Geophysical Journal...

Geophysical Journal InternationalGeophys. J. Int. (2015) 201, 652–661 doi: 10.1093/gji/ggv043

GJI Seismology

On the limitations of interstation distances in ambient noisetomography

Yinhe Luo,1,2 Yingjie Yang,3 Yixian Xu,1,2 Hongrui Xu,1 Kaifeng Zhao1 and Kai Wang1

1Hubei Subsurface Multi-scale Imaging Lab (SMIL), Institute of Geophysics and Geomatics, China University of Geosciences, Wuhan, Hubei 430074, China.E-mail: [email protected] Key Laboratory of Geological Processes and Mineral Resources, China University of Geosciences, Wuhan, Hubei 430074, China3ARC Centre of Excellence for Core to Crust Fluid Systems (CCFS)/GEMOC, Department of Earth and Planetary Sciences, Macquarie University, NorthRyde, NSW 2109, Australia

Accepted 2015 January 21. Received 2015 January 3; in original form 2014 July 21

S U M M A R YAmbient noise tomography (ANT) has recently become a popular tomography method tostudy crustal structures thanks to its unique capability to extract short-period surface waves.Empirically, in order to reliably measure surface wave dispersion curves from time-domaincross-correlations, interstation distances between a pair of stations have to be longer thantwo/three wavelengths. This requirement imposed a strong constraint on the use of ANTat the long-period end at local- and regional-scale tomography studies. In this study, weuse ambient noise data from USArray/Transportable Array recorded during 2007–2012 toinvestigate whether dispersion measurements from cross-correlations of ambient noise atshort interstation distances are consistent with those at long distances and whether the short-path dispersion measurements can be used in tomography, especially in local- and regional-scale tomography. Our results show that: (1) surface wave phase velocity dispersion curvesmeasured by a frequency-time analysis technique (FTAN) from time-domain cross-correlationsare consistent with those measured by a spectral method tracing the zero crossings of the realpart of cross-spectrum functions in frequency domain; (2) dispersion measurements fromtime-domain cross-correlations with short interstation distances, up to only one wavelength,are consistent with and also reliable as those with interstation distances longer than threewavelengths and (3) these short-path measurements can be included in ANT to improve pathcoverage and resolution.

Key words: Surface waves and free oscillations; Seismic tomography; Computationalseismology.

1 I N T RO D U C T I O N

It has been proved that surface wave Empirical Green’s function(EGF) between two receivers can be retrieved by cross-correlatingcontinuous ambient seismic noise recorded at the two receivers(Lobkis & Weaver 2001; Snieder 2004; Shapiro et al. 2005). Ambi-ent noise tomography (ANT) based on retrieved EGFs has becomea popular method to image crustal structures thanks to its uniquecapability to extract short-period surface waves (<40/50 s). Nu-merous studies of ANT have been performed across the globe tostudy seismic structures of the crust (e.g. in USA: Moschetti et al.2007; Bensen et al. 2008; Ekstrom et al. 2009; Ekstrom 2013; inChina: Yao et al. 2006, 2008; Zheng et al. 2011; Zhou et al. 2012;in Europe: Villasenor et al. 2007; Yang et al. 2007; Li et al. 2010;Verbeke et al. 2012; in Australia: Arroucau et al. 2010; Saygin &Kennett 2010; Young et al. 2011). In ANT, both group and phasevelocity dispersion curves of surface waves are measured from

cross-correlation functions (CCF) of ambient noise. Most of ANTstudies performed to date use a frequency-time analysis technique(FTAN; Dziewonski et al. 1969; Levshin et al. 1972; Levshin &Ritzwoller 2001; Bensen et al. 2007) to extract group and phasevelocities from CCFs performed in time domain (e.g. Bensen et al.2007).

Based on the experience of analysing a large number of data sets,Bensen et al. (2007) suggest that in order to measure group veloc-ities reliably and accurately from CCFs, the interstation distancebetween a pair of station needs to be longer than three wavelengths.Bensen et al. (2007) also show that phase velocity measurements aretypically more reliable and accurate than group velocity measure-ments because the uncertainty of phase measurement of a waveformis much smaller than that of group velocity measurement based onthe peak of waveform envelope. Thus, later on, a few studies ofANT have relaxed the distance cut-off from three wavelengths totwo wavelengths when measuring phase velocities from CCFs (e.g.

