnetostratigraphy indicate stepwise uplift of the Tien Shan; the exhumation of the South and North Tien Shan started at 25-15 Ma and 11-10 Ma, respectively [4, 11–15].

GPS measurements show that the Central Tien Shan moves northward relative to the Eurasian Plate ([4, 16]; Figure 1). The rate of the motion decreases linearly from ~20 mm/yr in the South Tien Shan to <5 mm/yr in the North Tien Shan. The shortening of the South Tien Shan (10 mm/yr) is significantly stronger than that of the North Tien Shan (2-6 mm/yr). Cenozoic crustal shortening is at a scale of ~200 km (e.g., [17]), which contributes only a small portion of the Indo-Asian convergence. However, the shortening estimated from the present GPS measurements is several-fold the levels of the slip rates in the Holocene, implying that the rate of mountain building has been accelerated several times during the past 10 Ma [8, 18].

Deep structures and rheology in the crust and upper mantle are important for understanding the mountain buildings. Seismic results reveal two local depressions of the Moho topography under the South and North Tien Shan where the Moho depths exceed 60 km (e.g., [19–22]). By contrast, the Moho depths are shallower than 50 km under the Tarim Basin, Kazakh Shield, and the Middle Tien Shan. High Vp/Vs ratios (>1.73) under the Central Tien Shan imply the partial melting within the crust, which argues for the magmatic intrusion of the hot upwelling mantle material [19]. Some inclined high velocity anomalies in the upper mantle are proposed to be the subducting Tarim lithosphere under the South Tien Shan and the Kazakh lithosphere under the North Tien Shan [23–28]. Besides, low velocity zones are imaged in the top 150 km under the Central Tien Shan (e.g., [23, 26]). A thicker mantle transition zone was revealed and interpreted to be caused by the detached lithosphere sunk into the transition zone [29–31].

The architecture of the lithosphere plays an essential role in the intracontinental orogens. But images of the lithosphere are unclear in seismic tomography due to the smoothing and smearing effect embedded in the inversion. S wave receiver function (SRF) is a good tool to determine accurate bottom of the lithosphere (i.e., the lithosphere-asthenosphere boundary, LAB hereafter) through S-to-P converted wave (Sp phase) at the LAB. It has been applied in various regions in the work and proven to be a useful method (e.g., [32–34]). A previous study revealed that the LAB depths vary significantly in the Tien Shan, Tarim Basin, and Kazakh Shield [35]. In this study, we determined the LAB of the Central Tien Shan with much improved data set. The improved LAB images provide new insight into the stepwise lithospheric delamination and pulsed uplifts of the Central Tien Shan in the Cenozoic.

We collected waveforms from 28 permanent and 99 temporary broadband stations deployed in the Tien Shan from 1997 to 2020 (Figure 2(a)). The 28 permanent stations belong to three networks: KC (2012-2020, 1 station), KN (2010-2020, 10 stations), and KR (2011-2020, 17 stations). The 99 temporary stations belong to three networks: XP (2005-2007, 40 stations), XW (1997-2001, 28 stations), and 8H (2015-2017, 31 stations). A total of 3299 events with epicentral distances of 40-85° and $magnitudes≥5.5$ were selected (Figures 2(b) and 2(c)), among which 601, 487, 280, and 2211 events were recorded by the XW, XP, 8H, and permanent (KC, KN, and KR) stations, respectively. Most of the events are located in western Pacific and Indonesian subduction zones with a back-azimuthal range of 30-160° (Figure 2(b)).

We calculated the SRFs to reveal the LAB beneath the Central Tien Shan by searching for the Sp phase at the LAB [33, 36]. The Sp phases arrive earlier than direct S waves and so are not contaminated by crustal multiples. Because of different velocity gradient at Moho and LAB, the Sp polarities for the Moho and LAB are opposite, i.e., positive for Moho and negative for LAB. To reveal clear and reliable Sp phases, we first removed the noise from seismograms by applying a 4^th order zero-phase bandpass Butterworth filter of 1-25 s. Then, we rotated the original Z-E-N components (Figure 3(a)) to the L-Q-T component with the theoretical back-azimuth and incident angles calculated in the IASP91 model [37]. Next, we used a wave-shaping time-domain deconvolution to calculate the SRF, i.e., the L component deconvolved by the corresponding Q component ([38] Figure 3(b)). At last, we used the common-conversion point stacking method to make the SRF profiles [33, 39]. We first projected the SRF in the time domain backward to the points that generate the Sp phases in depth domain; the amplitude at a grid node is obtained by averaging the amplitudes of all the SRFs within a specific distance around the node or along a specific profile. In this study, we further filtered the SRFs with two low-pass filters with the corner frequency of 0.125 Hz (8 s, low frequency) and 0.25 Hz (4 s, high frequency) before constructing the SRF cross-sections, which could reveal the first-order patterns and more details, respectively.

