Get 20M+ Full-Text Papers For Less Than $1.50/day. Start a 14-Day Trial for You or Your Team.

Learn More →

Lithospheric structure of the Arabian Shield from joint inversion of P- and S-wave receiver functions and dispersion velocities

Lithospheric structure of the Arabian Shield from joint inversion of P- and S-wave receiver... Al-Amri, A.M. 2015. Lithospheric structure of the Arabian Shield from joint inversion of P- and S-wave receiver functions and dispersion velocities. Acta Geologica Polonica, 65 (2), 229­245. Warszawa. New velocity models of lithospheric thickness and velocity structure have been developed for the Arabian Shield by three tasks: 1) Computing P-Wave Receiver Functions (PRFs) and S-Wave Receiver Functions (SRFs) for all the broadband stations within the Saudi seismic networks. The number of receiver function waveforms depends on the recording time window and quality of the broadband station. 2) Computing ambient noise correlation Green's functions for all available station pairs within the Saudi seismic networks to image the shear velocity in the crust and uppermost mantle beneath the Arabian Peninsula. Together they provided hundreds of additional, unique paths exclusively sampling the region of interest. Both phase and group velocities for all the resulting empirical Green's functions have been measured and to be used in the joint inversion. 3) Jointly inverted the PRFs and SRFs obtained in task 1 with dispersion velocities measured on the Green's functions obtained in task 2 and with fundamental-mode, Rayleigh-wave, group and phase velocities borrowed from the tomographic studies to precisely determine 1D crustal velocity structure and upper mantle. The analysis of the PRFs revealed values of 25­45 km for crustal thickness, with the thin crust next to the Red Sea and Gulf of Aqaba and the thicker crust under the platform, and Vp/Vs ratios in the 1.70­1.80 range, suggesting a range of compositions (felsic to mafic) for the shield's crust. The migrated SRFs suggest lithospheric thicknesses in the 80­100 km range for portions of the shield close to the Red Sea and Gulf of Aqaba and near the Arabian Gulf. Generally, the novelty of the velocity models developed under this paper has consisted in the addition of SRF data to extend the velocity models down to lithospheric and sub-lithospheric depths. Keywords: Receiver function; Crustal structure; Arabian shield; Ambient noise; Seismic velocities; Lithospheric thickness. INTRODUCTION The Arabian Shield comprises about one-third of the western Arabian Peninsula and is mostly composed of Precambrian metamorphic and plutonic rocks. It is clear example of late Precambrian plate tectonics, crustal growth, and a superb example of a recently rifted continent that follows the ideas of plate tecton- ics. Moreover, although the shield is composed of Precambrian terrains, it contains Tertiary and Quaternary volcanic rocks ­ the harrats ­ that related to the early stages of the Red Sea formation and possibly to plumerelated lithospheric erosion (Stoesser and Camp 1985; Camp and Roobol 1992). Mapping the subsurface structure of the Shield is therefore an important step towards improving our understanding of the continental 230 ABDULLAH M. AL-AMRI. growth processes and subsequent modifications through rifting and plume-lithosphere interaction. Current models of Precambrian crustal evolution (Durrheim and Mooney 1991) predict that Proterozoic terranes are underlain by fertile (FeO-rich) cratonic roots that will allow the production of mafic magmas and the under plating of the Proterozoic crust upon heating by a thermal event. Interestingly, seismic velocity models of the crust and uppermost mantle of the shield revealed crustal thicknesses under 40 km and lower crustal velocities below 4.0 km/s, suggesting the absence of a layer of mafic under plate under the Arabian terrains. Nevertheless, the surface volcanism points towards recent (<30 Ma) generation of magmas in the continental lithosphere. Therefore, imaging of the continental lithosphere under the Arabian Shield might provide a unique opportunity for imaging the initial stages of mafic crustal under plating postulated by Durrheim and Mooney (1991). In this paper, new velocity models of lithospheric thickness and velocity structure for the Arabian Shield by jointly inverting multiple data sets have been tried: P-wave receiver functions, S-wave receiver functions, surface-wave dispersion velocities, and ambient-noise cross-correlation dispersion velocities. These velocity models can provide robustly constrained images of the velocity structure or the Arabian lithosphere and determine the depth of seismic discontinuities bounding the top and bottom of Lower Velocity Zones (LVZs) associated with the migration of magma within the crust and lithospheric mantle. The velocity models that will be produced through the joint inversion of multiple datasets will be better constrained and have better resolution than other known models for the region. This is because we will try to fit four complimentary seismic datasets in our inversions. The joint inversion will take advantage of P-wave receiver functions to constrain depths to intra-crustal discontinuities and crustal thickness, S-wave receiver functions to constrain depths to lithospheric discontinuities and lithospheric thickness, and the dispersion velocities from ambient noise cross-correlation and surfacewaves to constrain velocities and velocity gradients in the shallow crust and continental lithosphere. The resolution of the velocity-depth profiles obtained by the joint inversion of receiver functions and dispersion velocities is a combination of the resolving power of the data sets. Our experience suggests that Swave velocities and layer thicknesses are generally constrained within 0.1 km/s and 2.5 km in the crust and uppermost mantle, and that the resolution degrades with depth (Juliá et al. 2003, 2008, 2009). Laterally, the resolution is at best within the size of one surface- wave tomographic cell (typically 1° × 1°). Teleseismic P-waveforms recorded at three-component seismic stations contain much information on the earthquake source, the structure near both the source and the receiver, and the mantle path in between. PRFs are obtained by removing the effects of the source and the path to isolate near-receiver structure (Langston 1979). At relatively sharp velocity discontinuities, several percent of an incoming P-wave's energy is converted to an S-wave; the amplitudes of the resulting P-to-S conversions carry important information on the velocity contrast across the discontinuity and the S-P travel-times of the conversions depend on the depth of the discontinuity as well as the velocity of the structure above it. Examining both the radial and tangential receiver functions can provide additional constraints on dipping interfaces and anisotropy (Cassidy 1992; Levin and Park 1997). The joint inversion of P-wave receiver functions (PRFs) and surface-wave dispersion velocities has proven an effective tool for determining the 1D velocity structure of the continental crust and its underlying uppermost mantle (Juliá et al. 2005, 2008, 2009; Dugda et al. 2007; Keranen et al. 2009; Kgaswane et al. 2009). Both PRFs and surface-wave dispersion are mostly sensitive to S-wave velocity, but the constraints conveyed by each data set are different. PRFs are mostly sensitive to subsurface seismic discontinuities, constraining S-P travel-times between the discontinuities and the free surface and the S-wave velocity contrasts across them. To date, most ambient seismic noise research has focused on group velocity measurements of the resulting surface waves (Moschetti et al. 2007; Shairo et al. 2005) although more recently, work has begun on measuring the phase velocities (Lin et al. 2008) and amplitudes (Prieto and Beroza 2008; Matzel 2008) and the new discoveries and methodologies are constantly being developed. SEISMOTECTONIC SETTING Text-fig. 1 shows the map of the Arabian Peninsula along with major tectonic features and earthquake occurrences. Active tectonics of the region is dominated by the collision of the Arabian Plate with the Eurasian Plate along the Zagros and Bitlis Thrust systems, rifting and seafloor spreading in the Red Sea and Gulf of Aden. Strike-slip faulting occurs along the Gulf of Aqabah and Dead Sea Transform fault systems. The framework of the shield is composed of Precambrian rocks and metamorphosed sedimentary and intruded 231 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD by granites. The fold-fault pattern of the shield, together with some stratigraphic relationships suggests that the shield have undergone two orogenic cycles. To the first order, the Arabian shield is composed of two layers, each about 20km thick, with average velocities of about 6.3 km/s and 7 km/s respectively (Mooney et al. 1985). The crust thins rapidly to less than 20 km at the western shield margin, beyond which the sediments of the Red Sea shelf and coastal plain are underlain by oceanic crust. The crustal thickness in the Arabian Shield varies from about 15 km in the Red Sea, to 20 km along the Red Sea coast to about 35­40 km in the in central Arabian Shield (Al-Amri 1998; Al-Damegh et al. 2005; Tkacic et al. 2006). Reports of large-scale seismic tomography (Debayle et al. 2001) suggest that a low-velocity anomaly in the upper mantle extends laterally beneath the Arabian Shield from the Red Sea in the west to the Shield-Platform boundary in the east. It is known that high-frequency regional phase behavior in the Arabian Plate is quite variable as demonstrated by Al-Damegh et al. (2005). They investigated the attenuation of Pn phase (QPn) for 1­2 Hz along the Red Sea, the Dead Sea fault system, within the Arabian Shield and in the Arabian Platform. Consistent with the Sn attenuation, they observed low QPn values