652 C© The Authors 2015. Published by Oxford University Press on behalf of The Royal Astronomical Society.

Dow

nloaded from https://academ

ic.oup.com/gji/article-abstract/201/2/652/573739 by M

acquarie University user on 17 M

ay 2019

mailto:[email protected]

Limitations of interstation distances in ANT 653

Lin et al. 2009; Porritt et al. 2011; Mordret et al. 2013). FollowingBensen et al. (2007) and considering the far-field approximationin the interpretation of CCFs as EGF (e.g. Snieder 2004), most ofANT studies, however, still adopt the three-wavelength requirementto select CCFs for tomography.

This two/three-wavelength requirement imposes strong con-straints on the period range where ANT can be performed for alocal or regional seismic array. For example, for a seismic arraywith an aperture of 100 km, the longest period at which ANT canbe carried out is ∼10 s with the three-wavelength requirement or∼15 s with the two-wavelength requirement if the phase velocity atthese period ranges is ∼3.0 km s−1. On the other hand, the require-ment also forces us to discard a significant portion of CCFs withshort separations, which may be able to provide good constraints onsmall-scale heterogeneities if they could be included in tomography.

To overcome this two/three-wavelength requirement and includeshort-separation paths in tomography, Ekstrom et al. (2009) havedeveloped a novel technique to perform cross-correlations in fre-quency domain rather than in time domain, and then extract phasevelocity measurements directly from the zero crossings of the realpart of the correlation spectrum functions (CSF; hereafter referredto as the spectral method).

They show that phase velocity can be measured at station separa-tions as short as one wavelength; and dispersion maps generated byusing short-path CSFs are very similar to the counterparts by usinglong-path CSFs.

Because most of ANT studies performed to date are based ontime-domain CCFs, a natural question one may ask is that whetherphase velocities measured from time-domain CCFs with short pathsless than two/three wavelengths are also reliable and could be in-corporated in ANT. Theoretically, Tsai & Moschetti (2010) haveillustrated that the frequency domain spectral method is indeed

equivalent to the time domain cross-correlation method. However,there is still not much empirical work using a large dataset to inves-tigate this question. With continuous ambient noise data availablefrom the USArray/Transportable Array (TA), a huge number ofCCFs with different station separations can be generated, whichprovide a unique opportunity to address this question. In this study,by using CCFs between station pairs from about 1000 TA stations,we empirically investigate (1) whether phase velocity dispersionmeasurements based on FTAN and the spectral method are consis-tent with each other at different interstation distances, (2) whetherthe dispersion curves of short-separation CCFs are consistent withthose of long-separation ones when both of them are measuredfrom time-domain CCFs using FTAN and (3) if they are consistent,whether these short-separation time-domain CCFs can be includedin ANT.

2 DATA

Continuous vertical-component seismic noise data recorded by∼1000 TA stations (Fig. 1) operating in 2007–2012 are collected toobtain interstation CCFs by cross-correlating ambient noise data.The data processing procedures applied here are very similar tothose described in detail by Bensen et al. (2007). Raw continuousseismic data are cut into daily segments after being decimated to 1sample per second. The mean, trend and instrument responses are re-moved from the daily data. The daily data are then band-pass filteredat 5–100 s. The pre-processed daily data are then normalized in timedomain and whitened in frequency domain to suppress earthquakesignals and instrumental irregularities prior to performing cross-correlation. Daily cross-correlations are computed between all sta-tion pairs and then stacked to produce stacked cross-correlations.To improve the signal-to-noise ratio (SNR) of surface wave signals,

Figure 1. Distribution of the selected TA stations. The solid blue lines represent ray paths M14A–M23A, T14A–P22A, J22A–P15A and F17A–X17A withtheir dispersion curves plotted in Fig. 2. The shaded area identifies the region where ambient noise tomography is performed as shown in Fig. 8.

Dow




ay 2019

654 Y. Luo et al.

the causal and acausal parts of each cross-correlation are stacked toobtain symmetric components.

Only those CCFs with SNR of surface waves larger than 10are retained for dispersion measurement. Period-dependent SNRis defined as the peak signal in a Rayleigh wave signal windowdivided by the rms of the trailing noise. The Rayleigh wave signalwindow is defined by the time window calculated using a groupvelocity range of 2–4 km s−1. We also discard a small number ofoutliers that are internally inconsistent with the majority of data. Theinternal consistence among dispersion measurements is examinedand identified during tomography as described in Section 5. Theoutliers with travel time misfits larger than half period of surfacewaves are discarded.

3 C O M PA R I S O N B E T W E E N F TA NA N D T H E S P E C T R A L M E T H O D

FTAN is applied to the retained CCFs to measure interstation phasevelocity dispersion curves. The details of the FTAN version usedhere can be found in Levshin & Ritzwoller (2001) and Bensenet al. (2007). Basically, phase velocities are measured by measuringinstantaneous phases of each cross-correlation at various periodsand unwrapping the instantaneous phases using a global referencephase velocity model GDM52 (Ekstrom 2011). We trace phasevelocity dispersion curves from long periods to shorter periodsbecause cycle skipping at long periods can be identified and avoidedmore easily given the interstation distances are typically less than3000 km in this study.