Following Zhao et al. [40], we selected the reliable SRFs automatically based on the signal-to-noise ratios (≥5.0), the shapes and amplitudes on the deconvolved Q and L components. The Q pulse could be well fit by a unit Gaussian function; the correlation efficient is larger than 0.9. The standard deviation and maximum of the amplitudes on the Q and L components before the Q pulse are smaller than 0.1 and 0.25, respectively. The procedure preserved 8295 (i.e., ~15%) high-quality SRFs from 56680 suitable measurements (Figure 2(c)).

Notable sidelobes may be caused by the filter and deconvolution in the data processing (e.g., [41]). Because the Q component was processed in the same way as the L component (i.e., the SRF), the sidelobes in the Q component can be used to estimate those in the L component [41]. We corrected the sidelobes in the SRF by referring to the Q component (Figure 4(a)). In the initial SRFs, the negative signal below the Moho is notable in the whole sections (Figures 4(b) and 4(d)). However, most of these negative signals are removed after the SRFs were corrected with the Q component except for those under the South Tien Shan (Figures 4(c) and 4(e)).

To estimate the uncertainty of the obtained SRFs, we used a bootstrapping method to estimate the uncertainties of the SRF images. We selected 60% of the data set randomly and calculated the corresponding SRFs 1000 times. The average and standard deviation of the 1000 SRFs were taken as the final SRF and its uncertainty, respectively. A LAB signal is reliable only when its upper limit is negative after the uncertainty is corrected from the mean value.

Figure 5 shows the stacked images along the NW-SE direction (Figure 2(a)); all SRFs are projected to the profile. As two most important discontinuities in the upper mantle, the Moho and 410 km discontinuities are clear (Figures 4 and 5). The Moho topography correlates well with that determined by the P wave receiver functions (PRF; [20, 22]) which has generally higher resolution thanks to the higher frequency. The clear Moho and 410 km discontinuities imply that the obtained Sp signals are reliable in the upper mantle, in particular those signals between the Moho and 410 km discontinuities. At the same time, the depths of the Moho and 410 km discontinuities provide proper estimate of the uncertainties of the LAB depths caused by the velocity anomalies in the upper mantle. Compared with the PRF results [20, 22, 31], the uncertainties of the Moho and 410 km discontinuities’ depths are ~5 km and ~10 km, respectively (Figure 5). Therefore, the uncertainties of the LAB depths in between are smaller than 5-10 km in general.

In the lithospheric domain (i.e., above 200 km), the SRF features vary laterally under different regions. The polarity of the LAB signals is negative because the velocity gradient is opposite to those at the Moho and 410 km discontinuities. Under the Tarim Basin, northward-dipped negative phases are found at ~120-170 km depth (T). Under the South Tien Shan, northward-dipped negative phases are imaged at 100-130 km depths (D1) right below the Moho. Under the Middle Tien Shan, two negative phases are revealed around 100 km (D2) and 200 km depth (Dx), respectively. The signal Dx dips northward slightly while the dipping feature of D2 is unclear. Under the North Tien Shan and Kazakh Shield, negative phases are imaged between 120 and 200 km depths. The negative phases dip southward in the low-frequency image (Figure 5(c)). However, they show apparent double phases in the high-frequency image, one at ~120-140 km depth (K1) while another at ~160-200 km depth (K2) (Figure 5(b)). A small negative phase (D3) is also imaged at 90 km under the North Tien Shan, but its lateral extent is limited.