of 22 and 15 along the western coast of the Arabian Plate and along the Dead Sea fault system, respectively, for a frequency of 1.5 Hz. Higher QPn values of the order of 400 were observed within the Arabian Shield and Platform for the same frequency. Their results based on Sn and Pn observations along the western and northern portions of the Arabian Plate imply the presence of an anomalously hot and thinned lithosphere in these regions that may be caused by the extensive upper mantle anomaly that appears to span most of East Africa and western Arabia. Pasyanos et al. (2009b) applied a technique to simultaneously invert amplitudes measurements of Pn, Pg, Sn and Lg to produce P-wave and S-wave attenuation models of the crust and upper mantle. Their attenuation is modeled as P-wave and S-wave attenuation surfaces for the crust, and similar set for the upper mantle. They used all of the phase amplitudes together by using different (source, geometrical-spreading, site, and attenuation) terms for each phase. Text-fig. 1. Seismotectonic map of the Arabian Peninsula and Arabian plate boundaries showing earthquake locations and volcanic centers 232 ABDULLAH M. AL-AMRI. ANALYTICAL METHODS Computation of P-wave receiver functions PRFs and SRFs have been computed for all the broadband stations within the Saudi seismic networks. We followed the approaches of Langston (1979) and Sodoudi (2005) for the computation of PRFs and SRFs, respectively, along with the iterative deconvolution method of Ligorría and Ammon (1999). The number of receiver function waveforms depended on the recording time window and quality of the broadband station. The data set for the computation of PRFs consisted of teleseismic P-waveforms for seismic sources located between 30° and 90° epicentral distance and with magnitude 5.5 mb or larger recorded at broadband seismic stations in Saudi Arabia (Text-fig. 2). The selected waveforms were processed to obtain Pwave receiver function estimates for all the stations (Langston 1979). First, the waveforms were windowed 10 s before and 110 s after the teleseismic P-wave arrival, detrended, and tapered with a cosine taper function of 5% width. The waveforms were then high-pass filtered with a 3-pole Butterworth filter at 20 s to remove long-period noise, and decimated to 10 samples per second after low-pass filtering below 8 Hz to avoid aliasing. The filtered waveforms were rotated into the ZRT cylindrical- coordinate system and the vertical component (Z) was then deconvolved from the radial (R) and transverse (T) components to produce radial and transverse receiver functions, respectively, for each event. The deconvolution was implemented through the iterative time-domain procedure of Ligorría and Ammon (1999) with 500 iterations, at two overlapping Gaussian widths of 2.5 (fc 1.25 Hz) and 1.0 (fc 0.5 Hz) to remove high-frequency noise. From the deconvolved waveforms we selected highquality receiver functions for our analysis. First, the Text-fig. 2. Topographic map of the Arabian Peninsula showing the broadband stations. The stations belong to three different networks: King Saud Center for Science and Technology (KACST), Saudi Geological Survey (SGS), and the Incorporated Research Institutes for Seismology (IRIS) 233 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD Table 1. Stations and waveform statistics for the selected dataset 234 ABDULLAH M. AL-AMRI. performance of the deconvolution process was assessed by computing the root-mean-square (RMS) misfit between the original radial waveform and the recovered radial waveform obtained after convolving the receiver function back with the vertical component. Receiver functions with percent recoveries under 85% were rejected. Second, transverse receiver functions were examined and the corresponding radial receiver functions were rejected if significant signal was observed. This step ensured that radial receiver functions were not significantly disrupted by small-scale lateral heterogeneities under the recording station. Finally, the remaining radial receiver functions were overlapped to visually identify outliers, which were removed from the data set. The final number of selected PRFs for each station and frequency band is reported in Table 1. Crustal thickness and bulk Vp/Vs ratios The hk-stacking technique of Zhu and Kanamori (2000) has been applied PRFs with frequencies (fc 1.25 Hz) to obtain thickness and bulk Vp/Vs ratios for the crust under each recording station in Saudi Arabia. The technique consists of a transformation from a time-domain representation of the PRFs into a new domain, which is a function of the depth to the interface (H) and Vp/Vs (k) ratio. The transformation is described as; s(H, k) = [w1 r fi (t1) + w2 r fi (t2) - w3 r fi (t3)], (3.2.1) where r fi (tj) are the receiver function amplitudes at t=tj, wi are a priori weights such that wi =1, and the summation extends to all the PRFs for a given station. The times tj are given by t1 = tPs ­ tP = H ( ­ ) t2 = tPpSs ­ tP = H ( + ) t3 = tPsPs­ tP = 2H , (3.2.2a) (3.2.2b) (3.2.2c) Computation of S-wave receiver functions The data set for the computation of SRFs consisted of teleseismic S-waveforms for seismic sources located between 60° and 82° epicentral distance and with magnitude 6.0 mb or larger recorded at broadband seismic stations in Saudi Arabia. At a given station, the converted Sp phase arrives earlier than the direct S phase, and S-to-P conversions upon refraction (Sp) across shallow discontinuities are best observed within the distance range mentioned earlier (Wilson et al. 2006). To compute the SRFs, the teleseismic S-waveforms were windowed 110 s before and 10 s after the teleseismic S-wave arrival, detrended, and tapered with a cosine taper function of 5% width. The waveforms were then high-pass filtered with a 3-pole Butterworth filter at 20 s to remove long-period noise, and decimated to 10 samples per second after low-pass filtering below 8 Hz to avoid aliasing. Following standard practice in SRF studies (Hansen et al. 2007), the filtered waveforms were rotated into the local ray-coordinate system (LQT system). First, a rotation from the ZNE system into the great-circle path (ZRT system) was performed using the event's back-azimuth. Second, a rotation was performed from the ZRT system into the LQT system, using the method of Sodoudi (2005). In this method, R-Z seismograms for a given event are rotated through a series of incidence angles to create sets of quasi-SV and quasi-P data. Each quasi-SV component is then deconvolved from the corresponding quasi-P component using Ligorría and Ammon (1999) iterative time domain method, with 500 iterations. The optimal incidence angle for rotation into the SH-SV-P coordinate system is the angle that minimizes the direct S-wave energy on the P- component. On the time-reversed receiver functions, the direct S arrival is at 0 s. Therefore, the receiver function of interest is the one which mean amplitude is closest to zero at zero time. The deconvolution was implemented at Gaussian widths of1.0 (fc 0.5 Hz) to remove high-frequency noise. While receiver functions are generated for all events at a given station, it is also important to highgrade the dataset using a number of quality control criteria. Since SRFs tend to be noisier than PRFs, the SRFs are compared to PRFs at the same station to identify the Moho conversion. Only SRFs that display a clear Moho conversion at the appropriate time are used for further analysis. In addition, the amplitude of the Moho conversion can be examined. Using assumed, average velocities for the crust and upper mantle, forward modeling can be used to predict the where is the vertical slowness (obtained through a trial k value and an a priori value for crustal P-wave velocity) and H is a trial depth. During the transformation, the amplitudes of P-to-S conversions at the computed times tj will interact constructively only when the trial H and Vp/Vs ratios coincide with the depth to and Vp/Vs ratio above an actual discontinuity. The results on crustal thickness and Vp/Vs ratio for the broadband stations in Saudi Arabia are displayed in Text-fig. 3 and listed in Table 2, along with 2 confidence bounds obtained after bootstrapping the original dataset with 200 replications (Efron and Tibshirai 1991) and other relevant information. 235 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD expected Moho amplitude. If the amplitude of the Moho conversion on the SRF is too large or too small, indicating an unrealistic velocity contrast across the crust-mantle boundary, the SRF is discarded. Generally, the percentage of useable, high-quality data obtained in SRFs analysis is comparable to that ob- tained in PRF analysis. However, fewer events overall are available for SRFs given the more restricted distance range used, so a large initial dataset is required. Some stations did not include enough data to compute stable SRF averages to be used in the joint inversion. Text-fig. 3. Map displaying variation in point measurements of crustal thickness (H) and Vp/Vs ratios in the Saudi Arabia. Each measurement is shown next to the station location as hh.h/v.vv, where hh.h is crustal thickness and v.vv is Vp/Vs ratio. Confidence bounds for the measurements can be found in Table 2. Note the stations are color-coded according to seismic network 236 ABDULLAH M. AL-AMRI. Station KACST NETWORK affs arss ayus bdas bljs djns haqs hass hils joms jmqs kbrs mzls nams qurs tats tays tbks ynbs IRIS NETWORK afif halm rani rayn soda taif uqsk SGS NETWORK sgs01 sgs02 sgs03 sgs08 sgs09 sgs12 sgs13 sgs16 sgs19 sgs21 sgs22 sgs23 sgs24 sgs25 sgs27 30 31 27 12 26 46 43 43 43 7 43 50 59 29 48 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.4 0.4 0.5 0.4 0.4 0.3 0.3 0.3 0.5 0.5 0.5 0.3 0.3 0.3 0.3 0.3 0.3 0.5 0.3 0.3 0.3 0.3 0.3 0.0 0.0 0.0 0.3 0.3 0.3 0.3 0.3 0.3 0.0 0.3 0.3 39.3±0.7 28.0±0.6 