Using time-domain CCFs, Luo et al. (2012) developed a revisedspectral method to obtain cross-spectrum functions (CSF) by sim-ply Fourier-transforming time-domain CCFs to frequency-domainCSFs and then extract phase velocity measurements from the zerocrossings of the real part of CSFs following Ekstrom et al. (2009).The revised spectral method of Luo et al. (2012) is different from thespectral method of Ekstrom et al. (2009) in that the spectral methodof Ekstrom et al. (2009) directly calculates CSFs in frequency do-main. However, Luo et al. (2012) have shown that CSFs obtainedby the two different procedures of data processing give almost samephase velocity measurements. Thus, in order to investigate whetherphase velocities calculated by FTAN and the spectral method areconsistent with each other, we measure phase velocities using FTANand the revised spectral method of Luo et al. (2012) both from thetime-domain CCFs rather than calculating all the CSFs in frequencydomain exactly following the method of Ekstrom et al. (2009).

Boschi et al. (2013) compared phase velocities measured byFTAN from time-domain CCFs and those by the spectral methodof Ekstrom et al. (2009). Their results show that phase velocitiesmeasured by these two methods are consistent with each other.

In this study, we use a much larger data set to compare themeasurements, and furthermore investigate whether the consistenceis similar for cross-correlations with different interstation distances.Four examples of phase velocity dispersion curves are presented inFig. 2 for CCFs between station pairs with their paths marked inFig. 1. Clearly, phase velocity dispersion curves measured by FTANand the spectral method are very close to each other with averagedifference less than 15 m s−1. For systematic comparison, we takeabout 350 000 dispersion measurements and divide them into threesubclasses based on their separations r relative to surface wavewavelength λ: (1) r ≥ 3λ, (2) 3λ < r ≤ 2λ and (3) 2λ < r ≤ λ.The reason why we do not include paths with r < λ for comparisonis that dispersion measurements from CCFs with r < λ have larger

Figure 2. Phase velocity dispersion curves between station pairs M14A–M23A, T14A–TP22A, J22A–P15A and F17A–X17A with their paths delin-eated in Fig. 1. The blue crosses represent the dispersion curves measuredby the spectral method and the red circles represent the dispersion curvesmeasured by FTAN.

uncertainties compared with those with r > λ, and the number ofthe returned measurements from these paths with r < λ is too smallfor statistical comparison.

One example of the histograms of differences between phasevelocities measured by FTAN and those by the spectral methodat 25 s period is plotted in Fig. 3. The histogram of differencesfor all the combined dispersion measurements at 10–50 s periodrange is also plotted in Fig. 4. In addition, Table 1 summarizesthe differences of phase velocity by listing the means and the stan-dard deviations of the differences at nine periods at 10–50 s periodrange.

The means of differences are almost close to zero for CCFs at allthe periods and at all the three distance ranges, having the largestmean of differences at only ∼2 m s−1 for paths with r ≤ 3λ. Thereis no noticeable systematic variation of the mean with period andinterstation distance. The standard deviations are about 10 m s−1

for paths with r ≥ 3λ and increase to ∼20 m s−1 for paths withλ < r ≤ 2λ. The slight increase of the standard deviations withdecreasing interstation distances is because that uncertainties ofphase velocities measured by both FTAN and the spectral methodslightly increase with decreasing distances. However, consideringthe phase velocities are around 3–4 km s−1 at the 10–50 s periodrange, the 20 m s−1 standard deviations are relatively small, onlyabout 0.5 per cent relative to the phase velocities. We do not noticethere is any systematic variation of the standard deviations with

Dow




ay 2019


Figure 3. Histograms of the differences between phase velocities measured by FTAN and those measured by the spectral method at 25 s period. The phasevelocity measurements are divided into three interstation distance ranges as indicated at the bottom of each panel. The differences for all measurementscombining the three ranges are presented in the top left-hand panel.

period. The high similarity of phase velocities measured by thesetwo methods indicates that FTAN and the spectral methods provideessentially consistent dispersion measurements from CCFs withseparations larger than one wavelength.