We also investigated the variation of the SRFs along the strike (i.e., WSW-ENE) of the Central Tien Shan (Figure 6). In all these profiles, the negative phases under the South Tien Shan (D1) are robust, in both high-frequency (4 s) and low-frequency images (8 s), confirming that they are generated by the actual structures rather than sidelobes. The negative phases under the Tarim Basin (T) and Kazakh Shield (K1 and K2) are also robust, being consistent with the regional features (Figure 5). The negative phases under the Middle Tien Shan (D2 and Dx) are more prominent in the high-frequency images than in the low-frequency images, but they are visible in all profiles. The shallow negative phases under the North Tien Shan (D3) are clearer in the high-frequency images, but they are almost absent in the low-frequency images. In a short summary, the along-strike variations of the SRFs in the Central Tien Shan are insignificant; the structures change across the orogen in general in the NW-SE direction.

The negative phases described above generally represent the LAB in Central Tien Shan. The LAB depths are consistent with those determined by Kumar et al. [35] under the South Tien Shan and the Tarim Basin (Figures 5(b) and 5(c)). By contrast, under the North and Middle Tien Shan, the LAB of Kumar et al. [35] is located at ~100 km, which is consistent with the weak negative signals in the high-frequency image (Figure 5(b)).

Another notable feature is the positive phase (L) under the LAB, which extends between 200 km and 300 km depths under the Kazakh Shield and Tien Shan and around 200 km depth under the Tarim Basin. They are parallel to the LABs (K2 and T) above, showing opposite trends toward the Tien Shan. The L phases may represent the Lehmann discontinuity that is the bottom of the asthenosphere [42].

The Cenozoic evolution of the Tien Shan is controlled by the distant effect of the Indo-Asian collision [4, 6–8, 43]. The Central Tien Shan is characterized by strong lateral variations in north-south direction. The southern part (i.e., South Tien Shan) has high topography, strain rate, and seismicity. In contrast, in the northern part (Middle and North Tien Shan), the topography includes various ridges and depressions; both the strain rate and seismicity are weaker than those in the South Tien Shan. In the time domain, the uplift started earlier in the South Tien Shan than in the North Tien Shan. The dominant features in the SRF images also change in north-south direction in general while the east-west variation is weak.

Under the South Tien Shan, the LAB is located at ~100-130 km depths (D1); it dips northward (Figure 5). Seismic tomography revealed a clear northward-dipped high-velocity body in the upper mantle that was generally interpreted as the Tarim lithosphere under the South Tien Shan [23–28, 44]. Detailed crustal structures favor northern intrusion or underthrust of the Tarim lithosphere rather than subduction [20, 45]. Our SRF images are consistent with a possible northward intrusion. This model may account for the initial uplift of the South Tien Shan in ~25-15 Ma [12, 14, 15].

However, the LABs are not continuous under the South Tian Shan and Tarim Basin; there is a ~50 km offset in depth (Figure 5). We propose that the lower part of the lithosphere under the South Tien Shan has been delaminated. P wave receiver functions revealed uplifted 410 km discontinuity and thickened mantle transition zone under the South Tien Shan [29–31], suggesting that the delaminated lithosphere has sunk into the transition zone. The delamination and subsequent mantle upwelling may cause another uplift of the South Tien Shan in ~10-8 Ma [12, 14, 15]. The upwelling explains the partial melting in the crust of the South Tien Shan that was expressed as low velocities and high Vp/Vs ratios [19, 22, 24, 46].

Under the Middle Tien Shan, the LAB structures are complex. Two plausible negative Sp phases are visible in the upper mantle, one weak phase at ~100 km depth (D2) while another at ~200 km depth (Dx) (Figure 5). Dx is deeper than the LAB under the Kazakh Shield and Tarim Basin, so it is not likely the LAB under the orogen. Here, we interpret that Dx represents the newly delaminated lithosphere while D2 is the new LAB after the lithospheric delamination.

Because the delamination is in an early stage, the response at surface is indistinctive as widespread mantle upwelling has not developed. However, two observations still argue that the delamination occurs under the Middle Tien Shan. First, the Bouguer gravity anomalies show strong negative values (Figure 6) that cannot be matched by regional compensation of a continuous elastic plate under the Tien Shan; mantle downwelling is required to explain the gravity anomalies [47, 48]. Thus, the initial-stage delamination without significant mantle upwelling is a good candidate to the measured gravity anomalies. Second, the delamination promoted the second exhumation of the North Tien Shan. Taking the speed of delamination of ~3-5 cm/yr [49], a 50-100 km distance of delamination (as compared with the adjacent LAB) implies that the delamination under the Middle Tien Shan started at ~3-1 Ma. The estimated time is consistent with the latest rapid exhumation of the North Tien Shan at 3-0 Ma and the South Tien Shan at ~3-2 Ma [4, 11, 12]. Therefore, we suggest that the lithospheric delamination under the Middle Tien Shan reduced the strength of the lithosphere so that the crustal deformation and uplift in the adjacent South and North Tien Shan were accelerated.