37.1±0.5 39.3±1.6 25.3±0.7 44.9±0.4 45.6±0.7 30.5±0.4 47.7±0.3 35.6±1.7 29.6±0.5 46.2±0.5 47.4±1.3 39.5±1.6 33.1±1.0 1.73±0.03 1.73±0.03 1.68±0.02 1.77±0.03 1.81±0.03 1.71±0.01 1.79±0.01 1.79±0.02 1.65±0.01 1.71±0.06 1.69±0.02 1.77±0.01 1.81±0.08 1.75±0.05 1.63±0.04 -93.4 -95.6 -95.3 -57.5 -87.0 -84.1 -93.2 -90.4 -55.6 -92.8 -87.50 -84.4 -98.8 -89.4 -86.1 53 74 30 456 55 11 31 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 36.1±0.3 36.1±0.3 37.1±0.8 40.2±0.1 38.2±0.3 37.2±0.6 36.3±0.4 1.73±0.01 1.76±0.01 1.75±0.03 1.76±0.00 1.73±0.01 1.75±0.03 1.76±0.02 -80.8 -96.7 -91.6 -86.0 -88.0 -91.5 -80.5 88 22 68 67 104 86 83 93 139 71 41 25 87 62 58 79 37 93 64 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.5 0.4 0.5 0.5 0.5 0.4 0.5 0.5 0.4 0.4 0.4 0.4 0.4 04 0.4 0.4 0.3 0.3 0.3 0.0 0.3 0.5 0.5 0.5 0.3 0.5 0.5 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 o.5 0.3 0.0 0.0 0.0 0.3 0.0 0.0 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 36.2±0.4 38.8±0.4 29.3±0.5 27.7±0.8 38.6±0.3 42.7±1.1 28.2±0.5 43.7±0.2 37.1±0.2 32.0±1.9 38.9±0.9 35.1±0.5 41.7±0.8 41.6±0.4 34.2±0.3 39.7±0.3 28.7±0.5 35.7±0.4 24.8±0.7 1.73±0.01 1.75±0.01 1.71±0.02 1.75±0.03 1.71±0.01 1.76±0.04 1.80±0.04 1.77±0.01 1.78±0.01 1.77±0.05 1.76±0.03 1.74±0.02 1.78±0.02 1.70±0.01 1.71±0.01 1.80±0.01 1.71±0.02 1.76±0.01 1.93±0.05 -84.4 -86.2 -86.7 -94.3 -85.1 -76.3 -82.1 -85.7 -76.2 -98.5 -80.4 -89.4 -90.5 -78.6 -68.5 -89.4 -90.6 -70.4 -88.8 PRFS VP(km/s) W1 W2 W3 H(km) VP /VS Cor(%) 237 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD Lithospheric Thickness SRFs have stacked and migrated into the depth domain in order to investigate possible occurrences of negative-polarity Sp conversions that may be sugges- tive of an S-to-P conversion at the base of the lithosphere. The lithosphere-asthenosphere boundary (LAB) is likely to have a thermal origin and therefore be characterized by a gradational transition rather than a sharp discontinuity (Karato 2003). This may reduce Text-fig. 4. Migrated SRF stacks for several stations in Saudi Arabia. The grey bands are bootstrapped 1s-confidence bounds. Note the clear Sp conversions at the Moho and the more subtle negative Sp amplitudes at lithospheric depths Table 2. Point measurements of crustal thickness (H) and VP/VS ratio from hk-stacking for broadband stations in Saudi Arabia 238 ABDULLAH M. AL-AMRI. the amplitude of the Sp conversion in the SRFs and make it not identifiable in individual SRF waveforms. First, phase-moveout curves were computed for each event-station pair for a range of Sp conversion depths ranging between 0 and 300 km at 2 km depth intervals. The phase-moveout curves were obtained after ray tracing through the ak135 global earth model using the TauP utility (Crotwell et al. 1999). Next, the amplitudes for the entire receiver functions were averaged for common Sp conversion depths along the computed phase-moveout curves. Finally, the migrated SRFs were produced by plotting the averaged amplitudes against conversion depth. Confidence bounds were also computed after bootstrapping across the receiver function data set with 200 replications (Efron and Tibshirani 1991). Text-fig. 4 displays the SRF stacks for those stations that had individual SRFs with a coherent crustal pattern. Some stations do display small-amplitude Sp conversions at depths that vary between 60 km and 200 km. In general, the stations that are located in the interior of the shield display a negative amplitude Sp conversion under 150 km depth (MZLS, RAYN) or no identifiable Sp conversion at all (AFFS, HILS, QURS), while those stations located along the Red Sea and Gulf of Aqaba (AYUS, BDAS, BLJS, HAQS), and also those near Arabian Gulf (HASS, SGS12, SGS15, SGS19) display negative Sp conversions at shallower depths around 60-80 km. High-frequency dispersion from ambient- noise cross-correlations Ambient noise method has been used to image the shear velocity in the crust and the uppermost mantle beneath the Arabian Peninsula. We calculated the ambient noise correlation Green's functions for all available station pairs within the Saudi seismic networks, which provide hundreds of characteristic paths exclusively sampling the region. We measured group velocities in conjunction with the joint-inversion of receiver functions and dispersion velocities. Separately, we inverted the Green's function waveforms for the best 1D model along each path and created a 3D model based on those results. Greatest resolution is determined by station density and thus it is greatest along the western part of the kingdom along the Red Sea. The raw data consisted of continuous waveform data for the month of January 2010 from 33 stations in the Saudi national network and an additional 6 stations from stations outside the network. Following the procedure outlined in Bensen et al. (2007) we cut the data into synchronous hour-long seg- ments, instrument corrected the traces to displacement, removed the mean and the trend and applied a 1% taper. The data were then whitened to equalize the frequency content, and finally converted to single-bit waveforms to minimize the weight of large earthquakes on the correlations (Larose et al. 2004). For each hour, every station was correlated with all the others creating an initial Green's function estimate. This results in a separate correlation waveform for each hour of each day for every pair of stations. By stacking all the hourly correlations, we recover the Green's function with the best signal to noise ratio (SNR). A complete 30-day record would provide us with 720 hours of data; however, incomplete traces reduce this, somewhat. Most of the ambient noise Green's functions in this paper contain more than 660 hours of continuous data with the maximum being 712 hours and the minimum including only 270 hours. If the source of the coherent noise were uniformly distributed throughout the Earth, the Green's functions would be symmetrical around zero time. Energy traveling from station A to station B would appear at positive time and is often referred to as "causal" energy. Energy traveling from station B to station A would appear at negative time and referred to as "a causal". Ambient noise is known to be in homogeneously distributed and furthermore, seasonal in its occurrence, so the correlations will typically be dominant on either the causal or a causal side. Additionally the source of the ambient noise varies with frequency, so that the waveform will appear to be dominated by causal energy in some bands and by acausal energy in others. In this study, we take every correlation trace, reverse it around zero time and stack it with the unreversed original. This combines the "causal" (t > 0) and "acausal" (t < 0) portions of the empirical Green's function. This step results in Green's functions that look symmetric around zero and allows us to perform all of our measurements on the causal trace. Excluding autocorrelations, we obtained Green's function estimates for 528 paths between stations within the Saudi network and 741 paths in total. Separation between stations ranged between 23 km and 3917 km. RESULTS AND INTERPRETATION 1D Model Inversions The correlation Green's functions have been used as data to invert for shear velocity using a partitioned waveform inversion scheme (Nolet 1990). The task is 239 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD to obtain the best 1D model along each path and then merge the paths together to create a 3D image. The optimization was defined to maximize the fit of a synthetic seismogram to the envelope of the Green's function. We used a very-fast simulated annealing technique (Sen and Stoffa 1995) to search efficiently through model space. Synthetics were created using the reflectivity method described by Randall (1994), which allows rapid calculation of seismograms for 1D-earth models. This allowed us to conduct several Text-fig. 5. Slices through a localized 3D model at 10, 20, 50 and 100 km 240 ABDULLAH M. AL-AMRI. thousand iterations and to obtain a separate best fitting model for each of the 741 paths. Because SNR is a function of distance and frequency, we ran two separate inversion runs: One for data filtered over frequency band between 0.05­0.25 Hz to optimize the strong signal from the short-range data and the other for a frequency band between 0.03­ 0.09 Hz to avoid the high noise content of the longrange correlations. All models for distances less than 500 km are the result of the 0.05­0.25 Hz run. All models for distances greater than 1100 km are the result of the 0.03­0.09 Hz run. Localized 3D tomography The best fitting 1D models were merged into a 3D model of the subsurface using a partitioned waveform inversion technique (Nolet 1990). Given the SNR quality of the long-range data, together with the differences in resolution due to station distribution, we focused on the most well resolved region, surrounding the densely instrumented western Kingdom. A grid restricted between 15°­30° latitude and 34°­44° with a node spacing of 0.2°. We also only used the paths for stations separated by less than 1150 km. Based on the data quality issues mentioned above, the result (Textfig. 