As demonstrated by Ekstrom et al. (2009), phase velocity mea-surements from short-separation CSFs, as short as one wavelength,based on their spectral method are robust and consistent with long-separation ones. It is, however, generally assumed that phase ve-locities measured from time-domain CCFs using FTAN are onlyreliable for those with separations longer than two/three wave-lengths. However, Table 1 and Figs 3 and 4 demonstrate that thetwo approaches actually provide consistent phase velocity measure-ments from CCFs at all the three distance ranges. Based on this,we suggest that the interstation distance cut-off of two/three wave-lengths imposed in selecting phase velocity dispersion curves fromtime-domain CCFs using FTAN may be relaxed to as short as onewavelength.

When CCFs of ambient noise contain both fundamental modeand higher mode surface waves, which are rare for continentalpaths but have been observed, especially in a sedimentary basinor on the seafloor at short periods (e.g. Savage et al. 2013; Takeoet al. 2014), measurements of fundamental-mode phase velocitieswill be affected by the presence of overtones. In these cases, timedomain CCFs need to be inspected prior and different mode of sur-face waves needs to be separated before applying the FTAN andthe spectral methods to accurately measure the fundamental-modephase velocity measurements.

4 E V I D E N C E B A S E D A T H R E E - S TAT I O NM E T H O D

To provide another line of evidence for the suggestion of relaxingthe interstation distance cut-off and meanwhile to further investi-gate if dispersion measurements using FTAN from short-separation

Dow




ay 2019

656 Y. Luo et al.

Figure 4. Same as Fig. 3, but for phase velocity measurements combining all the data at 10–50 s period range. Note the differences of counts at each panelbetween Figs 3 and 4.

Table 1. Summarized differences of phase velocity calculated by FTAN and the spectral method (evaluated asFTAN’s results minus the results of the spectral method). Slashed cells mean that the number of measurements atthat distance range and that period is too small for statistical comparison.

Period (s) Mean misfit (m s–1) Standard deviation (m s–1)All r ≥ 3λ 2λ ≤ r < 3λ λ ≤ r < 2λ All r ≥ 3λ 2λ ≤ r < 3λ λ ≤ r < 2λ

10–50 0.04 0.12 −1.96 −1.8 9.1 8.34 18.67 20.8410 0.02 / / / 3.95 / / /15 −0.14 −0.14 −0.86 / 4.13 4.01 17.76 /20 0.16 0.18 −0.8 −1.76 5.87 5.6 16.42 19.1225 0.26 0.32 −1.53 −1.68 9.0 8.47 18.48 19.4730 −0.10 −0.2 2.08 −1.82 11.62 10.91 20.31 20.8335 0.00 0.14 −2.01 −2.09 12.46 11.54 19.74 22.6640 −0.1 0.08 −2.01 −1.98 12.34 11.17 18.61 22.4445 −0.13 −0.17 2.42 1.72 11.29 9.52 17.5 19.6650 0.03 0.28 −1.97 −0.94 9.73 8.45 14.13 14.65

Dow




ay 2019


Figure 5. Histograms of the phase velocity differences Vdif at 10 s periodamong station-triples at three categories indicated at the bottom of eachpanel. All the phase velocity data are measured by FTAN.

Figure 6. Same as Fig. 5, but for phase velocity measurements at 25 speriod.

Dow




ay 2019

658 Y. Luo et al.

Table 2. Distribution of differences of the velocity and travel time between station pair a–c and those calculated by combining the two measurementsof a to b and b to c. The data used here are measured by FTAN.

Mean misfit Standard deviationThe distance ranges of the three Vr difference Time difference Time difference Vr difference Time difference Time differencestation pairs: a–b, b–c and a–c. (m s–1) (s) (per cent) (m s–1) (s) (per cent)

All three pairs: r ≥ 3λ −2.03 0.2 0.06 10.89 0.89 0.31Two pairs: r ≥ 3λ −2.55 0.2 0.07 15.42 0.92 0.43One pair: 2λ ≤ r < 3λ

Two pairs: r ≥ 3λ −1.87 0.14 0.05 17.48 0.94 0.47One pair: λ ≤ r < 2λ

Table 3. Same as Table 2, but for the data measured by the spectral method.