Under the North Tien Shan and Kazakh Shield, two negative Sp phases are located at ~120-140 km (K1) and ~160-200 km (K2) depths, respectively; they slightly dip southward (Figure 5). The horizontally consistent structures in these two regions indicate that the lithospheric mantle of the North Tien Shan has not been deformed in general as below the South and Middle Tien Shan. Therefore, the uplift of the North Tien Shan was caused solely by crustal deformation (e.g., [20, 45]). The slightly dipped plane may be formed by the crustal load as the North Tien Shan (e.g., [8, 11]. In fact, the two pulses of deformation in the North Tien Shan, i.e., 11-10 Ma and 3-0 Ma [11], are correlated with the lithospheric delamination occurring in the South and Middle Tien Shan as proposed above, being consistent with the model of crustal deformation in response of the lithospheric delamination under convergence (e.g., [43]).

Yet another two questions need to be addressed: what do the two negative phases (K1, K2) represent? Will the lithospheric delamination propagate northward further? They could be well answered if the upper negative phase (K1) represents the middle lithospheric discontinuity (MLD). The MLD is usually observed within the lithosphere under continents, especially the craton (e.g., [50, 51]). It may be a thin layer of partial melting that separate different layers of compositions or deformations in the lithosphere [50]. The MLD signal in our images is located at ~120-140 km, being deeper than the global averages [50, 51]. But if the upper mantle contains volumes of hydrous minerals, notable negative velocity gradients can be generated below 120 km depth [51]. Therefore, we regard the negative Sp phases at ~120-140 km as the MLD under the North Tien Shan and Kazakh Shield. The MLD is a weak layer, so the lower part of the lithosphere would be detached along the MLD [50]. In particular, as the delaminated lithosphere under the Middle Tien Shan sinks further, the asthenospheric material will rise and erode the adjacent lithosphere. Therefore, the delamination is expected to propagate northward under the North Tien Shan and Kazakh Shield in the future. Phase D3 provides another way for the potential delamination. If it is the same structure as phases D1 and D2 under the South and Middle Tien Shan, the whole lithosphere may be delaminated instead of the lower part of the lithosphere through the MLD (K1). However, the data coverage is poorer than those under the South Tien Shan and Tarim Basin (Figures 2(a) and 4–6); thus, improved observations are necessary to better understand the structures under the North Tien Shan and Kazakh Shield.

We imaged the lithospheric structures beneath the Central Tien Shan using S wave receiver functions (SRF). The SRF images show distinct features across the orogen in the north-south direction. A conceptual model is shown in Figure 7. The main findings are summarized as follows:

• (1)

  Delamination has occurred beneath the South and Middle Tien Shan. The delaminated lithosphere has sunk into the mantle transition zone under the South Tien Shan while the process is in the initial stage under the Middle Tien Shan. The lithosphere under the North Tien Shan is still stable, but the delamination may propagate northward as a plausible MLD is imaged in the lithosphere underneath

• (2)

  The uplift of the Central Tien Shan is closely related to the lithospheric delamination. While the initial uplift (~25-15 Ma) is caused by the lateral shortening due to the distant effect of the Indo-Asian collision, the latter two pulsed uplifts (~10 Ma and~3 Ma) coincide with delamination under the South and Middle Tien Shan. The delamination reduced the lithospheric strength significantly, so the uplift of the Tien Shan was accelerated following the delamination

The data were downloaded freely from the Incorporated Research Institutions for Seismology (IRIS) Data Center (10.7914/SN/KC, 10.7914/SN/KN, 10.7914/SN/KR, 10.7914/SN/XP_2005, and 10.7914/SN/XW_1997) and the GEOFON data centre of the GFZ German Research Centre for Geosciences (doi:10.14470/3U7560589977).

The authors declare that there is no conflict of interest.

This work is supported by the National Natural Science Foundations of China (41203011 and 42174056). ZH is also supported by the Deng-Feng Scholar Program of Nanjing University. The S wave receiver functions were calculated using Seismic Handler [52]. The numerical calculations in this paper have been done on the computing facilities in the High Performance Computing Center of Nanjing University. The figures were made using GMT [53].