5) should be considered preliminary, and will be enhanced by the joint inversion model. However, it illustrates the large degree of lateral heterogeneity that can be resolved using seismic interferometry in a densely instrumented region. Joint Inversion Models for the Arabian Shield The PRFs jointly inverted and SRFs obtained in task 1 with dispersion velocities measured on the Green's functions obtained in task 2 and with fundamental-mode, Rayleigh-wave, group and phase velocities borrowed from the tomographic studies of Pasyanos (2005) and Ekström et al. (1997). Using he receiver functions and dispersion measurements, we produced joint inversion models for the broadband stations in Saudi Arabia. Velocity models from the joint inversion of PRFs, SRFs, and dispersion velocities have been produced for 17 stations in Saudi Arabia, corresponding to stations for which stable SRF estimates could be obtained (Text-fig. 6). As explained earlier, stable SRF averages are harder to obtain than stable PRF averages due to the reduced epicentral distance range for telesismic S-wave sources and also due to the fact that Sp conversions occur within the coda of the teleseismic Pwave. Before the inversion, PRFs and SRFs wave- forms for each station were sorted by ray parameter and binned within ranges of 0.10 s/km and the receiver functions within each bin were averaged, if the bin included a minimum of three waveforms, to improve the signal-to-noise ratio. Binning around ray parameter values (or, equivalently, incidence angle) before averaging the PRFs and SRFs is important to prevent incoherent overlap of P-to-S and S-to-P conversions, respectively, due to different phase move-outs. Surface-wave dispersion velocities for the joint inversion were obtained from the independent surfacewave tomography study of Pasyanos (2005). The study measured quality fundamental-mode group velocities along 30,000 Rayleigh wave and 20,000 Love wave paths for periods between 7 and 100 s, and reported lateral variation in group-velocity across Eurasia and North Africa. Although the results of the tomographic study were reported in 1°×1° cells, the checkerboard tests showed that the average resolution of the study is around 4°×4° (for 50s periods Rayleigh waves). We identified the tomographic cell containing the selected stations in the Arabian Peninsula and extracted the corresponding dispersion curve, which was then regarded as the local dispersion estimate for that station. As the resolution of the Love waves is poorer than that of the Raleigh waves (due to fewer measurements and poorer signal-to-noise ratios), we only utilized the Rayleighwave dispersion curve. For the inversions, we considered a starting model consisting of a 40 km-thick crust with a gradational increase between 6.0 and 7.0 km/s in P-wave velocity and a uniform crustal Vp/Vs ratio of 1.75, overlying a flattened PREM model (Dziewonski and Anderson 1980). We found a uniform Vp/Vs ratio justified for the Arabian crust, as values obtained from the hk-stacking analysis do not deviate significantly from this average, in general (see Table 2); for the lithospheric and sublithospheric mantle a PREM-like, depth-dependent Vp/Vs value was assumed. The model was parameterized down to ~500 km depth as a stack of thin layers with variable thicknesses of 1.0 km between 0 and 5 km depth, 2.5 km between 5 and 60 km depth, 5 km between 60 and 200 km depth, and 10 km below. We inverted for S-wave velocity structure down to ~250 km depth and constrained the structure below that depth to be PREM-like in order to more real display of the partial sensitivity of the long-period surface-waves by wider upper mantle. Joint inversion models were obtained after 6 iterations, using an influence factor p=0.5 and a smoothing of =0.2. To equalize the datasets, values for the weights wb and ws (eq. 4) were obtained by multiplying the number of data points with average data variances of b2=5·10-4 s-2 and 241 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD s2=10-4 km2/s2 for the dispersion velocities and receiver functions, respectively. A detailed joint inversion of PRFs, SRFs, and dispersion velocities is shown in Text-fig. 6 for station AYUS, located near the Gulf of Aqaba. The velocity model consists of a ~28 km thick crust with crustal Svelocities well under 4.0 km/s that increase with depth (the low velocity zone in the uppermost crust is probably not well resolved due to the lack of short-period measurements in the dispersion velocity curve). The crust overlies a pretty thin (~15 km) and slow (~4.3 km/s) upper mantle lid which, in turn, overlies the top of a low-velocity zone with a minimum velocity around 4.0 km/s at ~60 km depth. The bottom of the low-velocity zone returns to PREM-like velocities at around 180 km depth. The predictions from this model display an excellent agreement with observed PRFs, SRFs and fundamental-mode, Rayleigh-wave, groupvelocities. A summary of all velocity models obtained at the seismic stations in Saudi Arabia is given in Text-fig. 7. The figure displays the joint inversion models down to 350 km depth (note the bottom 100 km is constrained to be PREM-like) along with stacked PRF and SRF waveforms migrated to depth. In general, the Moho is consistently seen in both PRF and SRF waveforms and Text-fig. 6. Joint inversion of PRFs, SRFs, and SW dispersion velocities at station AYUS. The left panels display observed (black) and predicted (red) receiver functions and the middle panel at the bottom displays observed (symbols) and predicted (lines) group velocities. Only fundamental-mode, Rayleigh waves were used during the inversion (red) but a prediction for the fundamental-mode, Love-wave, group velocity is shown (green). The inverted S-velocity model is shown to the right, with the crustal portion enlarged in the middle (top) panel 242 ABDULLAH M. AL-AMRI. as expected, accurately located in the joint inversion models. The lithosphere-asthenosphere boundary (LAB) and the bottom of the low-velocity channel, on the other hand, correspond to smaller velocity contrasts in the velocity models and are therefore harder to see in PRFs. For those discontinuities, the information from the SRFs is critical and has helped constrained lithospheric and sub-lithospheric structure accurately. Geologically, we would suggest that low velocity beneath the Gulf of Aqaba and southern Arabian Shield and Red Sea at depths below 200 km are related to mantle upwelling and seafloor spreading. Low velocities beneath the northern Arabian Shield below 200 km may be related to volcanism. The low velocity feature near the eastern edge of the Arabian Shield and western edge of the Arabian Platform could be related Text-fig. 7. Summary of joint inversion results for broadband stations in Saudi Arabia. Each panel displays the migrated SRF to the left, the migrated PRFs in the middle, and the joint inversion model (S-velocity) to the right. The grey bands indicate 2-confidence bounds for the stacks. The colored horizontal lines indicate the location of the Moho (blue), LAB (red), and bottom of the asthenosphere (green) 243 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD to mantle flow effects near the interface of lithosphere of different thickness. Additionally, the low-velocity region directly under the Shield can be attributed also to a 115­330 K thermal anomaly in the upper mantle, and this anomaly can be associated with the Cenozoic uplift of and volcanic centers on the Shield. New results for the lithosphere suggest that the mantle lithosphere is thin and the low velocity zone (LVZ) is significant near the Red Sea, where rifting is active. The mantle lid thickens away from the Red Sea in the Arabian interior. The upper mantle beneath the western portion of the Arabian Shield is anomalously slow, with velocities increasing towards the continental interior. These observations are attributed to thermal differences beneath Saudi Arabia and indicate much hotter mantle beneath the Red Sea than beneath the interior of the shield, which is consistent with plume flow directed beneath the rift. Generally, the resulting lithospheric structures indicate that the Arabian Platform has an average crustal thickness of 42 km, with relatively low crustal velocities without a strong velocity gradient. The Moho is shallower (36 km) and crustal velocities are 6 per cent higher (with a velocity increase with depth) for the Arabian Shield. Fast crustal velocities of the Arabian Shield result from a predominantly mafic composition in the lower crust. This is consistent with the island arc origin of the westernmost terranes of the Arabian Shield. Lower velocities in the Arabian Platform crust indicate a bulk felsic composition, consistent with orogenesis of this former active margin. The lower mantle velocities and higher Poisson's ratio beneath the Arabian Shield probably arise from a partially molten mantle associated with Red Sea spreading and continental volcanism. the velocity models developed under this paper has consisted in the addition of SRF data to extend the velocity models down to lithospheric and sub-lithospheric depths. It is perhaps a bit disappointing that, out of 50 stations considered, joint inversion velocity models have been developed for just 17 of them. The reason for this low number of selected stations has to be explained in the computation of stable SRF waveforms. Computation of stable SRF waveform averages requires a data set significantly larger than that required for computing stable PRF waveform averages. In summary, the 17 S-velocity models have been developed from the joint inversion of PRFs, SRFs and surface-wave (ballistic) dispersion velocities. These models will help better constrain the deep crustal, lithospheric, and sub-lithospheric structure under Saudi Arabia and, in turn, help further our understanding of regional wave propagation in tectonic evolution of the Arabian shield and platform in general. Acknowledgments I would like to express my thanks and gratitude to King Abdulaziz City for Science and Technology for funding the research project AR-30-36. Drs Jordi Julia and Eric Matzel, as project consultants, whose expert guidance and continuing advice made this work possible. Their willingness to devote their time greatly facilitated the completion of the project I owe them a deep debt of gratitude. I would like also to extend my sincerest thanks to Dr. Michael Pasyanos who helped in performing surface wave group velocity measurements and tomography. http://www.deepdyve.com/assets/images/DeepDyve-Logo-lg.png Acta Geologica Polonica de Gruyter