Mean misfit Standard deviationThe distance ranges of the three Vr difference Time difference Time difference Vr difference Time difference Time differencestation pairs: a–b, b–c and a–c. (m s–1) (s) (per cent) (m s–1) (s) (per cent)

All three pairs: r ≥ 3λ −2.7 0.25 0.08 11.17 0.96 0.32Two pairs: r ≥ 3λ −4.27 0.29 0.12 15.87 0.97 0.44One pair: 2λ ≤ r < 3λ

Two pairs: r ≥ 3λ −6.63 0.41 0.18 17.19 0.97 0.46One pair: λ ≤ r < 2λ

time-domain CCFs are consistent with those with long separations,we select a series of station-triples, which have three stations almostexactly aligning in a common great-circle path as Lin et al. (2008)and Ekstrom (2013) did before. Considering three stations in astation-triple named as a, b and c from one end of the great-circlepath towards the other end, we have three station pairs of a–b, b–cand a–c with their corresponding separations and phase velocitiesnamed as Dab, Dbc, Dac and Vab, Vbc, Vac, respectively. Two strictcriteria are adopted to ensure the three stations almost align in acommon great-circle path: (1) the interstation distances between thethree stations pairs need to satisfy Dab + Dbc − Dac < 0.1 km;and (2) the corresponding differences of azimuth angles betweenthe three paths need to be smaller than 1o.

If dispersion measurements from CCFs with short paths and longpaths are consistent with each other, for a station-triple, the phasevelocity Vac for a surface wave propagating from station a to cshould be almost the same as the combined phase velocity Vac′,defined as V ac′ = (Dab + Dbc)/(Dab/V ab + Dbc/V bc), for asurface wave propagating from a to b first and then from b to c.Thus, to evaluate the consistence, we examine the velocity dif-ference between Vac and the combined velocity Vac′, defined asV di f = V ac′ − V ac, and also the travel time difference betweenTac and the combined travel time adding Tab and Tbc, definedas T di f = (Dab/V ab + Dbc/V bc) − T ac. In order to normal-ize time difference for different periods and paths with differentdistances, we also calculate the percentage of traveltime differenceas T pdi f = T di f/T ac × 100 per cent. We examine station-triplesbelonging to three categories defined based on the station-pair sep-aration r and surface wave wavelength λ: Category I: all the threepairs with r > 3λ; Category II: one pair either a–b or b–c with 2λ

≤ r < 3λ and the rest two with r > 3λ; Category III: one pair eithera–b or b–c with λ ≤ r < 2λ and the rest two with r > 3λ.

The histograms of Vdif among the station-triples belonging toeach of the three categories are plotted separately in Figs 5 and 6for 10 and 25 s period surface waves respectively. The histogramsof Vdif for combined data including all the measurements at nineperiods, explicitly indicated at Table 2, are plotted in Fig. 7. Thedistributions of Vdif are nearly following a Gaussian distribution.Thus, we calculate the means and the standard deviations of thedifferences, which are shown at the top right corner of each panel.

Table 2 summarizes the comparison of Vdif, Tdif and Tpdif amongthe three categories of station-triples for the combined data of thenine periods.

Table 2 shows that the average means of Vdif, Tdif and Tpdif are∼2 m s−1, ∼0.2 s and ∼0.06 per cent, almost close to zero, andthe average standard deviations of them are ∼16 m s−1, ∼0.9 s and∼0.4 per cent. There are no significant variations of the means andstandard deviations of Vdif, Tdif and Tpdif among the three cate-gories of station-triples. We also calculate the differences for phasevelocity measurements based on the revised spectral method of Luoet al. (2012), which are presented in Table 3. Similar differencesof the measurements based on the revised spectral method are ob-served with those based on FTAN. Tables 2 and 3 and Figs 5–7confirm that phase velocity measurements made by FTAN and therevised spectral method from short-separation time-domain CCFs,as short as one wavelength, are consistent with those from long-separation ones. These results are consistent with those of Ekstromet al. (2009) and Ekstrom (2013).

5 COMPARISON OF PHASE VELOCITY MAPS

Base on the comparison of phase velocity measurements made byFTAN and the revised spectral method and the three-station methodpresented in the two preceding sections, we can conclude now thatsurface wave dispersion measurements by FTAN from time-domainCCFs are consistent with those measured by the spectral method,and dispersion measurements from short-path CCFs are as reliableand accurate as those from long-path CCFs. Because dispersionmeasurements from CCFs are finally used to generate phase velocitymaps by ANT, we investigate whether phase velocity maps derivedsolely from short-path dispersion measurements are consistent withthose solely based on long-path ones.

We divide dispersion measurements made by FTAN into two datasets according to their separations r relative to their surface wave-length λ: one with λ ≤ r < 3λ, and the other with 3λ ≤ r < 5λ.We perform tomography for both data sets using a tomographymethod of Barmin et al. (2001). The reason why we do not includedispersion measurements with distances longer than five wave-lengths is that long-path surface waves may be slightly influenced by

Dow




ay 2019


Figure 7. Same as Fig. 5, but for phase velocity measurements combing allthe data at 10–50 s period range.

no-straight-ray behaviour (Ekstrom 2013), such as the off-great-circle propagation. We do not include measurements with intersta-tion distances shorter than one wavelength because the number ofavailable measurements at that short distance is very small, and

the uncertainties of measurements are large as also mentioned inSection 3.