Lithospheric structure of the Arabian Shield from joint inversion of P- and S-wave receiver functions and dispersion velocities

Al-Amri, Abdullah M.
Acta Geologica Polonica , Volume 65 (2) – Jun 1, 2015
Download PDF
17 pages

Loading next page...
 
/lp/de-gruyter/lithospheric-structure-of-the-arabian-shield-from-joint-inversion-of-p-ZUy2Fz6D7b

References (47)

Publisher
de Gruyter
Copyright
Copyright © 2015 by the
ISSN
2300-1887
eISSN
2300-1887
DOI
10.1515/agp-2015-0009
Publisher site
See Article on Publisher Site

Abstract

Al-Amri, A.M. 2015. Lithospheric structure of the Arabian Shield from joint inversion of P- and S-wave receiver functions and dispersion velocities. Acta Geologica Polonica, 65 (2), 229­245. Warszawa. New velocity models of lithospheric thickness and velocity structure have been developed for the Arabian Shield by three tasks: 1) Computing P-Wave Receiver Functions (PRFs) and S-Wave Receiver Functions (SRFs) for all the broadband stations within the Saudi seismic networks. The number of receiver function waveforms depends on the recording time window and quality of the broadband station. 2) Computing ambient noise correlation Green's functions for all available station pairs within the Saudi seismic networks to image the shear velocity in the crust and uppermost mantle beneath the Arabian Peninsula. Together they provided hundreds of additional, unique paths exclusively sampling the region of interest. Both phase and group velocities for all the resulting empirical Green's functions have been measured and to be used in the joint inversion. 3) Jointly inverted the PRFs and SRFs obtained in task 1 with dispersion velocities measured on the Green's functions obtained in task 2 and with fundamental-mode, Rayleigh-wave, group and phase velocities borrowed from the tomographic studies to precisely determine 1D crustal velocity structure and upper mantle. The analysis of the PRFs revealed values of 25­45 km for crustal thickness, with the thin crust next to the Red Sea and Gulf of Aqaba and the thicker crust under the platform, and Vp/Vs ratios in the 1.70­1.80 range, suggesting a range of compositions (felsic to mafic) for the shield's crust. The migrated SRFs suggest lithospheric thicknesses in the 80­100 km range for portions of the shield close to the Red Sea and Gulf of Aqaba and near the Arabian Gulf. Generally, the novelty of the velocity models developed under this paper has consisted in the addition of SRF data to extend the velocity models down to lithospheric and sub-lithospheric depths. Keywords: Receiver function; Crustal structure; Arabian shield; Ambient noise; Seismic velocities; Lithospheric thickness. INTRODUCTION The Arabian Shield comprises about one-third of the western Arabian Peninsula and is mostly composed of Precambrian metamorphic and plutonic rocks. It is clear example of late Precambrian plate tectonics, crustal growth, and a superb example of a recently rifted continent that follows the ideas of plate tecton- ics. Moreover, although the shield is composed of Precambrian terrains, it contains Tertiary and Quaternary volcanic rocks ­ the harrats ­ that related to the early stages of the Red Sea formation and possibly to plumerelated lithospheric erosion (Stoesser and Camp 1985; Camp and Roobol 1992). Mapping the subsurface structure of the Shield is therefore an important step towards improving our understanding of the continental 230 ABDULLAH M. AL-AMRI. growth processes and subsequent modifications through rifting and plume-lithosphere interaction. Current models of Precambrian crustal evolution (Durrheim and Mooney 1991) predict that Proterozoic terranes are underlain by fertile (FeO-rich) cratonic roots that will allow the production of mafic magmas and the under plating of the Proterozoic crust upon heating by a thermal event. Interestingly, seismic velocity models of the crust and uppermost mantle of the shield revealed crustal thicknesses under 40 km and lower crustal velocities below 4.0 km/s, suggesting the absence of a layer of mafic under plate under the Arabian terrains. Nevertheless, the surface volcanism points towards recent (<30 Ma) generation of magmas in the continental lithosphere. Therefore, imaging of the continental lithosphere under the Arabian Shield might provide a unique opportunity for imaging the initial stages of mafic crustal under plating postulated by Durrheim and Mooney (1991). In this paper, new velocity models of lithospheric thickness and velocity structure for the Arabian Shield by jointly inverting multiple data sets have been tried: P-wave receiver functions, S-wave receiver functions, surface-wave dispersion velocities, and ambient-noise cross-correlation dispersion velocities. These velocity models can provide robustly constrained images of the velocity structure or the Arabian lithosphere and determine the depth of seismic discontinuities bounding the top and bottom of Lower Velocity Zones (LVZs) associated with the migration of magma within the crust and lithospheric mantle. The velocity models that will be produced through the joint inversion of multiple datasets will be better constrained and have better resolution than other known models for the region. This is because we will try to fit four complimentary seismic datasets in our inversions. The joint inversion will take advantage of P-wave receiver functions to constrain depths to intra-crustal discontinuities and crustal thickness, S-wave receiver functions to constrain depths to lithospheric discontinuities and lithospheric thickness, and the dispersion velocities from ambient noise cross-correlation and surfacewaves to constrain velocities and velocity gradients in the shallow crust and continental lithosphere. The resolution of the velocity-depth profiles obtained by the joint inversion of receiver functions and dispersion velocities is a combination of the resolving power of the data sets. Our experience suggests that Swave velocities and layer thicknesses are generally constrained within 0.1 km/s and 2.5 km in the crust and uppermost mantle, and that the resolution degrades with depth (Juliá et al. 2003, 2008, 2009). Laterally, the resolution is at best within the size of one surface- wave tomographic cell (typically 1° × 1°). Teleseismic P-waveforms recorded at three-component seismic stations contain much information on the earthquake source, the structure near both the source and the receiver, and the mantle path in between. PRFs are obtained by removing the effects of the source and the path to isolate near-receiver structure (Langston 1979). At relatively sharp velocity discontinuities, several percent of an incoming P-wave's energy is converted to an S-wave; the amplitudes of the resulting P-to-S conversions carry important information on the velocity contrast across the discontinuity and the S-P travel-times of the conversions depend on the depth of the discontinuity as well as the velocity of the structure above it. Examining both the radial and tangential receiver functions can provide additional constraints on dipping interfaces and anisotropy (Cassidy 1992; Levin and Park 1997). The joint inversion of P-wave receiver functions (PRFs) and surface-wave dispersion velocities has proven an effective tool for determining the 1D velocity structure of the continental crust and its underlying uppermost mantle (Juliá et al. 2005, 2008, 2009; Dugda et al. 2007; Keranen et al. 2009; Kgaswane et al. 2009). Both PRFs and surface-wave dispersion are mostly sensitive to S-wave velocity, but the constraints conveyed by each data set are different. PRFs are mostly sensitive to subsurface seismic discontinuities, constraining S-P travel-times between the discontinuities and the free surface and the S-wave velocity contrasts across them. To date, most ambient seismic noise research has focused on group velocity measurements of the resulting surface waves (Moschetti et al. 2007; Shairo et al. 2005) although more recently, work has begun on measuring the phase velocities (Lin et al. 2008) and amplitudes (Prieto and Beroza 2008; Matzel 2008) and the new discoveries and methodologies are constantly being developed. SEISMOTECTONIC SETTING Text-fig. 1 shows the map of the Arabian Peninsula along with major tectonic features and earthquake occurrences. Active tectonics of the region is dominated by the collision of the Arabian Plate with the Eurasian Plate along the Zagros and Bitlis Thrust systems, rifting and seafloor spreading in the Red Sea and Gulf of Aden. Strike-slip faulting occurs along the Gulf of Aqabah and Dead Sea Transform fault systems. The framework of the shield is composed of Precambrian rocks and metamorphosed sedimentary and intruded 231 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD by granites. The fold-fault pattern of the shield, together with some stratigraphic relationships suggests that the shield have undergone two orogenic cycles. To the first order, the Arabian shield is composed of two layers, each about 20km thick, with average velocities of about 6.3 km/s and 7 km/s respectively (Mooney et al. 1985). The crust thins rapidly to less than 20 km at the western shield margin, beyond which the sediments of the Red Sea shelf and coastal plain are underlain by oceanic crust. The crustal thickness in the Arabian Shield varies from about 15 km in the Red Sea, to 20 km along the Red Sea coast to about 35­40 km in the in central Arabian Shield (Al-Amri 1998; Al-Damegh et al. 2005; Tkacic et al. 2006). Reports of large-scale seismic tomography (Debayle et al. 2001) suggest that a low-velocity anomaly in the upper mantle extends laterally beneath the Arabian Shield from the Red Sea in the west to the Shield-Platform boundary in the east. It is known that high-frequency regional phase behavior in the Arabian Plate is quite variable as demonstrated by Al-Damegh et al. (2005). They investigated the attenuation of Pn phase (QPn) for 1­2 Hz along the Red Sea, the Dead Sea fault