The tomography method of Barmin et al. (2001) has been ap-plied for numerous ANT studies before (e.g. Moschetti et al. 2007;Bensen et al. 2008; Zheng et al. 2011). The details of this methodare documented in Barmin et al. (2001). Here we choose an areaextending from –113◦E to –96◦E and from 32◦N to 48◦N and pa-rameterize it to a 0.5◦ × 0.5◦ grid for tomography. We use sameregularization parameters in the tomography for the two data sets,which implies that any differences between the two sets of tomog-raphy results are purely resulting from the differences of data.

Fig. 8 shows the two sets of resulting phase velocity maps at25, 35 and 50 s periods based on the long-path data set (left-handcolumn) and the short-path one (middle column), respectively. Thehistograms of corresponding differences between the two sets ofphase velocity maps are also plotted in the right-hand column ofFig. 8. We do not show the comparison of phase velocity mapsat periods shorter than 25 s because the number of paths withdistances shorter than three wavelengths at this short-period rangeis too small for tomography. As shown by Fig. 8, the two data setsgenerate almost identical phase velocity maps with the means ofvelocity differences ranging from ∼2 to ∼7 m s−1 and the standarddeviations from ∼13 to ∼19 m s−1 at different periods.

These small differences between the two sets of phase velocitymaps again affirm that phase velocity dispersion measurements byFTAN from short-path time-domain CCFs are consistent with thosefrom long-path ones, and short-path dispersion measurements withdistance as short as one wavelength can be retained for ANT toincrease path coverage and subsequently improve resolution.

6 D I S C U S S I O N A N D C O N C LU S I O N S

This study has demonstrated that surface wave phase velocitydispersion curves measured by FTAN from time-domain cross-correlations of ambient noise is consistent with those measuredby the zero crossing of the real CSFs, and the dispersion measure-ments at short interstation distances, as short as one wavelength, areconsistent with and also reliable as those at distances longer thanthree wavelengths. These short-path dispersion measurements canbe included in ANT, which is typically discarded in previous ANTstudies. Short-path dispersion measurements are extremely usefulto constrain small-scale velocity anomalies in local and regionalANT (Ekstrom 2013).

The inclusion of short-path dispersion measurements in ANThelps to extend the range of the long-period end in ANT when theextent of a seismic array is limited. For instance, if a seismic arraycovers a region with an aperture of 300 km, the conventional require-ment of three-wavelength inter-station distance limits the longestperiod of tomography to ∼30 s. If we include the short-path datalike one wavelength, surface wave tomography could be extendedup to periods as long as 60/70 s, which provides constraints on up-permost mantle structures. Recently, there are a few studies showingthat long-period surface waves at periods up to 200/300 s can be ex-tracted from cross-correlations of ambient noise (e.g. Shen & Zhang2012; Yang 2014). The relaxation of the two/three wavelength lim-itation on interstation distances to one wavelength allows long-period surface waves from ambient noise to be included in regionaltomography, enabling ANT to image upper mantle structures whichis mostly constrained by earthquake surface wave data before.

The inclusion of short-path dispersion measurements from am-bient noise could also help to better resolve azimuthal anisotropy intomography. It is well known there is trade-off between azimuthal

Dow




ay 2019

660 Y. Luo et al.

Figure 8. Phase velocity maps at 25, 35 and 50 s periods generated using long-path dispersion measurements with 3λ ≤ r < 5λ (left-hand column) andshort-path dispersion measurements with λ ≤ r < 3λ (middle column) as well as the corresponding histograms of the differences between the two sets of maps(right-hand column). The numbers of paths used in tomography are indicated on the top of each phase velocity map.

anisotropy and lateral heterogeneities in surface wave tomography.The better constraints on small-scale heterogeneities and the betterlateral and azimuthal coverage obtained by including more short-separation paths are helpful to better resolve the trade-off. However,Harmon et al. (2010) have demonstrated that variations in phase

velocities with azimuth resulting from inhomogeneous distributionof noise sources could yield up to 1 per cent apparent peak-to-peakazimuthal anisotropy. As we mention in the preceding section, theuncertainties of phase velocity measurements increase slightly withthe decreasing interstation distances, the questions of whether and

Dow




ay 2019


how the inclusion of short-path dispersion measurements in ANTeventually improves the inversion of azimuthal anisotropy is stillneeded to be further investigated in future studies.

A C K N OW L E D G E M E N T S

Waveform data used in this study are obtained from the IRIS DataManagement System, and specifically the IRIS Data ManagementCenter. YL’s work is supported by the National Science Founda-tion of China (NSFC, #41374059), Seismic Professional ScienceFoundation (2014419013), and the Special Fund for Basic Scien-tific Research of Central Colleges, China University of Geosciences(Wuhan) (#CUG090106 and #CUGL100402). YY is supported byAustralian Research Council Discovery grants (DP120102372 andDP120103673) and Future Fellowship (FT130101220). This is con-tribution 579 from the ARC Centre of Excellence for Core to CrustFluid Systems (http://www.ccfs.mq.edu.au) and 986 in the GEMOCKey Centre (http://www.gemoc.mq.edu.au).

R E F E R E N C E S

Arroucau, P., Rawlinson, N. & Sambridge, M., 2010. New insight intoCainozoic sedimentary basins and Palaeozoic suture zones in southeastAustralia from ambient noise surface wave tomography, Geophys. Res.Lett., 37, L07303, doi:10.1029/2009GL041974.

Barmin, M.P., Levshin, A.L. & Ritzwoller, M.H., 2001. A fast and reliablemethod for surface wave tomography, Pure appl. Geophys., 158, 1351–1375.

Bensen, G.D., Ritzwoller, M.H., Barmin, M.P., Levshin, A.L., Lin, F.,Moschetti, M.P., Shapiro, N.M. & Yang, Y., 2007. Processing seismicambient noise data to obtain reliable broad-band surface wave dispersionmeasurements, Geophys. J. Int., 169, 1239–1260.

Bensen, G.D., Ritzwoller, M.H. & Shapiro, N.M., 2008. Broad-band ambientnoise surface wave tomography across the United States, J. geophys. Res.,113, B05306, doi:10.1029/2007JB005248.

Boschi, L., Weemstra, C., Verbeke, J., Ekstrom, G., Zunino, A. & Giardini,D., 2013. On measuring surface wave phase velocity from station–stationcross-correlation of ambient signal, Geophys. J. Int., 192, 346–358.

Dziewonski, A.M., Bloch, S. & Landisman, M., 1969. A technique for theanalysis of transient seismic signals, Bull. seism. Soc. Am., 59, 427–444.

Ekstrom, G., 2011. A global model of Love and Rayleigh surface wavedispersion and anisotropy, 25–250 s, Geophys. J. Int., 187(3), 1668–1686.

Ekstrom, G., 2013. Love and Rayleigh phase-velocity maps, 5–40 s, of thewestern and central USA from USArray data, Earth planet. Sci. Lett.,402, doi:10.1016/j.epsl.2013.11.022.

Ekstrom, G., Abers, G. & Webb, S., 2009. Determination of surface-wavephase velocities across USArray from noise and Aki’s spectral formula-tion, Geophys. Res. Lett., 36, L18301, doi:10.1029/2009GL039131.

Harmon, N., Rychert, C. & Gerstoft, P., 2010. Distribution of noise sourcesfor seismic interferometry, Geophys. J. Int., 183(3), 1470–1484.

Levshin, A.L. & Ritzwoller, M.H., 2001. Automated detection, extraction,and measurement of regional surface waves, Pure appl. Geophys., 158(8),1531–1545.

Levshin, A.L., Pisarenko, V.F. & Pogrebinsky, G.A., 1972. On a frequency–time analysis of oscillations, Ann. Geophys., 28, 211–218.

Li, H., Bernardi, F. & Michelini, A., 2010. Surface wave dispersion mea-surements from ambient seismic noise analysis in Italy, Geophys. J. Int.,180, 1242–1252.

Lin, F.-C., Moschetti, M.P. & Ritzwoller, M.H., 2008. Surface wave tomog-raphy of the western United States from ambient seismic noise: Rayleighand Love wave phase velocity maps, Geophys. J. Int., 173, 281–298.

Lin, F.-C., Ritzwoller, M.H. & Sneider, R., 2009. Eikonal tomography:surface wave tomography by phase front tracking across a regional broad-band seismic array, Geophys. J. Int., 177, 1091–1110.

Lobkis, O. & Weaver, R.L., 2001. On the emergence of the Green’s functionin the correlations of a diffuse field, J. acoust. Soc. Am. 110(6), 3011–3017.

Luo, Y., Xu, Y. & Yang, Y., 2012. Crustal structure beneath the Dabieorogenic belt from ambient noise tomography, Earth planet. Sci. Lett.,313–314, 12–22.

Mordret, A, Shapiro, N.M., Singh, S., Roux, P. & Barkved, O.I., 2013.Helmholtz tomography of ambient noise surface wave data to estimateScholte wave phase velocity at Valhall life of the field, Geophysics, 78(2),WA99–WA109.