system, within the Arabian Shield and in the Arabian Platform. Consistent with the Sn attenuation, they observed low QPn values of 22 and 15 along the western coast of the Arabian Plate and along the Dead Sea fault system, respectively, for a frequency of 1.5 Hz. Higher QPn values of the order of 400 were observed within the Arabian Shield and Platform for the same frequency. Their results based on Sn and Pn observations along the western and northern portions of the Arabian Plate imply the presence of an anomalously hot and thinned lithosphere in these regions that may be caused by the extensive upper mantle anomaly that appears to span most of East Africa and western Arabia. Pasyanos et al. (2009b) applied a technique to simultaneously invert amplitudes measurements of Pn, Pg, Sn and Lg to produce P-wave and S-wave attenuation models of the crust and upper mantle. Their attenuation is modeled as P-wave and S-wave attenuation surfaces for the crust, and similar set for the upper mantle. They used all of the phase amplitudes together by using different (source, geometrical-spreading, site, and attenuation) terms for each phase. Text-fig. 1. Seismotectonic map of the Arabian Peninsula and Arabian plate boundaries showing earthquake locations and volcanic centers 232 ABDULLAH M. AL-AMRI. ANALYTICAL METHODS Computation of P-wave receiver functions PRFs and SRFs have been computed for all the broadband stations within the Saudi seismic networks. We followed the approaches of Langston (1979) and Sodoudi (2005) for the computation of PRFs and SRFs, respectively, along with the iterative deconvolution method of Ligorría and Ammon (1999). The number of receiver function waveforms depended on the recording time window and quality of the broadband station. The data set for the computation of PRFs consisted of teleseismic P-waveforms for seismic sources located between 30° and 90° epicentral distance and with magnitude 5.5 mb or larger recorded at broadband seismic stations in Saudi Arabia (Text-fig. 2). The selected waveforms were processed to obtain Pwave receiver function estimates for all the stations (Langston 1979). First, the waveforms were windowed 10 s before and 110 s after the teleseismic P-wave arrival, detrended, and tapered with a cosine taper function of 5% width. The waveforms were then high-pass filtered with a 3-pole Butterworth filter at 20 s to remove long-period noise, and decimated to 10 samples per second after low-pass filtering below 8 Hz to avoid aliasing. The filtered waveforms were rotated into the ZRT cylindrical- coordinate system and the vertical component (Z) was then deconvolved from the radial (R) and transverse (T) components to produce radial and transverse receiver functions, respectively, for each event. The deconvolution was implemented through the iterative time-domain procedure of Ligorría and Ammon (1999) with 500 iterations, at two overlapping Gaussian widths of 2.5 (fc 1.25 Hz) and 1.0 (fc 0.5 Hz) to remove high-frequency noise. From the deconvolved waveforms we selected highquality receiver functions for our analysis. First, the Text-fig. 2. Topographic map of the Arabian Peninsula showing the broadband stations. The stations belong to three different networks: King Saud Center for Science and Technology (KACST), Saudi Geological Survey (SGS), and the Incorporated Research Institutes for Seismology (IRIS) 233 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD Table 1. Stations and waveform statistics for the selected dataset 234 ABDULLAH M. AL-AMRI. performance of the deconvolution process was assessed by computing the root-mean-square (RMS) misfit between the original radial waveform and the recovered radial waveform obtained after convolving the receiver function back with the vertical component. Receiver functions with percent recoveries under 85% were rejected. Second, transverse receiver functions were examined and the corresponding radial receiver functions were rejected if significant signal was observed. This step ensured that radial receiver functions were not significantly disrupted by small-scale lateral heterogeneities under the recording station. Finally, the remaining radial receiver functions were overlapped to visually identify outliers, which were removed from the data set. The final number of selected PRFs for each station and frequency band is reported in Table 1. Crustal thickness and bulk Vp/Vs ratios The hk-stacking technique of Zhu and Kanamori (2000) has been applied PRFs with frequencies (fc 1.25 Hz) to obtain thickness and bulk Vp/Vs ratios for the crust under each recording station in Saudi Arabia. The technique consists of a transformation from a time-domain representation of the PRFs into a new domain, which is a function of the depth to the interface (H) and Vp/Vs (k) ratio. The transformation is described as; s(H, k) = [w1 r fi (t1) + w2 r fi (t2) - w3 r fi (t3)], (3.2.1) where r fi (tj) are the receiver function amplitudes at t=tj, wi are a priori weights such that wi =1, and the summation extends to all the PRFs for a given station. The times tj are given by t1 = tPs ­ tP = H ( ­ ) t2 = tPpSs ­ tP = H ( + ) t3 = tPsPs­ tP = 2H , (3.2.2a) (3.2.2b) (3.2.2c) Computation of S-wave receiver functions The data set for the computation of SRFs consisted of teleseismic S-waveforms for seismic sources located between 60° and 82° epicentral distance and with magnitude 6.0 mb or larger recorded at broadband seismic stations in Saudi Arabia. At a given station, the converted Sp phase arrives earlier than the direct S phase, and S-to-P conversions upon refraction (Sp) across shallow discontinuities are best observed within the distance range mentioned earlier (Wilson et al. 2006). To compute the SRFs, the teleseismic S-waveforms were windowed 110 s before and 10 s after the teleseismic S-wave arrival, detrended, and tapered with a cosine taper function of 5% width. The waveforms were then high-pass filtered with a 3-pole Butterworth filter at 20 s to remove long-period noise, and decimated to 10 samples per second after low-pass filtering below 8 Hz to avoid aliasing. Following standard practice in SRF studies (Hansen et al. 2007), the filtered waveforms were rotated into the local ray-coordinate system (LQT system). First, a rotation from the ZNE system into the great-circle path (ZRT system) was performed using the event's back-azimuth. Second, a rotation was performed from the ZRT system into the LQT system, using the method of Sodoudi (2005). In this method, R-Z seismograms for a given event are rotated through a series of incidence angles to create sets of quasi-SV and quasi-P data. Each quasi-SV component is then deconvolved from the corresponding quasi-P component using Ligorría and Ammon (1999) iterative time domain method, with 500 iterations. The optimal incidence angle for rotation into the SH-SV-P coordinate system is the angle that minimizes the direct S-wave energy on the P- component. On the time-reversed receiver functions, the direct S arrival is at 0 s. Therefore, the receiver function of interest is the one which mean amplitude is closest to zero at zero time. The deconvolution was implemented at Gaussian widths of1.0 (fc 0.5 Hz) to remove high-frequency noise. While receiver functions are generated for all events at a given station, it is also important to highgrade the dataset using a number of quality control criteria. Since SRFs tend to be noisier than PRFs, the SRFs are compared to PRFs at the same station to identify the Moho conversion. Only SRFs that display a clear Moho conversion at the appropriate time are used for further analysis. In addition, the amplitude of the Moho conversion can be examined. Using assumed, average velocities for the crust and upper mantle, forward modeling can be used to predict the where is the vertical slowness (obtained through a trial k value and an a priori value for crustal P-wave velocity) and H is a trial depth. During the transformation, the amplitudes of P-to-S conversions at the computed times tj will interact constructively only when the trial H and Vp/Vs ratios coincide with the depth to and Vp/Vs ratio above an actual discontinuity. The results on crustal thickness and Vp/Vs ratio for the broadband stations in Saudi Arabia are displayed in Text-fig. 3 and listed in Table 2, along with 2 confidence bounds obtained after bootstrapping the original dataset with 200 replications (Efron and Tibshirai 1991) and other relevant information. 235 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD expected Moho amplitude. If the amplitude of the Moho conversion on the SRF is too large or too small, indicating an unrealistic velocity contrast across the crust-mantle boundary, the SRF is discarded. Generally, the percentage of useable, high-quality data obtained in SRFs analysis is comparable to that ob- tained in PRF analysis. However, fewer events overall are available for SRFs given the more restricted distance range used, so a large initial dataset is required. Some stations did not include enough data to compute stable SRF averages to be used in the joint inversion. Text-fig. 3. Map displaying variation in point measurements of crustal thickness (H) and Vp/Vs ratios in the Saudi Arabia. Each measurement is shown next to the station location as hh.h/v.vv, where hh.h is crustal thickness and v.vv is Vp/Vs ratio. Confidence bounds for the measurements can be found in Table 2. Note the stations are color-coded according to seismic network 236 ABDULLAH M. AL-AMRI. Station KACST NETWORK affs arss ayus bdas bljs djns haqs hass hils joms jmqs kbrs mzls nams qurs tats tays tbks ynbs IRIS NETWORK afif halm rani rayn soda taif uqsk SGS NETWORK sgs01 sgs02 sgs03 sgs08 sgs09 sgs12 sgs13 sgs16 sgs19 sgs21 sgs22 sgs23 sgs24 sgs25 sgs27 30 31 27 12 26 46 43 43 43 7 43 50 59 29 48 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.5 0.5 0.5 0.4 0.4 0.4 0.4 0.4 0.4 0.5 0.4 0.4 0.3 0.3 0.3 0.5 0.5 0.5 0.3 0.3 0.3 0.3 0.3 0.3 0.5 0.3 0.3 