Moschetti, M.P., Ritzwoller, M.H. & Shapiro, N.M., 2007. Surface wavetomography of the western United States from ambient seismic noise:Rayleigh wave group velocity maps, Geochem. Geophys. Geosys., 8,Q08010, doi:10.1029/2007GC001655.

Porritt, R.W., Allen, R.M., Boyarko, D.C. & Brudzinski, M.R., 2011. Inves-tigation of Cascadia segmentation with ambient noise tomography, Earthplanet. Sci. Lett., 309(1), 67–76.

Savage, M.K., Lin, F-C. & Townend, J., 2013. Ambient noise cross-correlation observations of fundamental and higher-mode Rayleigh wavepropagation governed by basement resonance, Geophys. Res. Lett.,40(14), 3556–3561.

Saygin, E. & Kennett, B.L.N., 2010. Ambient noise tomography for theAustralian Continent, Tectonophys, 481, 116–125.

Shapiro, N.M., Campillo, M., Stehly, L. & Ritzwoller, M.H., 2005. High-resolution surface wave tomography from ambient seismic noise, Science,307, 1615–1618.

Shen, Y. & Zhang, W., 2012. Full-wave ambient noise tomography of theEastern Hemisphere, in Proceedings of the 2012 IRIS Annual Meeting,Boise, Idaho, June 13–15, 2012.

Snieder, R., 2004. Extracting the Green’s function from the correlation ofcoda waves: a derivation based on stationary phase, Phys. Rev. E, 69,046610.

Takeo, A., Forsyth, D.W., Weeraratne, D.S. & Nishida, K., 2014. Estimationof azimuthal anisotropy in the NW Pacific from seismic ambient noise inseafloor records, Geophys. J. Int., 199(1), 11–22.

Tsai, V.C. & Moschetti, M.P., 2010. An explicit relationship between time-domain noise correlation and Spatial Autocorrelation (SPAC) results,Geophys. J. Int., 182, 454–460.

Verbeke, J., Boschi, L., Stehly, L., Kissling, E. & Michelini, A., 2012. High-resolution Rayleigh-wave velocity maps of central Europe from a denseambient-noise dataset, Geophys. J. Int., 188, 1173–1187.

Villasenor, A., Yang, Y., Ritzwoller, M.H. & Gallart, J., 2007. Ambi-ent noise surface wave tomography of the Iberian Peninsula: implica-tions for shallow seismic structure, Geophys. Res. Lett., 34, L11304,doi:10.1029/2007GL030164.

Yang, Y., 2014. Application of teleseismic long-period surface waves fromambient noise in regional surface wave tomography: a case study in west-ern USA, Geophys. J. Int., 198(3), 1644–1652.

Yang, Y., Ritzwoller, M.H., Levshin, A.L. & Shapiro, N.M., 2007. Ambientnoise Rayleigh wave tomography across Europe, Geophys. J. Int., 168(1),259–274.

Yao, H., van der Hilst, R.D. & de Hoop, M.V., 2006. Surface-wave ar-ray tomography in SE Tibet from ambient seismic noise and twosta-tion analysis – I. Phase velocity maps. Geophys. J. Int., 166(2), 732–744.

Yao, H., Beghein, C. & Van der Hilst, R.D., 2008. Surface-wave arraytomography in SE Tibet from ambient seismic noise and two-stationanalysis: II – crustal and upper mantle structure, Geophys. J. Int., 173(1),205–219.

Young, M.K., Rawlinson, N., Arroucau, P., Reading, A.M. & Tkalcic, H.,2011. High-frequency ambient noise tomography of southeast Australia:new constraints on Tasmania’s tectonic past, Geophys. Res. Lett., 38,L13313, doi:10.1029/2011GL047971.

Zheng, Y., Shen, W., Zhou, L., Yang, Y., Xie, Z. & Ritzwoller, M.H., 2011.Crust and uppermost mantle beneath the North China Craton, northeasternChina, and the Sea of Japan from ambient noise tomography, J. geophys.Res., 116, B12312, doi:10.1029/2011JB008637.

Zhou, L., Xie, J., Shen, W., Zheng, Y., Yang, Y., Shi, H. & Ritzwoller, M.H.,2012. The structure of the crust and uppermost mantle beneath SouthChina from ambient noise and earthquake tomography, Geophys. J. Int.,189(3), 1565–1583.

Dow




ay 2019

Date post:	25-Aug-2020
Category:	Documents
Upload:	others
View:	5 times
Download:	0 times

Geophysical Journal...

Documents