0.3 0.3 0.3 0.0 0.0 0.0 0.3 0.3 0.3 0.3 0.3 0.3 0.0 0.3 0.3 39.3±0.7 28.0±0.6 37.1±0.5 39.3±1.6 25.3±0.7 44.9±0.4 45.6±0.7 30.5±0.4 47.7±0.3 35.6±1.7 29.6±0.5 46.2±0.5 47.4±1.3 39.5±1.6 33.1±1.0 1.73±0.03 1.73±0.03 1.68±0.02 1.77±0.03 1.81±0.03 1.71±0.01 1.79±0.01 1.79±0.02 1.65±0.01 1.71±0.06 1.69±0.02 1.77±0.01 1.81±0.08 1.75±0.05 1.63±0.04 -93.4 -95.6 -95.3 -57.5 -87.0 -84.1 -93.2 -90.4 -55.6 -92.8 -87.50 -84.4 -98.8 -89.4 -86.1 53 74 30 456 55 11 31 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.4 0.4 0.4 0.4 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 36.1±0.3 36.1±0.3 37.1±0.8 40.2±0.1 38.2±0.3 37.2±0.6 36.3±0.4 1.73±0.01 1.76±0.01 1.75±0.03 1.76±0.00 1.73±0.01 1.75±0.03 1.76±0.02 -80.8 -96.7 -91.6 -86.0 -88.0 -91.5 -80.5 88 22 68 67 104 86 83 93 139 71 41 25 87 62 58 79 37 93 64 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 6.5 0.4 0.4 0.4 0.5 0.4 0.5 0.5 0.5 0.4 0.5 0.5 0.4 0.4 0.4 0.4 0.4 04 0.4 0.4 0.3 0.3 0.3 0.0 0.3 0.5 0.5 0.5 0.3 0.5 0.5 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 o.5 0.3 0.0 0.0 0.0 0.3 0.0 0.0 0.3 0.3 0.3 0.3 0.3 0.3 0.3 0.3 36.2±0.4 38.8±0.4 29.3±0.5 27.7±0.8 38.6±0.3 42.7±1.1 28.2±0.5 43.7±0.2 37.1±0.2 32.0±1.9 38.9±0.9 35.1±0.5 41.7±0.8 41.6±0.4 34.2±0.3 39.7±0.3 28.7±0.5 35.7±0.4 24.8±0.7 1.73±0.01 1.75±0.01 1.71±0.02 1.75±0.03 1.71±0.01 1.76±0.04 1.80±0.04 1.77±0.01 1.78±0.01 1.77±0.05 1.76±0.03 1.74±0.02 1.78±0.02 1.70±0.01 1.71±0.01 1.80±0.01 1.71±0.02 1.76±0.01 1.93±0.05 -84.4 -86.2 -86.7 -94.3 -85.1 -76.3 -82.1 -85.7 -76.2 -98.5 -80.4 -89.4 -90.5 -78.6 -68.5 -89.4 -90.6 -70.4 -88.8 PRFS VP(km/s) W1 W2 W3 H(km) VP /VS Cor(%) 237 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD Lithospheric Thickness SRFs have stacked and migrated into the depth domain in order to investigate possible occurrences of negative-polarity Sp conversions that may be sugges- tive of an S-to-P conversion at the base of the lithosphere. The lithosphere-asthenosphere boundary (LAB) is likely to have a thermal origin and therefore be characterized by a gradational transition rather than a sharp discontinuity (Karato 2003). This may reduce Text-fig. 4. Migrated SRF stacks for several stations in Saudi Arabia. The grey bands are bootstrapped 1s-confidence bounds. Note the clear Sp conversions at the Moho and the more subtle negative Sp amplitudes at lithospheric depths Table 2. Point measurements of crustal thickness (H) and VP/VS ratio from hk-stacking for broadband stations in Saudi Arabia 238 ABDULLAH M. AL-AMRI. the amplitude of the Sp conversion in the SRFs and make it not identifiable in individual SRF waveforms. First, phase-moveout curves were computed for each event-station pair for a range of Sp conversion depths ranging between 0 and 300 km at 2 km depth intervals. The phase-moveout curves were obtained after ray tracing through the ak135 global earth model using the TauP utility (Crotwell et al. 1999). Next, the amplitudes for the entire receiver functions were averaged for common Sp conversion depths along the computed phase-moveout curves. Finally, the migrated SRFs were produced by plotting the averaged amplitudes against conversion depth. Confidence bounds were also computed after bootstrapping across the receiver function data set with 200 replications (Efron and Tibshirani 1991). Text-fig. 4 displays the SRF stacks for those stations that had individual SRFs with a coherent crustal pattern. Some stations do display small-amplitude Sp conversions at depths that vary between 60 km and 200 km. In general, the stations that are located in the interior of the shield display a negative amplitude Sp conversion under 150 km depth (MZLS, RAYN) or no identifiable Sp conversion at all (AFFS, HILS, QURS), while those stations located along the Red Sea and Gulf of Aqaba (AYUS, BDAS, BLJS, HAQS), and also those near Arabian Gulf (HASS, SGS12, SGS15, SGS19) display negative Sp conversions at shallower depths around 60-80 km. High-frequency dispersion from ambient- noise cross-correlations Ambient noise method has been used to image the shear velocity in the crust and the uppermost mantle beneath the Arabian Peninsula. We calculated the ambient noise correlation Green's functions for all available station pairs within the Saudi seismic networks, which provide hundreds of characteristic paths exclusively sampling the region. We measured group velocities in conjunction with the joint-inversion of receiver functions and dispersion velocities. Separately, we inverted the Green's function waveforms for the best 1D model along each path and created a 3D model based on those results. Greatest resolution is determined by station density and thus it is greatest along the western part of the kingdom along the Red Sea. The raw data consisted of continuous waveform data for the month of January 2010 from 33 stations in the Saudi national network and an additional 6 stations from stations outside the network. Following the procedure outlined in Bensen et al. (2007) we cut the data into synchronous hour-long seg- ments, instrument corrected the traces to displacement, removed the mean and the trend and applied a 1% taper. The data were then whitened to equalize the frequency content, and finally converted to single-bit waveforms to minimize the weight of large earthquakes on the correlations (Larose et al. 2004). For each hour, every station was correlated with all the others creating an initial Green's function estimate. This results in a separate correlation waveform for each hour of each day for every pair of stations. By stacking all the hourly correlations, we recover the Green's function with the best signal to noise ratio (SNR). A complete 30-day record would provide us with 720 hours of data; however, incomplete traces reduce this, somewhat. Most of the ambient noise Green's functions in this paper contain more than 660 hours of continuous data with the maximum being 712 hours and the minimum including only 270 hours. If the source of the coherent noise were uniformly distributed throughout the Earth, the Green's functions would be symmetrical around zero time. Energy traveling from station A to station B would appear at positive time and is often referred to as "causal" energy. Energy traveling from station B to station A would appear at negative time and referred to as "a causal". Ambient noise is known to be in homogeneously distributed and furthermore, seasonal in its occurrence, so the correlations will typically be dominant on either the causal or a causal side. Additionally the source of the ambient noise varies with frequency, so that the waveform will appear to be dominated by causal energy in some bands and by acausal energy in others. In this study, we take every correlation trace, reverse it around zero time and stack it with the unreversed original. This combines the "causal" (t > 0) and "acausal" (t < 0) portions of the empirical Green's function. This step results in Green's functions that look symmetric around zero and allows us to perform all of our measurements on the causal trace. Excluding autocorrelations, we obtained Green's function estimates for 528 paths between stations within the Saudi network and 741 paths in total. Separation between stations ranged between 23 km and 3917 km. RESULTS AND INTERPRETATION 1D Model Inversions The correlation Green's functions have been used as data to invert for shear velocity using a partitioned waveform inversion scheme (Nolet 1990). The task is 239 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD to obtain the best 1D model along each path and then merge the paths together to create a 3D image. The optimization was defined to maximize the fit of a synthetic seismogram to the envelope of the Green's function. We used a very-fast simulated annealing technique (Sen and Stoffa 1995) to search efficiently through model space. Synthetics were created using the reflectivity method described by Randall (1994), which allows rapid calculation of seismograms for 1D-earth models. This allowed us to conduct several Text-fig. 5. Slices through a localized 3D model at 10, 20, 50 and 100 km 240 ABDULLAH M. AL-AMRI. thousand iterations and to obtain a separate best fitting model for each of the 741 paths. Because SNR is a function of distance and frequency, we ran two separate inversion runs: One for data filtered over frequency band between 0.05­0.25 Hz to optimize the strong signal from the short-range data and the other for a frequency band between 0.03­ 0.09 Hz to avoid the high noise content of the longrange correlations. All models for distances less than 500 km are the result of the 0.05­0.25 Hz run. All models for distances greater than 1100 km are the result of the 0.03­0.09 Hz run. Localized 3D tomography The best fitting 1D models were merged into a 3D model of the subsurface using a partitioned waveform inversion technique (Nolet 1990). Given the SNR quality of the long-range data, together with the differences in resolution due to station distribution, we focused on the most well resolved region, surrounding the densely instrumented western Kingdom. A grid restricted between 15°­30° latitude and 34°­44° with a node spacing of 0.2°. We also only used the paths for stations separated by less than 1150 km. Based on the data quality issues mentioned above, the result (Textfig. 5) should be considered preliminary, and will be enhanced by the joint inversion model. However, it illustrates the large degree of lateral heterogeneity that can be resolved using seismic interferometry in a densely instrumented region. Joint Inversion Models for the Arabian Shield The PRFs jointly inverted and SRFs obtained in task 1 with dispersion velocities measured on the Green's functions obtained in task 2 and with fundamental-mode, Rayleigh-wave, group and phase velocities borrowed from the tomographic studies of Pasyanos (2005) and Ekström et al. (1997). Using he receiver functions and dispersion measurements, we produced joint inversion models for the broadband stations in Saudi Arabia. Velocity models from the joint inversion of PRFs, SRFs, and dispersion velocities have been produced for 17 stations in Saudi Arabia, corresponding to stations for which stable SRF estimates could be obtained (Text-fig. 6). As explained earlier, stable SRF averages are harder to obtain than stable PRF averages due to the reduced epicentral distance range for telesismic S-wave sources and also due to the fact that Sp conversions occur within the coda of the teleseismic Pwave. Before the inversion, PRFs and SRFs wave- forms for each station were sorted by ray parameter and binned within ranges of 0.10 s/km and the receiver functions within each bin were averaged, if the bin included a minimum of three waveforms, to improve the signal-to-noise ratio. Binning around ray parameter values (or, equivalently, incidence angle) before averaging the PRFs and SRFs is important to prevent incoherent overlap of P-to-S and S-to-P conversions, respectively, due to different phase move-outs. Surface-wave dispersion velocities for the joint inversion were obtained from the independent surfacewave tomography study of Pasyanos (2005). The study measured quality fundamental-mode group velocities along 30,000 Rayleigh wave and 20,000 Love wave paths for periods between 7 and 100 s, and reported lateral variation in group-velocity across Eurasia and North Africa. Although the results of the tomographic study were reported in 1°×1° cells, the checkerboard tests showed that the average resolution of the study is around 4°×4° (for 50s periods Rayleigh waves). We identified the tomographic cell containing the selected stations in the Arabian Peninsula and extracted the corresponding dispersion curve, which was then regarded as the local dispersion estimate for that station. As the resolution of the Love waves is poorer than that of the Raleigh waves (due to fewer measurements and poorer signal-to-noise ratios), we only utilized the Rayleighwave dispersion curve. For the inversions, we considered a starting model consisting of a 40 km-thick crust with a gradational increase between 6.0 and 7.0 km/s in P-wave velocity and a uniform crustal Vp/Vs ratio of 1.75, overlying a flattened PREM model (Dziewonski and Anderson 1980). We found a uniform Vp/Vs ratio justified for the Arabian crust, as values obtained from the hk-stacking analysis do not deviate significantly from this average, in general (see Table 2); for the lithospheric and sublithospheric mantle a PREM-like, depth-dependent Vp/Vs value was assumed. The model was parameterized down to ~500 km depth as a stack of thin layers with variable thicknesses of 1.0 km between 0 and 5 km depth, 2.5 km between 5 and 60 km depth, 5 km between 60 and 200 km depth, and 10 km below. We inverted for S-wave velocity structure down to ~250 km depth and constrained the structure below that depth to be PREM-like in order to more real display of the partial sensitivity of the long-period surface-waves by wider upper mantle. Joint inversion models were obtained after 6 iterations, using an influence factor p=0.5 and a smoothing of =0.2. To equalize the datasets, values for the weights wb and ws (eq. 4) were obtained by multiplying the number of data points with average data variances of b2=5·10-4 s-2 and 241 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD s2=10-4 km2/s2 for the dispersion velocities and receiver functions, respectively. A detailed joint inversion of PRFs, SRFs, and dispersion velocities is shown in Text-fig. 6 for station AYUS, located near the Gulf of Aqaba. The velocity model consists of a ~28 km thick crust with crustal Svelocities well under 4.0 km/s that increase with depth (the low velocity zone in the uppermost crust is probably not well resolved due to the lack of short-period measurements in the dispersion velocity curve). The crust overlies a pretty thin (~15 km) and slow (~4.3 km/s) upper mantle lid which, in turn, overlies the top of a low-velocity zone with a minimum velocity around 4.0 km/s at ~60 km depth. The bottom of the low-velocity zone returns to PREM-like velocities at around 180 km depth. The predictions from this model display an excellent agreement with observed PRFs, SRFs and fundamental-mode, Rayleigh-wave, groupvelocities. A summary of all velocity models obtained at the seismic stations in Saudi Arabia is given in Text-fig. 7. The figure displays the joint inversion models down to 350 km depth (note the bottom 100 km is constrained to be PREM-like) along with stacked PRF and SRF waveforms migrated to depth. In general, the Moho is consistently seen in both PRF and SRF waveforms and Text-fig. 6. Joint inversion of PRFs, SRFs, and SW dispersion velocities at station AYUS. The left panels display observed (black) and predicted (red) receiver functions and the middle panel at the bottom displays observed (symbols) and predicted (lines) group velocities. Only fundamental-mode, Rayleigh waves were used during the inversion (red) but a prediction for the fundamental-mode, Love-wave, group velocity is shown (green). The inverted S-velocity model is shown to the right, with the crustal portion enlarged in the middle (top) panel 242 ABDULLAH M. AL-AMRI. as expected, accurately located in the joint inversion models. The lithosphere-asthenosphere boundary (LAB) and the bottom of the low-velocity channel, on the other hand, correspond to smaller velocity contrasts in the velocity models and are therefore harder to see in PRFs. For those discontinuities, the information from the SRFs is critical and has helped constrained lithospheric and sub-lithospheric structure accurately. Geologically, we would suggest that low velocity beneath the Gulf of Aqaba and southern Arabian Shield and Red Sea at depths below 200 km are related to mantle upwelling and seafloor spreading. Low velocities beneath the northern Arabian Shield below 200 km may be related to volcanism. The low velocity feature near the eastern edge of the Arabian Shield and western edge of the Arabian Platform could be related Text-fig. 7. Summary of joint inversion results for broadband stations in Saudi Arabia. Each panel displays the migrated SRF to the left, the migrated PRFs in the middle, and the joint inversion model (S-velocity) to the right. The grey bands indicate 2-confidence bounds for the stacks. The colored horizontal lines indicate the location of the Moho (blue), LAB (red), and bottom of the asthenosphere (green) 243 LITHOSPHERIC STRUCTURE OF THE ARABIAN SHIELD to mantle flow effects near the interface of lithosphere of different thickness. Additionally, the low-velocity region directly under the Shield can be attributed also to a 115­330 K thermal anomaly in the upper mantle, and this anomaly can be associated with the Cenozoic uplift of and volcanic centers on the Shield. New results for the lithosphere suggest that the mantle lithosphere is thin and the low velocity zone (LVZ) is significant near the Red Sea, where rifting is active. The mantle lid thickens away from the Red Sea in the Arabian interior. The upper mantle beneath the western portion of the Arabian Shield is anomalously slow, with velocities increasing towards the continental interior. These observations are attributed to thermal differences beneath Saudi Arabia and indicate much hotter mantle beneath the Red Sea than beneath the interior of the shield, which is consistent with plume flow directed beneath the rift. Generally, the resulting lithospheric structures indicate that the Arabian Platform has an average crustal thickness of 42 km, with relatively low crustal velocities without a strong velocity gradient. The Moho is shallower (36 km) and crustal velocities are 6 per cent higher (with a velocity increase with depth) for the Arabian Shield. Fast crustal velocities of the Arabian Shield result from a predominantly mafic composition in the lower crust. This is consistent with the island arc origin of the westernmost terranes of the Arabian Shield. Lower velocities in the Arabian Platform crust indicate a bulk felsic composition, consistent with orogenesis of this former active margin. The lower mantle velocities and higher Poisson's ratio beneath the Arabian Shield probably arise from a partially molten mantle associated with Red Sea spreading and continental volcanism. the velocity models developed under this paper has consisted in the addition of SRF data to extend the velocity models down to lithospheric and sub-lithospheric depths. It is perhaps a bit disappointing that, out of 50 stations considered, joint inversion velocity models have been developed for just 17 of them. The reason for this low number of selected stations has to be explained in the computation of stable SRF waveforms. Computation of stable SRF waveform averages requires a data set significantly larger than that required for computing stable PRF waveform averages. In summary, the 17 S-velocity models have been developed from the joint inversion of PRFs, SRFs and surface-wave (ballistic) dispersion velocities. These models will help better constrain the deep crustal, lithospheric, and sub-lithospheric structure under Saudi Arabia and, in turn, help further our understanding of regional wave propagation in tectonic evolution of the Arabian shield and platform in general. Acknowledgments I would like to express my thanks and gratitude to King Abdulaziz City for Science and Technology for funding the research project AR-30-36. Drs Jordi Julia and Eric Matzel, as project consultants, whose expert guidance and continuing advice made this work possible. Their willingness to devote their time greatly facilitated the completion of the project I owe them a deep debt of gratitude. I would like also to extend my sincerest thanks to Dr. Michael Pasyanos who helped in performing surface wave group velocity measurements and tomography.

Journal

Acta Geologica Polonica – de Gruyter

Published: Jun 1, 2015

There are no references for this article.